Part-III 整车零部件台架耐久性试验及其试验

nCode 疲劳耐久性工程
高级培训班
整车/零部件台架耐久性试验
及其试验加速技术零部件台架模拟试验
是一体化解决疲劳问题的策
略中的重要一环！
重新设计优化
重新设计优化零部件模拟
试验零部件模拟试验计算机辅助
疲劳寿命模拟计算机辅助
疲劳寿命模拟用户使用情况用户使用情况实测载荷实测载荷应力分析应力分析材料性能材料性能产品寿命
产品寿命
关联产品寿命
产品寿命加速的
Sign-off 试验加速的Sign-off 试验
台架试验的好处
•可重复的试验环境
•易对试验进行监控
•通过比较试验对设计
参数变化进行评价
•早期检验零部件的性能
•可能能实现试验加速
•验证理论模型•...
根据实测的载荷及响应信号，在实验室里
重现实际工况，模拟零部件的性能及寿命
进行台架试验的好处：
台架试验时也可进行数据采集！
做哪些台架试验？
•能模拟道路试验的台架试验
•能一定程度评价零部件耐久性能的试验•模拟损伤最严重的，要看车辆的特点
•。

不要做和实际工况无关的试验！
实验室负责人的困境
•预先预测台架试验大概需要多长时间，费用大概多少？•高效地利用试验台
•判断被要求做的试验是否是一个不合理的试验
•在试验开始前，使用疲劳编辑技术“合法地”加速试验•“合法地”过滤掉试验台不能模拟的高频
•…
要求的疲劳寿命要长，
但试验结果必须尽快出来!
疲劳分析能够帮助...
How should we test in lab?
From P.G. or Field to Test Rig
Customer Usage Test Track
Individual
Surfaces, Events
Test Rig
What Do We Want From A Durability Test?•Durability test that ’s suitable for the item in question: a component, sub-assembly, or a whole vehicle
•Test must replicate the same failure
mechanisms as seen in the real world
•Test should be representative of the real
loading environment
•Test should be accelerated where possible to reduce project time scales and costs
•Test specification can be used in FE based virtual test or real physical test
Test Synthesis –Route Map
Deterministic
Random Quasi-static Dynamic Uniaxial Multiaxial Test Synthesis Frequency Domain
Time Domain
Peak-Valley Domain
Dynamic Time Series Load Scaling Uniaxial Signal / Fatigue Fatigue Editing Accelerated Testing •PSD random •Sine sweep •Sine on random •Peak valley extraction
•Block load sequence
•Statistical exceedence
•Constant amplitude
•Is it proportional i.e. dominant plane?
•Multiaxial peak valley
extraction •Buffered fatigue editing •Remote parameter simulation test •Proving ground •Increase frequency of Time Series
Vibration Load Scaling Deterministic and Stochastic
Deterministic Stochastic
Load Scaling
Load Scaling
•Scaling up the load will reduce the test duration exponentially.•Target life is influenced by endurance limit and onset of local plasticity as well as dynamic response of component
•Scaling should be used with extreme care to avoid local yielding and changing the load paths
•Not suitable for most inertia reacted tests
Scaled Range
Original Range
Real Duration
Test Duration
Where b is the Basquin Exponent (gradient of SN curve)
This is only approximate!Load Scaling
Positive •Maintains Sequence •Maintains Phase between multiple channels •Maintains Frequency
Content
Negative
•Amplitude is not maintained (may affect the failure mechanism)Increasing Load Frequency
Increase loading frequency •Doubling the frequency will half the test time
•Limit acceleration to max 1/
3first mode natural
frequency
•Not suitable for inertia reacted tests
1/3 * natural frq Increase loading frequency
Positive
•Maintains Amplitude •Maintains Sequence •Maintains Phase between multiple channels Negative •Frequency is not maintained (cannot
consider dynamic
response)
Peak valley extraction
Peak valley extraction
What is it?
•Remove non peak or valley points in the signal and reduce the length of the signal in order to accelerate the fatigue test
•Frequency may be re-adjusted after peak-valley extraction
“峰谷”点
非“峰谷”点
360 Points
36 Points
Peak valley extraction
What is the method based on?
•Fatigue damage is calculated by cycles which are constituted by peak and valley points
“峰谷”点
非“峰谷”点
Peak valley extraction
Positive
•Maintains Amplitude •Maintains Sequence
•Test can be accelerated
significantly, typical 90%
reduction in signal length
•‘Gate’small cycles on range, rainflow or fatigue contribution •Be care with slew rates, etc. Negative:
•Frequency is not maintained (cannot consider dynamic
response)
•Phase between multiple
channels is not maintained (so, not suitable for multiaxial)
Constant amplitude Cantilever: actual load change Cantilever: simple test load
Constant amplitude
What is it?
