Position Estimation And Speed Control Computer Science Essay

Published Date: 02 Nov 2017

S. Antony Viyani Arasu Dr. R. Arulmozhiyal

PG Student, Associate Professor,

Department of Electrical and Electronics Engineering, Department of Electrical and Electronics Engineering,

Sona College of Technology, Salem, India. Sona College of Technology, Salem, India.

[email protected] [email protected]

Abstract—Brushless DC motors are widely used in the applications of industrial automation, automotive, aerospace, instrumentation and appliances. So the precise control with quick response time is required. For the good performance of the Brushless DC motor drive in sensorless method, accurate knowledge of the rotor position is essential. Good rapidity, stability and robustness of BLDC motor speed regulation system are necessary for effective speed control. Adaptive Neuro-Fuzzy Inference System (ANFIS) is proposed in this paper for estimating the accurate position of rotor and control of variable speed Brushless DC motor (BLDC) drive in a closed-loop. ANFIS provides a nonlinear modeling of motor drive system and the motor speed can accurately track the reference signal. ANFIS has the advantages of employing expert knowledge from the fuzzy inference system and good learning power of neural networks. The ANFIS controller is designed by computer simulation through the MATLAB/SIMULINK software. The obtained results show the effectiveness of the proposed controller.

Keywords—Brushless DC motor (BLDC); PID (Proportional Integral Derivative) Controller; Adaptive Neuro Fuzzy Interference System (ANFIS);

INTRODUCTION

Brushless DC motors are widely used in the applications of industrial automation, automotive, aerospace, instrumentation and appliances. Brushless DC motor has permanent magnet in the rotor and winding in the stator. In Brushless DC motor trapezoidal back emf is produced because of stator winding arrangement and the location of the permanent magnets on the rotor. So it is also known as Brushless Trapezoidal Permanent Magnet Motor. Brushless DC motor requires drive to supply commutated current to the motor windings synchronized to the rotor position. In other words, some kind of feedback position sensors is necessary to commutate brushless DC motor.

For the good performance of the Brushless DC motor drive, accurate knowledge of the rotor position is required. In industrial applications of Brushless DC motor, highly reliable rotor position is very important. For effective control of speed and torque, the excitations of BLDC phases need to be properly synchronized with the rotor position. An encoder, resolver, or Hall Effect position sensors are usually employed to determine the rotor position. These discrete position sensors causes additional cost, complexity and less reliable of drive. To avoid all these drawbacks, the indirect method of identifying the rotor position is appreciated.

The relationship between flux linkage, phase current and rotor position is needed to get the rotor position of BLDC motor. In sensorless scheme, to obtain the accurate rotor position, a terminal measurement of voltage and current is required without any additional hardware. In this paper, ANFIS based rotor position estimation in sensorless BLDC motor drive is explained. The major advantage of ANFIS is, it does not need a complex mathematical model and gives accurate and reliable result [1].

The speed control of the BLDC motor drive involves a complex process. Recently, various modern control solutions are proposed for the speed control of BLDC motor. However, Conventional PID controller algorithm is simple, stable, easy adjustment and high reliability [2] [3]. But it is difficult to achieve the optimal state under field conditions in the actual production. So, intelligent control algorithm has been largely employed so as to improve rapidity, stability and robustness of BLDC motor speed regulation system [4]. Fuzzy logic control technique for BLDC motor is employed to overcome the conventional PID controller drawbacks [4] [5]. However, rules of fuzzy control algorithm have some defects in time-varying nonlinear system when considering practical use, and parameters for controller lack self-adjustment capability.

ANFIS has functions of neural network, like learning and optimization ability and connection structure. In addition, human-simulated rules and expert knowledge can be applied to fuzzy system [6]. And there is no need to set the model structure according to features of system variables in advance. ANFIS can optimize control rules and membership function to improve performance through information storage and learning ability of neural network, so it can be used in nonlinear system control. In this paper, effective speed control of BLDC motor is achieved by using Adaptive neural-fuzzy inference system (ANFIS).

POSITION ESTIMATION AND SPEED CONTROL OF BLDC MOTOR

Fig. 1. Block Diagram

The complete block diagram of the proposed system is shown in Fig. 1. The voltage and current of the BLDC motor is measured by using the measurement devices. Flux estimator produces flux linkage by using the voltage and current inputs. ANFIS controller is designed to estimate the position of rotor by using the estimated flux and current.

