Perry Y. Li Department of Mechanical Engineering, University of Minnesota,

合集下载

MechanicalEngineering-PME-MSc

The programme includes analysis, design and implementation of solutions, using analytical models, computational methods and experimental work to reach new performance and understanding. With this focus on the ‘ultimate in mechanical engineering’ the program confronts students with the daunting conceptual and design challenges of developing (and utilising) tools for precision mechanical engineering. Although the emphasis is on high-tech equipment and instrumentation, the same knowledge and methodology applies to energy systems, medical equipment, automotive and aerospace design and many other fields of mechanical engineering, enabling these future engineers to address the needs of our modern society.
The subjects of the graduation projects of some of the past students are listed below. Topology optimization of Dynamic structures Determination of critical loads for aeroplane wing optimisation New finite element formulations for computational contact mechanics Design of an improved 6DOF positioning stage with nanometre stability for AFM measurements Development of Distributed on Demand Supply for Air Conveyor Systems Deformable gripper design using distributed actuators Static balancing of reconfigurable mechanisms AFM Hollow Cantilevers Integrated with Actuating Systems Polymeric temperature micro sensor for wound monitoring A non-linear identification tool for extracting mechanical properties of graphene sheets High-resolution on-demand growth of nanodiamond and carbon nanotubes using femtopipette

Design and demonstration of policy-based management in a multi-hop ad hoc network,” Ad Hoc

Design and demonstration of policy-based managementin a multi-hop ad hoc networkKaustubh S.Phanse*,Luiz A.DaSilva,Scott F.MidkiﬀBradley Department of Electrical and Computer Engineering,Virginia Polytechnic Institute and State University,Alexandria,VA22314,USAReceived16May2003;received in revised form4September2003;accepted24September2003Available online29November2003AbstractIn this paper,we propose a policy-based framework for the management of wireless ad hoc networks and brieﬂy describe a characteristics-based taxonomy that provides a platform to analyze and compare diﬀerent architectural choices.We develop a solution suite that helps achieve our goal of a self-organizing,robust and eﬃcient management system.One of the main contributions of this work is the prototype implementation and testing of the mechanisms and protocols comprising our framework in a multi-hop ad hoc network environment.Experiments are conducted using both an emulated ad hoc network testbed and a true wireless testbed.Degradation in management system performance is observed as the number of hops between a policy server and client increases.Our proposed k-hop clustering algorithm alleviates this problem by limiting the number of hops between a server and client.We demonstrate the operation of our prototype implementation,illustrating QoS management in a multi-domain ad hoc network environment using the proposed cluster management,redirection,and policy negotiation mechanisms.Ó2003Elsevier B.V.All rights reserved.Keywords:Policy-based networking;Resource management;Service discovery;Ad hoc networks1.IntroductionMobile ad hoc networks are characterized by dynamic topologies,bandwidth-constrained vari-able capacity links,limited physical security and survivability,and nodes with limited battery life, processing power and storage capacity.These characteristics pose signiﬁcant challenges to the management of such networks.Policy-Based Network Management(PBNM)is one network management approach that has met with considerable interest in the networking com-munity[1,2].Unlike legacy network management, which generally involves conﬁguring and manag-ing each network entity individually,PBNM con-ﬁgures and controls the network as a whole, providing the network operator with simpliﬁed, logically centralized and automated control over the entire network.PBNM can simplify adminis-tration of complex operational characteristics of a network,including Quality of Service(QoS),ac-cess control,network security,and IP address allocation.However,so far,the work on policy-based network management[1–4]has focused on largeﬁxed networks such as enterprise networks,*Corresponding author.E-mail address:kphanse@(K.S.Phanse).1570-8705/$-see front matterÓ2003Elsevier B.V.All rights reserved.doi:10.1016/j.adhoc.2003.09.013Ad Hoc Networks3(2005)389–401/locate/adhoccontent provider networks,and Internet service provider(ISP)networks.In this paper,we discuss how to extend and apply the policy-based approach for managing QoS in ad hoc networks.QoS requires mechanisms to support service diﬀerentiation as well as mech-anisms for QoS management.The former concerns itself with how to achieve performance objectives of heterogeneousﬂows;the latter concerns itself with who should be entitled to preferential treat-ment by the network.QoS architectures such as SWAN[5],INSIGNIA[6]and dRSVP[7]address the means for service diﬀerentiation,including marking,classiﬁcation,scheduling and others,and often assume some external mechanism such as pricing will determine who should have access to what level of service.However,the control struc-ture(in support of authentication,authorization, and dynamically changing policies)required for QoS provisioning and management in ad hoc networks is not yet well understood,and is the focus of our research.The dynamic nature of ad hoc networks calls for a control structure that al-lows automated QoS management and one that supports dynamic admission control or bandwidth allocation based on diﬀerent policies such as bandwidth availability,traﬃc ownership(e.g.,the identity of the user,application,or organization from which the traﬃc originates),and temporal elements(e.g.,time of day).A policy-based approach addresses most of the key requirements of an ad hoc network manage-ment system,namely automation,self-organizing capability,robustness and eﬃciency(for a survey of related work and detailed discussion of how these requirements are met by a policy-based management system,the reader is referred to[8]). The fundamental challenge in extending the policy-based approach to ad hoc networks is to adapt this conceptually centralized approach to a distributed, infrastructure-independent environment.In this paper,we describe a PBNM framework for ad hoc networks consisting of a suite of solu-tions that together help meet the challenges out-lined in[8].The key components of this solution suite are automated service discovery,k-hop cluster-ing,and Dynamic Service Redundancy(DynaSeR).We also address policy inter-operability issues in a multi-domain ad hoc inter-network––formed by a multi-organization consortium,such as the US NavyÕs Coalition Wide Area Network (C-WAN)[9].We propose and demonstrate a signaling mechanism for inter-domain policy negotiation that makes seamless QoS feasible in such networks.Unlike many research eﬀorts in theﬁeld of ad hoc networks that are based solely on simulations, we implement and test our management scheme in a multi-hop ad hoc network testbed.We believe that the challenges of operating in an actual ad hoc network environment are not always exposed in a simulation environment or through theoretical analysis;experimental evaluation of proposed solutions is critical.Prior experiences of research-ers(e.g.,[10,11])support our concerns.We implement a prototype of our PBNM system and illustrate its operation in a multi-hop ad hoc net-work testbed.2.Policy-based management framework for wireless ad hoc networksAn automated,intelligent,eﬃcient and robust management structure is needed to manage ad hoc networks.In this section,we describe the various modules that constitute the framework.In partic-ular,we focus on the policy architectural and distribution,resource discovery and policy provi-sioning aspects of the framework;the underlying techniques we propose––k-hop cluster manage-ment,the DynaSeR solution,service discovery and signaling for inter-domain policy negotiation––are presented.Using a systems approach,we decompose the framework into functional blocks.This approach highlights the inter-dependencies among the vari-ous components and the complex functional tasks that need to be carried out by a management system.A comprehensive representation of a management system is lacking in most prior pub-lished research(which focuses mainly on network monitoring,e.g.,[12,13])and is one of the contri-butions of this work.390K.S.Phanse et al./Ad Hoc Networks3(2005)389–401The seven key modules that constitute the framework are as follows.2.1.Policy speciﬁcationThe policy speciﬁcation is a mapping of the overall network goals(e.g.,QoS speciﬁcation)into network-wide policies.Typically,the high-level policies are reasonably static,while lower-level policies may change according to network utiliza-tion or time of day.2.2.Policy architecture and distribution2.2.1.Types of architecturesWe have proposed a characteristics-based tax-onomy of policy architectures in[14].Here we provide a brief overview of the taxonomy.The taxonomy is based on four characteristics:locus of control,locus of information,policy distribution mode,and tiers of control.The policy architecture taxonomy is summarized in Table1.The taxonomy broadly classiﬁes the various architectures into three categories based on the policy distribution model used.These categories are:outsourced(all policy decisions are outsourced by a client to a remote server),provisioned(clients are conﬁgured to make policy decisions locally), and hybrid(combination of the outsourcing and provisioning models)architectures,which are then further classiﬁed as shown in Table1.The taxonomy provides a systematic way to analyze and compare the applicability of one or more architectures to the network environment of interest.We have identiﬁed the hybrid architecture as the most promising for implementing our management framework in an ad hoc network environment[14].The provisioning of clients al-lows them to make decisions locally minimizing overhead,while outsourcing provides support for dynamic policies and inter-domain mobility.2.2.2.Protocol for policy distributionSeveral mechanisms exist for policy distribution in a network[1]:using a command-line script, using management frameworks(e.g.,based on CORBA),using web servers,and using protocols such as Common Open Policy Service(COPS), Simple Network Management Protocol(SNMP), or Lightweight Directory Access Protocol (LDAP).We choose the COPS for PRovisioning(COPS-PR)[15],an extension of the COPS protocol[16], for policy distribution.COPS-PR integrates the outsourcing and provisioning models,thus allow-ing theﬂexibility to support a hybrid architecture.Some of the features that make COPS-PR a promising choice are:event-driven control(i.e., there is no polling)and asynchronous notiﬁcation, structured row-level access and atomic transac-tional model,support for fault tolerance and security mechanisms,and reliable transport usingTable1Policy architecture taxonomy matrixFeaturesArchitectures Locus of control Locus of informa-tionPolicy distribution Tiers of controlOutsourced CCO Centralized Centralized Outsourcing1DCO Distributed Centralized Outsourcing1 Provisioned DDO Distributed Distributed Outsourcing1DDP Distributed Distributed Provisioning2DDP-hierarchical Distributed Distributed Provisioning>2Hybrid DDOP Distributed Distributed Outsourcing andprovisioning2DDOP-hierarchical Distributed Distributed Outsourcing andprovisioning >2K.S.Phanse et al./Ad Hoc Networks3(2005)389–401391persistent TCP connections.COPS-PR may also co-exist and interact with other management pro-tocols such as SNMP.2.2.3.Automated and self-organizing control structureA self-organizing and automated management system is the key for eﬀective management in an ad hoc network environment.We propose a suite of solutions and techniques that address these requirements.The components of this solution suite include k-hop cluster management,service discovery,and dynamic service redundancy.During initial deployment of a wireless ad hoc network,we assume that a certain number of policy servers are present in the network and ini-tially serve as cluster heads.Other nodes can eventually become PDPs(and,hence,cluster heads)through the process of delegation,de-scribed below.All the policy clients within k-hops from a server are eligible for service from that policy server.Each policy server along with its clients forms a cluster,as shown in Fig.1.Whenever a proactive ad hoc routing protocol is adopted,the k-hop cluster management can be implemented by taking advantage of the routing information available with the routing daemon. Using this topology information,the policy servers can track nodes moving in and out of the clusters with minimal additional clustering protocol over-head.We implement the proposed method by interfacing our PBNM application with the underlying Optimized Link State Routing(OLSR) protocol daemon(see Section4).Due to deployment of an insuﬃcient number of policy servers or due to node mobility,it is possi-ble that one or more nodes may not be within k-hops of any of the policy servers.To increase policy-based service availability and to allow pol-icy servers to eﬃciently keep track of network nodes as they move,we further enhance our clus-tering algorithm using what we call Dynamic Ser-vice Redundancy(DynaSeR).The DynaSeR solution consists of two techniques:redirection and delegation.When a client moves out of a cluster, the server managing the client gathers relevant topology information(from the routing daemon) to detect whether the client is within k-hops of any other servers.If it is,then it redirects the client to the appropriate server.The client is now a part of the new cluster managed by the‘‘Redirected PDP.’’This can be termed as a server-centric way of implementing cluster management.If a policy client is not within k hops from any of the existing servers,an existing policy server delegates an appropriate network node to assume serving responsibilities for that client.For both delegation and redirection,the decision of which node should act as a server may be based on factors such as connectivity,processing load,bandwidth avail-ability,and remaining battery life.Such‘‘dele-gated’’servers or policy decision points(PDPs) form another tier of control below the super-PDPs,creating a hierarchical control architecture. Simulation results reported by us in[17]indicate that this delegation scheme considerably improves the service coverage of the PBNM system,while allowing the use of smaller cluster sizes.We propose and implement a lightweight ser-vice discovery mechanism to facilitate automated discovery of policy servers in the network.Two types of messages are used:Service Advertisement (SA)and Client Service Request(CSRQ).A policy server periodically advertises itself via a limited k-hop broadcast of the SA message.A client that does not receive an SA message within a certain time interval broadcasts a CSRQ message.The server that may have moved within k-hops of the client responds with a unicast SA message.Alter-392K.S.Phanse et al./Ad Hoc Networks3(2005)389–401natively,a client node that is currently being ser-viced,upon hearing a CSRQ message,may vol-unteer to act as a delegated server.The main motivation for this type of clustering is to limit the number of wireless hops between a client and a server.As will be seen from our results (Sections3.1and3.2),this considerably improves the performance of the management system––keeping the policy response time low and reducing the unpredictability in the performance.Further, fewer hops between a client and a server means fewer resources(e.g.,bandwidth and battery life) are used at intermediate nodes for forwarding control messages.The trade-oﬀis generally in the poorer service coverage for smaller cluster sizes k [17].2.3.Resource discoveryA policy-based management system translates high-level policies(Policy Speciﬁcation)into de-vice-speciﬁc conﬁguration to dictate the use of network resources.To achieve this,the manage-ment framework must be aware of the available network resources.This includes active network devices and their capabilities,network topology, bandwidth utilization,etc.This is even more crit-ical in ad hoc networks,where it is crucial for the policy system to keep updated knowledge about the dynamic network topology.Resource discov-ery calls for additional signaling and/or computa-tion;thus,the trade-oﬀis generally between eﬃciency(minimal signaling)and accuracy of the information maintained by the management sys-tem.2.4.Policy provisioningPolicy provisioning can be viewed as the phase after policies are distributed,consisting of install-ing and implementing the policies using device speciﬁc mechanisms(e.g.,marking,classiﬁcation and queuing).Thus,policy provisioning directly aﬀects the way in which the various traﬃcﬂows in the network are treated.In our implementation, we use a DiﬀServ-like architecture to provision QoS policies.2.5.Policy-based routingA routing approach that honors the deﬁned network policies is called policy-based routing [18,19].These policies may involve end-users, temporal policies,access control,resource alloca-tion,etc.While policy-based routing has been studied and deployed extensively in wireline net-works,its applicability to ad hoc networks is open for further investigation.2.6.Policy monitoringTo provide robust management of a network, an independent policy monitoring process must ensure that the network meets the high-level goals or speciﬁcations.Policy monitoring can be achieved using active(e.g.,dummy transactions or sending probe packets)or passive(e.g.,measure-ment-based estimation)methods[1].2.7.Adaptation logicGiven the dynamic nature of ad hoc networks, it is necessary for a policy system to incorporate dynamic and state-dependent policies that allow the control structure to adapt to the current state of the ing feedback(e.g.,policy monitoring)and resource discovery mechanisms,a management system can make intelligent decisions and adapt to the changing network environment. Further discussion of this module is outside the scope of this paper.3.Experimental evaluationIn[8],we presented our preliminary experi-mental results to validate our qualitative analysis of policy architectures based on our taxonomy. However,those experiments involved single-hop communications.In an ad hoc network,the number of hops between a policy client and a policy server may change over time and it is important to characterize the performance of the management system as a function of the number of hops.K.S.Phanse et al./Ad Hoc Networks3(2005)389–401393In this section,we describe the experiments conducted in a multi-hop ad hoc network testbed,and present our results.We are interested in characterizing the eﬀect of available bandwidth on management system performance,in particular under constrained bandwidth availability typical of many wireless ad hoc networks.We used the Intel âCOPS client software development kit (SDK)[20]to implement a COPS-based management system.Dummy policy re-quests were generated by running a script at each policy enforcement point (PEP).The metrics used in this set of experiments are the policy response time and the inter-decision time.The policy response time is the diﬀerence between the time at which the policy request was sent by a PEP and the time at which the corre-sponding policy decision was received at the PEP.The inter-decision time is the diﬀerence between consecutive decisions received by the PEP.These are illustrated in Fig.2.3.1.Multi-hop ad hoc network (wired testbed)We emulate diﬀerent ad hoc network topologies using a software developed at Virginia Tech [21]that allows emulation of dynamic topologies,variable packet drop rates and low bandwidth.We implemented a CCO architecture (see Table 1).Intermediate nodes used the Optimized Link State Routing protocol daemon (olsrd)[22,23]to route packets between the policy server (PDP)and client (PEP).A low end-to-end bandwidth of 64kb/s was emulated using the Diﬀserv on Linux tool [24].We measured thepolicy response time and the inter-decision time as a function of the number of hops between the PEP and PDP.Results were collected using multiple iterations for each case;95%conﬁdence intervals were calculated using the method of batch means [25].As seen in Fig.3,the policy response time and inter-decision time increase exponentially with the number of hops,indicating the desirability of an upper bound on the number of hops between a policy client and server.These results largely motivated our proposal of a k -hop clustering algorithm (Section 2)that controls the number of hops between a PEP and PDP.The conﬁdence intervals for the two metrics are shown in Table2.394K.S.Phanse et al./Ad Hoc Networks 3(2005)389–4013.2.Multi-hop ad hoc network (wireless testbed)To gain insight into the performance of a pol-icy-based management system in an actual multi-hop wireless ad hoc network and as a proof of concept,we ported our experiments to a wireless ptops with IEEE 802.11b wireless PC cards were used in this set of experiments.The wireless cards are based on the Intersil Prism chipset,which supports transmitter power control.Fig.4shows the placement of laptops in our workarea.A 4-hop wireless network topology is illus-trated.To set up multi-hop topologies in a small work area,we reduced the transmitter power of the wireless cards using the Wireless Tools for Linux [26]package.In addition,we used a crude,but eﬀective,way to further reduce the transmitter power.We wrapped the antenna portion of the wireless cards with aluminum foil acting as an attenuator as shown in Fig.5.In an otherwise static setup,intermittent loss of routes between end nodes was observed,especially with the increase in the number of hops.Clausen et al.[10]reported a similar problem when evalu-ating the OLSR protocol using an experimental testbed.Further investigation indicated that the transmissions of control packets by the OLSR daemon at two or more nodes became synchro-nized resulting in packet collisions.To alleviate this problem,we have implemented random jitter (in the range suggested in [22])by modifying the current OLSR daemon.Results from this set of experiments are shown in Table 3;the policy response time with the con-ﬁdence intervals is plotted in Fig.6.Increasing the number of hops resulted in considerable increaseinyout of the area where we conduct our wireless experiments.Placement of nodes for a 4-hop wireless ad hoc network topology is shown.Nodes A and E are the policy server and client,respectively.Fig.5.Antenna portion of the wireless PC card wrapped with aluminum foil ‘‘attenuator.’’Table 3Policy response time and inter-decision time vs.number of hops (wireless testbed)Number of hops 95%conﬁdence interval Policy response time (ms)Inter-decision time (ms)1425.026±1.34215.452±0.00082938.346±12.72532.629±0.039931374.268±763.49458.817±5.652845181.985±17221.85165.466±20.717Table 2Policy response time and inter-decision time vs.number of hops (wired testbed)Number of hops 95%conﬁdence interval Policy response time (ms)Inter-decision time (ms)1399.082±0.167417.156±0.00032743.591±1.79831.308±0.003231340.473±15.58457.503±0.034744595.428±601.453132.537±0.7254K.S.Phanse et al./Ad Hoc Networks 3(2005)389–401395the response time and inter-decision time.The policy response time results collected after numer-ous runs exhibited high variance for the3-hop and 4-hop scenarios.This indicates that the average values in these two cases are not a good indicator of system performance.A look at the instantaneous values revealed intermittent occurrence of large spikes.For example,values as high as5and15s were observed for policy response time in the3-hop and4-hop scenarios,respectively.The increase in the response times and the unpredictability of the system with an increase in the number of hops bolsters our proposal of using k-hop cluster management to control the number of hops between a policy server and client in a wireless ad hoc environment.4.Policy-based QoS managementIn the previous section,we discussed our experimental results that characterized the policy management architecture performance in a static multi-hop ad hoc network environment.In this section,we brieﬂy describe implementation of our proposed schemes and illustrate our PBNM sys-tem at work in a multi-domain mobile ad hoc network.4.1.ImplementationTo implement our proposed schemes,we used the open source COPS API made available by Telia Research[27].We implemented a policy server and client to incorporate our proposed cluster management,redirection,delegation and service discovery mechanisms.In addition,we propose and implement an extension to the COPS-PR protocol,for inter-PDP communication and policy negotiation.This is of particular importance in a multi-domain network environ-ment[9].•Inter-domain policy negotiation:In a wireless mobile ad hoc network formed by a consortium of diﬀerent organizations,nodes may move across domains1administered by the policies of each individual organization.In general,a nodeÕs movement into a foreign domain may have several implications on its operation and performance.From a QoS perspective,if the foreign domain does not have policies to handlea particular‘‘visiting’’node,the service guaran-tees enjoyed by that node may degrade consid-erably.Speciﬁcally,time-sensitive mission critical data and real-time applications may be rendered impractical.To account for mobility of nodes across do-mains,we deﬁne a new object called the‘‘Home PDP Address’’in the COPS protocol.The for-mat of the‘‘Home PDP Address’’object is similar to the‘‘Last PDP Address’’object de-ﬁned in the COPS protocol standard[16].A client embeds both these objects in the COPS OPEN message sent to the server for estab-lishing a new COPS connection.The policy negotiation signaling is shown in Fig.7.When a ‘‘visiting’’client(PEP H)establishes a COPS connection with a policy server(PDP F)in a foreign domain,the server searches for policies for that client.If it does notﬁnd any relevant policies in its Policy Information Base(PIB),it gathers the address of the clientÕs home domain policy server(PDP H)from the‘‘Home PDP Address’’object.PDP F then acts as a client (with a new‘‘COPS Negotiation’’client-type) and establishes a COPS connection with PDP H.1We deﬁne the administrative domain to which a nodebelongs as its‘‘home’’domain,while any other domain isreferred to as a‘‘foreign’’domain.396K.S.Phanse et al./Ad Hoc Networks3(2005)389–401It then sends a COPS request(REQ)message to PDP H to download policies relevant to the ‘‘alien’’client(PEP H).PDP F then adapts itspolicies to reﬂect the service level agreement between the domains.For policy provisioning,we use a partial implementation of the Diﬀerentiated Services PIB[28]that provides a simple if-then mapping between the node/domain addresses and appli-cationﬂows based on source and destination ports and the corresponding policies for band-width allocation.•Integration with OLSR:The olsrquery tool,a part of the INRIA OLSR implementation[23],allows a user to access the OLSR routingtable maintained at the various network nodes.We modiﬁed the olsrquery tool to direct its out-put to a textﬁle in a desired format.When a policy server is initialized,a separate thread is created and dedicated to interact with the underlying olsrd routing daemon.Every10 s,the thread calls a function to execute the olsrquery command that generates an output ﬁle containing the required topology informa-tion(routing information gathered from the policy servers in the network).The policy server then stores(or updates)the information from theﬁle in a linked list,and uses it for cluster management.Fig.8shows a snapshot of apolicy server exhibiting updated topology information maintained by it.•Cluster management:We discussed our k-hop clustering algorithm in Section2.Here,we de-scribe its implementation and deployment using the COPS protocol.As mentioned earlier,a pol-icy server maintains topology information rele-vant to its clients.When‘‘signiﬁcant’’topology changes occur,i.e.,one or more clients move out of the k-hopcluster,the policy server is alerted about these changes,and functions for accessing the topol-ogy information(from the linked list)and, hence,for cluster management are invoked.Currently we have implemented the redirection mechanism assuming that a client that moves out of the k-hop cluster of its policy server, moves within k-hops of at least one other policy server.The delegation mechanism,applicable toa more general case of node movements,is partof our ongoing research.The COPS protocol has some inherent sup-port for the redirection mechanism.We take advantage of the‘‘PDP Redirect Address’’ob-ject in the COPS‘‘Client-Close’’message deﬁned in the COPS protocol standard[16].Whenevera er-interface of a policy server showing topology information gathered from underlying OLSR routing daemon, and implementation of1-hop cluster management.K.S.Phanse et al./Ad Hoc Networks3(2005)389–401397。