•Replace variable load with a constant amplitude sinusoidal load, and test the component for a
fixed number of cycles or until failure occurs •Sometimes called “bogey test”
Constant amplitude
What is the method based on?
• a concept of equivalent damage
•Normally increase the load to accelerate the test
Constant amplitude
Equivalent damage based S-N curve
ΔS1
ΔS2
Νf1Νf2D1=1/Nf1 D2=1/Nf2
Nf1 cycles of ΔS1 range is equivalent to Nf2 cycles of ΔS2 range.
Both make the component failure
Constant amplitude
How can calculate equivalent amplitude or lading cycles?•Assume the slope of SN curve
•Estimate the accumulated damage in a component over the life of the vehicle
•Specify the number of test cycle for sinusoidal load, and use equivalent damage to calculate the load amplitude •Or specify the amplitude of the sinusoidal load, and use equivalent damage to calculate the number of cycle Constant amplitude
Determine amplitude from specified cycle number
D=sum(Di)
D=N ×D ΔS
Nf ΔS N cycles 1 repeat
ΔS?D ΔS = 1/Nf Constant amplitude
Determine cycle number from specified amplitude
D=sum(Di)
D=N ×D ΔS
Nf ΔS N? cycles 1 repeat
ΔS D ΔS = 1/Nf
Constant amplitude
Frequency of test load
•As quick as possible, say 10Hz? Constant amplitude
Positive
•Total damage is
maintained
•Test is simple
•Test can be accelerated significantly Negative
•Damage distribution is not maintained (may change
failure mode)•Frequency is not
maintained (cannot
consider dynamic
response)
•Not suitable for multiaxial
An example
Case Study 1: Durability Test
Background
•Need: Create a durability test
•Analysis: Create durability test specification for chassis component testing based on proving ground data
–Steering knuckles and control arms
–Constant amplitude lab test
–Equivalent damage
•How many cycles?
•What size cycles?
•Current Process: Infield with Excel; very manual and step-by-step •Challenges
–Reduce time required to analyze proving ground loads data.
–Promote standard processes for analyzing those loads.
–Make sure input data are clean.
•Solution: GlyphWorks Signal
Case Study 1: Durability Test
Results
•Report
–Contains results and
user inputs traceability
–Archivable as Word
document or Web page
•Lab test
–100,000 cycles @ +/-load
–Equivalent damage
Case Study 1: Durability Test
Value
Value:
•Results are given in an archivable report.
•Process is easily repeated by non-experts.
•Massive reduction in analysis time and effort:
Engineer:“Took the analysis time required from 2 days to 5 minutes.”Engineer:“Get the right answers without all the manual processing, and all the results are given in archivable reports.”
Manager:“90% reduction in time; this kind of time savings is unheard of. The time that was eliminated was all the tedious work that made it hard to focus on the engineering.”
Block loading Cantilever: actual load change Cantilever: block loading
block loading
What is it?
•Replace variable load with a block loading that
consists of several constant sinusoidal load with different amplitude, and test the component for a fixed number of cycles or until failure occurs
block loading
What is the method based on?
•Rainflow cycle counting (a concept of equivalent damage and equivalent damage distribution)•Normally can accelerate the test without
increasing load (only gating out small amplitude loading cycles)
block loading
Principle: rainflow cycle counting
block loading
How can we obtain block loading time history?•Rainflow cycle counting with several bins (say, 8)•Gating out small amplitude cycles (damage
calculation is needed for appropriate gate)•Use constant amplitude cycles to represent
cycles for each bin. Frequency can be as high as possible
•Combine all time histories obtained for each bin (normally in the order from small to large
amplitudes)
block loading
No gating
block loading
Gating out the first bin
block loading
Positive
•Total damage is
maintained
•Damage distribution is maintained
•Test is simple
•Test can be accelerated significantly with gating
out small amplitude cycles Negative
•Frequency is not
maintained (cannot
consider dynamic
response)
•Cycle sequence is not maintained
•Not suitable for multiaxial
Histogram editing Cantilever: actual load change
Histogram editing
What is it?
•Replace variable load with an equivalent re-constructed time history load for testing
Histogram editing
What is the method based on?