The error generated by comparing the reference speed with the actual speed is taken as input for the speed controller. The ANFIS controller produces torque value in accordance with error inputs. Based on this torque value the triggering of IGBT is adjusted to get good speed control.

CONTROLLER DESIGN

Review of ANFIS

Neural networks can learn from data. Neural networks have good fault tolerance characteristics and extremely fast parallel computation. Neural Network technique gives good estimate of the speed and robust to parameter variation. Fuzzy logic is a good technique in the field of control. It can be used for controlling various parameters of the real time systems. The merged technique of the neural network and fuzzy logic created a hybrid technique called as Neuro-Fuzzy networks.

The ANFIS system determines a control action by using a neural network which implements a fuzzy inference. The combination of the advantage of learning power of neural network and knowledge representation of fuzzy logic is used here. ANFIS controller has two states, a learning state and a controlling state. In the learning state, the performance evaluation is carried out according to the feedback which represents the process state. In controlling state, the controlling process is carried out based on the information got in the learning state.

Position Estimation

The ANFIS is a Sugeno adaptive network based fuzzy inference system. In this paper, the fuzzy inference system for estimating the rotor position has inputs of flux linkage (ψ), current (i) and output of rotor position (ϴ). Each input has 7 membership functions. Then the rule base contains 49 fuzzy if-then rules of Takagi and Sugeno’s type. The ANFIS network is formed with five layers. This architecture is shown in Fig. 2.

Layer 1: This layer is an input layer. It represents node. Each node is a function. Each input has 7 membership functions. The output of flux linkage membership function is Ok1=µAk(ψ) and the output of current member ship function is Ok2=µBk(i). Ak and Bk are the linguistic variables (mf1, mf2,..mf7) associated with the node functions. In this work, the triangular shaped membership functions µAk(ψ) and µBk(i) are used with a maximum of 1 and a minimum of 0. The generalized triangular membership function of the flux linkage is given by

0 ψ ≤ ak

ak ≤ ψ ≤ bk

µAk(ψ) = bk ≤ ψ ≤ ck

ck ≤ ψ

0 i ≤ ak

ak ≤ i ≤ bk

µBk(i) = bk ≤ i ≤ ck

0 ck ≤ i

Where ak , bk and ck are the adaptable variables known as parameters.

Layer 2: This layer is a fuzzification layer. It does the fuzzy AND operator. This provides the strength of each rule.

Wk = µAk(ψ) x µBk(i)

Where k = 1, 2, 3…7;

i current in amps;

ψ flux linkage in volts/s.

Layer 3: This is a normalization layer. The no of layers in this layer is equal to previous layer. It computes each rule’s firing strengths.

Wk =

Layer 4: This layer is an adaptive layer. The output of adaptive layer a linear combination of the inputs multiplied by firing strength.

Ok4 = Wk fk = Wk (mk ψ+ nk i+ rk)

Fig. 2. ANFIS Architecture

Where Wk is the output of layer 3 and the modifiable variables mk, nk and rk are known as consequent parameters.

Layer 5: This layer is output layer.output of output layer is the summation of output of the nodes in the previous layer. The output is theta(Ï´).

Ok5 = =

Speed Controller

The fuzzy inference system for the speed controller has inputs of error (e) and change in error (Δe) and the output of Torque (ϴ). Each input has 7 membership functions. Then the rule base contains 49 fuzzy if-then rules of Takagi and Sugeno’s type. The ANFIS network is formed with five layers.

e = Wref – W

Δe = en – en-1

Layer 1: This layer is an input layer. It represents node. Each node is a function. Each input has 7 membership functions. The output of member ship function of error is Ok1=µAk(e) and the output of membership function of change in error is Ok2=µBk(Δe). Ak and Bk are the linguistic variables (mf1, mf2,.. mf20) associated with the node functions. In the speed controller also the triangular shaped membership functions µAk(e) and µBk(Δe) are used with a maximum of 1 and a minimum of 0. The generalized triangular membership function of the speed controller is given by

0 e ≤ ak

ak ≤ e ≤ bk

µAk(e) = bk ≤ e ≤ ck

0 ck ≤ e

0 Δe ≤ ak

ak ≤ Δe ≤ bk

µBk(Δe) = bk ≤ Δe ≤ ck

0 ck ≤ Δe

Where ak , bk and ck are the adaptable variables known as parameters.