Design and test of a high-performance piezoelectric micropump for drug delivery

College of Mechanical Science and Engineering, Jilin University, Renmin Road 142, Changchun 130025, China Received 26 June 2004; received in revised form 3 December 2004; accepted 4 December 2004 Available online 21 January 2005
Keywords: Piezoelectric micropump; PZT actuator; Drug delivery; Cantilever valve; Natural frequency
1. Introduction
Most drugs have a range of concentrations of greatest efﬁcacy in the body, above which they are toxic and below which they have no therapeutic beneﬁt [1]. Conventional drug delivery routes such as oral tablets or injections are not easily able to control the rate of drug delivery or the target area of the drug. Consequently, initial concentration of the drug in the blood peaks above the level of toxicity and then gradually decreases over time to an ineffective level and the patients have to take the drug frequently [2,3]. In order to control drug release better, drug delivery systems (DDS) are necessary. The general advantages of a dominated DDS are the ability for the drug to act directly when needed and not at any ﬁxed time or deliver a localized dosage to reduce the side effects of medication [4,5]. Moreover, with the help of an

纳米颗粒尺寸、形状以及界面效应对介电和击穿场强等因素的影响

Society Chem.Mater.2010,22,1567–15781567DOI:10.1021/cm902852hNanoparticle,Size,Shape,and Interfacial Effects on Leakage Current Density,Permittivity,and Breakdown Strength of MetalOxide-Polyolefin Nanocomposites:Experiment and TheoryNeng Guo,†Sara A.DiBenedetto,†Pratyush Tewari,‡Michael nagan,*,‡Mark A.Ratner,*,†and Tobin J.Marks*,††Department of Chemistry and the Materials Research Center,Northwestern University,Evanston, Illinois60208-3113and‡Center for Dielectric Studies,Materials Research Institute,The Pennsylvania State University,University Park,Pennsylvania16802-4800Received September11,2009.Revised Manuscript Received December2,2009A series of0-3metal oxide-polyolefin nanocomposites are synthesized via in situ olefin polymeriza-tion,using the following single-site metallocene catalysts:C2-symmetric dichloro[rac-ethylenebisindenyl]-zirconium(IV),Me2Si(t BuN)(η5-C5Me4)TiCl2,and(η5-C5Me5)TiCl3immobilized on methylaluminoxane (MAO)-treated BaTiO3,ZrO2,3-mol%-yttria-stabilized zirconia,8-mol%-yttria-stabilized zirconia, sphere-shaped TiO2nanoparticles,and rod-shaped TiO2nanoparticles.The resulting composite materials are structurally characterized via X-ray diffraction(XRD),scanning electron microscopy(SEM), transmission electron microscopy(TEM),13C nuclear magnetic resonance(NMR)spectroscopy,and differential scanning calorimetry(DSC).TEM analysis shows that the nanoparticles are well-dispersed in the polymer matrix,with each individual nanoparticle surrounded by polymer.Electrical measurements reveal that most of these nanocomposites have leakage current densities of∼10-6-10-8A/cm2;relative permittivities increase as the nanoparticle volume fraction increases,with measured values as high as6.1. At the same volume fraction,rod-shaped TiO2nanoparticle-isotactic polypropylene nanocomposites exhibit significantly greater permittivities than the corresponding sphere-shaped TiO2nanoparticle-isotactic polypropylene nanocomposites.Effective medium theories fail to give a quantitative description of the capacitance behavior,but do aid substantially in interpreting the trends qualitatively.The energy storage densities of these nanocomposites are estimated to be as high as9.4J/cm3.IntroductionFuture pulsed-power and power electronic capacitors will require dielectric materials with ultimate energy storage den-sities of>30J/cm3,operating voltages of>10kV,and milli-second-microsecond charge/discharge times with reliable operation near the dielectric breakdown limit.Importantly, at2and0.2J/cm3,respectively,the operating characteristics of current-generation pulsed power and power electronic capacitors,which utilize either ceramic or polymer dielectric materials,remain significantly short of this goal.1An order-of-magnitude improvement in energy density will require the development of dramatically different types of materials, which substantially increase intrinsic dielectric energy den-sities while reliably operating as close as possible to the die-lectric breakdown limit.For simple linear response dielectric materials,the maximum energy density is defined in eq1,U e¼12εrε0E2ð1Þwhereεr is the relative dielectric permittivity,E the dielec-tric breakdown strength,andε0the vacuum permittivity (8.8542Â10-12F/m).Generally,metal oxides have large permittivities;however,they are limited by low breakdown fields.While organic materials(e.g.,polymers)can provide high breakdown strengths,their generally modest permit-tivities have limited their application.1Recently,inorganic-polymer nanocomposite materials have attracted great interest,because of their potential for high energy densities.2By integrating the complementary*Authors to whom correspondence should be addressed.E-mail addresses: mxl46@(M.T.L.),ratner@(M.A.R.),and t-marks@(T.J.M.).(1)(a)Pan,J.;Li,K.;Li,J.;Hsu,T.;Wang,Q.Appl.Phys.Lett.2009,95,022902.(b)Claude,J.;Lu,Y.;Li,K.;Wang,Q.Chem.Mater.2008, 20,2078–2080.(c)Chu,B.;Zhou,X.;Ren,K.;Neese,B.;Lin,M.;Wang,Q.;Bauer,F.;Zhang,Q.M.Science2006,313,334–336.(d) Cao,Y.;Irwin,P.C.;Younsi,K.IEEE Trans.Dielectr.Electr.Insul.2004,11,797–807.(e)Nalwa,H.S.,Ed.Handbook of Low and High Dielectric Constant Materials and Their Applications;Academic Press:New York,1999;V ol.2.(f)Sarjeant,W.J.;Zirnheld,J.;MacDougall,F.W.IEEE Trans.Plasma Sci.1998,26,1368–1392.(2)(a)Kim,P.;Doss,N.M.;Tillotson,J.P.;Hotchkiss,P.J.;Pan,M.-J.;Marder,S.R.;Li,J.;Calame,J.P.;Perry,J.W.ACS Nano 2009,3,2581–2592.(b)Li,J.;Seok,S.I.;Chu,B.;Dogan,F.;Zhang, Q.;Wang,Q.Adv.Mater.2009,21,217–221.(c)Li,J.;Claude,J.;Norena-Franco,L.E.;Selk,S.;Wang,Q.Chem.Mater.2008,20, 6304–6306.(d)Gross,S.;Camozzo,D.;Di Noto,V.;Armelao,L.;Tondello,E.Eur.Polym.J.2007,43,673–696.(e)Gilbert,L.J.;Schuman,T.P.;Dogan,F.Ceram.Trans.2006,179,17–26.(f)Rao,Y.;Wong,C.P.J.Appl.Polym.Sci.2004,92,2228–2231.(g)Dias,C.J.;Das-Gupta,D.K.IEEE Trans.Dielectr.Electr.Insul.1996,3,706–734.(h)Mammone,R.R.;Binder,M.Novel Methods For Preparing Thin,High Permittivity Polymerdielectrics for Capacitor Applica-tions;Proceedings of the34th International Power Sources Symposium, 1990,Cherry Hill,NJ;IEEE:New York,1990;pp395-398./cmPublished on Web01/05/2010 r2010American Chemical1568Chem.Mater.,Vol.22,No.4,2010Guo et al.properties of their constituents,such materials can simul-taneously derive high permittivity from the inorganic in-clusions and high breakdown strength,mechanical flexibility,facile processability,light weight,and tunability of the properties(polymer molecular weight,comonomer incorporation,viscoelastic properties,etc.)from the poly-mer host matrix.3In addition,convincing theoretical argu-ments have been made suggesting that large inclusion-matrix interfacial areas should afford greater polarization levels,dielectric response,and breakdown strength.4 Inorganic-polymer nanocomposites are typically pre-pared via mechanical blending,5solution mixing,6in situ radical polymerization,7and in situ nanoparticle syn-thesis.8However,host-guest incompatibilities intro-duced in these synthetic approaches frequently result in nanoparticle aggregation and phase separation over largelength scales,9which is detrimental to the electrical prop-erties of the composite.10Covalent grafting of the poly-mer chains to inorganic nanoparticle surfaces has alsoproven promising,leading to more effective dispersionand enhanced electrical/mechanical properties;11how-ever,such processes may not be optimally cost-effective,nor may they be easily scaled up.Furthermore,thedevelopment of accurate theoretical models for the di-electric properties of the nanocomposite must be accom-panied by a reliable experimental means to achievenanoparticle deagglomeration.In the huge industrial-scale heterogeneous or slurryolefin polymerization processes practiced today,SiO2isgenerally used as the catalyst support.12Very large localhydrostatic pressures arising from the propagating poly-olefin chains are known to effect extensive SiO2particlefracture and lead to SiO2-polyolefin composites.12Based on this observation,composite materials with enhancedmechanical properties13have been synthesized via in situpolymerizations using filler surface-anchored Ziegler-Natta or metallocene polymerization catalysts.14There-fore,we envisioned that processes meditated by rationallyselected single-site metallocene catalysts supported onferroelectric oxide nanoparticles15might disrupt ubiqui-tous and problematic nanoparticle agglomeration,16toafford homogeneously dispersed nanoparticles within thematrix of a processable,high-strength commodity poly-mer,already used extensively in energy storage capaci-tors.17Moreover,we envisioned that the methylalumino-xane(MAO)co-catalyst14i layer applied to the metaloxide nanoparticle surfaces would,after polymer workupunder ambient conditions,serve as an effective precursorfor a thin Al2O3layer to moderate the large anticipated(3)(a)Nelson,J.K.;Hu,Y.J.Phys.D:Appl.Phys.2005,38,213–222.(b)Tanaka,T.;Montanari,G.C.;M€u lhaupt,R.IEEE Trans.Dielectr.Electr.Insul.2004,11,763–784.(c)Lewis,T.J.IEEE Trans.Dielectr.Electr.Insul.1994,15,812–825.(d)Newnham,R.E.Annu.Rev.Mater.Sci.1986,16,47–68.(4)(a)Saha,S.K.Phys.Rev.B2004,69,1254161–125464.(b)Nelson,J.K.;Utracki,L.A.;MacCrone,R.K.;Reed,C.W.IEEE Conf.Electr.Insul.Dielectr.Phenomena2004,314–317.(c)Li,J.Phys.Rev.Lett.2003,90,217601/1–4.(5)(a)Subodh,G.;Deepu,V.;Mohanan,P.;Sebastian,M.T.Appl.Phys.Lett.2009,95,062903.(b)Dang,Z.;Wu,J.;Fan,L.;Nan,C.Chem.Phys.Lett.2003,376,389–394.(6)(a)Goyal,R.K.;Jagadale,P.A.;Mulik,U.P.J.Appl.Polym.Sci.2009,111,2071–2077.(b)Afzal,A.B.;Akhtar,M.J.;Nadeem,M.;Hassan,M.M.J.Phys.Chem.C2009,113,17560–17565.(c)Huang, X.Y.;Jiang,P.K.;Kim,C.U.J.Appl.Phys.2007,102,124103.(d) Parvatikar,N.;Ambika Prasad,M.V.N.J.Appl.Polym.Sci.2006, 100,1403–1405.(e)Badheka,P.;Magadala,V.;Gopi Devaraju,N.;Lee,B.I.;Kim,E.S.J.Appl.Polym.Sci.2006,99,2815–2821.(f)Xie, S.;Zhu,B.;Xu,Z.;Xu,Y.Mater.Lett.2005,59,2403–2407.(g) Schroeder,R.;Majewski,L.;Grell,M.Adv.Mater.2005,17,1535–1539.(h)Bai,Y.;Cheng,Z.;Bharti,V.;Xu,H.;Zhang,Q.Appl.Phys.Lett.2000,76,3804–3806.(7)(a)Andou,Y.;Jeong,J.-M.;Nishida,H.;Endo,T.Macromolecules2009,42,7930–7935.(b)Thomas,P.;Dwarakanath,K.;Varma,K.B.R.Synth.Met.2009,159,2128–2134.(c)Chen,Y.-M.;Lin,H.-C.;Hsu,R.-S.;Hsieh,B.-Z.;Su,Y.-A.;Sheng,Y.-J.;Lin,J.-J.Chem.Mater.2009,21,4071–4079.(d)He,A.;Wang,L.;Li,J.;Dong,J.;Han,C.C.Polymer2006,47,1767–1771.(e)Ginzburg,V.V.;Myers, K.;Malowinski,S.;Cieslinski,R.;Elwell,M.;Bernius,M.Macro-molecules2006,39,3901–3906.(f)Mizutani,T.;Arai,K.;Miyamoto, M.;Kimura,Y.J.Appl.Polym.Sci.2006,99,659–669.(g)Xiao,M.;Sun,L.;Liu,J.;Li,Y.;Gong,K.Polymer2002,43,2245–2248.(h)R.Popielarz,R.;Chiang,C.K.;Nozaki,R.;Obrzut,J.Macromolecules 2001,34,5910–5915.(8)(a)Balan,L.;Jin,M.;Malval,J.-P.;Chaumeil,H.;Defoin,A.;Vidal,L.Macromolecules2008,41,9359–9365.(b)Lu,J.;Moon,K.S.;Xu,J.;Wong,C.P.J.Mater.Chem.2006,16,1543–1548.(c)Yogo, T.;Yamamoto,T.;Sakamoto,W.;Hirano,S.J.Mater.Res.2004,19, 3290–3297.(9)(a)Vaia,R.A.;Maguire,J.F.Chem.Mater.2007,19,2736–2751.(b)Mackay,M.E.;Tuteja,A.;Duxbury,P.M.;Hawker,C.J.;Van Horn,B.;Guan,Z.;Chen,G.;Krishnan,R.S.Science2006,311,1740–1743.(c)Lin,Y.;Boeker,A.;He,J.;Sill,K.;Xiang,H.;Abetz,C.;Li,X.;Wang,J.;Emrick,T.;Long,S.;Wang,Q.;Balazs,A.;Russell,T.P.Nature2005,434,55–59.(10)(a)Stoyanov,H.;Mc Carthy,D.;Kollosche,M.;Kofod,G.Appl.Phys.Lett.2009,94,232905.(b)Chen,G.;Davies,A.E.IEEE Trans.Dielectr.Electr.Insul.2000,7,401–407.(c)Khalil,M.S.IEEE Trans.Dielectr.Electr.Insul.2000,7,261–268.(11)(a)Zhang,Y.;Ye,Z.Macromolecules2008,41,6331–6338.(b)Maliakal,A.;Katz,H.E.;Cotts,P.M.;Subramoney,S.;Mirau,P.J.Am.Chem.Soc.2005,127,14655–14662.(c)Rusa,M.;Whitesell,J.K.;Fox,M.A.Macromolecules2004,37,2766–2774.(d)Bartholome,C.;Beyou,E.;Bourgeat-Lami,E.;Chaumont,P.;Zydowicz,N.Macro-molecules2003,36,7946–7952.(e)Corbierre,M.K.;Cameron,N.S.;Sutton,M.;Mochrie,S.G.J.;Lurio,L.B.;R€u hm,A.;Lennox,R.B.J.Am.Chem.Soc.2001,123,10411–10412.(f)von Werne,T.;Patten,T.E.J.Am.Chem.Soc.2001,123,7497–7505.(g)Nuss,S.;B€o ttcher,H.;Wurm,H.;Hallensleben,M.L.Angew.Chem.,Int.Ed.2001,40, 4016–4018.(12)(a)Kaminsky,W.;Funck,A.;Wiemann,K.Macromol.Symp.2006,239,1–6.(b)Li,K.-T.;Kao,Y.-T.J.Appl.Polym.Sci.2006,101, 2573–2580.(c)du Fresne von Hohenesche,C.;Unger,K.K.;Eberle,T.J.Mol.Catal.A:Chem.2004,221,185–199.(d)Fink,G.;Steinmetz,B.;Zechlin,J.;Przybyla,C.;Tesche,B.Chem.Rev.2000,100,1377–1390.(13)(a)Dubois,P.;Alexandre,M.;J e r^o me,R.Macromol.Symp.2003,194,13–26.(b)Kaminsky,W.Macromol.Chem.Phys.1996,197, 3907–3945.(14)For recent reviews of single-site olefin polymerization,see:(a)Amin,S.B.;Marks,T.J.Angew.Chem.,Int.Ed.2008,47,2006–2025.(b)Marks,T.J.,ed.Proc.Natl.Acad.Sci.,U.S.A.,2006,103, 15288-15354,and contributions therein(Special Feature on Poly-merization).(c)Suzuki,anomet.Chem.2005,8,177–216.(d)Alt,H.G.Dalton Trans.2005,20,3271–3276.(e)Kaminsky,W.J.Polym.Sci.Polym.Chem.2004,42,3911–3921.(j)Wang,W.;Wang, L.J.Polym.Mater.2003,20,1–8.(f)Delacroix,O.;Gladysz,J.A.mun.2003,6,665–675.(g)Kaminsky,W.;Arndt-Rosenau, M.Applied Homogeneous Catalysis with Organometallic Com-pounds,2nd Edition;Wiley-VCH Verlag GmbH:Weinheim, Germany,2002.(h)Lin,S.;Waymouth,R.M.Acc.Chem.Res.2002,35,765–773.(i)Chen,E.Y.-X.;Marks,T.J.Chem.Rev.2000,100,1391–1434.(15)For recent reviews of single-site heterogeneous catalysis,see:(a)Thomas,J.M.;Raja,R.;Lewis,D.W.Angew.Chem.,Int.Ed.2005, 44,6456–6482.(b)Cop e ret,C.;Chabanas,R.;Petroff Saint-Arroman, R.;Basset,J.-M.Angew.Chem.,Int.Ed.2003,42,156–181.(c) Hlatky,G.G.Chem.Rev.2000,100,1347–1376.(d)Reven,L.J.Mol.Catal.1994,86,447–477.(16)Kim,P;Jones,S.C.;Hotchkiss,P.J.;Haddock,J.N.;Kippelen,B.;Marder,S.R.;Perry,J.W.Adv.Mater.2007,19,1001–1005. (17)Rabuffi,M.;Picci,G.IEEE Trans.Plasma Sci.2002,30,1939–1942.Article Chem.Mater.,Vol.22,No.4,20101569polyolefin -ferroelectric permittivity contrast.If too large,such contrasts are associated with diminished breakdown strength and suppressed permittivity.18,19In a brief preliminary communication,we reported evidence that high-energy-density BaTiO 3-and TiO 2-isotactic polypropylene nanocomposites could be pre-pared via in situ propylene polymerization mediated by anchoring/alkylating/activating C 2-symmetric dichloro-[rac -ethylenebisindenyl]zirconium(IV)(EBIZrCl 2)on the MAO-treated oxide nanoparticles (see Scheme 1).20The resulting nanocomposites were determined to have rela-tively uniform nanoparticle dispersions and to support remarkably high projected energy storage densities ;as high as 9.4J/cm 3,as determined from permittivity and dielectric breakdown measurements.In this contribution,we significantly extend the inorganic inclusion scope to include a broad variety of nanoparticle types,to investi-gate the effects of nanoparticle identity and shape on the electrical/dielectric properties of the resulting nanocom-posites,and to compare the experimental results with theoretical predictions.We also extend the scope of metallocene polymerization catalysts (see Chart 1)and olefinic monomers,with the goal of achieving nanocom-posites that have comparable or potentially greater pro-cessability and thermal stability.Here,we present a full discussion of the synthesis,microstructural and electrical characterization,and theoretical modeling of these nano-composites.It will be seen that nanoparticle coating with MAO and subsequent in situ polymerization are crucial to achieving effective nanoparticle dispersion,and,simul-taneously,high nanocomposite breakdown strengths (as high as 6.0MV/cm)and high permittivities (as high as 6.1)can be realized to achieve energy storage densities as high as 9.4J/cm 3.Experimental SectionI.Materials and Methods.All manipulations of air-sensitive materials were performed with rigorous exclusion of O 2and moisture in flamed Schlenk-type glassware on a dual-manifold Schlenk line or interfaced to a high-vacuum line (10-5Torr),or in a dinitrogen-filled MBraun glovebox with a high-capacity recirculator (<1ppm O 2and H 2O).Argon (Airgas,pre-purified),ethylene (Airgas,polymerization grade),and propy-lene (Matheson or Airgas,polymerization grade)were purified by passage through a supported MnO oxygen-removal column and an activated Davison 4A molecular sieve column.Styrene (Sigma -Aldrich)was dried sequentially for a week over CaH 2and then triisobutylaluminum,and it was freshly vacuum-transferred prior to polymerization experiments.The monomer 1-octene (Sigma -Aldrich)was dried over CaH 2and was freshly vacuum-transferred prior to polymerization experiments.To-luene was dried using activated alumina and Q-5columns,according to the method described by Grubbs,21and it was additionally vacuum-transferred from Na/K alloy and stored in Teflon-valve sealed bulbs for polymerization experiments.Ba-TiO 3and TiO 2nanoparticles were kindly provided by Prof.Fatih Dogan (University of Missouri,Rolla)and Prof.Thomas Shrout (Penn State University),respectively.20ZrO 2nanopar-ticles were purchased from Sigma -Aldrich.The reagents 3-mol %-yttria-stabilized zirconia (TZ3Y)and 8-mol %-yttria-stabilized zirconia (TZ8Y)nanoparticles were purchased from Tosoh,Inc.TiO 2nanorods were purchased from Reade Ad-vanced Materials (Riverside,RI).All of the nanoparticles were dried in a high vacuum line (10-5Torr)at 80°C overnight to remove the surface-bound water,which is known to affect the dielectric breakdown performance adversely.22The deuteratedScheme 1.Synthesis of Polyolefin -Metal OxideNanocompositesChart 1.Metallocene polymerization catalysts andMAO.(18)(a)Li,J.Y.;Zhang,L.;Ducharme,S.Appl.Phys.Lett.2007,90,132901/1–132901/3.(b)Li,J.Y .;Huang,C.;Zhang,Q.M.Appl.Phys.Lett.2004,84,3124–3126.(19)Cheng,Y.;Chen,X.;Wu,K.;Wu,S.;Chen,Y.;Meng,Y.J.Appl.Phys.2008,103,034111/1–034111/8.(20)Guo,N.;DiBenedetto,S.A.;Kwon,D.-K.;Wang,L.;Russell,M.T.;Lanagan,M.T.;Facchetti,A.;Marks,T.J.J.Am.Chem.Soc.2007,129,766–767.(21)Pangborn,A.B.;Giardello,M.A.;Grubbs,R.H.;Rosen,R.K.;Timmers,anometallics 1996,15,1518–1520.(22)(a)Hong,T.P.;Lesaint,O.;Gonon,P.IEEE Trans.Dielectr.Electr.Insul.2009,16,1–10.(b)Ma,D.;Hugener,T.A.;Siegel,R.W.;Christerson,A.;M artensson,E.;€Onneby,C.;Schadler,L.S.Nano-technology 2005,16,724–731.(c)Ma,D.;Siegel,R.W.;Hong,J.;Schadler,L.S.;M artensson,E.;€Onneby,C.J.Mater.Res.2004,19,857–863.1570Chem.Mater.,Vol.22,No.4,2010Guo et al. solvent1,1,2,2-tetrachloroethane-d2was purchased fromCambridge Isotope Laboratories(g99at.%D)and was usedas-received.Methylaluminoxane(MAO;Sigma-Aldrich)waspurified by removing all the volatiles in vacuo from a1.0Msolution in toluene.The reagents dichloro[rac-ethylenebisin-denyl]zirconium(IV)(EBIZrCl2),and trichloro(pentamethyl-cyclopentadienyl)titanium(IV)(Cp*TiCl3)were purchasedfrom Sigma-Aldrich and used as-received.Me2Si(t BuN)(η5-C5Me4)TiCl2(CGCTiCl2)was prepared according to publishedprocedures.23nþ-Si wafers(root-mean-square(rms)roughnessof∼0.5nm)were obtained from Montco Silicon Tech(SpringCity,PA),and aluminum substrates were purchased fromMcMaster-Carr(Chicago,IL);both were cleaned according to standard procedures.24II.Physical and Analytical Measurements.NMR spectra were recorded on a Varian Innova400spectrometer(FT400 MHz,1H;100MHz,13C).Chemical shifts(δ)for13C spectra were referenced using internal solvent resonances and are reported relative to tetramethylsilane.13C NMR assays of polymer microstructure were conducted in1,1,2,2-tetrachlor-oethane-d2containing0.05M Cr(acac)3at130°C.Resonances were assigned according to the literature for isotactic polypro-pylene,poly(ethylene-co-1-octene),and syndiotactic polystyr-ene,respectively(see more below).Elemental analyses were performed by Midwest Microlabs,LLC(Indianapolis,IN). Inductively coupled plasma-optical emission spectroscopy (ICP-OES)analyses were performed by Galbraith Laboratories, Inc.(Knoxville,TN).Powder X-ray diffraction(XRD)patterns were recorded on a Rigaku DMAX-A diffractometer with Ni-filtered Cu K R radiation(λ=1.54184A).Pristine ceramic nanoparticles and composite microstructures were examined with a FEI Quanta sFEG environmental scanning electron microscopy(SEM)system with an accelerating voltage of30 kV.Transmission electron microscopy(TEM)was performed on a Hitachi Model H-8100TEM system with an accelerating voltage of200kV.Samples for TEM imaging were prepared by dipping a TEM grid into a suspension of nanocomposite powder in acetone.Polymer composite thermal transitions were mea-sured on a temperature-modulated differential scanning calori-meter(TA Instruments,Model2920).Typically,ca.10mg of samples were examined,and a ramp rate of10°C/min was used to measure the melting point.To erase thermal history effects, all samples were subjected to two melt-freeze cycles.The data from the second melt-freeze cycle are presented here.III.Electrical Measurements.Metal-insulator-metal (MIM)or metal-insulator-semiconductor(MIS)devices for nanocomposite electrical measurements were fabricated by first doctor-blading nanocomposite films onto aluminum(MIM)or nþ-Si(MIS)substrates,followed by vacuum-depositing top gold electrodes through shadow masks.Specifically,a clean substrate was placed on a hot plate heated to just below the polymer-nanocomposite melting point.A small amount of the polymer nanocomposite powder was placed in the center of the substrate and left until the powder began to melt.Once in this phase,the polymer nanocomposite is spread over the center of the sub-strate using a razor blade.The sample was removed from the heat,cooled,and then pressed in a benchtop press to ensure uniform film thicknesses and smooth surfaces.Gold electrodes 500A thick were vacuum-deposited directly on the films through shadow masks that defined a series of different areas (0.030,0.0225,0.01,0.005,and0.0004cm2)at3Â10-6Torr(at 0.2-0.5A/s).Electrical properties of the films were character-ized by two probe current-voltage(I-V)measurements using a Keithley Model6430Sub-Femtoamp Remote Source Meter, operated by a local LABVIEW program.Triaxial and low triboelectric noise coaxial cables were incorporated with the Keithley remote source meter and Signatone(Gilroy,CA)probe tip holders to minimize the noise level.All electrical measure-ments were performed under ambient conditions.For MIS devices,the leakage current densities(represented by the symbol J,given in units of A/cm2)were measured with positive/negative polarity applied to the gold electrode to ensure that the nþ-Si substrate was operated in accumulation.A delay time of1s was incorporated into the source-delay-measure cycle to settle the source before recording currents.Capacitance measurements of the MIM and MIS structures were performed with a two-probe digital capacitance meter(Model3000,GLK Instruments,San Diego,CA)at(5and24kHz.Several methods have been developed to measure the dielectric breakdown strength of polymer and nanocomposite films.1a,25In this study,various methods were examined(e.g.,pull-down electrodes25),and the two-probe method was used to collect the present data because the top gold electrodes had already been deposited for leakage current and capacitance measurements.The dielectric break-down strength of the each type of composite film was measured in a Galden heat-transfer fluid bath at room temperature with a high-voltage amplifier(Model TREK30/20A,TREK,Inc., Medina,NY)with a ramp rate of1000V/s.26The thicknesses of the dielectric films were measured with a Tencor P-10step profilometer,and these thicknesses were used to calculate the dielectric constants and breakdown strengths of the film sam-ples(see Table2,presented later in this work).IV.Representative Immobilization of a Metallocene Catalyst on Metal Oxide Nanoparticles.In the glovebox,2.0g of BaTiO3 nanoparticles,200mg of MAO,and50mL of dry toluene were loaded into a predried100-mL Schlenk reaction flask,which was then attached to the high vacuum line.Upon stirring,the mixture became a fine slurry.The slurry was next subjected to alternating sonication and vigorous stirring for2days with constant removal of evolving CH4.Next,the nanoparticles were collected by filtration and washed with fresh toluene(50mLÂ4) to remove any residual MAO.Then,200mg of metallocene catalyst EBIZrCl2and50mL of toluene were loaded in the flask containing the MAO-coated nanoparticles.The color of the nanoparticles immediately became purple.The slurry mixture was again subjected to alternating sonication and vigorous Table1.XRD Linewidth Analysis Results for the Oxide-PolypropyleneNanocompositespowder2θ(deg)full width athalf maximum,fwhm(deg)crystallitesize,L(nm)a BaTiO331.4120.25435.6 BaTiO3-polypropylene31.6490.27132.8 TiO225.3600.31727.1 TiO2-polypropylene25.3580.36123.5a Crystallite size(L)is calculated using the Scherrer equation:L=0.9λ/[B(cosθB)whereλis the X-ray wavelength,B the full width at half maximum(fwhm)of the diffraction peak,andθB the Bragg angle.(23)Stevens,J.C.;Timmers,F.J.;Wilson,D.R.;Schmidt,G.F.;Nickias,P.N.;Rosen,R.K.;Knight,G.W.;Lai,S.Eur.Patent Application EP416815A2,1991.(24)Yoon,M.-H.;Kim,C.;Facchetti,A.;Marks,T.J.J.Am.Chem.Soc.2006,128,12851–12869.(25)Claude,J.;Lu,Y.;Wang,Q.Appl.Phys.Lett.2007,91,212904/1–212904/3.(26)Gadoum,A.;Gosse,A.;Gosse,J.P.Eur.Polym.J.1997,33,1161–1166.Article Chem.Mater.,Vol.22,No.4,20101571stirring overnight.The nanoparticles were then collected by filtration and washed with fresh toluene until the toluene remained colorless.The nanoparticles were dried on the high-vacuum line overnight and stored in a sealed container in the glovebox at-40°C in darkness.V.Representative Synthesis of an Isotactic Polypropylene Nanocomposite via In Situ Propylene Polymerization.In the glovebox,a250-mL round-bottom three-neck Morton flask, which had been dried at160°C overnight and equipped with a large magnetic stirring bar,was charged with50mL of dry toluene,200mg of functionalized nanoparticles,and50mg of MAO.The assembled flask was removed from the glovebox and the contents were subjected to sonication for30min with vigorous stirring.The flask was then attached to a high vacuum line(10-5Torr),the catalyst slurry was freeze-pump-thaw degassed,equilibrated at the desired reaction temperature using an external bath,and saturated with1.0atm(pressure control using a mercury bubbler)of rigorously purified propylene while being vigorously stirred.After a measured time interval,the polymerization was quenched by the addition of5mL of methanol,and the reaction mixture was then poured into800 mL of methanol.The composite was allowed to fully precipitate overnight and was then collected by filtration,washed with fresh methanol,and dried on the high vacuum line overnight to constant weight.VI.Representative Synthesis of a Poly(ethylene-co-1-octene) Nanocomposite via In Situ Ethyleneþ1-Octene Copolymeriza-tion.In the glovebox,a250-mL round-bottom three-neck Morton flask,which had been dried at160°C overnight and equip-ped with a large magnetic stirring bar,was charged with50mL of dry toluene,200mg of functionalized nanoparticles,and 50mg of MAO.The assembled flask was removed from the glo-vebox and the contents were subjected to sonication for30min with vigorous stirring.The flask was then attached to a high vacuum line(10-5Torr),the catalyst slurry was freeze-pump-thaw degassed,equilibrated at the desired reaction temperature using an external bath,and saturated with1.0atm(pressure control using a mercury bubbler)of rigorously purified ethylene while being vigorously stirred.Next,5mL of freshly vacuum-transferred1-octene was quickly injected into the rapidly stirred flask using a gas-tight syringe equipped with a flattened spraying needle.After a measured time interval,the polymerization was quenched by the addition of5mL of methanol,and the reaction mixture was then poured into800mL of methanol.The com-posite was allowed to fully precipitate overnight and was then collected by filtration,washed with fresh methanol,and dried on the high vacuum line overnight to constant weight.Film fabri-cation of the composite powders into thin films for MIS electrical testing was unsuccessful due to the high incorporation level of1-octene.VII.Representative Synthesis of a Syndiotactic Polystyrene Nanocomposite via In Situ Styrene Polymerization.In the glove-box,a250-mL round-bottom three-neck Morton flask,which had been dried at160°C overnight and equipped with a large magnetic stirring bar,was charged with50mL of dry toluene, 200mg of functionalized nanoparticles,and50mg of MAO.The assembled flask was removed from the glovebox and the con-tents were subjected to sonication for30min with vigorous stirring.The flask was then attached to a high vacuum line(10-5 Torr)and equilibrated at the desired reaction temperature usingTable2.Electrical Characterization Results for Metal Oxide-Polypropylene Nanocomposites aentry compositenanoparticlecontent b(vol%)melting temperature,T m c(°C)permittivity dbreakdownstrength e(MV/cm)energy density,U f(J/cm3)1BaTiO3-iso PP0.5136.8 2.7(0.1 3.1 1.2(0.1 2BaTiO3-iso PP0.9142.8 3.1(1.2>4.8>4.0(0.6 3BaTiO3-iso PP 2.6142.1 2.7(0.2 3.9 1.8(0.2 4BaTiO3-iso PP 5.2145.6 2.9(1.0 2.7 1.0(0.3 5BaTiO3-iso PP 6.7144.8 5.1(1.7 4.1 3.7(1.2 6BaTiO3-iso PP13.6144.8 6.1(0.9>5.9>9.4(1.37s TiO2-iso PP g0.1135.2 2.2(0.1>2.8>0.8(0.1 8s TiO2-iso PP g 1.6142.4 2.8(0.2 4.1 2.1(0.2 9s TiO2-iso PP g 3.1142.6 2.8(0.1 2.8 1.0(0.1 10s TiO2-iso PP g 6.2144.8 3.0(0.2 4.7 2.8(0.211r TiO2-iso PP h 1.4139.7 3.4(0.3 1.00.40(0.35 12r TiO2-iso PP h 3.0142.4 4.1(0.70.90.22(0.09 13r TiO2-iso PP h 5.1143.7 4.9(0.40.90.23(0.0814ZrO2-iso PP 1.6142.9 1.7(0.3 1.50.1815ZrO2-iso PP 3.9145.2 2.0(0.4 1.90.3216ZrO2-iso PP7.5144.9 4.8(1.1 1.00.2017ZrO2-iso PP9.4144.4 6.9(2.6 2.0 1.02(0.7318TZ3Y-iso PP 1.1142.9 1.1(0.1N/A N/A19TZ3Y-iso PP 3.1143.5 1.8(0.2N/A N/A20TZ3Y-iso PP 4.3143.8 2.0(0.2N/A N/A21TZ3Y-iso PP 6.7144.9 2.7(0.2N/A N/A22TZ8Y-iso PP0.9142.9 1.4(0.1 3.8 1.07(0.04 23TZ8Y-iso PP 2.9143.2 1.8(0.1 2.80.5924TZ8Y-iso PP 3.8143.2 2.0(0.2 2.00.4125TZ8Y-iso PP 6.6146.2 2.4(0.4 2.20.61a Polymerizations performed in50mL of toluene under1.0atm of propylene at20°C.b From elemental analysis.c From differential scanning calorimetry(DSC).d Derived from capacitance measurements.e Calculated by dividing the breakdown voltage by the film thickness,which is measured using a Tencor p10profilometer.f Energy density(U)is calculated from the following relation:U=0.5ε0εr E b2,whereε0is the permittivity of a vacuum,εr the relative permittivity,and E b the breakdown strength.g The superscripted prefix“s”denotes sphere-shaped TiO2nanoparticles.h The superscripted prefix“r”denotes rod-shaped TiO2nanoparticles.。

