VIEWS: 327 PAGES: 7 CATEGORY: Emerging Technologies POSTED ON: 12/4/2010
(IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 Intelligent Controller for Networked DC Motor Control B.Sharmila N.Devarajan Department of EIE, Sri Ramakrishna Engineering College Department of EEE, Government College of Tech. Coimbatore, India Coimbatore, India sharmi.rajesh@gmail.com Abstract—This paper focuses on the feasibility of Neural Network The novelty of this paper lies in comparison of the application controller for Networked Control Systems. The Intelligent of NARMA-L2 Controller and Mamdani Fuzzy Logic Controllers has been developed for controlling the speed of the Controller with conventional PID controller for the Networked DC Motor by exploiting the features of Neural improvement of the performance of networked control DC Networks and Fuzzy Logic Controllers. The major challenges in motor. Networked Control Systems are the network induced delays and data packet losses in the closed loop. These challenges degrade There are two approaches to utilize a data network as the performance and destabilize the systems. The aim of the Hierarchical Structure and Direct Structure as shown in Fig. 1 proposed Neural Network Controller and Fuzzy Logic Controller and Fig. 2 respectively. In the hierarchical structure the dc schemes improve the performance of the networked DC motor motor is controlled by its remote controller at remote station and also compare the results with the Zeigler-Nichols tuned whereas in direct structure the central controller is used for Proportional-Integral-Derivative Controller. The performance of controlling the speed of dc motor. Since the hierarchical the proposed controllers has been verified through simulation structure has a poor interaction between central and remote using MATLAB/SIMULINK package. The effective results show unit, direct structure is preferred. that the performance of networked dc motor is improved by Recently the stability analysis and control design for NCS using Intelligent Controller than the other controllers. have attracted considerable research interest [3], [4], [6] and Keywords- Networked Control Systems (NCS); Network [11]. The work of Nesic and Teel [2] presents an approach for Challenges; Tuning; Proportional – Integral - Derivative stability analysis of NCS that decouples the scheduling Controllers (PID); Fuzzy Logic Controller (FLC); Artificial Neural protocol from properties of network free nominal closed-loop Networks (ANN). system. Nesic and Tabbara [3] extended [2] by stochastic deterministic protocols in the presence of random packet I. INTRODUCTION dropouts and inter transmission time and they also proposed wireless scheduling protocol for non-linear NCS in [6]. The Networked Control System is the adaptation of communication network for information exchange between networked predictive control scheme for forward and feedback controllers, sensors and actuators to realize a closed control channels having random network delay was proposed in [4], and [5] addresses the problems of how uncertain delays are loop. Networks reduce the complexity in wiring connections smaller than one sampling period which affects the stability of and the costs of Medias. They are easy to maintain and also the NCS and how these delays interact with maximum enable remote data transfer and data exchanges among users. allowable transfer interval and the selected sampling period. Because of these benefits, many industries and institutions has shown interest in applying different types of networks for their Robust feedback controller design for NCS with uncertainty in remote industrial control and automation. Regardless of the the system model and the network induced delay has been addressed in [7]-[8], whereas [9] handles networked induction types of networks, the overall performance of NCS is affected motor speed control by using linear matrix n equality (LMI) by two major challenges as networked induced delay and data method. Ref. [1] measure the networked vehicle control losses. The challenges of networked DC motor are generally performance using an H infinity norm with linear matrix controlled by Conventional Proportional – Integral - Derivative Controllers, since they are less expensive with inequalities conditions and markovian jumping parameters in inexpensive maintenance, designed easily, and very effective. communication losses. In case of time varying transmission times, model based NCSs has been proposed for stabilization But mathematical model of the controller and tuning of PID problem of NCS. The stability analysis and controller parameters are difficult and generally not used for non-linear synthesis problems are investigated in [11] for the NCSs with systems. Hence to overcome these challenges auto-tuning and random packet losses by using H infinity control and linear adaptive PID Controller was developed with few mathematical calculations. The Intelligent controllers as Fuzzy Logic matrix inequalities. A moving horizon method was developed by [12], which was applied as a quantized NCS in a practical Controller and Artificial Neural Networks were used to context. Since these methods transmit data specifying only a overcoming the challenges. Thus this paper proposes region in which the measurements lie, it will reduce the Intelligent Controller for the compensation of the challenges. 