Design of Embedded Control System Based on ARM9 Microcontroller

Design of Embedded Control System Based on ARM9 MicrocontrollerLei Zhang and Zhengfeng WangDepartment of Electrical & Electronic EngineeringHenan University of Urban ConstructionPing Ding Shan, Henan Province, Chinazhl@Abstract：This paper mainly focuses on the design and realization of embedded control system based on ARM kernel. Firstly, hardware structure, configuration and signal detection, processing of the control system based on ARM are described. Secondly, the system application programs, including main control program, user operation interface program, keyboard program and development of drivers for some S3C2410X related functional modules are described. Lastly the closed-loop control of parameter is achieved by way of PID and NN. As practical experience shows, after introducing the control system into the parameter control for microbiology fermentative process, the design of the system reaches expected effect, realizes relevant functions and shows a broad market prospect. Keywords：ARM kernel， embedded system， Linux， biology fermentationI.I NTRODUCTIONThe process of microbiology fermentation has found extensive application in such sectors as medicine, foodstuff, light industry, agriculture, environmental protection, etc. The key technology lies in how to realize intelligent control of fermentative process. In some countries with leading edge intelligent technology, this technology has been widely applied, developed and converted into mature products. In China, however, intelligent level of fermentative process has far lagged behind some developed countries. What’s more, a great many relatively advanced facilities in use are mostly imported from other countries. Therefore, development of microbiology fermentation control system where we have our own independent intellectual right is virtually of great importance and should be promoted.Along with the continued penetration of embedded control into industrial fields[1-3], following the arrival of 21st century, embedded integrated industrial control systems based on ARM structure and Linux operation system (OS) has become a mainstream for development of industrial control. ARM chips boast strong performance in addition to compact size and low power consumption[2]. Linux has found extensive application in embedded systems due to its advantages of being open source, stability and low cost. Against such a background, the system uses ARM9-kernel-based embedded system-on-a-chip (SoC) S3C2410X as the core to develop embedded control system for fermentative process, while its OS is transplanted from real-time OS - Linux[4]. Specific to such characteristics of microbiology fermentation control process as time-variability, big lag and randomicity, the system uses a control strategy that combines both conventional PID and neural network (NN) [5]. In this way, it solves to some extent the shortage of conventional PID - difficulty in online real-time setting of parameters and that in effective control over the system, thus improving the self-adaptability and control effect of the control system. As proven by practice, such systems can realize a relatively satisfactory control effect. Moreover, such a control method can be applied to controlled objects with such characteristics as big lag and time-variability.II.H ARDWARE D ESIGN OF S YSTEMThe whole system uses ARM9 hardware platform with embedded Linux OS so as to meet the system’s high requirement on real-time performance, facilitate change, expansion, upgrade and other needs on both software and hardware function, shorten development cycle and lower R&D cost.Operating principle of intelligent control system for microbiology fermentative process: After startup of the system, in the mode of automatic operation, ARM9 microprocessor enables sampling function of sensors and, based on the gap (deviation) between the sample value and the set value, respectively controls each actuator according to NN PID control strategy. In order to improve accuracy of measurement, measures are taken from both hardware and software aspects: Nonlinear compensation is provided for output signals of sensors while digital filter is used to further remove interference and improve accuracy. During manual operation, ARM9 microprocessor controls all actuators according to control instructions output by the user. As signals output by ARM9 microprocessor are not sufficient to drive such actuators as solenoid valve and peristaltic pump, the system also comes with drive IC. Meanwhile, the actual sample value is transmitted via serial interface IC to the computer for processing. In this way, intelligent control over fermentative process is achieved. Figure 1 shows the structural block diagram of intelligent control system for fermentative process.A.ARM Embedded Microprocessor S3C2410XThe system uses 16/32 bit RISC embedded microprocessor based on ARM920T (ARM9TDMI) kernel[6]. It is primarily used in scenarios requiring cost-effectiveness and low power consumption. The operating frequency is generally 203 Hz. It offers abundant SoC resources: 16 K cache, 1 LCD controller (supporting STN and TFT touch screens on LCD); SDRAM controller; 3-channel UART; 4-channel DMA; 4 timers with PWM function and 1 internal clock; 8-channel 10-bit ADC; interface for touch screen; I2S bus interface; 2 USB ports on main unit and 1 USB port for equipment; 2 SPI interfaces; SD interface and MMC interface; watchdog timer counter; 117-bit universal I/O interface and 24-bit external interrupt source; 8-channel 10-bit AD controller. As the system uses many2010 International Conference on Electrical and Control Engineeringfunctional modules of the processor, the system structure is compact and the complexity of the system is reduced. The system uses S3C2410X’s 8-channel 10-bit ADC, LCD controller, 3-channel UART, 