机械应力对锂离子电池性能的影响

EFFECTS OF MECHANICAL STRESSES ON LITHIUM ION BATTERIES
byHyoncheol KmA dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Mechanical Engineering) in The University of Michigan 2009
Doctoral Committee: Professor Ann Marie Sastry, Chair Professor James R. Barber Associate Professor Christian M. Lastoskie Associate Professor Wei Lu
©
Hyoncheol Kim 2009 All Rights Reserved
€
To My Parents
ii
ACKNOWLEDGEMENTS
First of all, I would like to thank Prof. Ann Marie Sastry for her guidance and support. I was lucky to have her as my advisor because I have learned not only how to solve problems in science and technology from her great expertise, but also and more importantly how we can contribute to the world with what we can do for the better future. Her enthusiasm toward life and work is the most valuable thing I have learned at school and will always guide me throughout my life and career. Many thanks go to my committee members, Prof. James R. Barber, Prof. Christian M. Lastoskie, and Prof. Wei Lu for serving on my committee and providing your valuable advices on my thesis. I am very grateful to all of our current and former group members, Dr. Myounggu Park, Mr. Xiangchun Zhang, Mr. Myoungdo Chung, Mr. Min Zhu, Mr. Sangwoo Han, Mr. Dong Hoon Song, Ms. Qiuye Jin, Dr. Chia-Wei Wang, Dr. Munish V. Inamdar, Dr. Fabio Albano, Dr. Yenhung Chen, Dr. Kimberly Cook, Dr. Yun-Bo Yi, Dr. Taeyong Kim, Ms. Tahira Reid, and Mr. Peter Verhees for their support and sharing the wonderful moments for the past years. I also would like to thank my friends, Jiwon, Chi-Shan, Jung Kap and Jong-Hoon for their help and encouragement. Most importantly, I would like to thank my parents Ip Kwon Kim and Bup Jin Song, and my sister Hyun Hee for their love and support. I also would like to thank my parents in-law, Jeong Hong Chung and Myung Sook Lee, and

应用总应变_应变能区分法预测热机械疲劳寿命

第25卷第1期燃气涡轮试验与研究Vol.25,No.1 2012年2月Gas Turbine Experiment and Research Feb.,2012严严严严严严严严严严严严严严严严严严严严严严严严严严收稿日期：2011-06-15；修回日期：2011-11-25作者简介：胡绪腾(1980-)，男，江苏沛县人，讲师，博士，主要从事结构强度、耐久性、损伤容限等方面的研究。

摘要：简要介绍了总应变-应变能区分法(TS-SEP)的基本假设和基本方程，应用TS-SEP法对三种金属材料(304不锈钢、1Cr-1Mo-0.25V钢和2.25Cr-1Mo钢)的热机械疲劳试验数据进行了分析和预测，初步评估了TS-SEP法对热机械疲劳数据的相关和预测能力。

研究结果表明：TS-SEP法与总应变-应变范围区分法(TS-SRP)，对三种金属材料的热机械疲劳试验数据具有相当的相关和预测能力，寿命预测分散带均在2倍范围内。

关键词：热机械疲劳；蠕变-疲劳；寿命预测；总应变-应变能区分法；应变能区分法；应变范围区分法中图分类号：O346.2文献标识码：A文章编号：1672-2620(2012)01-0014-03Life Prediction for Thermomechanical Fatigue Using TotalStrain Version of Strain Energy PartitioningHU Xu-teng，SONG Ying-dong(State Key Laboratory of Machinery Structural Mechanics and Control，Nanjing University of Aeronautics and Astronautics，Nanjing210016，China) Abstract：The basic concepts and equations of Total Strain Version of Strain Energy Partitioning(TS-SEP) are introduced briefly.Then the correlative and predictive capabilities of TS-SEP to thermomechanical fa⁃tigue(TMF)data of three metals(Type304Stainless Steel,1Cr-lMo-O.25V Steel and2.25Cr-lMo Steel)in the literature are evaluated.The result indicates that TS-SEP and Total Strain Version of Strain Range Par⁃titioning(TS-SRP)have comparative correlative and predictive capabilities with the TMF data of three steels.The maximum life prediction scatter bands of two methods are both less than2.0.Key words：thermomechanical fatigue；creep-fatigue；life prediction；TS-SEP；SEP；SRP1引言许多热机械结构的疲劳失效本质上是热机械疲劳所致，例如航空发动机涡轮叶片在使用中即承受着复杂的热机械循环载荷作用。

商务英语中的委婉表达与翻译

（3）酌情使用一些常用套语，如：贵，乞谅，承蒙，为盼，赐复等。Dear Sir(s)/ Madam尊敬的阁下 /先生/女士/谨启者/执事先生/台鉴
Yours faithfully /truly/sincerely
谨上、敬上、顺致敬意
汉语常用敬辞有： “贵方、贵公司、阁下、惠请、惠顾、奉告、
is doubtlessly a pride of your family.