131 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 network stabilization of the NCS. However, this method could reduce the stability of the control system by introducing uncertainty in the control system. The issues of limited bandwidth, time delay and data dropouts was taken into consideration when NCSs controllers were designed in [12] – [14]. The networked control system performance depends on the control algorithm and the network conditions. Several network conditions such as bandwidth, end-to-end delay, and packet loss rate are major impacts on networked control systems. Depending upon the control algorithm and network conditions the overall performance of the networked system may vary and hence the stability of the system. II. MODELLING Figure 3. An overall real-time networked control system. A networked control system can be divided into the remote unit, the central controller and the data network. Fig. 3 where u=ea is the armature winding input voltage; eb =Kbω is shows the general block diagram of the networked control the back-electromotive-force (EMF) voltage; L is the armature system under investigation. In order to focus our discussion on winding inductance; ia is the armature winding current; R is the performance of networked closed loop control system with the armature winding resistance; Kb is the back-EMF constant network conditions (delay, data loss), a networked dc motor and ω is the rotor angular speed. Based on Newton’s law the control system has been illustrated. mechanical-torque balance equation is A. Remote Unit dω J + Bω + Tl = Kia The Remote Unit consists of the plant (dc motor), sensor dt (2) and an interfacing unit. Via the network the remote unit can J is the system moment of inertia; B is the system damping send measurements like motor speed, current, temperature, coefficient; K is the torque constant and Tl is the load torque. and local environment information, back to the central By letting x1 = ia and x2 = ω, the electromechanical dynamics controller. The electro-mechanical dynamics of the dc motor of the dc motor can be described by the following state-space can be described by the loop equation as first order differential description: equations. • R K 1 di x1 (t ) = − x1 − b x 2 + u u (t ) = ea = L a + Ria + eb L L L dt (1) (3) • K B 1 x 2 (t ) = x1 − x 2 + Tl J J L (4) The parameters of the motor Table 1 are used for determine the state space model of dc motor. TABLE 1. DC MOTOR PARAMETERS J Moment of Inertia 42.6 e-6 Kg-m2 L Inductance 170 e-3 H Figure 1. Hierarchical Structure. R Resistance 4.67 Ω B Damping Coefficient 47.8 e-6 Nm-sec/rad K Torque Constant 14.7 e-3 Nm/A Kb Back EMF constant 14.7 e-3 Vsec/rad B. Central Controller The central controller will provide the control signal uC(t) to the remote systems. The central controller will monitor the network conditions of the remote unit link and provide appropriate control signals to each remote unit. Similarly the output responses are taken as feedback signal yR(t) to the central controller. The proposed Intelligent Controllers will compensate the network-induced delays, data losses and external disturbances. The data losses and disturbances occur due to missing or disturbances in input reference signal, control signal and feedback signal. Figure 2. Direct Structure. 132 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 C. Data Network There are different ways to define network conditions for point-to-point (from the central control to a specific remote unit). Two of the most popular network measures are the point-to-point network throughput and maximal delay bound of the largest data. One factor of interest is the sampling time. To keep the illustration simple, the remote unit receives the data sent from the central controller as uR(t), which can be mathematically expressed as uR (t) = uc (t −τ R ) (5) where τR is the time delay to transmit the control signal uC(t) Figure 4. Neural Network Controller for NCS. from the central controller to the remote unit. The remote unit also sends the sensors signals yR(t) of the remote system back TABLE 2. ANN PLANT SPECIFICATION to the central controller yC(t), and these two signals are related No. of Inputs 3 