4-channel DMA and other modules.B.I/O ChannelAccording to design requirements, the system should realize online measurement on PH value, dissolved oxygen (DO) concentration, temperature, pressure, flow rate and so on of the fermentation broth. After comparison between the data obtained from the measurement and set values, fuzzy NN processes them and feeds them back to the fermentation system so as to achieve effective control over these parameters. As shown in Figure 1, data measured by PH sensor, DO sensor, temperature sensor, flow instrument and pressure gage are all analog signals. Therefore, we have to convert these analog signals into voltage signals. A piece of CD4501 serves as multi-selector switch. Moreover, as the signal generated is relatively weak and is not sufficient to meet the requirements for accurate measurement, we use the high-precision variable gain amplifier AD526 to achieve pre-amplification of signals, thus regulating amplitude of analog voltage signals. A/D conversion module serves to convert signals acquired into digital signals identifiable by the CPU; the digital signals then become input for fuzzy controller. Therefore, design of A/D converter is critical to the whole system. In this system, 10-bit A/D conversion module supplied with the chip is used. The module can convert the input analog signals into 10-bit binary digital codes; with a 2.5 MHz A/D converter clock, the maximum conversion rate can reach 500 ksps. It also supports on-chip acquisition and holding functions as well as power-down mode.In the output channel, the system exerts control over solenoid valve, peristaltic pump, immersion heater, electric control valve and other actuators. We can output control signals via 8255 extended parallel port. Furthermore, since such actuators as solenoid valve need 24 V AC for driving, drive IC has to be furnished. After being amplified by triode, the output of 8255 drives solid relay and further controls make and break of 24 V AC, thus driving solenoid valve and other actuators.C.Human-machine Interaction Channel1)LCD Display ModuleIn the system, color LCD LM7M632 (STN) supplied by Sharp Corporation is used. It offers a resolution of 640*240 and supports 256-color display.In this display system, the display is achieved by directly operating the contents in the display SDRAM; via DMA, LCD controller accesses data in display SDRAM without needing the intervention of CPU. In the 256-color display mode, each byte of data in the display SDRAM represents information for the color of one dot on the LCD. Therefore, the size of SDRAM should be 640*240*1 bytes. RGB data format for each byte should consist of three bits for red (Bit7~Bit5), three for green (Bit4~Bit2) and two for blue (Bit1~Bit0). LCD controller and LCD hardware should be connected as illustrated in Figure 2,3:Figure 2 Diagram of Internal Structure of ControllerT ABLE 1D EFINITION OF S IGNAL W IRE OF I NTERFACE OF LCD M ODULE(LM7M632)Figure 3 Connection between LCD Controller and LCD Hardware2)Keyboard CircuitHD7297 chip for keyboard is used to design the keyboard circuit. HD7297 is an intelligent drive chip coming with serial interface. It can be connected to keyboard matrix of up to 64 keys and it has built-in debounce circuit. When a key is pushed down, its key value will be automatically saved in the register; keyboard’s key number can be obtained by accessing this register. Such a design facilitates programming for there’s no need to use conventional software such as those for keyboard scan and calculation of key value. Connection between HD7297 and S3C2410X is very simple; just connect the four wires as follows to GPIO of S3C2410X: , CLK, DATA and KEY.D.Other Basic ModulesOther basic modules include power management system, reset circuit, clock circuit, SDRAM data memory, flash program memory and buzzer, etc. Memory module is for Pin Pin description Pin Pin description1 YD: scan starts 8 VDD: logic power supplied(3. 3V)2 M: alternation signal 9 Vcon: contrast regulatingvoltage3 LP: input latching 10 Vss: ground4 Vss:ground 11~14D0~D3: the lowest 4 bits ofdata cable5 Xck: clock 15 Vss: ground6 Vss:ground 16~19D4~D7: the highest 4 bits ofdata cable7 DISP: display switch 20 Vss: groundrecording information of the system, fermentative process, accumulative fermentation time, etc. Overrun alarm module functions by means of a buzzer. When the actual sample value accidentally exceeds the set value ceiling, the buzzer will sound to give alarms. UART module communicates with its upper computer by way of MAX232 level conversion and RS-232 so as to make use of PC resources. JTAG system is connected to the PC via a simulator. It uses the online debugging module built in S3C2410X chip for debugging programs on the upper computer. Power management system and clock circuit provide power and clock for operation of the system. The reset circuit is for realizing reset function of the system.III. PID C ONTROL S TRATEGY OF NNTo overcome the non-linearity of the fermentative process control system and meet the high accuracy requirement of the system, use of a good control strategy is a must. PID control has been widely applied due to its simple algorithm and proven technology. To achieve a good control effect for PID control, it is necessary to make a good adjustment for three control functions, namely proportion, integration and differentiation so as to form among them a relationship where control parameters coordinate with and check each other. The random non-linear expression ability of NN can realize PID control of optimal combination by way of learning the system performance. The system uses NN PID control strategy to successfully solve non-linearity of the system and achieves a good control effect [7-9]. Single-neuron network is the core to neuron self-adapting PID control [4]; its structural model is shown in Figure 4.x()r tFigure 