We are very happy for you and wish to
extend to you our utmost congratulations. I believe
that the knowledge ha has acquired will enable
世界。在未来的岁月里，让我们在记忆的彩虹里永远珍藏着对他的怀念。
确实，没有他的生活无论如何都不会再是
原样，但我们一定要尽可能努力地生活下去，并期望在天国有重逢的一天。

爱你的，

邦妮
4）商务信函的常用翻译技巧：
（1）重复、省略、否定、转换等技巧的灵活使用。
（2）掌握分寸，体现原文文体，尽力保持“公事公办”的慎重态度
敬礼！

ห้องสมุดไป่ตู้
你诚挚的

奥尔登.艾德
Dear Mrs. Taylor,

We hear that your son Paul has graduated
from the notable M.I.T with honors. You must be
quite happy with his splendid accomplishment. He

ImportanceofthePre-RequisiteSubject

Importance of the Pre-Requisite SubjectK.Kadirgama, M.M.Noor, M.R.M.Rejab, A.N.M.Rose, N.M. Zuki N.M., M.S.M.Sani, A.Sulaiman,R.A.Bakar, Abdullah IbrahimUniversiti Malaysia Pahang,***************.myABSTRACTIn this paper, it describes how the pre-requisite subjects influence the student’s performance in Heat transfer subject in University Malaysia Pahang (UMP). The Pre-requisite for Heat transfer in UMP are Thermodynamics I and Thermodynamics II. Randomly 30 mechanical engineering students were picked to analysis their performance from Thermodynamics I to Heat transfer. Regression analysis and Neural Network were used to prove the effect of prerequisite subject toward Heat transfer. The analysis shows that Thermodynamics I highly affect the performance of Heat transfer. The results show that the students who excellent in Thermodynamics I, their performance in Thermodynamics II also the same and goes to Heat transfer. Those students who scored badly in their Thermodynamics I, the results for the Thermodynamics II and Heat transfer are similar to Thermodynamics I. This shows the foundation must be solid, if the students want to do better in Heat transfer.INTRODUCTIONPre-requisite means course required as preparation for entry into a more advanced academic course or program [1]. Regression analysis is a technique used for the modeling and analysis of numerical data consisting of values of a dependent variable (response variable) and of one or more independent variables (explanatory variables). The dependent variable in the regression equation is modelled as a function of the independent variables, corresponding parameters ("constants"), and an error term. The error term is treated as a random variable. It represents unexplained variation in the dependent variable. The parameters are estimated so as to give a "best fit" of the data. Most commonly the best fit is evaluated by using the least squares method, but other criteria have also been used [1].Regression can be used for prediction (including forecasting of time-series data), inference, hypothesis testing, and modelling of causal relationships. These uses of regression rely heavily on the underlying assumptions being satisfied. Regression analysis has been criticized as being misused for these purposes in many cases where the appropriate assumptions cannot be verified to hold [1, 2]. One factor contributing to the misuse of regression is that it can take considerably more skill to critique a model than to fit a model [3].However, when a sample consists of various groups of individuals such as males and females, or different intervention groups, regression analysis can be performed to examine whether the effects of independent variables on a dependent variable differ across groups, either in terms of intercept or slope. These groups can be considered from different populations (e.g., male population or female population), and the population is considered heterogeneous in that these subpopulations may require different population parameters to adequately capture their characteristics. Since this source of population heterogeneity is based on observed group memberships such as gender, the data can be analyzed using regression models by taking into consideration multiple groups. In the methodology literature, subpopulations that can be identified beforehand are called groups [4, 5].Model can account for all kinds of individual differences. Regression mixture models described here are a part of a general framework of finite mixture models [6] and can be viewed as a combination of the conventional regression model and the classic latent class model [7, 8]. It should be noted that there are various types of regression mixture models [7], but this only focus on the linear regression mixture model. Thefollowing sections will first describe some unique characteristics of the linear regression mixture model in comparison to the conventional linear regression model, including integration of covariates into the model. Second, a step-by-step regression mixture analysis of empirical data demonstrates how the linear regression mixture model may be used by incorporating population heterogeneity into the model.Ko et al. [9] have introduced an unsupervised, self-organised neural network combined with an adaptive time-series AR modelling algorithm to monitor tool breakage in milling operations. The machining parameters and average peak force have been used to build the AR model and neural network. Lee and Lee [10] have used a neural network-based approach to show that by using the force ratio, flank wear can be predicted within 8% to 11.9% error and by using force increment, the prediction error can be kept within 10.3% of the actual wear. Choudhury et al. [11] have used an optical fiber to sense the dimensional changes of the work-piece and correlated it to the tool wear using a neural network approach. Dimla and Lister [12] have acquired the data of cutting force, vibration and measured wear during turning and a neural network has been trained to distinguish the tool state.This paper will describe the influence of prerequisite subject toward Heat transfer. The analysis will be done using regression method and Neural Network.REGRESSION METHODIn linear regression, the model specification is that the dependent variable, yi is a linear combination of the parameters (but need not be linear in the independent variables). For example, in simple linear regression for modelling N data points there is one independent variable: xi, and two parameters, β0 and β1 [2]:Results from the 30 mechanical engineering students were collected. There are mixed between female and male, no age different, different of background and all the students from same class. Regression analysis was done to check the most dominant variables (Thermodynamics I and Thermodynamics II) effect towards response (Heat transfer). Table 1 shows the marks of the students.Table 1: Marks for the subjects.Student Thermodynamics1 Thermodynamics2Heat transfer1 85 83 852 51 50 533 67 65 694 55 61 555 44 51 516 64 63 557 42 50 498 54 63 609 58 50 5810 52 61 6011 69 77 7712 58 64 6813 57 61 6814 71 68 6015 61 70 7316 53 66 6217 60 71 5918 45 55 5719 47 60 5620 62 77 6921 45 60 53(1)22 40 52 3723 53 70 6224 53 61 7025 56 60 7326 51 63 6927 44 62 5728 40 58 5229 62 80 7130 47 63 46MULTILAYER PERCEPTIONS NEURAL NETWORKIn the current application, the objective is to use the supervised network with multilayer perceptrons and train with the back-propagation algorithm (with momentum). The components of the input pattern consist of the control variables used in the student performance (Thermodynamics I and Thermodynamics II), whereas the components of the output pattern represent the responses from sensors (Heat transfer). During the training process, initially all patterns in the training set were presented to the network and the corresponding error parameter (sum of squared errors over the neurons in the output layer) was found for each of them. Then the pattern with the maximum error was found which was used for changing the synaptic weights. Once the weights were changed, all the training patterns were again fed to the network and the pattern with the maximum error was then found. This process was continued till the maximum error in the training set became less than the allowable error specified by the user. This method has the advantage of avoiding a large number of computations, as only the pattern with the maximum error was used for changing the weights. Fig.1 shows the neural network computational mode with 2-5-1 structure.Fig. 1: Neural Network with 2-5-1 structure. Heat transferThermodynamic IIRESULTS AND DISCUSSIONThe regression equation as below:Heat transfer = 8.04 + 0.498 Thermodynamics I + 0.408 Thermodynamics II (2)Equation 2 shows that Thermodynamics I is more dominant compare with Thermodynamics II. One can notice that, increase in Thermodynamics I and Thermodynamics II it will increase the result in Heat transfer. Table 2 show that Thermodynamics really significantly effect the heat transfer. It means, those have a very good foundation in Thermodynamics I, they can do better in Heat transfer. The p-value in the Analysis of Variance Table 2 (0.000) indicates that the relationship between Thermodynamics I and Thermodynamics II is statistically significant at an a-level of 0.05. This is also shown by the p-value for the estimated coefficient of Thermodynamics I, which is 0.008 as shown in Table 3.Table 2: Analysis of VarianceFPSSMSSource DF22.64936.93Regression 2 1873.87Residual Error 27 1117.6 41.39Total 29 2991.47Table 3: Estimated coefficientTCoefPSEPredictor Coef0.920.3678.771Constant 8.043Thermodynamics1 0.498 0.1734 2.870.008Thermodynamics2 0.4079 0.2032 2.01 0.055Fig. 2 shows the sensitivity test. The test shows that Thermodynamics I is the main effect for the heat transfer. The results for the sensitivity test and regression analysis show the same results.Fig.2: Sensitivity TestCONCLUSIONThe regression analysis and Neural Network is very useful tool to do analysis in term of measure student performance and importance of prerequisite subject. The results prove that Thermodynamics I effect lot the student performance in Heat transfer. The foundation subject must be very strong, if the students want to perform better in Thermodynamics II and Heat transfer. ACKNOWLEDGEMENTThe authors would like to express their deep gratitude to Universiti Malaysia Pahang (UMP) for provided the financial support.REFERENCESRichard A. Berk, Regression Analysis: A Constructive Critique, Sage Publications (2004)David A. Freedman, Statistical Models: Theory and Practice, Cambridge University Press (2005)R. Dennis Cook; Sanford Weisberg "Criticism and Influence Analysis in Regression", Sociological Methodology, Vol. 13. (1982), pp. 313-361.Lubke, G. H., & Muthén, B. (2005). Investigating population heterogeneity with factor mixture models. Psychological Methods, 10(1), 21-39.Muthen, B. O., & Muthen, L. K. (2000). Integrating person-centered and variable-centered analyses: Growth mixture modeling with latent trajectory classes. Alcoholism: Clinical and Experimental Research, 24, 882-891.Nagin, D., & Tremblay, R. E. (2001). Analyzing developmental trajectories of distinct but related behaviors: A group-based method. Psychological Methods, 6, 18-34.Lazarsfeld, P. F., & Henry, N. W. (1968). Latent structure analysis. Boston: Houghton Mifflin Company.McCutcheon, A. L. (1987). Latent class analysis. Thousand Oaks, CA: Sage Publications, Inc.T. J .Ko, D. W Cho, M. Y. Jung,” On-line Monitoring of Tool Breakage in Face Milling: Using a Self-Organized Neural Network”, Journal of Manufacturing systems, 14(1998), pp. 80-90.J.H. Lee, S.J. Lee,” One step ahead prediction of flank wear using cutting force”, Int. J. Mach. Tools Manufact, 39 (1999), pp 1747–1760.S.K. Chaudhury, V.K. Jain, C.V.V. Rama Rao,” On-line monitoring of tool wear in turning using a neural network”; Int. J. Mach. Tools Manufact, 39 (1999), pp 489–504.D.E. Dimla, P.M. Lister,” On-line metal cutting tool condition monitoring. II: tool state classification using multi-layer perceptron neural network”, Int. J. Mach. Tools Manufact ,40 (2000), pp 769–781。

Mechanical Behavior of Materials

SAN JOSE STATE UNIVERSITYDepartment of Chemical and Materials EngineeringMatE 195 Fall 2000 W. Richard Chung Engr.-385EMechanical Behavior of MaterialsObjectives: The course is designed to help materials engineering seniors understand the basic mechanical responses of engineering materials. Emphasis will beplaced on how to perform various mechanical tests, how to apply statisticalmethods to the analysis of mechanical properties data, and how mechanicalbehavior influences the load-bearing limit for a selected material in a givenapplication.Prerequisites: CE 99, MatE 115 and MatE 141.Class Hours: Lecture on Mondays from 0830 to 1020, in IS113Laboratory on Wednesdays from 0830 to 1120, in E-225Office Hours: Mondays & Wednesdays: 1300-1500, other times by appointment only. Office Room: E-385EOffice Phone: (408) 924-3927E-mail address: wrchung@Textbook: Norman E. Dowling, Mechanical Behavior of Materials, Prentice-Hall, Upper Saddle River, New Jersey, 2nd Edition, 1999. (ISBN0-13-905720-X) References: S.D. Antolovich, R.O. Ritchie, and W.W. Gerberich (editors), Mechanical Properties and Phase Transformations in Engineering Materials, APublication of the Metallurgical Society, Warrendale, Pennsylvania, 1986.(ISBN 0-87339-012-1) TA 401.3 M4155Craig R. Barrett, William D. Nix, and Alan S. Tetelman, The Principles ofEngineering Materials, Prentice-Hall, Englewood Cliffs, New Jersey,1973. (ISBN 0-13-709394-2) TA403.B24.David Broek, Elementary Engineering Fracture Mechanics, MartinusNijhoff Publishers, Hingham, Massachusetts, 3rd Edition, 1984. (ISBN 90-247-2656-5)Thomas H. Courtney, Mechanical Behavior of Materials, McGraw Hill,New York, 2nd Edition, 2000. (ISBN0-07-028594-2) TA405.C859George E. Dieter, Mechanical Metallurgy, McGraw Hill, New York, 3rdEdition, 1986. (ISBN 0-07-016853-8) TA405.D53W.A. Green and M. Micunovic (editors), Mechanical Behavior ofComposites and Laminates, Elsevier Applied Science Publishing, NewYork, 1986. (ISBN 1-85166-144-1) TA418.9C6 E976James M. Gere and Steven P. Timoshenko, Mechanics of Materials,PWS-KENT Publishing, Boston, Massachusetts, 3rd Edition, 1990. (ISBN0-534-92174-4) TA405.G44Richard W. Hertzberg, Deformation and Fracture Mechanics ofEngineering Materials, John Wiley & Sons, New York, 3 Edition, 1989.(ISBN 0-471-63589-8) TA417.6H46Donald Peckner (editor) The Strengthening of Metals, ReinholdPublishing,New York, 2nd Edition, 1967.Grading Basis: There will be two midterm examinations and one final examination.Examinations are comprehensive; including subjects from all assignedreadings, lectures, laboratory activities, and classroom demonstrations.Homework assignments will consist of essay questions and problemsolving cases. The laboratory component affects 25% of the course grade.A term project must be completed and submitted by November 29th. Thedetails will follow.Homework assignments………………………………………………….15%Two midterm exams at 15% each………………..……………..…….….30%Laboratory activities.….…………………………………………………25%Term project with oral presentation……………………………………...15%Final Examination………………………………………………………..15%Total: 100% For all graded work, course letter grades will be assigned according to thecorresponding ranges of cumulative averages listed below.A+ 97 -- 100 A 94 -- 96 A- 90 -- 93B+ 87 -- 89 B 84 -- 86 B- 80 -- 83C+ 77 -- 79 C 74 -- 76 C- 70 -- 73D+ 67 -- 69 D 64 -- 66 D- 60 -- 63F below 60Add/Drop Policy:Students wanting to enroll in the class must sign the roster and receive an enrollment code, provided space is available. Students may drop this classfrom now until September 15 without “W” grade assigned.Important Dates:Midterm examination dates: October 11 and November 15Final exam date: Tuesday, December 19, 0715-0930Term project report submission date: Nov. 29Reserve Desk:The Reserve Desk is located by the book checkout area in the ClarkLibrary. To help your study in the course material, the instructor hasreserved some reference books, technical articles, and supplemental lecturenotes.Homework: Work the homework problems on one side of a sheet of paper only. Youneed to number all the pages if more than one page is submitted. On top ofeach page write down your name, the course number, the semester, and thesubmission date. List the problem numbers in the Dowling textbook andrestate the statement of the problem including simple sketches, ifapplicable. Show your working steps and circle the numeric solutions. It isvery important to have one or two sentences describing your conclusions.This is a brief statement used to state the physical significance orimplication of your answer. Underline them and don’t forget units! Theinstructor will pay additional attention to this requirement. The homeworkassignments are collected in class on Sept. 25, Oct. 23, and Nov. 20. Nolate assignments will be accepted, as the problem solutions will be postedimmediately after the class due date.Laboratory: A Laboratory Activity Logbook will be purchased and kept by eachstudent. It must be brought to be checked and initialed by the instructorbefore beginning the first lab exercise. The logbook must be 8 1/2 x 11inches, NOT spiral bound. You will use it to record a detailed log of all labactivities, data, sketches of experimental setups and results, and records ofreferences used for class projects. Mark it clearly on the cover with yourname, group number, class and section, instructor, and semester. Eachpage should be numbered and dated and each lab activity labeled.Record your partners’ names, phone numbers, and schedules inside thecover.Term Project: A Term Project must be completed. The term project will involve amaterial's testing activity, which you design, initiate, and conduct in agroup (not more than three members) and must be supported by costanalysis, technical drawings, and related references. A few topics will bediscussed in class at a later date. Completed term project reports will befrom 10 pages in length, double spaced not including illustrations orappendices, and will follow the class format. A writing format will beprovided at a later date. An oral presentation on the term project must beconducted in the end of class (November 27).A group's oral presentation on the term project is expected to last at least20 minutes, followed by a 5-minute discussion period. All members of agroup project must present together, but are graded separately. Thepresentation should be technical and include view graphs or visual aidsrelated to the chosen subject area. Transparencies, films, LCD projector,and/or VCR recordings (VHS) can be used, but may not replace spokenreporting. A guideline with tips of presentation requirements will bedistributed at a later date.Mat E 195 Course Activity OutlineFinal Examination on Tuesday, December 19th from 0715 to 0930 hoursWeek StartingDate Reading: Chapter #Homework AssignmentsLaboratory1 Aug. 28 Ch.1 IntroductionProb. 1.1, 1.3, and 1.5 Lab Tour/Safety 2 Sept. 4 Labor Day --No Class on Sept. 4 Prob. 5.2, 5.5, 5.10Types of Material Failure3Sept. 11Ch.5 Stress-Strain Relationships Prob. 5.20, 5.22, 5.23 Tension Tests 4 Sept. 18Ch.5 Stress-Strain RelationshipsCh. 4 Mechanical Testing (Tension Test)Prob. 5.26, 5.30, 5.32 Mechanical Tests(Impact, Izod,Hardness, etc.) 5 Sept. 25 Ch. 4 Mechanical Testing (Tension Test)Prob. 4.4, 4.5, 4.7Problem Solving6 Oct. 2 Ch. 4 Mechanical Testing (Other Tests)Prob. 4.16, 4.18, 4.28 DMA 7 Oct. 9 Ch. 4 Mechanical Testing (Other Tests)Prob. 4.31, 4. 34, 4.37, 4.38 Exam 1 – Oct.11 8 Oct. 16 Ch.6 Complex and Principal States of Stress and StrainProb. 6.1, 6.7, 6.10Mohr’s Circles/ Term project9 Oct. 23 Ch.6 Complex and Principal States of Stress and Strain Prob. 6.14, 6.16, 6.28 SEM Fractography 10 Oct. 30 Ch. 8 Fracture of Cracked MembersProb. 8.3, 8.4, 8.7 2024-TT6 Aluminum 11 Nov. 6 Ch. 9 Fatigue of Materials Prob. 9.6, 9.7, 9.14 Fatigue Test 12 Nov. 13 Ch. 9 Fatigue of Materials Prob. 9.16, 9.20 Exam 2 –Nov. 15 13 Nov. 20 Ch. 11 Fatigue Crack Growth Prob. 11.4, 11.8, 11.13Polymers/Composites 14 Nov. 27 Ch. 12 Plastic DeformationBehavior and Models for Materials Prob. 12.1, 12.13 Term Paper Due (Nov. 29) 15 Dec. 4 Ch. 15 Time-Dependent Behavior: Creep and Damping Prob. 15.2, 15.6, 15.16 Problem Solving 16Dec. 11Review of the CourseLast Day of InstructionMatE 195 Course Learning ObjectivesUpon the completion of this course, the student will be able to:1. Understand the basic test methods to characterize the mechanical behavior of engineering materials: tension, compression, hardness, impact, fatigue, and creep.2. Apply the basic theory of elasticity and plasticity and the importance of brittle-ductile transformation and elastic and plastic behavior of materials to industrial applications.3. Learn the concept of fracture mechanics and its application to product design, manufacturing method, and service reliability.4. Perform the mathematical calculation of a multi-axial or complex stress state and relate it to the uni-axial stress state and the yielding condition.5. Describe and predict the mechanical behavior of crystalline solids using the concepts of dislocation theory and a micro-mechanical approach.6. Improve fracture toughness and deflect a crack’s propagation through the understanding of the microstructural alignment and the associated mechanical anisotropy.。

清华校友中的美国名校教授

【致谢】感谢参与编写和提供信息的以下校友：IhateBS strong根据美国US News 2013年美国大学排名前100的顺序统计，缺漏错误之处难免，请大家补充分为Full Professor（正教授）、Associate Professor（副教授）和Assistant Professor三类，其中前两类一般具有终身教职，不含解放前的清华学子，也含有极少数非tenure track教职人员,和已离职的人员，只列出，不计入统计总数。

共有在职Tenure track系列教职的共552人，其中95名正教授。

副教授207人另有部分著名实验室全职研究人员54人，非常不全。

----------------------------------------------------------------------------1,哈佛大学7林希虹Professor, 生物统计系，美国统计学会会士，统计学最高奖COPSS总统奖获得者1989年清华大学数学系学士，1994年华盛顿大学博士/xihong-lin/刘军Professor, 生物统计系，美国统计学会会士，统计学最高奖COPSS总统奖获得者1985－1986年清华大学数学系研究生/~junliu/Yiling ChenAssociate Professor of Computer Science1996年人民大学经济学学士，1999年清华大学金融硕士，2005年宾州州立大学信息科学技术博士/Wei-Dong YaoAssistant Professor of Psychiatry清华大学物理系学士，1998年爱荷华大学博士/Profiles/display/Person/4435Hesheng LiuAssistant Professor of Radiology at Harvard Medical School2003年清华大学生医工程系博士/martinos/people/showPerson.php?people_id=636Jinjun ShiAssistant Professor of AnaesthesiaThe Harvard Clinical and Translational Science Center2000年清华大学化学学士，德州农机大学博士/Profiles/display/Person/43651李全政Assistant Professor , Harvard medical school1992年浙江大学生医工程学士，1997年清华大学生医工程硕士，南加州大学博士/CAMIS/?page_id=1628----------------------------------------------------------------------------1,普林斯顿大学2琚诒光Robert Porter Patterson Professor，普林斯顿大学机械与航空工程系1986年清华大学工程力学系学士,1988年硕士，1994年日本东北大学博士/People_files/Ju-resume.htm//2014年7月开始王梦迪Assistant Professor in Operations and Financial Engineering2008年清华自动化学士，2013年MIT博士/research/news/archive/?id=11152********施一公（已辞职）Professor, 分子生物系，全球蛋白质学会Irving Sigal青年科学家奖获得者。