as No. of Outputs 2 y C (t ) = y R (t − τ C ) (6) No. of Hidden Layers 2 No. of Training Samples 1000 where τC is the time delay to transmit the measured signal No. of Training Epochs 200 from the remote unit to the central controller. There are also processing delays as τPC and τPR, at the central and remote unit, The error signals are trained for number of epochs by using respectively which could be approximate small constants or the NARMA-L2 controller and the control signal are generated even neglected because these delays are usually small for any challenges in the network. compared to τC and τR. The functions of network variables such as the network B. Fuzzy Logic controller throughput, the network management/policy used, the type In general, fuzzy logic control is used for the control of a and number of signals to be transmitted, the network protocol plant where the plant modeling is difficult. For such systems used, and the controller processing time, and the network that are difficult to model, fuzzy logic controller has been traffic congestion condition are taken as the current network successful by Mamdani. The basic principle of fuzzy logic lies conditions n(t) and let z-t be a time delay operator which in the definition of a set where any element can belong to a set defines the signals as with a certain degree of membership. Using this idea, the u R (t ) = u c ( z − t R , n(t )) knowledge of an expert can be expressed in a relatively simple (7) −t c form and the inference for given inputs can be implemented y c (t ) = y R ( z , n(t )) very efficiently. Due to these advantages, fuzzy logic control (8) In this paper, we have chosen sampling time as 0.5 ms and is an attractive method for NCS whose modeling is very simulations are done. difficult because of the stochastic and discrete nature of the network. Fig. 5 shows the structure of FLC for a single input III. MODELLING CONTROLLER DESIGN FOR NCS single output plant. In Fig. 5 r(t) is the reference input, y(t) is the plant output, e(t) is the error signal between the reference In this session the proposed Neural Network Controller and input and plant output and uC(t) is the control signal. Fuzzy Logic Controller as the central controller is described The FLC consists of three parts as 1) Fuzzifier that and the results are compared with the PID controller. converts the error signal into linguistic values, 2) Inference engine that creates the fuzzy output using fuzzy control rules A. Neural Network Controller generated from expert experience and 3) Defuzzifier that The proposed scheme utilizes the neural-network calculate the control input to the plant from the inferred NARMA-L2 Controller. The Neural Network Controller is results. The input and output signals to the FLC are error designed to take the error as the input and computes the output signal e(t) and control signal uC(t) respectively. In this paper, stabilizing signal depending on the input error signal. The the trapezoidal fuzzy members are selected for membership block diagram of Neural Network Controller for NCS is functions. Three fuzzy linguistic variables, i.e., Small, shown in Fig. 4. Medium and Large are defined. The coefficients of the The NARMA-L2 controller, a multilayer neural network membership function depend upon the set point and are has two steps involved as system identification stage and determined by several trial and error experiments with the control design stage. In system identification stage, a neural plant without the network. In order for faster execution of the network model of the plant which has to be controlled is fuzzy logic controller, the Mamdani’s min-max inference developed and in later stage the neural network plant model method and the central average defuzzifier are used. has been designed to train the controller. The ANN plant The rules used in this paper are as specification has been shown in Table 2. If e(t) is small then uC(t) is small 133 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 If e(t) is medium then uC(t) is medium IV. SIMULATION SETUP AND RESULTS If e(t) is large then uC(t) is large In the simulation scenario, the direct structure of the C. PID Controller networked DC motor control system is simulated using MATLAB/ SIMULINK under fully controlled environments It is used to compute the control signal to the remote dc for Neural Network Controller, Fuzzy Logic Controller and motor for step tracking, based on the monitored system signals PID Controller. Equations (3) - (4) are used as the main sent from the remote unit via the network link as in Fig. 6. The Proportional-Integral-Derivative (PID) controller used is model, and it is controlled by the controller with the insertions t of network delays according to (5) - (6). The delays are varied de(t ) according