Layer 2: This layer is a fuzzification layer. It does the fuzzy AND operator. This provides the strength of each rule.

Wk = µAk(e) x µBk(Δe)

Where k = 1, 2, 3… 7 ;

e error in rpm;

Δe change in error in rpm.

Layer 3: This is a normalization layer. The no of layers in this layer is equal to previous layer. It computes each rule’s firing strengths.

Wk =

Layer 4: This layer is an adaptive layer. The output of adaptive layer a linear combination of the inputs multiplied by firing strength.

Ok4 = Wk fk = Wk (mk e+ nk Δe + rk)

where Wk is the output of layer 3 and the modifiable variables mk, nk and rk are known as consequent parameters.

Layer 5: This layer is output layer.output of output layer is the summation of output of the nodes in the previous layer. The output is Torque(T).

Ok5 = =

SIMULATION RESULTS & DISCUSSION

The overall simulink diagram of the proposed system is shown in Fig.3. The input supply DC is inverter by the three phase IGBT/DIODE inverter. The input current and voltage of the BLDC motor is measured by using measurement devices. For estimating the rotor position, the value of flux linkage is needed. So to find this, flux estimator is used. It takes voltage and current value as input and find the value of flux linkage by using the relation between them. In ANFIS controller two set of 7 membership functions are used as input for phase current and flux linkage. The output of ANFIS controller is the rotor angle. The ANFIS controller used for speed controlling purpose has iputs speed error and change in error. Based on the output of ANFIS speed controller, current controller adjusts the triggering signal of the three phase inverter.

ANFIS model needs training data and checking data for training the neuron. These data got from the experimental results. The experimented data are imported from the files to ANFIS editor. After the fis file is generated. For generating fis file there are three ways. If the fis file is available, that has to be loaded, if it is not available,grid partition or sub. Clustering is used. In training the fis file also hybrid and backpropagations methods are there. In this case, backpropagation is used. The structure of ANFIS controller is shown in fig.4. It has two set of inputs such as flux linkage and phase current in the input layer. In the fuzzification layer both the inputs have 7 membership functions. This ANFIS structure shows the overall process of it.

Fig. 3. Simulink of overall proposed system

Fig. 4. ANFIS structure of position estimation

The simulation results achived in obtaining the rotor position is shown in Fig.5. The effective estimation of rotor results smooth and stable rotation. The estimated rotor angle using ANFIS controller is compared with the actual rotor angle. The error is very very less. This shows the effecitive estimation of the proposed controller.

The structure of ANFIS controller for the speed control is shown in fig.6. It has two set of inputs such as error and change in error in the input layer. In the fuzzification layer both the inputs have 7 membership functions. This ANFIS structure shows the overall process of it. The training data

Fig. 5. Simulation result of rotor angle estimation

Fig. 6. ANFIS structure of speed controller

Fig. 7. Speed Response of BLDC

for the ANFIS controller is taken from the simulation result of the Proportional Integral(PI) speed control of BLDC motor. The testing data is taken from the experimental results of speed control of BLDC motor. After importing these files, fis file is generated. To train the fis backpropagation method is used. Trainin and testing have to be done to get the complete output.

The simulation results achived in obtaining the speed control is shown in fig.7. The reference speed is compared with output speed achieved by the use of ANFIS controller. The motor speed achieves the reference speed in very less time and maintain the stablility. This shows the stability and the quick response of the proposed controller. To achieve more accurate result the number of training data in parameter and the the number of membership function has to increase. This will increase good speed response with very quick tracing of reference speed.

CONCLUSION

A systematic approach of estimating the rotor position and achieving the speed control of brushless dc motor drive by means of adaptive neuro fuzzy inference control strategy has been investigated in this paper. The simulated rotor angle waveform is compared with the actual rotor position and the speed response of the motor is compared with the reference speed. The results show that the proposed ANFIS based sensorless scheme is estimating the rotor position with a very low error. Due to the incorporation of the ANFIS controller in closed loop with the BLDC motor, it was observed that the motor reaches the rated speed very quickly in a lesser time. The ANFIS scheme is computationally efficient, works well with linear techniques. The results show that the ANFIS controller provides very good dynamic response, faster settling times and good stabilization.