布里斯托大学-校园平面图

45 44
62 61 60
59 58
57
55
56
106
ST MICHAEL’S
PARK
41
40
43
42 TYNDALL
AVENUE
37
36
ELTON ROAD
33
32
Bristol
34
Grammar
SOUTHWELL STREET
P
ROYAL FORT RD
பைடு நூலகம்
38 35 St Mi8c5hael’s H o s p i?t?ailt a l 31
KINGSDOWN PARADE
ALEXANDRA ROAD LEIGH ROAD
London, S. Wales, Midlands and The West
99
BELGRAVE
ROAD
77 76
PORTLAND ST
Steps Pedestrian walkways
ALFRED PLACE
OAKFIELD ROAD
Road _________________________________________________________ 68 Advanced Composites Centre for Innovation and Science (ACCIS), Queen's Building 20 Advanced Studies, Institute for, Royal Fort House _________________________ 30 Aerospace Engineering, Queen’s Building _______________________________ 20 ALS

阿尔法拉維（Alfa Laval）三花型（Tri-Clover）和三綜合型（Tri-Weld）適用於

connections.Alfa Laval offers a full line of UltraPure Fittings that aremanufactured in compliance with the current ASME BPE Standard.All BPE items are individually capped and bagged in clear6mil.Poly bags.All product is labeled with a bar code,product informationand manufacturing date.This provides the optimum identification and ensures that the product arrives to the job site in a clean orbital weld condition.Technical DataWide Range of Surface Finish offering-Alfa Laval offers a rangeof Mechanical Polish as well as Electropolish finishes.Mechanical polishing is achieved by using a progressive series of abrasives,from low to high grit.This allows a consistent internal finish and both optimal and economical cleaning.Electropolishing is a further process that promotes a chromium-enriched surface layer that maximizes corrosion resistance as well as minimizing bacterial buildup on surface cavities. Metallurgy-Incoming raw material goes through a stringent inspection process to ensure its chemistry will be ideal for both weldability and electropolishing Quality Control Methods-Our manufacturing facilities operate under an approved ISO9001quality standard.Wall thickness integrity is maintained through the use of fabrication grade minimum wall tubing for all cold-formed tubular products.Our BPE fittings are designed for use with all current orbital welding equipment.After cold forming, our tube product is resized to ensure that the ovality falls within the prescribed BPE tolerances.End facing is provided with a machined square-cut method.This allows for the most accurate and consistent orbital weld result.All fittings are put through100%visual inspection and ovality and squareness tolerances are inspected with calibrated equipment.Surface finish is inspected with a calibrated profilometer to ensure the Roughness average(Ra)maximum is not exceeded. Hygenic fittings identified with this symbol on the following pages are accepted as meeting the3A Hygenic standards by the appropriate committees of the International Association of Milk,Food and Environmental Sanitarians,U.S.Public Health Service,and Dairy Industry Committee.ID or Product Contact SurfaceMaximum Surface Roughness (Ra)Finish code Microinches (µ-inch)Micrometers (µm)ASME BPE Finish CodePolishing Method OD or Product Non-contact Surface#1UnpolishedUnpolished #3320.8-----Mechanical polished Unpolished#7320.8-----Mechanical polished Polished to Ra,32µ-inch/0.8µm PC 200.5SF1Mechanical polishedUnpolished PD 150.4SF4Mechanical polished and electropolished UnpolishedPL 200.5SF1Mechanical polishedPolished to Ra,32µ-inch/0.8µm PM150.4SF4Mechanical polished and electropolishedPolished to Ra,32µ-inch/0.8µmService Rating of Tri-Clamp ®ConnectionsService Ratings*(PSI)Size Tube OD½&¾inch1&1½inch 2inch 2½inch 3inch4inch 6inch 13MHLA (Screw tightened to maximum)at 70°F --150150150150100--at 250°F --12512512512575--13MHHM (Wing nut tightened to 25in.lb.of torque)at 70°F --500450400350300150at 250°F --3003002001951507513MHHS (Wing nut tightened to 25in.lb.of torque)at 70°F 2200600550450350300--at 250°F 1200300275225175150--13MHP (Bolts tightened to 20ft.lb.of torque)at 70°F --1500100010001000800300at 250°F --1200800800800600200A13MO (1-3"nuts tightened to 20in.lb.,4"to 30in.lb.)at 70°F --50035030020010075at 250°F --25020015010010050A13MHM (Wing nut tightened to 25in.lb.of torque)at 70°F --500450400350300150at 250°F--30025020017515075*Service ratings are based on hydrostatic tests using standard-molded Buna-N material gaskets,with proper installation of ferrules,assembly of joints and absence of shock pressure.Contact Tri-Clover ®for service of other type and material gaskets,and for ratings at higher temperatures.All ratings shown are dependent upon related components within the systems and proper installation.For temperatures above 250°F,we recommend using only 13MHP clamps.This information is only valid if Tri-Clover ®clamps,ferrules,and gaskets are used.Tri-Clamp ®Gasket MaterialsCharacteristicBuna-N (U)EPDM (E)Fluoro-elastomer(SFY)Silicone (X)PTFE (G)Hardness,Shore A 70707070---Tensile Strength,psi 1875165012121340---Original Physical Properties Elongation,%340317272260---Temperature Range-65to 200°F -60to 300°F -20to 350°F -40to 400°F -40to 200°F *Acid Resistance Good Good to Excel.Good to Excel.Poor to Good Good to Excel.Alkali ResistanceFair to Good Good to Excel.Poor to Good Poor to Fair Excellent Resistance to Fats/Oils Good to Excel.Poor Good to Excel.Poor to GoodExcellent Abrasion Resistance ExcellentGoodGood to Excel.Poor Fair ResistanceCompression Set ResistanceGoodFairGood to Excel.Good to Excel.Cold Flows*Note:PTFE materials tendency to "cold flow"and incompressibility,limit its max.temperature to 200°F due to possible leaking problems.Basic Dimensions of Tri-Clamp®Connection for Hygenic OD-TubingOD Outer Diameter(Inches)ID Inner Diameter(Inches)Wall Thickness(Inches/Gauge)A Ferrule Face(Inches)½0.370.065/16ga.0.984¾0.620.065/16ga.0.98410.870.065/16ga. 1.9841½ 1.370.065/16ga. 1.9842 1.870.065/16ga. 2.5162½ 2.370.065/16ga. 3.0473 2.870.065/16ga. 3.5794 3.870.083/14ga. 4.682 Hygenic Tube InformationTube OD Tube ID Wall Thickness Volume Weight Dry Weight withWaterFlow(GPM)at a Mean VelocityInches Inches Inches Gal/100ft Lbs/100ft Lbs/100ft5fps7fps10fps ½0.370.0650.5630.635.3 1.7 2.3 3.4¾0.620.065 1.5748.261.3 4.7 6.69.410.870.065 3.0965.891.59.313191½ 1.370.0657.66100.9164.82332462 1.870.06514.27136.1255.14360862½ 2.370.06522.92171.2362.469961383 2.870.06533.6206.4486.71011412024 3.8340.08359.97351.8851.91802523606 5.7820.109136.39694.71832.240957381887.7820.109247.07930.62991.174110381482Technical InformationPipe Schedule and Chemical CompositionSchedule5PipeSize OD Inches ID Inches Wall Thickness ⅛0.4050.3350.035¼0.5400.4420.049⅜0.6750.5770.049½0.8400.7100.065¾ 1.5000.9200.0651 1.315 1.1850.0651¼ 1.660 1.5300.0651½ 1.900 1.7700.0652 2.375 2.2450.0652½ 2.875 2.7900.0833 3.500 3.3340.0833½ 4.000 3.8340.0834 4.500 4.3340.0835 5.563 5.3450.1096 6.625 6.4070.109 88.6258.4070.109Chemical Composition%304316LC0.0800.030MN 2.000 2.000P0.0450.045S0.0300.030*Si 1.000 1.000Cr18.0-20.016.0-18.0Ni8.0-13.010.0-14.0Mo- 2.0-3.0*The sulfur content for316L ASME BPE fittings is0.005-0.017%for all weld ends.Material Test Reports(MTRs)Easy Online Access to Comprehensive FittingsInformationA5-alpha character serial ID is stenciled on to each new316SS fitting As one of the most comprehensive and technologically advanced reports in the market,our new Material Test Reports(MTRs)provide detailed information that takes traceability and validation to a new level. Alfa Laval has established a new standard as all MTRs are available24 hours a day,7days a week online at .Simply type a5-alpha character code(e.g.AAABC)called the serial ID,which you can find stenciled on each new316SS fitting,to access the following information:•All heat certification numbers used to manufacture the fitting •Date the fitting was manufactured•The fitting’s part number and description•View and print any MTR and the above informationThis web site will even allow you to print the MTR or original heat certification from the raw material supplier.If you do not know the actual number,MTRs can be searched by either MTR serial ID or heat certification number.Go to and follow these simple steps to access MTRs: Step1.Once at our website,click on the MTRlinkStep2.On the MTR page,click"View Material TestReportsStep3.Enter or search for the SerialIDConnection TypesClamp FittingsTri-Clamp HDI-Line H-LineA connection is made up of a plain ferrule,a clamp,and a gasket.Tees,elbows and reducers are available with Tri-Clamp connections.All three styles are in compliance with3A standards for C.I.P.(clean in place).The three types of clamp fittings are designed for use in Food,Dairy,Pharmaceutical and Chemical Industries.•Tri-Clamp connections are the industry standard,having nueter-style ferrules to simplify design and installation.•H-Line and HDI-Line male/female ferrules self-align during tightening so joints are quick and easy to assemble or take apart.•H-Line uses the same series of clamps as the Tri-Clamp.Threaded FittingsBevel Seat John Perry DCA connection is made up of a plain ferrules,a threaded ferrule,a nut and a gasket.The faces on Bevel Seat fittings are angled to create a metal to metal sealing surface.A John Perry fitting consistes of a flat-faced threaded ferrule,a flat-faced plain ferrule and a profiled gasket.These joints are particularly useful with swing connections and flow diverter panels.A DC fitting utilizes the Bevel Seat plain ferrule and a threaded ferrule with a grooved face to retain a gasket.The three types of threaded fittings are designed for use in the Food,Dairy,and Beverage processing industries.Bevel Seat Joints are in compliance with3A standards for manual cleaning.Both John Perry and DC fittings are in compliance with3A standards for C.I.P.(clean-in-place).•Bevel Seat•John Perry•DCLoss of head pressure due to friction.Loss is shown in feet ofhead.Loss through tubing is for1ft.of tubeCapacity O.D.1"O.D.1½"O.D.2"O.D.2½"O.D.3"O.D.4"in U.S.I.D0.902"I.D. 1.402"I.D. 1.870"I.D. 2.370"I.D. 2.870"I.D. 3.834"G.P.M.Tubing Elbow Tee Tubing Elbow Tee Tubing Elbow Tee Tubing Elbow Tee Tubing Elbow Tee Tubing Elbow Tee20.010.010.140.0250.020.250.0350.0250.25100.120.060.40.020.010.150.0050.0150.1150.250.10.80.040.020.250.0130.020.15200.430.22 1.50.060.030.30.020.0250.20.0050.020.10.0030.020.06250.660.4 2.30.080.040.40.0250.030.250.0060.030.150.0040.030.08300.930.7 3.30.1050.060.550.0350.050.30.0080.050.20.0050.040.135 1.22 1.25 5.20.1350.090.80.040.060.40.0110.060.250.0060.050.13400.170.11 1.00.050.080.50.0150.070.30.0070.060.15450.210.16 1.30.0630.10.60.020.090.350.0080.0650.18500.250.2 1.60.0730.120.70.0220.10.40.010.070.2600.340.35 2.20.10.180.90.030.120.450.0150.080.25800.570.76 3.70.160.3 1.50.050.150.550.020.10.41000.85 1.35 5.80.230.44 2.30.0750.180.60.030.110.50.0080.040.1 120 1.18 2.059.10.320.64 3.30.1050.21 1.00.040.130.60.010.050.15 1400.420.85 4.50.140.23 1.250.050.160.80.0130.060.2 1600.54 1.13 5.80.170.28 1.60.070.2 1.10.0150.070.25 1800.67 1.457.40.2050.31 2.00.080.21 1.30.020.080.3 2000.81 1.829.00.2450.35 2.50.10.26 1.60.0250.090.4 2200.95 2.2211.00.290.41 3.00.120.3 1.90.0280.10.5 240 1.10 2.6313.50.340.48 3.70.140.33 2.20.0350.110.55 2600.390.53 4.50.1650.39 2.50.040.1150.6 2800.450.61 5.30.190.42 2.80.0450.120.65 3000.5150.7 6.20.220.5 3.10.050.130.7 3500.68 1.058.50.280.67 4.10.070.150.9 4000.86 1.5511.00.360.88 5.20.0850.18 1.2 450 1.05 2.2513.50.44 1.1 6.60.1050.2 1.5 5000.54 1.48.00.130.23 1.75 5500.64 1.79.50.150.27 2.1 6000.75 2.0510.20.1750.3 2.5 6500.87 2.4113.00.20.34 2.8 700 1.0 2.815.00.230.4 3.4 7500.260.43 3.8 8000.30.5 4.4 8500.330.56 5.0 9000.370.62 5.7 9500.410.7 6.3 10000.450.87.0 11000.53 1.068.6NOTES: 1.For elbows-R/D=1.5 2.Flow thru teesFlow A to B3.Test medium-water at70°F Port C capped offPrepared by members of the hygenic pump subgroupof the natl.assn.of dairy equipment manufacturers.ESE00301ENUS1507Alfa Laval reserves the right to change specifications without priornotification.ALFA LAVAL is a trademark registered and owned by Alfa LavalCorporate AB.©Alfa LavalHow to contact Alfa LavalContact details for all countriesare continually updated on our website. Please visit toaccess the information direct.。

HYDRO-MECHANICAL HYDRAULIC HYBRID DRIVE TRAIN WIT

专利名称：HYDRO-MECHANICAL HYDRAULIC HYBRID DRIVE TRAIN WITH INDEPENDENT WHEELTORQUE CONTROL发明人：LI, Perry, Y.,OLSON, Michael,TRUE, Kyle,TOY, Charles申请号：US2008004618申请日：20080410公开号：WO08/133805P1公开日：20081106专利内容由知识产权出版社提供摘要：A hydro-mechanical hybrid drive train (100, 200, 300) is provided The drive train includes a prime mover (102) that has an output (104 A first hydraulic pump/motor unit (110) is operably coupled to the output (104) of the prime mover (102) The first hydraulic pump/motor unit (110) has a mechanical output (112) A mechanical transmission (114) has a plurality of gear ratios, an input shaft and a mechanical output (116) The input shaft is coupled to the mechanical output (112) of t4e first hydraulic pump/motor unit (110) A planetary differential (118) is mechanically coupled to the output (116) of the transmission (114) and provides first and second mechanical outputs A second hydraulic pump/motor unit (130) is coupled to the first mechanical output of the planetary differential (118) At least one drive element (126) is operably coupled to the second mechanical output of the planetary differential (118)申请人：VAN DE VEN, James, D.,LI, Perry, Y.,OLSON, Michael,TRUE, Kyle,TOY, Charles 地址：Office for Technology Commercialization 1000 Westgate Drive, Suite 160 St. Paul, MN 55114 US,111 Church St. SE Minneapolis, MN 55455 US,111 Church St. SEMinneapolis, MN 55455 US,111 Church St. SE Minneapolis, MN 55455 US,111 Church St. SE Minneapolis, MN 55455 US,111 Church St. SE Minneaplis, MN 55455 US 国籍：US,US,US,US,US,US代理机构：CHRISTENSON, Christopher, R更多信息请下载全文后查看。

Maximum Power Tracking for PVs with Improved Track

Journal of Energy and Power Engineering 6 (2012) 972-977Maximum Power Tracking for PVs with Improved Tracking Accuracy at Irradiance TransientsAthmi Jayawardena, Dilini Weerasinghe, Sanchala Coorera, Sunil Abeyratne and Nimal RathnayakeDepartment of Electrical and Electronic Engineering, Faculty of Engineering, University of Peradeniya, Peradeniya 20400, Sri LankaReceived: February 01, 2011 / Accepted: August 15, 2011 / Published: June 30, 2012.Abstract: PV (photovoltaic) systems need MPPT (maximum power point tracking) techniques to harness maximum power from PV arrays. P&O (perturb & observe), and incremental conductance methods are two basic MPPT algorithms applied to PV systems with fixed and variable step sizes. However, the existing variable step MPPT method exhibits complications which occur in the algorithm due to sudden, large irradiance changes which result in dips in power extraction. This paper proposes a modification to the existing variable step MPPT method to avoid such complications and hence improve the tracking accuracy under irradiance transients. The proposed technique is experimentally verified under sudden irradiance disturbances using a solar array. The results are compared with the existing variable step method. The superiority of the proposed technique is demonstrated through a laboratory prototype. Key words: MPPT, variable step, irradiance transient, solar array.1. IntroductionElectrical power generation from PV (photovoltaic) arrays has achieved a significant position among the other renewable energy conversion technologies. The solar array has an inherent nonlinear characteristic with a peak power point which varies with the solar irradiance, temperature and aging of the PV array. Therefore, it is essential to use a MPPT (maximum power point tracking) algorithm to track the maximum power point automatically, thereby harnessing the maximum available power from the PV array.In the recent past, many MPPT technologies have been developed which vary in complexity, speed, cost, effectiveness and hardware etc. [1-10]. Wide variety of MPPT methods are being used in PV applications, starting from the basic P&O (perturb & observe) method to Fuzzy logic controls, with their own advantages and disadvantages [1].Corresponding author: Sunil Abeyratne, Ph.D., senior lecturer, research fields: power electronics, electrical machine anddrives.E-mail:************.ac.lk.In the literature, several approaches using fixed step sizes have been proposed by improving P&O [1, 3] and incremental-conductance [1, 2] methods. When tracking the MPP, a small fixed step size will cause the tracking speed to decrease and tracking accuracy of the MPP will decrease due to large fixed step size and sudden large irradiance changes.To overcome most of the above disadvantages, a variable step method combined with incremental conductance method has been proposed [2], which calculates a variable step at each operating point according to its position on the characteristic curve. This variable step method tracks the MPP faster than the fixed step methods [2], but the authors have found that sudden large irradiance changes cause the system to oscillate due to incorrect calculations. The problem is further explained in this paper with a proposed solution. The proposed algorithm is verified by experiment using a solar panel.2. Theory of Variable Step MethodOutput characteristics of a solar array are illustratedll Rights Reserved.Maximum Power Tracking for PVs with Improved Tracking Accuracy at Irradiance Transients973in Fig. 1.0/<dv dp In voltage source region 0/=dv dp At MPP0/>dv dp In current source region()()dv di v i dv v i d dv dp ⨯+=⨯= (1)Duty cycle of the pulse applied to the boost converter is calculated using Eq. (2).new previous r DutyCycle DutyCycle N =+ (2)where, N r is the calculated variable step size using instantaneous voltage, current and dv/di values of the solar array, (Appendix) [2]. Here, limitations occur when calculating variable steps (N rmin , N rmax ). There is a minimum step size due to the resolution of the IC used, noise and switching frequency. As N approaches -∞, a maximum step size must be defined to avoid instability in the voltage source region.3 Effects of Sudden Large Irradiance ChangesConsider the present operating point to be as point A on the curve 1 in Fig. 2. If the irradiance changes with this constant load (i.e. with no duty cycle change of the converter), operating point will move to B on curve 2. However, in order to calculate the next variable step size, the duty cycle needs to be altered. For this, dv/di relevant to point A will be taken into account, consequently shifting the load curve to curve 3. With the irradiance change, next operating point will be point D instead of point C. In order to move from point D, the existing algorithm will use the value of dv/di indicated by anarrow in Fig. 2, which is incorrect. This will cause the system to have unnecessary power variations. The accurate dv/di must be calculated at point D along the curve 2.4. Proposed MPPT Algorithm4.1 Behavior of Solar Array Operating Point under Constant IrradianceDue to variable step sizes, the operating point gets closer to the MPP quickly. When the operating point is closer to the MPP, it will come to a region whereFig. 1Output characteristics of a solar array.Fig. 2 Wrong calculation of dv/di due to sudden irradiance change.the calculated Nr is less than the minimum Nr . Hence, the operating point will be moved by the minimum Nr within this region (within load curve 1 and load curve 2 in Fig. 3) until an irradiance change occurs. The dotted curves in Fig. 3 represent the variations of load curve seen by the solar array due to the variation of duty cycle by minimum value.4.2 Behavior of Solar Array Operating Point under Large Irradiance ChangesFor the characteristic curve of the solar array, the gradient of the curve at each operating point is negative.0<di dv (3) Hence, 1N < (4) This condition is used to identify a large irradiance change. If the solar array is operating in the current source region, the duty cycle will cause the operating point to be shifted to the right and if it is in the voltagell Rights Reserved.Maximum Power Tracking for PVs with Improved Tracking Accuracy at Irradiance Transients974Fig. 3 Variations of load curve by minimum step size around the MPP.source region, the operating point will be shifted towards left.Since the variable step is calculated using the N value (Nr = C *N ) (Appendix), the largest step size for the current source region will be calculated at the short circuit condition. After a sudden, large irradiance change, the operating point will be moved to a different characteristic curve (Fig. 2). But, when N > 1 (due to dv/di > 0), a large step size will becalculated which will cause the operating point to beshifted far away from the MPP. Hence, to track theirradiance change accurately, the condition N > 1 or dv/di > 0 is checked inside the proposed algorithm. As detailed in Fig. 4, after identifying a sudden, large irradiance change, the operating point will be forced to shift by N r min step along the new irradiance curve. This is done in order to ensure that the calculated dv/di is on the new characteristic curve of the solar array. This principal is applied to both current source and voltage source regions.4.3 Behavior of Solar Array Operating Point under Small Irradiance ChangesConsider Fig. 4 for irradiance changes in current source region. The horizontal and vertical lines drawn through the present operating point P shows the limits where the condition di/dv > 0 can be applied to track large irradiance changes. In areas A-B and A’-B’, dv/di is less than 0. Thus, to track the MPP in those regions, di/dv > 0 or N> 1 criterion can not be applied.Fig. 4 Variations of the operating point under large irradiance changes.Instead, as explained in Fig. 5, the MPP will be tracked by checking the previous step size and present step size inside the algorithm. The same principle is applied to the voltage region as well.Fig. 5 shows the detailed flow chart of the proposed MPPT algorithm. This new method can identify a sudden large irradiance change, so that generating incorrect variable step size will be avoided.5. Implementation Details of the Experimental SetupFig. 6 illustrates the details of the experimental setup which was used to test the proposed MPPT method. A solar array which has a 6 V maximum open circuit voltage was used. The system used a boost converter and a 6 V battery. 230 V DC supply was used to illuminate filament bulbs to simulate the sunlight. Different irradiances were obtained by switching different number of bulbs. The converter was switched by 150 kHz signal.6. Experimental ResultsFig. 7 shows the unnecessary duty cycle variations in the existing variable step method, under sudden large irradiance changes. But, this problem does not occur in the proposed method.When the calculated duty cycle is inaccurate, the solar array output power will also vary significantly. Fig. 8 shows the corresponding power variations of the solar array due to duty cycle variations for threell Rights Reserved.Maximum Power Tracking for PVs with Improved Tracking Accuracy at Irradiance Transients 975Fig. 5 Flow chart of the proposed MPPT algorithm.Maximum Power Tracking for PVs with Improved Tracking Accuracy at Irradiance Transients 976Fig. 6 Experimental setup.Fig. 7 Comparison of duty cycle variations due to sudden large irradiance changes.Fig. 8 Comparison of power variations of the existing variable step method and the proposed algorithm. different irradiance levels. When using the existing variable step method, a considerable reduction in output power can be seen during a sudden large irradiance change. But, it is apparent from the results that this error does not occur in the proposed algorithm. Both increase and decrease in irradiance levels are considered for the verification of the proposed method. Current and voltage are fed to the microcontroller and converted from analog to digital signals. Thus, the power levels in Fig. 8 are calculated using those digitized current and voltage levels.7. ConclusionsIn this paper, an improved variable step MPPT algorithm has been introduced, which can avoid unnecessary duty cycle variations and incorrect tracking that would result in reduced power capturing during sudden large irradiance changes. It has been experimentally tested and compared with the variable step MPPT algorithm.Verification has been done using a solar panel, a boost converter, a 6 V battery and a DC supplied bulb panel. Through the experimental results, it has been proven that the proposed algorithm avoids the confusion occurring in variable step MPPT method due to sudden large irradiance changes.References[1]T. Esram, P.L. Chapman, Comparison of photovoltaicarray maximum power point tracking techniques, IEEETransaction on Energy Conversion 22 (2) (2007) 439-449.[2]J.H. Lee, H.S. Bae, B.H. Cho, Advanced incrementalconductance MPPT algorithm with a variable step size, in:12th International Power Electronics and Motion ControlConference, 2006, pp. 603-607.[3]M.A.S. Masoum, H. Dehbonei, E.F. Fuchs, Theoreticaland experimental analyses of photovoltaic systems withvoltage- and current-based maximum power-point tracking, IEEE Transaction on Energy Conversion 17 (4)(2002) 514-522.[4]J.A. Jiang, T.L. Huang, Y.T. Hsiao, C.H. Chen,Maximum power tracking for photovoltaic power systems, Tamkang Journal of Science and Engineering 8(2) (2005) 147-153.ll Rights Reserved.Maximum Power Tracking for PVs with Improved Tracking Accuracy at Irradiance Transients977[5] M. Azab, A new maximum power point tracking forphotovoltaic systems, Proceedings of World Academy of Science, Engineering and Technology 34 (2008) 2070-3740.[6] A. Yafaoui, B. Wu, R. Cheung, Implementation ofmaximum power point tracking algorithm for residential photovoltaic systems, in: 2nd Canadian Solar Buildings Conference, 2007.[7] X. Wang , A.P. Hu, An improved maximum power pointtracking algorithm for photovoltaic systems, in: Australasian Universities Power Engineering Conference (AUPEC), 2004.[8] E. Koutroulis, K. Kalaitzakis, N.C. Voulgaris,Development of a microcontroller-based, photovoltaic maximum power point tracking control system, IEEE Transaction Power Electronics 16 (2001) 46-54.[9] D. Sera, R. Teodorescu, J. Hantschel, M. Knoll,Optimized maximum power point tracker for fast-changing environmental conditions, in: IEEE International Symposium on Industrial Electronics, 2008, pp. 2401-2407.[10] W. Xiao, W.G. Dunford, P.R. Palmer, A. Capel,Application of centered differentiation and steepest descent to maximum power point tracking, IEEE Transaction on Industrial Electronics 54 (5) (2007) 2539-2549.Appendix: Variable step calculationIf,()()didv v N +=110<<N In current source region0=N At MPP0<<∞-N In voltage source regionThe variable step Nr is defined as follows,C N N r ⨯=where C is a user defined conversion factor.ll Rights Reserved.。