to different effects of interests. The disturbance and U PID (t ) = K p e(t ) + K I ∫ e(t )dt + K D loss of input signal, control signal and the feedback signal 0 dt (9) were made for few milliseconds at each stage and the results where KP is the proportional gain; KI is the integral gain; KD is were studied. The system setup is illustrated in Fig.4, Fig.5 the derivative gain; r(t) is the reference signal for the system to and Fig.6. Using (3)-(4) and Table 1, the state model of the dc track; y(t) is the system output; and e(t) is the error function. motor is obtained. Then the results of the ANN and FLC are In our case, y = ω is the motor speed, and UPID(t) is the input compared with the PID controller. voltage to the motor system. Output Responses of the system are obtained for all The results of model system with ANN, FLC and PID controllers used in this paper. Fig. 7 shows the comparison of Controllers for network induced delays, losses and disturbance the system performance for all controllers without delays and are simulated and compared. data losses. Fig. 8 - 10 shows the response of the system for the controllers with different network induced delays and the comparison of these performances are tabulated in Table 3. The system responses with delay and data losses are obtained as in Fig. 11. From the simulation results as in Fig. 7 - 11, the overall system performance with Intelligent Controllers as ANN and FLC are better than the PID controller. Output Response 2000 1800 1600 Figure 5. Fuzzy Logic Controller for NCS. 1400 M o to r S p e e d (rp m ) 1200 Fuzzy Logic Controller; ts = 7 ms; Mp = 3% PID Controller; ts = 10 ms; Mp = 23% 1000 Set Point = 1500 rpm Neural Network Controller; ts = 7 ms; Mp = 3 % 800 600 400 200 0 0 2 4 6 8 10 12 14 16 18 20 Time (ms) Figure 7. Comparison of System Responses for ANN, FLC and PID Figure 6. ZN Tuned PID Controller for NCS Controller without delay and losses. TABLE 3. COMPARISON OF PERFORMANCE OF THE NETWORKED DC MOTOR CONTROL SYSTEM WITH DELAY IN ANN, FLC AND PID CONTROLLER. (Set point = 1500 rpm; Sampling Time = 0.5ms) Time delay (ms) Maximum overshoot (%) Settling Time (ms) Feedforward path Feedback path PID FLC ANN PID FLC ANN 0.5 1 3.3 3.3 3 30 7 7 1 1 3.3 3.3 3 40 8 7 2 2 6.6 3.3 3 62 9 8 2 3 8 3.3 3 70 9 8 3 2 9 3.3 3 75 10 9 134 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 Output Response 1800 1600 1400 Set point = 1500 rpm 1200 FF: 0.5ms; FB: 1ms M to S e d (rp ) m FF:1 ms; FB:1ms 1000 FF:2ms; FB:1ms o r pe FF: Feedforward Delay 800 FB: Feedback Delay 600 400 200 0 0 5 10 15 20 25 30 35 40 45 50 Time (ms) Figure 8. Response of the System using PID Controller with varying delays in forward and feedback path of NCS. Output Response 1600 1400 1200 1000 M to S e d (rp ) m FF: 0.5ms; FB: 1ms; Mp 3.3%:ts: 7 ms o r pe Sep Point: 1500 rpm 800 FF:1 ms; FB:1ms;Mp: 3.3%;ts: 8 ms FF:2 ms; FB:1ms; Mp:3.3%; ts: 9 ms 600 FF: FeedForward Delay FB: FeedBack Delay 400 200 0 0 5 10 15 20 25 30 35 40 45 50 Time (ms) Figure 9. Response of the System using FLC with varying delays in forward and feedback path of NCS. Output Response 1600 1400 Motor Speed (rpm) 1200 1000 Set Point=1500 rpm 800 FF = 0.5 ms; FB = 1 ms; Mp = 3%; ts = 7 ms FF = 1 ms; FB = 1 ms; Mp = 3%; ts = 7 ms 600 FF = 2 ms; FB = 1 ms; Mp = 3%; ts = 8 ms Feedforward Delay = FF 400 Feedback Delay = FB 200 0 0 5 10 15 20 25 30 35 40 45 50 Time (ms) Figure 10. Output Response of the System using ANN with varying delays in forward and feedback path of NCS. 135 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 Output Response 2500 Set Point = 1500 rpm PID Controller Neural Network Controller 2000 Fuzzy Logic Controller M o to r S p eed (rpm ) 1500 Feedforward = Feedback Delay = 2 ms 1000 500 Loss of Input Signal 0 0 50 100 150 Time (ms) Figure 11. Comparison of system responses of ANN, FLC and PID Controllers with delay and losses. [8] D.Yue, Q.Han, and J.Lam, “Network-based robust H∞ control of V. CONCLUSION systems with uncertainty,” Automatica, vol. 41, pp. 999-1007, June 2005. Networks and their applications play a promising role for [9] J.Ren, C.Wen Li De, and Z. Zhao, “Linearizing Control of Induction real-time high performance networked control in industrial Motor Bsed on Networked Control Systems,” International Journal of applications. The major concerns are the network induced Automation and Computing, vol. 6, pp. 192-197, May 2009. delays and data losses that are provided by the network which [10] L.A.Montestruque, and P. Antsaklis, “Stability of Model-Based affects the performance of the networked control systems. This Networked Control Systems with Time-Varying Transmission Times,” paper has describes and formulates the Intelligent Controllers IEEE Trans. Autom. Control, vol. 49, pp. 1562–1571, September 2004. as Neural