4 Model of Single-Neuron Network(1X k (2X k (3X k ()kFigure 5 NN PID ControlAs shown in Figure 4, single neuron is a multi-input and single-output non-linear processing unit; i X ()1,2,3i =refers to input of neuron. iW ()1,2,3i = is the connectionstrength and is called weight. Using certain operation to combine the functions of input signals to get an overall effect is called “net input” and is expressed with i net . Based on different ways of operation, the expression of net input comesin many ways. Among them, the simplest one is linear weighted summation. The effect leads to status change of the neuron i , and the neuron output i u is the function of the current status ()*g which is called activation function [4].The mathematical expression of the model is:i net =1nij j j w x =∑−i θ (1)i u =()i net g (2)In the expression, i θ is the threshold of neuron i ; i net is the net input, i.e. the overall effect combining all input signals by way of certain operation.Discrete formula for conventional PID is:()()()()()01kp i di e k e k u k K k K e i T K T =−−=++∑ (3)Among them: p K , i K and d K are respectivelyproportion, integration and differentiation; e(k) is the deviationbetween actual output of control system and expected output at the time of k; u(k) is the output of control parameters at the time of k; T is the sampling interval of the system.In the single-neuron network model, make the thresholdθ = 0,()()1X k e k =, ()()20ki X k e i T ==∑,()()()31e k e k X k T−−=.Then the output of neuron isk U ＝()()11X k W k ＋()()22X k W k ＋()()33X k W ki.e. forming the model for NN PID control, as shown in Figure 5.Based on the deviation between the actual output of controlled parameters and expected output during fermentative process, and by applying certain rules, NN proceeds with self-adapting learning and keeps regulating the weight vector i W until it becomes the optimal value *W . The process of weight adjustment can be regarded as the self-organizing of the PID control parameters. Therefore, the whole control system is provided with self-adaptability. According to practical experience, compared to conventional PID control, the above-mentioned algorithm can reduce overshoot by 2% and steadystate error by 0.1%. As a result, both the anti-interference ability and control accuracy are improved.IV. D ESIGN OF S YSTEM S OFTWAREThe design is made with the currently prevalent embedded system development technology. Real-time OS Linux is embedded; mixed compilation of ARM and Thumb instruction sets are made to optimize code density. Firstly, transplant the real-time OS Linux to S3C2410X embedded microprocessor. The system divides the function to be achieved into several core tasks; Linux’s real-time kernel proceeds with dispatching to realize parallel execution of multiple tasks. In this way, both reliability and real-time performance of the system are greatly enhanced.Based on functions to be provided by the system, the system software is divided into several parallel existing tasks. Preemptive OS dispatches tasks according to their priority. Priority of all tasks of the system, in the descending order, is as follows: system monitoring, keyboard scan, LCD display, NN PID control algorithm, control parameter output and asynchronous serial communication. Data acquisition runs in the timer interrupt program, i.e., controlled parameters are acquired and saved every two seconds. System monitoring task is for monitoring other tasks. When a monitored task makes any error during execution, the system monitoring task will process it according to the preset processing list so that the task in question will restore proper operation, thus enhancing the reliability of the system operation. The Figure 6 below shows the block diagram of system operation. Figure 7 shows block diagram of data acquisition.Figure 6 Block Diagram of System OperationFigure 7 Block Diagram of Data AcquisitionV. C ONCLUSIONUse of embedded control system based on ARM kernel not only reduces peripheral devices for the system and improves reliability of the system, but also realizes accurate control over controlled parameters as NN PID control strategy is employed. Additionally, transplantation of Linux OS greatly facilitates programming, shortens development period of software and improves development efficiency. As practical experience shows, by introducing the control system into the environmental parameter control for microbiology fermentation process, the system can meet the overall performance requirements and produce a good control effect.R EFERENCES[1] Sun Yanchun, Li Changyou. Development of Optical Power MonitorBased on Embedded System[J]. Instrument Technique and Sensor, 2006 (8): 7-9.[2] Sun Tianze, Yuan Wenju. Guide to Design and Driver Development ofLinux (2nd Edition)[M]. Beijing: Publishing House of Electrics Industry, 2006.[3] Jia Zhiping, Zhang Ruihua. Principles of Embedded System andInterface Technology[M].Beijing: Tsinghua University Press, 2005: 149-233．[4] Gong Linan, Lv Hongwu, Wang Hongzhi. Embedded Web Server Basedon ARM9[J].Journal of Changchun University of Techonology,2009,30(3):323-326.[5] Simon Haykin, translated by Ye Shiwei, etc. NeuralNetworks[M].Beijing: China Machine Press, 2004: 81-94.[6] Wang Ligan, Mu Haiwei, Liu Shiqing. Research and Design ofEmbedded Indoor Intelligent Temperature Controller[J].JOURNAL OF NORTHEAST FORETRY UNIVERSITY,2009,37(6):64-66.[7] Zhou Yanyan, Wu Xiaoyan, Wei Gang. Study of Model VerificationMethod Based on Neural Network[J]. JOURNAL OF AIR FORCE ENGINEERING UNIVERSITY,2009,10(6):64-67.[8] Wang Chunhua, Tang Yan, Self-Adaptive PID Control Based on NeuralNetwork for Active Suspension Vehicle[J].COMPUTER SIMULATION,2009,26(5):274-277.[9] Ma Lingling, Zheng Bin, Ma Yuanyuan, Study of PID CompositeIntelligent Controller Based on Neural Network[J].METROLOGY & MEASUREMENT TECHNIQUE,2009,36(3):17-19.。