TomaszWierzbicki

Tomasz WierzbickiProfessor of Applied MechanicsDirector, Impact and Crashworthiness LaboratoryRoom 5-218AMassachusetts Institute of Technology77 Massachusetts AvenueCambridge MA 02139-4307Phone: 617-253-2104Fax: 617-253-8125Email: *************Administrative Contact:Barbara SmithRoom 5-320Phone: 617-253-0137Email: **************Education:Ph.D. in Applied Mechanics, 1965Institute of Fundamental Technological Research, Warsaw, PolandS.M. in Engine Design, 1960Warsaw Technical University, Warsaw, PolandB.S. I took a unified program that led directly to the Master of Science.MIT Service:1983 to date: Professor of Applied Mechanics, Department of Ocean Engineering, MIT; currently the Department of Mechanical EngineeringPrincipal Publications in last five years: (Selected from last two years)1.Elham Sahraei1, Rich Hill, Tomasz Wierzbicki, (2011) Calibration and finite elementsimulation of pouch lithium-ion batteries for mechanical integrity, Journal of PowerSources, Volume 201, 1 March 2012, Pages 307–3212.Luo, M., Wierzbicki, T. (2010). “Numerical failure analysis of a stretch-bending test ondual-phase steel sheets using a phenomenological fract ure model”, International Journal of Solids and Structures, Volume 47, Issue 22-23, Pages 3084-3102.3.Li, Y., Wierzbicki, T., Sutton, M. et al., (2010). “Mixed mode stable tearing of thin sheetAl6061-T6 specimens: experimental measurements and finite element simulations usinga Modified Mohr-Coulomb fracture criterion”, International Journal of Fracture, 168(1),53-71.4.Li, Y., Luo, M., Gerlach, J. and Wierzbicki T. (2010). “Prediction of Shear-InducedFracture in Sheet Metal Forming“, Journal of Materials Processing Technology, Volume 210, Issue 14, Pages 1858-1869.5.Li Y., Wierzbicki T., 2010. “Prediction of plane strain fracture of AHSS sheets with post-initiation softening”, International Journal of Solids and Structures, Volume 47, Issue 17, Pages 2316-2327.6.Beese, A. M., Luo M., Li, Y., Bai, Y., Wierzbicki, T, "Partially coupled anisotropicfracture model for aluminum sheets", Engineering Fracture Mechanics, Volume 77, Issue 7, Pages 1128-1152 (2010).7.Bai, Y., Wierzbicki, T., “Application of the extended Coulomb-Mohr model to ductilefracture”, International Journal of Fracture, Vol. 161, p.1-20 (2010)..8.Y. Bao and T. Wierzbicki (2005), "On the cut-off value of negative triaxiality forfracture", Engineering Fracture Mechanics, 72(7): 1049-1069.9.Bao, Y.B. and Wierzbicki, T. “On fracture locus in the equivalent strain and stresstriaxiality space,” International Journal of Mechanical Sciences, 2004, 46(1): 81-98. Scientific & Professional Societies:Society of Naval Architects and Marine EngineersAmerican Society of Mechanical EngineeringInternational Society of DEHonors & Awards:Maximilian T. Huber Award for the best work in Mechanics, Polish Academy of Sciences, 1974 Chairman of the Euromech Colloquium No. 121 on "Dynamics and Crushing of Plastic Structures", 1978Chairman of the Summer School on "Dynamics of Plastic Structures", International Center for Mechanical Sciences, Udine, Italy, 1979Polish Academy of Sciences award for the book "Design of Structures to Dynamic Loads", 1979 Co-chair, First International Symposium, "Structural Crashworthiness", UK, 1983Co-chair, Second International Symposium, "Structural Failure", Cambridge, MA, 1988 Alexander von Humboldt Foundation, Senior US Scientist Award, 1988-1989Co-chair, Third International Symposium, "Structural Crashworthiness and Failure", UK, 1993 Member of the Editorial Boards of the International Journal of Impact Engineering and International Journal of Vehicle Design。

Out-of-round railway wheels-a literature survey

/Transit Engineers, Part F: Journal of Rail and RapidProceedings of the Institution of Mechanical/content/214/2/79The online version of this article can be found at:DOI: 10.1243/09544090015313512000 214: 79Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit J. C. O. Nielsen and A JohanssonOut-of-round railway wheels-a literature surveyPublished by: On behalf of:Institution of Mechanical Engineers can be found at:Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid TransitAdditional services and information for/cgi/alerts Email Alerts:/subscriptions Subscriptions: /journalsReprints.nav Reprints:/journalsPermissions.nav Permissions:/content/214/2/79.refs.html Citations:What is This?- Mar 1, 2000Version of Record >>Out-of-round railway wheelsÐa literature surveyJ C O Nielsen1Ãand A Johansson21CHARMEC,Department of Solid Mechanics,Chalmers University of Technology,GoÈteborg,Sweden2Frontec Research and Technology AB,GoÈteborg,SwedenAbstract:This literature survey discusses the state-of-the-art in research on why out-of-round railway wheels are developed and on the damage they cause to track and vehicle components.Although the term out-of-round wheels can be attributed to a large spectrum of different wheel defects,the focus here is on out-of-round wheels with long wavelengths,such as the so-called polygonalization with1±5harmonics (wavelengths)around the wheel circumference.Topics dealt with in the survey include experimental detection of wheel/rail impact loads,mathematical models to predict the development and consequences of out-of-round wheels,criteria for removal of out-of-round wheels and suggestions on how to reduce the development of out-of-round wheels.Keywords:railway wheel out-of-roundness,polygonalization,wheel/rail impact load detectors,removal criteria1INTRODUCTIONImperfections on the wheel tread can have a detrimental influence on both track and vehicle components such as sleepers,rails,wheelsets and bearings.Examples of imperfections are isolated wheelflats,causing severe re-peated high-frequency impact loads,and polygonal wheels with irregularity wavelengths of approximately1m,lead-ing to an increased low-frequency component of the dynamic wheel/rail contact force.These are both regarded here as types of wheel out-of-roundness(OOR).The disastrous Eschede accident in Germany in June1998may have started with a fatigue crack in the wheel rim caused by the fluctuating contact force on a non-round wheel tread.The OOR also leads to impact noise and/or increased rolling noise.Thus,in order to minimize costs for repair and maintenance and to meet noise legislation,there is a large economic incentive for detecting and replacing non-round wheels in time.Also,the cause of OOR should be investigated in order to find suitable countermeasures. Current Swedish criteria for determining if a railway wheel should be replaced or not are based on the length of a wheelflat.However,recent studies have shown that there may be other wheel defects causing large impact forces that are not covered by the present criteria for wheel removal. For example,the depth of a flat may be more important than its length.Since it is desirable to remove all wheels that cause additional damage to trains and tracks,new criteria need to be developed as well as new methods for detection of wheel defects.The objective of this literature survey is to describe the state-of-the-art in research on why out-of-round railway wheels are developed and on the damage they cause to track and vehicle components.The present survey,with56 references,stems from an original technical report[1] where119references are cited.Focus is on wheel defects with irregularity wavelengths in a range from about0.5m up to the full wheel circumference.2CLASSIFICATION OF WHEEL TREAD IRREGULARITIES AND THE PROPOSEDORIGIN OF THE DIFFERENT PHENOMENAThe sections below classify different types of defects on railway wheel treads.Emphasis is put on defects with longer wavelengths,but local defects and defects with shorter wavelengths are also dealt with.For the classifica-tion,reports[2]to[4]have been used.2.1EccentricityEccentricity is caused by misalignment in the fixation of the wheel during profiling or reprofiling,and it is present to some extent on all railway wheels.2.2Discrete defectThis is a deviation of the wheel radius that is present over a small part of the tread.The deviation may be caused by a wheelflat or by inhomogeneous material properties.Plastic79The MS was received on14June1999and was accepted after revision forpublication on4January2000.ÃCorresponding author:CHARMEC,Department of Solid Mechanics,Chalmers University of Technology,SE-41296GoÈteborg,Sweden.F01199#IMechE2000Proc Instn Mech Engrs Vol214Part Fdeformation is common in conjunction with this wheel tread defect.2.3Periodic non-roundnessThis type of OOR has a periodic irregularity around the wheel circumference superimposed on the constant wheel radius.The wavelength of the irregularity ranges from 14cm to approximately one wheel circumference,while the amplitude is of the order of1mm.This defect has been detected only on disc-braked wheelsets.According to Zacher[5],examples of periodic OOR with one,three and four periods around the wheel circumference have been found on wheels from ICE trains in Germany. Investigations carried out at DB AG by Rode et al.[6]state that the fixation(claw clamping)of the wheel during reprofiling may be a cause of a triple-shaped polygon.This initially small OOR is amplified during rolling.Figures1and2are two examples taken from Pallgen[7], showing periodic OOR on ICE wheelsets.The wavelength contents of the different wheel tread defects are also given in the figures.A conclusion from the investigation is that the third harmonic dominates for solid steel wheels,while the second harmonic dominates for rubber sprung wheels. Based on experiments performed by an international workgroup(UNRA)on the Gotthard line,MuÈller et al.[8] state that OOR may be caused by inhomogeneous material properties around the wheel circumference.Another ex-perimental and theoretical investigation of periodic OOR has been carried out by Werner[9].Here,a coupling to the frequencies of natural vibration of the railway wheelset is used to explain the OOR.2.4Non-periodic(stochastic)non-roundnessThis type of OOR may be caused by unbalances in the wheelset or by inhomogeneous material properties of the wheel.As for periodic non-roundness,this defect has been observed only on disc-braked wheelsets.So-called stochastic OOR has been detected on ICE wheels in Germany[7].An example of a non-periodic (stochastic)OOR is shown in Fig.3,where the wavelength content of the OOR is also illustrated.From the figure it can be concluded that the stochastic shape contains several different harmonics.2.5CorrugationThis defect appears on wheel treads that are block braked. The dominating circumferential wavelength of this type of OOR is3±6cm,while the amplitude is smaller than 10ìm.Experimental and theoretical studies(numerical simulations)of the development of this type of defect have been performed by Vernersson[10,11].The proposed and verified hypothesis is that,during block braking,some regions on the wheel tread become warmer(formation of hot spots owing to a thermoelastic instability,TEI)than neighbouring regions.The heated regions protrude from the wheel surface owing to thermal expansion and they are therefore subjected to more wear than the other parts oftheFig.1Detected OOR of a solid steel wheel with dominantly three harmonics around wheel circumference.The bars indicate the distribution of different harmonics of the OOR.(From reference[7])Proc Instn Mech Engrs V ol214Part F F01199#IMechE2000 80J C O NIELSEN AND A JOHANSSONwheel tread surface.When the wheel cools down,the volume of material at these hot spots decreases (valleys are formed),which results in a corrugation pattern.It is noted that corrugation is a main source of rolling noise.2.6RoughnessThe circumferential wavelength of this defect is in the order of magnitude of 1mm,while the amplitude is of the order of 10ìm.Fig.2Detected OOR of a rubber sprung wheel with dominantly two harmonics around wheel circumference.Thebars indicate the distribution of different harmonics of the OOR shape.(From reference [7])Fig.3Detected non-periodic OOR.The bars indicate the distribution of different harmonics of the OOR shape.(From reference [7])F01199#IMechE 2000Proc Instn Mech Engrs Vol 214Part FOUT -OF-ROUND RAILWA Y WHEELS ÐA LITERATURE SUR VEY 812.7FlatsThis type of defect is due to unintentional sliding(without rolling)of the wheel on the rail.The primary cause is that the braking force is too high in relation to the available wheel/rail friction.The reason for this may be that the brakes are poorly adjusted,frozen or defective.Another reason may be that there are regions where wheel/rail friction incidentally and locally becomes low.2.8SpallingSpalling is the term used for the rolling contact fatigue phenomenon occurring when surface cracks of thermal origin meet,resulting in part of the wheel coming away from the wheel tread.The thermal cracks may arise in the hard and brittle martensite that is developed owing to heating and rapid cooling of the wheel tread during and after block braking.2.9ShellingShelling is a term normally used for all types of subsurface induced cracks.It is manifested by loss of flakes of material from the wheel tread.Excessive vertical wheel/rail contact forces with respect to the diameter of the wheel is the primary cause for this particular form of rolling contact fatigue.3DETECTION AND SIMULATION OF OUT-OF-ROUND WHEELSImpact loads due to wheel defects may cause rail fracture as discussed in reference[12].The risk of fracture increases at low temperatures.The most severe type of wheel defect is a newly developed wheelflat with sharp edges.Older wheelflats with rounded edges may also damage sleepers and ballast.Different types of wheel defects may also cause high-cycle fatigue of wheels and other vehicle components, such as bearing failures[12].3.1Experimental detection of impact loadsThe two most common approaches to detect impact loads are based on the use of either strain gauges or acceler-ometers.The Association of American Railroads(AAR) uses both methods.Details on the function of the two detector systems are given in report[13]:1.The first approach is known as WILD(Wheel Impact Load Detector).This system is composed of a series of strain gauges in a shear gauge load circuit configuration on the web of the rail.Ten vertical load circuits are installed on each rail.The coverage of this system is not complete,since wheels with different diameters cause maximum impact loads at different positions on the rail.2.In the second approach,seven accelerometers are placed on each rail.This system offers a coverage of nearly100 per cent for all wheel diameters.However,the registered acceleration does not give a quantitative measure of the size of the impact load.Therefore,approach1is more widely used in North America.In1991,a series of wheel impact tests was performed at the Transportation Test Centre in Pueblo,Colorado,United States.These tests are discussed by Kalay et al.[14]and by Stone et al.[15].The two types of detectors described above were tested.Both accidental and machined wheel defects were considered.The investigation showed that both types of detectors clearly identified wheels with tread defects,but,because of long-wavelength OOR,the repeat-ability of measurements was not quite satisfactory.In Figs 4and5,the detection uncertainties of the different detectors,for a wheel with a long(0.5m)wavelength defect,are illustrated.Impact loads and rail accelerations were found to increase with increasing depth of the wheel defects and as functions of train speed.The scatter in the measurements increased with increasing speed and increasing size of the wheel defects.The optimum impact load threshold level, where out-of-round wheels with long-wavelength defects were most easily identified by the WILD system,was found to be in the range60±80kips(267±356kN).The train speed at which most defective wheels were most easily identified was found to be40mile=h(64.4km=h). The corresponding values for the accelerometer-based detection system were200±300g and50mile=h (80.5km=h).Increased acceleration levels could also be measured on the rail opposite that hit by the defective wheel.Further,since loaded and non-loaded cars with the same wheel defects produced approximately the same acceleration levels,the same threshold limit can be adopted in both cases.Kalay et al.have investigated impact loads as a function of train speed[16].Data are given for different lengths of wheelflat and depths of the longer-wavelength defects in Figs6and7respectively.Loads increase with the length of the wheelflat and with the depth of the long-wavelength defect.Other examples of available techniques to measure wheel/rail contact forces include strain gauges mounted on wheelset axles or strain gauges applied to the wheel web (see reference[17]).Ohtani[18]reports that the East Japan Railway Company has developed a system for detecting wheelflats.The principle of the computer-based method is to detect shock waves in the rail caused by a rolling wheel with a wheelflat.Another method for detection of wheelflats and corruga-tion defects is to analyse the frequency spectrum of the measured rail acceleration.A description of this method based on the so-called cepstrum function in conjunction with Fourier analysis is given by Braccialli et al.[19,20].Proc Instn Mech Engrs V ol214Part F F01199#IMechE2000 82J C O NIELSEN AND A JOHANSSON3.2Numerical simulation of the influence of out-of-round wheelsSeveral mathematical models for simulation of dynamic wheel/rail interaction in the presence of imperfections on the railhead and wheel tread have been developed.These are,for example,described in the literature surveys by Knothe and Grassie [21]and Nielsen [22].Imperfections on the wheel tread,such as pits,flats andthermally affected zones,may lead to large impact forces owing to dynamic interaction of the wheel and track.The studies by Jenkins et al.[23]and Newton and Clark [24]are contributions in this field that had a significant influence on the early understanding of the effects of out-of-round wheels.Jenkins et al.[23]carried out a theoretical and experimental study,where different types of impact forces were treated and suggestions of an improved wheel design were given.Experimental andtheoreticalFig.4Impact loads due to a long-wavelength (0.5m)wheel defect measured with a WILD detector.1kip 4:45kN and 1mile =h 1:609km =h.(From reference [15])Fig.5Measured accelerations due to a long-wavelength (0.5m)wheel defect.1g 9:8m =s 2and1mile =h 1:609km =h.(From reference [14])F01199#IMechE 2000Proc Instn Mech Engrs Vol 214Part FOUT -OF-ROUND RAILWA Y WHEELS ÐA LITERATURE SUR VEY 83approaches to investigate impact loading due to wheelflats were described by Newton and Clark [24].Results from calculations with three different mathematical models were given and compared with results from experiments.It was concluded that the models are applicable in differentfrequency intervals of the impact loads,i.e.for different train speeds.Impact loads present on the Northeast Corridor high-speed track in North America are dealt with in papers by Ahlbeck and Hadden [25,26].The studies reportbothFig.6Experimental measurements of impact loads from wheelflats versus train speed.The different lines showthat the impact load increases with the size of the flat.1kip 4:45kN,1mile =h 1:609km =h and 1inch 25:4mm.(From reference [16])Fig.7Experimental measurements of impact loads from wheels with longer-wavelength defects (18±22in longdefects around the wheel circumference)versus train speed.The different lines show that the impact load increases with the depth of the wheel defect.1kip 4:45kN,1mile =h 1:609km =h and 1mil 0:0254mm.(From reference [16])Proc Instn Mech Engrs V ol 214Part FF01199#IMechE 200084J C O NIELSEN AND A JOHANSSONexperimental work and numerical investigations.The influ-ence of sleeper bending modes on impact loads is examined in a mathematical model and the loading on bearings is discussed.The wheel defects that were investigated were 25±40cm long and2±4mm deep.These caused peak impact loads whose amplitude was greater than400kN [25].It was concluded that wheel defects with long wavelengths often lead to large impact loads,and that these wheel defects are not always easily detected by visual inspection of the wheel.Therefore,other methods need to be used[26].Ahlbeck and Harrison[27]measured wheel profiles and adopted a mathematical model to predict impact loads from these defects.It was concluded that high-frequency impact loads at the wheel/rail interface are substantially attenuated by the wheelset mass.However,longer-wavelength tread irregularities leading to lower-frequency excitation may result in significant loads on the bearings.It was found that these loads increase with the ratio of depth to wavelength of the OOR.An early experimental and theoretical investigation of the effects of out-of-round railway wheels on railway bridges was carried out by FryÂba[28].Results from an investigation with the purpose of specifying geometry limits on allowable wheel irregularities are presented by Grassie[29].Predictions made by an adopted mathematical track model were found to correspond rather well to experimental data.It was found that the amplitude of measured and calculated responses for a wide variety of defects found in operational service varied essentially in proportion to speed.Cai and Raymond[30]have developed a theoretical model for simulating dynamic wheel/rail interaction. Various types of wheel defects(wheelflat,randomly worn wheel)are studied.The authors conclude that the wheel/ rail impact behaviour is highly dependent on train speed and that one defective wheelset can also lead to large impact loading on the adjacent wheelset.The effect of loss of contact between wheel and rail is also covered in the numerical simulation.In reference[31],the influence of an impact load caused by,for example,a wheelflat on deflections,accelerations,stresses and strains in rail and sleepers and on ballast pressures is computed. Theoretical investigations on wheel/rail impact loads and comparisons of different mathematical train/track models have been carried out by Dong et al.[32].Non-linear effects such as loss of wheel/rail contact and sleeper lift-off from the ballast are taken into account.It is concluded that axle load and train speed determine the magnitude of the impact loads caused by rge impact forces are obtained when the length of the flat in conjunction with train speed excites the fundamental eigenfrequency of the coupled wheelset/track system.Impact forces transferred from rail to sleeper are strongly influenced by pad stiffness and sleeper mass.The authors claim that,in order to detect wheelflats,it is preferable to position accelerometers on the rail,since smaller flats are not always detected by strain gauges on the rail.Dong and Sankar conclude that the factors that influence the impact loads the most are the shape and size of the wheel defects,axle load,train speed and railpad stiffness[33].4CRITERIA FOR REMOV AL OF OUT-OF-ROUND WHEELSThe use of an impact load detecting system has offered the opportunity to define criteria for removal of railway wheels that are not only based on visual inspection of wheel tread defects but also on the impact loads that are measured by the detectors.In reference[13],a review of changes to North American criteria for removal of out-of-round wheels is given.From January1996,a wheel shall be replaced if it causes a peak impact load larger than90kips (400kN).The allowable length of the wheel flat was increased from2in(50.8mm)to2.5in(63.5mm).Inves-tigations have shown that only half of the wheels that caused impact loads of100kips(445kN)had visual defects that were unacceptable,and also that the depth of a flat is a better criterion for condemning wheels than its length.Although different North American railway admin-istrations use different criteria,the limit for replacing railway wheels is approximately100kips(445kN)for most administrations.A conceptual framework for investigating the economic consequences of high-impact wheels is proposed in reference[13].The objective is to determine at which impact load level it is economically beneficial to remove a defective wheel.It is concluded that,for North American conditions,wheels should be removed from service when they cause impact loads greater than85kips(378kN). Kalay and Hargrove discuss wheel tread defects in reference[34].An economic analysis is also given,the authors concluding that a large sum of money can be saved each year by developing proper removal criteria based on impact load detection.The tests that led to the new impact load-based AAR wheel removal criteria are described by Kalay et al.[16]and by Tajaddini and Kalay[35],along with an economic motivation for wheel removal criteria.In Sweden,the criteria for wheel repair are as follows [36]:1.If the length of the defect is40±60mm,or if there exists a material build-up but with a height smaller than 1mm,the train has to go to the nearest workshop for repair.On such an occasion and at temperatures below À108C,the train speed must not be higher than 10km=h.At higher temperatures,there are no restric-tions other than that the speed interval15±45km=h should be avoided since the risk of damaging the rails is largest at these speeds.2.If the length of the damage is larger than60mm,or if the height of a material build-up is larger than1mm,F01199#IMechE2000Proc Instn Mech Engrs Vol214Part FOUT-OF-ROUND RAILWA Y WHEELSÐA LITERATURE SUR VEY85the train must go to the nearest manned station at a speed not higher than10km=h.For freight wagons,the measured wheel impact loads indicating that the length of a wheelflat is within one of the intervals specified in the above criteria are290and320kN respectively[12].5METHODS TO PREDICT OUT-OF-ROUNDNESS BY NUMERICAL SIMULATIONAccording to Meinke and Meinke[37],two important features introduced by modern high-speed trains,as com-pared with conventional trains,are as follows:1.The rotational speed of the wheels is higher since the wheel diameter is still of the same size.2.Higher speeds lead to larger kinetic energies and require more brake power to stop the train.The wheelsets are thus equipped with more disc brakes,typically four discs instead of two.Most mathematical models adopted to predict railway wheel OOR include:(a)a model of the dynamic interaction between wheelsetand track to determine forces and creepages at the wheel/rail contact point,(b)a wear model to account for the long-term wearprocess of the wheel tread.A review of the development of numerical methods for prediction of wear on the wheel and rail is given by Zobory [38].One basic approach in most models addressing the development of OOR is the assumption of so-called multi-ple time-scales.In the dynamic interaction model,the time-scale of the vibrations can be expressed in seconds,while an order of108wheel revolutions is considered in the wear model.For the dynamic interaction model,this means that the geometries of the wheel tread and rail can be treated as constant and that a controlled motion of the wheelset can be simulated with given conditions on speed,load and track.The contact forces and the slip lead to wear,but the geometry of the running surfaces changes in a very slow process.The coupling of the two models is often illustrated by a feedback loop,such as the one in Fig.8.An initial out-of-round profile,together with model parameters of train and track and disturbance parameters such as unbalanced rotating masses,is taken as input to the simulation.Contact forces and wear power in the contact patch are calculated by use of the interaction model.Material excavation versus location on the wheel tread is then calculated on the basis of a wear hypothesis.The out-of-round shape is modified by the wear and then included in the new input data.By this procedure,the long-term wear is monitored iteratively. Certain model parameters such as speed and track proper-ties can be varied from one iteration to another in order to simulate more realistic operating conditions.The dynamics of a high-speed wheelset is to a large extent a matter of rotor dynamics.The effects of rotatory inertia and gyroscopic moments are therefore important [37].In numerical simulations used to predict longer-wavelength defects,rigid body dynamics in combination with a wear model is often adopted.Morys et al.have investigated a rigid body model of an ICE carriage on an elastic track model[39±43].The wheel/ rail contact is modelled by use of a simplified theory according to Kalker[44].The adopted wear model is based on the assumption that the rate of mass excavation is proportional to wear power in the contact patch.The significant variations in vertical wheel/rail contact forces caused by the out-of-round profile lead to an excitation of the lower wheelset bending modes[40].At frequencies below200Hz,the wheels can be treated as stiff and rigidly coupled to the axle.Thus,the bending oscillation of the axle leads to lateral slip and material excavation at the contact patches.Longitudinal slip and spin play a minor role.Vertical resonances of the coupled train/track system lead to peaks in the vertical contact force at certain train speeds.For a stiffer track,the resonance train speed is higher.The amplitude of the dominating lateral wear energy within the contact patch is mainly determined by lateral slip,lateral contact force and vertical contact force.In Fig. 9,an example of calculated wear energy,vertical contact force and wheel radius deviation versus time is illustrated. The phase shift between wear energy maxima and radius deviation maxima is important for whether a certain OOR will be enlarged or not.Depending on this phase shift,three significant ranges of excitation frequency can be defined. In the low-and high-frequency ranges,the excavation maxima are located at the falling and rising slopes of the radius deviation curve away from the maxima and minima. An enlargement of the existing OOR harmonic order will not occur in these two frequency ranges.However,for medium frequencies,the maximum excavation occurs approximately at the maximum and minimum of the OOR shape.Because of the higher excavation at the minimum radius,the OOR enlarges rapidly.Next to thedominant Fig.8Scheme for numerical simulation of the growth of wheel OOR.Connection between short-term train/track dy-namics and long-term wear process is illustrated.(Fromreference[39])Proc Instn Mech Engrs V ol214Part F F01199#IMechE2000 86J C O NIELSEN AND A JOHANSSONoriginal OOR shape,higher harmonic orders develop.No precise limits of frequency ranges can be given because they strongly depend on track properties.In the case of an ICE running on a stiff track,the medium range is approximately 50±80Hz.The development of a certain OOR order is dependent on train speed and track conditions.In Fig.10,three typical qualitative results of long-time wear simulations are shown.For all simulations,various speeds,both driving directions and stiff track properties were assumed.In Fig.10a a slow change from an eccentricity (first-order OOR)to a third-order OOR is observed.In Fig.10b a change from a second-order to a fourth-order OOR is shown.In Fig.10c an enlargement of a third-order OOR without changes in shape or phase is illustrated.It is interesting to note that,within the investigated high-speed range and based on the assumed track properties,only the third-order OOR harmonic increases,whereas all other orders develop into higher harmonic orders.Unbalances in the wheelsets may be another cause of out-of-round wheels.This topic has been investigated by Meinke et al .[37,45]and Morys [41].The unbalances are modelled as point masses distributed on the wheelset at different radii on the wheels and disc brakes.Meinke suggests that dynamic unbalances have a stronger influence than static ones [37].According to Morys,dynamic unbalances cause large vibrations of the wheels and small vibrations of the disc brakes,whereas the opposite condi-tions hold for static unbalances [41].A complete locomotive vehicle is simulated by Soua and Pascal [46]in order to investigate the initiation and evolution of three different wheel wear shapes with 1,2and 4harmonic OOR orders.The authors state that wheelset axle torsional vibrations in combination with lateral motion of the whole wheelset explain the generation and evolution of the wear pattern.Numerical simulations of wheel polygonalization are presented by V ohla et al.[47±49].One hypothesis is thatexcitation of the wheel eigenmodes may play an important role for the development of OOR since the number of nodal diameters in the wheel eigenmodes coincides with some periodic irregularities found on worn wheels.Also,Frischmuth and Langemann [50,51]havecarriedFig.9Example of calculated lateral wear energy W R ,vertical contact force F N and wheel radius deviation r OOR(train speed 70m =s,third-order OOR,peak-to-peak OOR amplitude r OOR 0:3mm,stiff track).(From reference [40])Fig.10Calculated long-term wear development as a conse-quence of small initial radius deviations caused by,for example,manufacturing tolerances.(From reference [40])F01199#IMechE 2000Proc Instn Mech Engrs Vol 214Part FOUT -OF-ROUND RAILWA Y WHEELS ÐA LITERATURE SUR VEY 87。