Network Controller and Fuzzy Logic Controller in a [11] Z.Wang, F.Yang, W.C.H.Daniel and X.Liu, “Robust H∞ Control for networked DC motor control. The numerical result are obtained Networked Systems with Random Packet Losses,” IEEE Trans. Sys. Man Cybernetics-Part B, vol. 37, pp. 916-923, August 2007. and compared for Neural Network Controller, Fuzzy Logic [12] G.C.Goodwin, H.Haimovich, D.E.Quevedo, and J.S.Welsh, “A Moving Controller and PID Controller. The effective results show that Horizon approach to networked control system design,” IEEE Trans. the performance of networked control DC motor is improved Autom. Control, vol. 49, pp. 1427–1445, September 2004. by using Intelligent Controller than the convention controller in [13] K.Li, and J.Baillieul, “Robust quantization for digital finite all network variations and deteriorations. The analysis on using communication bandwidth (DFCB) control,” IEEE Trans. Autom. intelligent controls improves and strengthens the networked Control, vol. 49, pp. 1573–1584, September 2004. control systems concepts in the future. [14] R.C.Luo, and T.M.Chen, “Development of a multibehaviour -based mobile robot for remote supervisory control through the internet,” IEEE. Trans. Mechatron., vol. 5, pp. 376-385, October 2000. REFERENCES [15] J.P.Hespanha, P.Naghshtabrizi, and Y.Xu, “A survey of recent results in [1] P.Seiler, and R. Sengupta, “An H∞ Approach to Networked Control,” networked control systems,” Proc. IEEE. , vol. 95, pp. 138-162, January IEEE Trans. Autom. Control, vol. 50, pp. 356-364, March 2005.. 2007. [2] D.Nesic, and A.R.Teel, “Input-Output stability properties of networked [16] Y.Tipsuwan, and M.Y.Chow, “Control methodologies in networked control systems,” IEEE Trans. Autom. Control, vol. 49, pp. 1650-1667. control systems,” Control Eng. Practice, vol. 11, pp. 1099-1111, October 2004. Feburary 2003. [3] M.Tabbara, and D. Nesic, “Input-Output Stability of Networked Control [17] K.Ogata, Modern Control Engineering, Englewood Cliffs, NJ: Prentice Systems With Stochastic Protocols and Channels,” IEEE Trans. Autom. Hall, 1990. Control, vol. 53, pp. 1160-1175, June 2008. [18] J.G.Ziegler, and N.B.Nichols, “Optimum settings for automatic [4] G.P.Lin, Y. Xia, J.Chen, D.Rees, and W.Hu, “Networked Predictive controllers,” Trans. ASME, vol. 64, pp. 759-768, November 1942. Control of Systems With Random Network Delays in Both Forward and [19] C.C.Lee, “Fuzzy logic in control systems: fuzzy logic controller-Part I, Feedback Channels,” IEEE Trans. Ind. Electron., vol. 54, pp. 1282-1297, ”IEEE Trans. Syst., Man, Cybern., vol. 20, pp. 404-418, Mar/Apr 1990. June 2007. [20] C.C.Lee, “Fuzzy logic in control systems: fuzzy logic controller-Part II,” [5] D.S. Kim, Y.S. Lee, W.H. Kwon, and H.S.Park, “Maximum allowable IEEE Trans. Syst., Man, Cybern., vol. 20, pp. 419-435, Mar/Apr 1990. delay bounds of networked control system,” Control Eng. Practice, vol. 11, pp. 1301–1313, 2003. AUTHORS PROFILE [6] M.Tabbara, C. Nesic, and A.Teel, “Stability of wireless and wireline networked control systems,” IEEE Trans. Autom. Control, vol. 52, pp. Dr.N.Devarajan received B.E (EEE) and M.E (Power Electronics) from GCT 1615-1630, September 2007. Coimbatore in the year 1982 and 1989 respectively. He received Ph.D in the area of control systems in the year 2000. He is currently working as Assistant [7] D. Yue, Q. Han, and P. Chen, “State feedback controller design of Professor in the department of EEE at Government College of Technology, networked control systems,” IEEE Trans. Circuits Systems II, vol. 51, Coimbatore. He published 135 papers in the national and international pp. 640-644, November 2004. 136 http://sites.google.com/site/ijcsis/ ISSN 1947-5500 (IJCSIS) International Journal of Computer Science and Information Security, Vol. 8, No.8, 2010 conferences. He published 37 papers in international journals and 12 in from Maharaja College of Engineering, Coimbatore in the year 2004. She is national journal. Under his supervision currently 10 research scholars are currently working as Senior Lecturer in the department of EIE at Sri working and 7 scholars completed their Ph.D. His areas of interests are control Ramakrishna Engineering College, Coimbatore. She is a Ph.D. research systems, electrical machines and power systems. He is a member of system scholar and published 2 papers in international journals and also presented 4 society of India, ISTE and IE(India). papers in national and international conference. Her areas of interests are networked control system and intelligent controllers. She is a member of IEEE B.Sharmila completed B.E (EIE) from Tamilnadu College of Engineering, and ISTE. Coimbatore in the year 2000. She completed her M.E (Applied Electronics) 137 http://sites.google.com/site/ijcsis/ ISSN 1947-5500