•equivalent damage and equivalent damage distribution
•Normally can accelerate the test without
increasing load
基于损伤的直方图编辑技术思路
•用应变或应力法，对循环直方图进行疲劳分析，获得对应的损伤直方图
•比较循环和损伤直方图，在循环直方图中将那些无损伤循环移走(将循环数置0 即可)
•根据编辑后的循环直方图重构一个只有“峰谷”的随机时域信号，作为载荷控制信号
•对重构的信号再进行疲劳计算，比较编辑前后的疲劳寿命
基于损伤的直方图编辑技术思路
重构的应变信号
00
00.050.10.150.2
00
00000
应变 (uE)RESPREG.DAC
Sample = 204.8Npts = 44
Max Y = 681.9Min Y = -1152
时间 (秒)
z 重构时域信号:
-Range-mean 雨流矩阵
-Max-min 雨流矩阵
-Markov 矩阵
-不规则因子
组合录自不同时间的
时域信号
信号重构
Histogram editing
Positive •Total damage is maintained •Damage distribution is maintained •Test can be accelerated significantly Negative
•Frequency is not maintained (cannot consider dynamic response)•Cycle sequence is not
maintained •Not suitable for multiaxial
Multi-channel Peak Valley
Extraction
Multi-axial peak valley extraction
•Maintains phase
relationship between
multiple channels by
keeping points that
correspond with a peak or
valley in a different
channel
•Ordinary peak valley
would apply all peaks /
valleys simultaneously
therefore changing the
load paths
•‘Gate’small cycles
多通道信号的峰谷值抽取
编辑前
•一个应变响应信号response
•四个驱动加速度信号
g01,g02,g03,g04
•信号点数：8000
多通道信号的峰谷值抽取
编辑后
•不设门槛值
•编辑后信号点数：7449
多通道信号的峰谷值抽取
•采用45.3%门槛值
•编辑后信号点数：598
Resultant / Critical Plane Analysis •Proportional multi-axial, or cases
with a dominant fatigue plane
•Establish critical plane
•Eliminate non-damaging channels
•Determine a single drive channel
with fixed proportions between
inputs or align component on the
uniaxial test rig at a given angle Resultant Load Plane
Multi-axial peak valley extraction Pos:•Maintains Amplitude •Maintains Sequence •Maintains Phase between multiple channels Neg:
•Frequency is not maintained (cannot consider dynamic response)
•Only suitable for
proportional multi-axial
loads
Time-domain damage editing ＝
编辑前编辑后
试验加速技术的基本原理
•根据响应信号预估疲劳寿命
•用响应信号的损伤结果编辑驱动信号
•编辑原理为原始响应信号和浓缩后的
响应信号的损伤值和损伤分布（损伤
直方图）保持不变
时间关联损伤编辑技术
•基于时间-疲劳损伤概念
•计算应变响应信号的时间-损伤分布图
•损伤可用S-N 或e-N 方法计算
•用时间-损伤分布图同步移去响应和驱
动信号中的无损伤或损伤小于某一门
槛值的信号段
•插入一个递减或连接信号，避免在连
接处有一个突然的信号跳跃
•可用于单通道或多通道加载
时间-疲劳损伤图
•每一循环周产生的损伤值对分至组成
循环周的“波峰”和“波谷”
•叠加所有的循环获取损伤分布图
应变时域信号（4个循环）
时间-损伤图
1
122334
4
时间-疲劳损伤图
时间-损伤图
应变时域信号
同步移去响应和驱动信号中的无损伤信号段
•用时间-损伤图同步
移去响应和驱动信
号中的无损伤或损
伤小于某一门槛值
的信号段
•插入一个递减或连
接信号，避免在连
接处有一个突然的
信号跳跃
编辑前后的驱动信号比较
•编辑前信号长度39 秒
•编辑后试验时间8 秒
Fatigue damage editing
Positive
•Maintains Amplitude
•Maintains Sequence
•Maintains Phase between multiple channels •Maintains Frequency content
•Typical acceleration 50-80% depending on amount of damage to be retained and number of failure
locations assessed
•Can be used with uniaxial or multiaxial fatigue solvers
一个疲劳加速实例
经过疲劳编辑原先需要114 天的车体疲劳模拟试验被缩短到15 天，节省试验费用$235,000 美元!疲劳试验加速分析
做台架试验需要什么？
•所要模拟的道路谱（这很重要！）•道路谱数据处理软件
•台架试验加速软件
•台架试验控制软件
•。

Frequency-domain accelerated
testing
Example –Head lamp on car
The car Manufacturer
knows what vibrations
are transmitted to the
vehicle through the
chassis
Question What vibration level is transmitted to the light?