系统最佳化简介PPT资料优选版

15.057 春季 03 范德維特
安裝Premium規劃求解器
• 首先規劃求解器一定要安裝 • 執行光碟片 • 路徑:\html\PremiumSolver • 沒有絕對需要但是建議安裝此求解器
15.057 春季 03 范德維特
帶手提電腦來上課
我們會用電腦處理許多實作模型星期四
➢01指派模型 (AssignmentModel.xls) ➢0最佳化簡介
約翰‧范德維特
15.057 春季 03 范德維特
概括重點介紹
• 藉著中間的抽象模型，應用數學分析技巧以設計與分析系統。
15.057 春季 03 范德維特
自我介紹
057 春季 03 范德維特
約翰‧范德維特 02運輸模型 (TransportationModel.
Carnegie Mellon University（卡內基美隆大學） 057 春季 03 范德維特
• 期末考
25%
• 作業
25%
• 課程專案
25%
15.057 春季 03 范德維特
安裝試算表規劃求解器
工具|增益集
➢ 選擇規劃求解增益集程式
尚須一些安裝的元件方法一
➢ 安裝的光碟片
方法二
➢ 複製 solver.xla 及兩個檔案到下面的路徑 ..\Microsoft Office\Office\Library\Solver\
15.057 春季 03 范德維特
課程組織
教學主軸
➢ 網路模型 ➢ 線性規劃 ➢ 整數規劃 ➢ 非線性規劃 ➢ 啟發式演算法
增加複雜度增加次方數
15.057 春季 03 范德維特
課程組織
工具及教科書
➢試算表的最佳化求解與Eppen等等

1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。

Perry Y.Li Department of Mechanical Engineering,University of Minnesota,111Church St.SE,Minneapolis MN55455e-mail:pli@ Toward Safe and Human Friendly Hydraulics:The Passive Valve Hydraulic systems,as power sources and transmissions,offer many advantages over electromechanical or purely mechanical counterparts in terms of power density,ﬂexibil-ity,and portability.Many hydraulic systems require touching and contacting the physical environments;and many of these systems are directly controlled by humans.If hydraulic systems are passive,they would be both safer to interact with,and easier for humans to control.In this paper,it is shown that a critical hydraulic component,the directional control valve,is not passive.However,the directional valve,as a one-port or a two-port device can become passive if appropriate spool valve dynamics are imposed.Methods to passify the valve for bothﬁrst-order and second-order spool dynamics are considered.In the case of second-order spool dynamics,a passive method that relies on hardware modiﬁcation,and an active feedback method,are proposed.͓S0022-0434͑00͒01803-7͔I IntroductionHydraulic systems,as power sources and transmissions,offermany advantages over electromechanical or purely mechanicalcounterparts in terms of power density,ﬂexibility,and portability.Many hydraulic systems require touching and contacting thephysical environments.It is critical that these systems can safelyinteract with the environments without unintentionally damagingthe environments or themselves.The environment and the hydrau-lic system form a closed-loop system when they interact with eachother.The closed loop system can be unstable even if the envi-ronment and the hydraulic system are individually stable.Many ofthese systems are also directly controlled by humans in the workenvironment.A typical example is a construction worker operat-ing a hydraulic boom-and-bucket to perform an earth digging task.It is,therefore,important that these systems are natural and easyfor the human operator to control.Both the safety and the human friendliness aspects of theseapplications can be enhanced if the system can be shown to bepassive.Roughly speaking,a passive system behaves as if it doesnot generate energy,but only stores,dissipates,and releases it.Apassive system is inherently safer than a nonpassive system be-cause the amount of energy that it can impart on the environmentis limited.The well-known passivity theorem͓1͔ensures that apassive system can interact stably with any strictly passive sys-tem.The latter includes a wide variety of physical objects andenvironments.The inherent safety that passive systems afford hasbeen exploited in machines that interact with humans such assmart exercise machines͓2,3͔,bilateral teleoperated manipulators ͓4,5͔,and Cobots͓6͔.It is also exploited in the passive velocity ﬁeld control͑PVFC͒methodology͓7–10͔,which by ensuring thatthe closed loop system is passive,enables mechanical systems tobecome coordinated͑such as for contour following in machiningoperation or multiple robots cooperating͒and at the same timeinteract safely with the often ill characterized environment.Passive systems are also potentially easier to control.One rea-son for this is that some form of stable haptic feedback of theenvironment to the human user will be quite natural to achieve fora passive system͓4͔.Moreover,users will be able to use thefamiliar concept of‘‘power’’while executing a manipulation task.Based on the observation that passive systems form the basis ofalmost all artiﬁcial learning and adaptive schemes͑see͓11͔andreferences therein͒,it can also be argued that passive tools arepotentially easier than non-passive ones for users to learn to use.A key premise of many passivity based control techniques isthat the underlying system possesses some passivity property.While it is true that mechanical systems are passive,the passivityproperties of hydraulic devices have not previously been investi-gated.The objective of this paper is to investigate the passivityproperty of a key component in a hydraulic system-the directionalcontrol valve,and to propose modiﬁcation and control so that thevalve becomes passive.Speciﬁcally,it will be shown that thedirectional control valve,as a one-port,is not passive with respectto the hydraulic power input at the hydraulic ing an analo-gous electrical circuit,the source of the nonpassivity is identiﬁed.Subsequently,we will show that the valve can become passive ifappropriateﬁrst or second order spool dynamics are imposed.This passiﬁcation method can be extended to the two-port valve ͑the extra port is for the command signal͒.In the case of the second order spool dynamics,the valve can be passiﬁed by mak-ing hardware modiﬁcation to the valve,or by using active feed-back compensation.The rest of this paper is organized as follows.Passivity is de-ﬁned in Section II.In Section III,we show that the one-port di-rectional control valve is not passive.In Sections IV and V,weproposeﬁrst-order spool dynamics so that the one-port valve andthe two-port valve are passive.Methods to preserve the passivityof the valve in the presence of second order spool dynamics arepresented in Section VI.Simulation results are presented in Sec-tion VII.Sections VIII and IX contain discussion and concludingremarks.II Passivity and Supply RatesConsider a system with input u and output y.A supply rate fora system is some functions:͑u,y͒ۋs͑u,y͒෈R.For physical systems,useful supply rates are those associated withpower input into the system.For example,if the pair of input andoutput variables for a system are the effort andﬂow variables,then a supply rate which is the power input is the inner productbetween the effort andﬂow variables.Following͓12͔,a system is said to be passive with respect tothe supply rate s(u,y)if,for a given initial condition,there existsa constant c so that for all time t and for all input u(•)͵0t s͑u͑␶͒,y͑␶͒͒d␶уϪc2.(1)If s(u,y)is the power input into the system,then Eq.͑1͒says that no matter how one manipulates the input u(t),the maximumContributed by the Dynamic Systems and Control Division for publication in theJ OURNAL OF D YNAMIC S YSTEMS,M EASUREMENT,AND C ONTROL.Manuscriptreceived by the Dynamic Systems and Control Division July15,1999AssociateTechnical Editor:S.Nair.402ÕVol.122,SEPTEMBER2000Copyright©2000by ASME Transactions of the ASMEamount of energy that one can extract from the system is limited by the constant c 2which can be interpreted to be the initial energy stored in the system.III Four-Way Directional Control Valves are not PassiveFour-way directional valves are central to electro-hydraulic technology.A typical four-way valve is shown in Fig.1.A hy-draulic device,such as a hydraulic actuator ͑piston ͒or a hydraulic motor,is connected to device ports A and B.The constant supply pressure P s Ͼ0is typically supplied by a pump and an accumula-tor,and the return line is connected to the reservoir at pressure P 0Ϸ0.By stroking the spool in the four-way valve,ﬂow is me-tered into and out of the device ports A and B.For example,when the spool is stroked to the left ͑as shown in Fig.1͒,ﬂow is me-tered into port A from the pump and out of port B back to the reservoir.When the spool is stroked to the right,ﬂow in the re-verse direction occurs.For simplicity,assume that the valve will be connected to de-vices that cannot store ﬂuid volume.This assumption excludes single ended actuators that have different cap side and rod side areas.However,it allows us to analyze the interaction between the valve and the device as a one-port with the pressure difference,P L ϭP A ϪP B and the ﬂow rate Q ϭQ 1ϭQ 2being the correspond-ing effort and ﬂow variables.We further assume that the four-way valve is critically lapped,matched,and symmetric.Under these assumptions,it easy to show that the pressure at the two terminals P A and P B are symmetrically located from P s /2͓13͔,in the sense thatP s ϭP s ϪP 0ϭP A ϩP B ;P A ϪP B ϭP L .This means that P s ϩP L ϭ2P A and P s ϪP L ϭ2P B so that as P L increases from 0,P A and P B increases and decreases symmetri-cally from P s /2.Under these assumptions,the relationship between the ﬂow rate and the pressure difference can be readily derived using the stan-dard oriﬁce relationship ͓13͔:Q ϭQ L ͑x v ,P L ͒ªΆC dͱ␳wx vͱP s Ϫx v͉x v ͉P L;sgn ͑x v ͒P L ϽP s ϪC dͱ␳wx vͱx v͉x v ͉P LϪP s ;sgn ͑x v ͒P L уP s(2)where C d Ͼ0is the oriﬁce coefﬁcient,␳Ͼ0is the ﬂuid density,w Ͼ0is the gradient of the oriﬁce area with respect to the spool position,and x v is the spool displacement from the center posi-tion.In normal operation,͉P L ͉ϽP s so that only the ﬁrst case statement in Eq.͑2͒is normally used.We include the abnormal case P s рsgn(x v )P L for mathematical completeness.The graph of Q L (x v ,P L )is depicted in Fig.2.A natural supply rate to consider for the four-way valve iss ͑P L ,Q ͒ªϪP L Q(3)which is the hydraulic power input exerted by the hydraulic de-vice on the valve.Proposition 1.For each nonzero ﬁxed spool position x v 0,the four-way directional valve is not passive with respect to the sup-ply rate given in Eq.(3).Proof:Let x v 0be a nonzero spool position.Choose a con-stant port pressure trajectory P L (t )ϭP c such that 0Ͻsgn(x v )P c ϽP s .From Eq.͑2͒,deﬁne p 0ªP L (t )•Q (t )ϭP c •Q L (x v ,P c )Ͼ0,͵tϪP L ͑␶͒Q ͑␶͒d ␶ϭϪp 0•t .Thus,any amount of energy M Ͼ0can be extracted by waiting for a long enough time t ϾM /p 0because then,͵tϪP L ͑␶͒Q ͑␶͒d ␶ϽϪM .Therefore,there is no constant c Ͼ0that will satisfy Eq.͑1͒for all t Ͼ0.This proves that the valve is not passive.᭿Notice that p 0in the proof above is the power withdrawn from the valve.This propositionshows that for each constant spoolFig.1Four-way,three land directional valveFig.2Flow Õpressure relationship of a typical matched,criti-cally lapped,four-way directional valve.Top:normal case sign …x v …P L ÏP s ;bottom:exceptional case:sign …x v …P L ÐP s .Ps Ä3000psi,w Ä0.3␲in,C d Ä0.66,and ␳Ä1000kg Õm 3.Journal of Dynamic Systems,Measurement,and ControlSEPTEMBER 2000,Vol.122Õ403position x v ,it is possible for a device connected to the valve to extract an inﬁnite amount of energy from the valve.A component which allows an inﬁnite amount of energy to be extracted from it is clearly not passive with respect to the supply rate given by power input into the component.We will now deﬁne an analogous electrical circuit for the four-way directional valve which will be useful in understanding the rest of this paper.Let us rewrite Eq.͑2͒asQ L ͑x v ,P L ͒ϭQ 0͑x v ͒ϪK¯t ͑x v ,P L ͒P L (4)where(5)is the ﬂow rate through the valve when P L ϭ0,andK¯t ͑x v ,P L ͒ªϪ͵1ץQ LץP L ͑x v,l •P L ͒dl ϭ͵1C d w ͉x v ͉2ͱ␳͉P s Ϫsign ͑x v ͒P L ͉•dl(6)is the mean gradient of the ﬂow with respect to the port pressure.Notice that K¯t (x v ,P L )is the negative slope of the line joining the two points (0,Q 0(x v ))and (P L ,Q L (x v ,P L ))in Fig.2.Notice also that Eq.͑4͒is exact and it is not a linearization of Eq.͑2͒.Lemma 1.For each x v ,the mapping P L ۋQ L (x v ,P L )is mono-tone non-increasing.Thus,for any (x v ,P L ),K¯t ͑x v ,P L ͒у0.Moreover ,K¯t (x v ,P L )ϭ0if and only if x v ϭ0.Proof:The validity of this key lemma can be seen graphically from Fig.2;or analytically by noticing that ץQ L /ץP L р0and ץQ L /ץP L ϭ0if and only if x v ϭ0.᭿Because of Lemma 1,we can interpret Eq.͑4͒using an analo-gous electrical circuit shown in Fig.3͑A ͒which consists of an ideal current source Q 0(x v )ϭK q x v and a nonlinear output shuntconductance K¯t (x v ,P L ).The output current and voltage of the circuit are Q and P L ,respectively.If we deﬁne the supply rate to this system to be the negative of the power output of the circuit,then it is equivalent to the supply rate Eq.͑3͒for the four-way valve.From Fig.3͑A ͒,it is clear that it is the current source,not the nonlinear shunt resistance,which is responsible for the circuit not being passive.We remark that although possible,the Thev-enin’s equivalent circuit ͓14͔involving a voltage source ͑with voltage x v K q /K¯t ͒and output resistance (1/K ¯t )is not as useful for interpreting Eq.͑4͒because of the way that P L would affect the voltage source and the output resistance.IV Passifying the One-Port ValveWe now propose a method to make the four-way valve passive.The key idea is to make the spool position x v ͑or the current source ͒dynamically related to the port pressure P L .In essence,the current source will be replaced by an inductor.Consider the ideal spool dynamicsBx ˙v ϭϪAP L .(7)We can think of this as the dynamics of a light and heavilydamped spool acted on by a force ϪAP L .One way to generate the force ϪAP L is to utilize the actual pressure forces to act on the spool via direct structural feedback.This can be achieved by making the spool lands have different areas ͑see Fig.4͒.For ex-ample,if we make the center land area A in smaller than the out-lying land area A out ,so that A ϭA out ϪA in ,then the pressure force acting on the spool would be ϪAP L .Alternatively,this force can be generated by an actuator such as a second stage valve or a linear motor that acts on the spool,using pressure measurements in the two chambers.In this section,we will show that the spool dynamics Eq.͑7͒passiﬁes the four-way valve.We postpone,to a later section,the detailed discussion of how to deal with the reality that the spool dynamics are in fact second order.To see intuitively how Eq.͑7͒may successfully make the four-way valve passive,notice the compensatory effect of the spool dynamics when P L increases:an increase in P L would normally extract more power from the valve if Q were to remain the same.However,with the spool dynamics in Eq.͑7͒,the magnitude of x v ,hence the ﬂow rate Q ,are decreased.The spool dynamics in Eq.͑7͒therefore limits the power .As we shall see,the spool dynamics also limits the available energy .Theorem 1.The nonlinear four-way valve Eq.(4)with the spool dynamics given by Eq.(7)is passive with respect to the supply rate s (P L ,Q )ϭϪP L Q ,the power input into the valve by the hydraulic device.Proof:Consider the storage function candidateW ͑t ͒ϭ12Bx v2͑t ͒.Using Eq.͑7͒,its time derivative is given byW˙͑t ͒ϭϪx v AP L ϭϪAK q ͓Q 0͑x v ͒P L͔where in the second equality,we utilized ͑5͒to express x v in terms of Q 0(x v ).Applying Eq.͑4͒,we getW˙͑t ͒ϭϪA K q ͓QP L ϩK ¯t ͑x v ,P L ͒P L 2͔рAK q͓ϪQP L ͔.(8)Integrating this expression,we obtainFig.3…A …Analogous electrical circuit for the four way valve.…B …Circuitis passiﬁed by replacing the current source by an inductor.Fig.4Structural pressure feedback using different land areas404ÕVol.122,SEPTEMBER 2000Transactions of the ASME͵ts ͑P L ͑␶͒,Q ͑␶͒͒d ␶уK q A ͑W ͑t ͒ϪW ͑0͒͒уϪK qAW ͑0͒.This shows that the nonlinear four way valve with the spool dy-namics given by Eq.͑7͒is passive with respect to the supply rategiven by s (P L ,Q )ϭϪP L •Q .᭿The effect of the spool dynamic in Eq.͑7͒on the analogous electrical circuit is to replace the current source by an inductor with inductance L ϭB /AK q ͑Fig.3͑B ͒͒.The dynamics of the new circuit isLdi LdtϭϪP L Q ϭi L ϪK¯t ͑x v ,P L ͒P L where the spool position x v is identiﬁed with the inductor currentvia i L ϭK q x v .It is obvious that the new circuit is passive.V Passivity of the Two-Port ValveA passive one-port hydraulic valve is not very useful since it only reacts to the environment but cannot be commanded.We now introduce a passive two-port valve in which the addition port can be used as a port for external control,such as the direct or teleoperated control by a human.We modify the spool dynamics Eq.͑7͒so that it can interact with an additional external force F xBx ˙v ϭϪAP L ϩF x (9)Q ϭQ 0͑x v ͒ϪK¯t ͑x v ,P L ͒P L (10)where Q 0(x v )and K¯t (x v ,P L )are given by Eqs.͑5͒and ͑6͒and B Ͼ0is the damping.Notice that the spool is now affected by both the hydraulic port pressure P L and the additional control command F x .The valve is now a two-port device.The ﬁrst port of this system interacts with the hydraulic device via the effort and ﬂow variables (P L ,Q ).The additional port interacts with a control unit via the variables (F x ,x v ).For this two-port system we can think of P L and F x as the input variables,and Q and x v as the output variables.Proposition 2.The modiﬁed two-port nonlinear valve given by Eqs.͑9͒–͑10͒is passive with respect to the supply rate given bys ͑͑P L ,F 1͒,͑Q ,x v ͒͒ϭAK q͓ϪP L Q ͔ϩF x x v (11)where K q ϭC d w ͱP s /␳and A is the pressure feedback constant in ͑9͒.Proof:Using the storage function candidate W (t )ϭ1/2Bx v 2,we obtainW˙ϭϪAP L x v ϩF x x v .Using a similar procedure in the proof of Theorem 1to rewrite theﬁrst term on the right-hand side,we have W˙͑t ͒ϭϪA K q ͓QP L ϩK ¯t ͑x v ,P L ͒P L 2͔ϩF x x v рAK q ͓ϪQP L ͔ϩF x x v ϭs ͑͑P L ,F 1͒,͑Q ,x v ͒͒where K¯t (x v ,P L )is given in Eq.͑6͒.The passivity property en-sued on integration.᭿VI Passivity of Valves With Second Order Spool Dy-namicsSpools have inertia,so their dynamics are necessarily second order.In this section,we consider spool dynamics of the formͩx ˙v⑀x ¨v ͪϭͩ010ϪBͪͩx v x ˙vͪϩͩF x ϪAP L ͪ,(12)where ⑀Ͼ0is the spool inertia.These dynamics can be realized by using a force motor to generate the force F x ,and using either structural pressure feedback or sensor/actuator feedback to realize AP L .The second order dynamics generate an approximation of the desired spool dynamics Eq.͑9͒because as ⑀→0,the slow dynamics is given byBx ˙v ϭF x ϪAP Lwhich is the desired spool dynamics in Eq.͑9͒.The question we must now address is how the passivity prop-erty that results from the ideal spool dynamics Eq.͑9͒can be preserved when the spool dynamics becomes Eq.͑12͒.We propose two methods to ensure the passivity of the hydrau-lic valve in the presence of nonzero spool inertia.In the ﬁrst method,we utilize an output redeﬁnition technique,which amounts to physically introducing an additional ﬂuid leakage and a stiff interaction spring.In the second method,active feedback control is applied to enforce the passivity.A Passive Passiﬁcation Method.In this method,we make two modiﬁcations to the two-port valve.First,we add an extra leakage between the ﬂuid chambers ͑Fig.5͒such that q leak ϭ␦•sgn(P L )ͱ͉P L ͉with ␦being a leakage coefﬁcient.This can be done by simply adding an oriﬁce across the outlets,or by punch-ing a hole between the two middle chambers in the valve.The second modiﬁcation to the valve system is to place a spring with spring constant K between the spool and the actuator that provides the force F x .The outputs of the two-port valve are correspond-ingly modiﬁed to:Q mod ϭK q x v ϪK¯t ͑x v ,P L ͒P L Ϫ␦P L͉P L ͉ͱ͉P L ͉(13)y ªx v ϩ1K F x.(14)Remark 1.1y,called the interaction displacement for the force F x ,isthe displacement of the end of the spring at which F x is applied (Fig.5).2It is possible to bound the nonlinear leakage ␦by a linear leakage if ͉P L ͉is known to be bounded.Let ␦M (␦,P L )ª␦/ͱP L .Then whenever ͉P L ͉ϽP L ,͉q leak ͉ϭ␦•ͱ͉P L ͉у␦M ͉P L ͉.3As ␦→0and 1/K →0,the modiﬁed outputs Q mod and y inEqs.(13)–(14)converge to Q and x v ,respectively,which are the port ﬂow variables for the two-port valve in Section V with the ideal ﬁrst-order dynamics (9).The following theorem states that with suitable leakage ␦and spring constant K ,the modiﬁed valve is passive with respect to a modiﬁed supply rate.Moreover,thebehavior of the modiﬁed valve approaches the behavior of the valve with the ﬁrst order spool dynamics when the inertia of the spool is small.Fig.5Valve with extra leakage and added springJournal of Dynamic Systems,Measurement,and Control SEPTEMBER 2000,Vol.122Õ405Theorem2.Consider the four-way directional valve which is modiﬁed by a leakage␦,and an interaction spring K,and whose output equations are given by Eqs.(13)–(14).Suppose that the spool dynamics are second order and are given by Eq.(12).The two-port valve has the following properties:1It is passive with respect to the supply rates new͑͑P L,F x͒,͑Q mod,y͒͒ϭAK q͓ϪP L Q mod͔ϩF x y,(15)if the matrixNϭͩB2⑀A1A A␦M K q0101K ͪ,where␦M is the linear leakage as deﬁned in Remark1,is positive deﬁnite.Thus,the compliance1/K and leakage␦M(and hence␦) can be chosen to be arbitrarily small when spool inertia⑀→0. 2Let Q i(t)ϭQ(t)and x v,i(t)ϭx v(t)be theﬂow rate and the spool displacement trajectories of the two-port valve withﬁrst-order ideal spool dynamics,described by Eqs.(9)–(10),given a pair of bounded pressure and interaction force trajectories input (P L(t),F x(t)).Let Q mod(t)and y(t)be the modiﬁedﬂow and interaction displacement of the two-port valve given by Eqs.(12) and(13)–(14),which has been modiﬁed by leakage␦and inter-action spring K,under the same inputs(P L(t),F x(t)).Assume that x v(0)ϭx v,i(0)and x˙v(0)ϭ0.Under these conditions,as⑀, 1/K,␦→0,ʈy͑•͒Ϫx v,i͑•͒ʈϱ→0,ʈQ mod͑•͒ϪQ i͑•͒ʈϱ→0.Proof:1.Let us deﬁne a newﬂow variable,Q linªK q x vϪ␦M P L(16) and a new supply rate,s mod͑͑P L,F x͒,͑Q lin,y͒͒ªAK q͓ϪP L Q in͔ϩF x y.Notice that s mod((P L,F x),(Q lin,y))рs new((P L,F x),(Q mod,y)) because of Remark1.Thus,to show passivity with respect to s new((P L,F x)(Q mod,y)),it sufﬁces to show that͵0t s mod͑͑P L,F x͒,͑Q lin,y͒͒dtуϪc2.Using a set of new coordinates(z1,z2):ͩz1z2ͪϭͩ1⑀B0Ϫ⑀Bͪͩx v x˙vͪ,ͩx v x˙vͪϭͩ110ϪB⑀ͪͩz1z2ͪ,the spool dynamics and the output equations͑after scaling͒becomez˙1ϭ1B͑F xϪAP L͒(17)z˙2ϭϪB⑀z2Ϫ1B͑F xϪAP L͒(18)(19)(20)Now consider the storage functionWϭ12z12ϩ12z22.Differentiating with respect to time,W˙ϭϪB⑀z22Ϫz21B͑F xϪAP L͒ϩz1B͑F xϪAP L͒.(21)Inspired by the formulation of the KYP lemma with a directfeedthrough term͓1͔,we rewrite the multipliers toϪP L z2andF x z2in Eq.͑21͒in terms of the coefﬁcients that multiply z2inEqs.͑19͒and͑20͒,respectivelyϪABϭC12Ϫ2␣,Ϫ1BϭC22Ϫ2␤.(22)Hence,␣ϭA/B and␤ϭ1/B.Substituting Eq.͑22͒into Eq.͑21͒,and subsequently replacingC12z2and C22z2using͑19͒,͑20͒,we haveBW˙ϭϪ͑z2ϪP L F x͒ͩB2⑀A1A A␦M K q0101Kͪͩz2ϪP L F xͪϩs mod͑͑P L,F x͒,͑Q lin,y͒͒(23)So thatBW˙рs new͑͑P L,F x͒,͑Q mod,y͒͒.The passivity property is attained after integration.2.Because z1(0)ϭx v,i(0)ϭ0and z1(t)and x v,i(t)satisfy thesame differential equation͑Eq.͑9͒and Eq.͑17͒͒,x v(t)ϭx v,i(t)ϩz2(t).However,from Eq.͑18͒,ʈz2͑•͒ʈϱр⑀B2ʈF x͑•͒ϪAP L͑•͒ʈϱ.Thus,as⑀/B2→0,x v(•)→x v,i(•).Moreover,as␦→0,the leakedﬂow Q mod in͑13͒approaches Q in͑4͒.Similarly,as1/K→0,y(•)→x v(•)in͑14͒which in turn approaches the ideal spooldisplacement x v,i(•).᭿The advantage of the passive passiﬁcation method is that thestructural pressure feedback,as well as the extra leakage and in-teraction spring,can all be implemented by using hardware com-ponents.Sensing/actuation is not needed.However,the extra leak-age can lead to inefﬁciency,and the extra interaction spring mayintroduce sluggishness when using the interaction displacement yas feedback signal.To get an idea of how much leakage is necessary,consider a0.0508m͑2in.͒long steel spool with diameter d406ÕVol.122,SEPTEMBER2000Transactions of the ASMEϭ0.00762m (0.3in.).Suppose that C d /ͱ␳ϭ0.021m 2/N 1/2/s (70in.2/ͱlbf/s.͒and w ϭ␲d .The mass of the spool is ⑀ϭ0.0056kg.Suppose also that P s ϭ20685KPa (3000Psi)and the pressure feedback gain A ϭ0.1␲(d /2)2.For passivity,we must have ␦M (␦)уAK q /⑀(⑀/B )2ϭ1.88ϫ10Ϫ3m 4•s •(⑀/B )2.We get ␦Ͼ1.88ϫ10Ϫ3(⑀/B )2ͱP ¯L in S.I.units.As a comparison,if P ¯L ϭ689kPa ͑100psi ͒and B /⑀ϭ1000rad/s,the required leak-age is ␦ϭ1.56ϫ10Ϫ6.This corresponds to an oriﬁce area of 7.43ϫ10Ϫ5m 2,which is highly signiﬁcant.Notice however that the leakage ␦can be signiﬁcantly reduced if the maximum load pressure or the time constant ⑀/B decrease.B Active Passiﬁcation.The passive passiﬁcation technique has several drawbacks.For example,a possibly large additional leakage,which introduces inefﬁciency,is needed.Moreover,the damping B must be sufﬁciently large.This also implies that the area differential A for the pressure feedback need to be corre-spondingly increased to maintain sensitivity.If the pressure feed-back is to be achieved using structural feedback,then the spool size must be increased.The spool inertia may then have to in-crease even more.To overcome these difﬁculties,an active feedback method is proposed.We deﬁne the spool dynamics to be⑀x ¨v ϭϪBx ˙v ϪAP L ϩF x ϩF act(24)where F act is the active passiﬁcation control to be deﬁned.Letz ªx ˙v Ϫ1B ͑F xϪAP L ͒.Then the spool dynamics become:ͩBx ˙vz ˙ͪϭͩ0B 0ϪB /⑀ͪͩx vz ͪϩͩF x ϪAP LF act ⑀Ϫd dt ͑F x ϪAP L ͒Bͪ.(25)Now deﬁne the control F act to be of the form:F act ϭϪ␥Bx v ϩ⑀B ͓F ˙xϪAP ˙L Ϫg 2͑t ͒sgn ͑z ͔͒(26)where ␥is a positive constant,g 2(t )is some dominating gain to be chosen,and • denotes the best estimate of the argument.Theorem 3.Consider the spool dynamics given by (24)with the control law F act given by Eq.(26).If g 2(t )in Eq.(26)is deﬁned so thatg 2͑t ͒Ͼsgn ͑z ͑t ͕͓͒͒F˙x ͑t ͒ ϪAP ˙L ͑t ͒ ͔Ϫ͓F ˆx ͑t ͒ϪAP ˙L ͑t ͔͖͒,then the four-way two-port valve is passive with respect to thesupply rates ͑͑P L ,F x ͒,͑Q ,x v ͒͒ªAK q͓ϪP L Q ͔ϩF x x v .(27)Notice that this supply rate is exactly the one in Eq.͑11͒for the valve in Section V with the ﬁrst order spool dynamics.Proof:Deﬁne the storage function W ϭ1/2Bx v 2ϩ⑀/2␥z 2and taking its time derivative W˙ϭx v ͑F x ϪAP L ͒ϩx v Bz ϪB ␥z 2ϩ␥Ϫ1z ͩF act ⑀Ϫ͑F˙x ϪAP ˙L ͒Bͪ.Thus,if F act is chosen as given in the theorem,W˙рx v F x ϩA ͑ϪP L x v ͒ϪB ␥z 2.Using the fact thatK q x v ϭQ 0͑x v ͒ϭQ ϩK t ͑x v ,P L ͒P L ,we get,W˙рA K q ͓ϪP L Q ͔ϩF x x v ϪA K qK t ͑x v ,P L ͒P L 2from which the passivity result is obtained.᭿In the case of the active passiﬁcation method,there is no need for B to be large.It is however required that the derivative of F x ϪAP L can be well estimated or bounded.The spool dynamics resulting from Eqs.͑24͒and ͑26͒are no longer the ideal ﬁrst-order spool dynamics in Eqs.͑9͒.Indeed,if the estimates of the deriva-tives of P L and F x are good,the transfer function from F x ϪAP L to x v will be given by:H ͑s ͒ϭs ϩB /⑀B ͓s ͑s ϩB /⑀͒ϩ␥B /⑀͔.Notice that ␥introduces a spring force which has the largest effect at low frequencies.At high frequencies,the H (s )approximates 1/Bs which is the ﬁrst order dynamics in Eq.͑9͒.VII SimulationTo illustrate the property of the passiﬁed four-way valve,we simulate the situation when the valve is used to operate a piston.We investigate only the nonpassive valve and the valvepassiﬁedFig.6Four way valve controlling a piston that interacts with awallFig.7Pressure trajectory that occurs when a piston con-trolled by a nonpassive four-way valve interacts with a wallJournal of Dynamic Systems,Measurement,and ControlSEPTEMBER 2000,Vol.122Õ407。