The supplier wants to reproduce it on shaker table for sign off tests!
Measure base acceleration
Light Mechanical link
Simple PSD Scaling
Real Duration Test Duration
2/ b
PSDs from Mission Envelope for worst response Scale loads to accelerate test Vibration test specification •Test specified in (US) MIL STD 810 and (UK) DEF STAN 0035•Test only valid for PSD input, no facility for transient loads •Better suited to a single event
•No account of component’s frequency response characteristics
Stochastic Loading
Parameters T reduced How does Test Synthesis Work?PSD Transient Loading Time Series Test Specification
Validation
T reduced
M Relative displacement z(t)Vehicle displacement x(t)Acceleration input )(t x &&Vehicle Platform Component Damping c Stiffness K 050100
5
1015
Frequency of excitation
A
m
p
l
i
t
u d
e
o
f
d
i
s p l
a c e m
e
n
t
z
Q
ω
o
How does Test Synthesis Work?
•Consider a simple vibrating
component mounted on the
vehicle platform
•The component receives the most critical loads when
excited at it’s natural
frequency
The Extreme Response Spectrum
2.) Filter using a Single Degree of Freedom
Response and find maximum amplitude
of response
1.) Input PSD or Time Signal
4.) Find envelope of maximum responses
3.) Repeat for next frequency interval
PSD input = Extreme Response Spectrum (ERS)
Time input = Shock Response Spectrum (SRS)0
The Fatigue Damage Spectrum
2.) Filter using a Single Degree of Freedom
Response and calculate the fatigue
damage
1.) Input PSD or Time Signal
4.) Join the damage points to create the
Fatigue Damage Spectrum (FDS)
3.) Repeat for next frequency interval
Mission Profiling
Calculate FDS
Calculate ERS Calculate FDS Calculate SRS FDS
ERS
SRS FDS ΣFDS
Mission events consisting:•several time
histories with
#repeats •several PSDs
with duration P
S
D
s T
i m
e H i s t
o r
i
e s Calculate sum of
all FDS Calculate envelope of all
SRS & ERS
Calculate FDS Calculate ERS Calculate FDS Calculate SRS FDS
ERS
SRS
FDS
ΣFDS Duty cycle consisting:•several time histories with #repeats •several PSDs with duration P S
D s
T i
m e
H
i s
t
o r i
e s
Calculate sum
of all FDS Calculate envelope
of all SRS & ERS
Generating a Synthesized Test Signal
Guarantee Coefficients:
* k safety
* k test
P d
Loading
Strength
Synthesized Data from the total lifetime damage Calculate test ERS and compare with
Mission ERS and SRS.
Ensure test does not over scale the loading
beyond what is reasonably expected in real
life
Calculate the synthesized PSD
Test PSD over
reduced duration
Measured PSDs 024********
02000400060008000
Synthesised Test
Original Duty Cycle
Extreme Response Spectra Comparison Natural Frequency [Hz]E R S
Aim: ensure that Synthesized Specification
contains the necessary damage and is still representative:ERS(Synthesized) >= ERS(OriginalData)
Validate the synthesized signal
Also: ensure that Synthesized Specification
does not exceed maximum shock loads seen
in practice :
ERS(Synthesized) < SRS(OriginalData)Case Study Exhaust Muffler • A supplier is required to
prove the durability of a bus
exhaust muffler. •The OEM specifies repeats of proving data equivalent to 53 days of continuous testing.
•Can we create a valid,
accelerated shaker test with
equivalent damage to 53
days?
Front Chassis Bracket Rear Chassis Bracket
Case Study –Exhaust Muffler
Accelerated PSD,
72 hours per axis (x, y & z)
6 times quicker than
original test spec
x axis y axis z axis
工程应用实例
四通道道路模拟机
四通道道路模拟机加速试验数据准备
•Instrumentation
•Data acquisition of RLD
•Analysis & editing
•Response files
•Iteration
•Drive file creation
Acquisition of road load data
Channels
•Four vertical acceleration of spindle
•Several strains at critical locations in body •Displacement from wheel to body
Road surface
• A proving ground
Hardware used
•Somat eDAQ
Raw data
加速度应变
信号长度约2200 秒
Data preparation
Software used
•nCode GlyphWorks
Data processing
•Pre-reporting
•Data cleaning
•Data reduction (fatigue editing) Pre-reporting
Fatigue assessment
Fatigue editing
Deduced data for road simulator
加速度应变
编辑后信号长度约100 秒
（原信号长度约2200 秒）
振动台损伤等效试验谱制定。