CHAPTER 3 PERFORMANCE ANALYSIS WITH PI, ANFIS AND SLIDING MODE CONTROLLERS

61 CHAPTER 3 PERFORMANCE ANALYSIS WITH PI, ANFIS AND SLIDING MODE CONTROLLERS 3.1 INTRODUCTION The Permanent Magnet Synchronous Motor (PMSM) has a sinusoidal back emf and requires sinusoidal stator currents to produce constant torque while the permanent magnet brushless dc (PMBLDC) motor has a trapezoidal back emf and requires rectangular stator currents to produce constant torque (Miller 1989). The system is becoming increasingly attractive in high-performance variable-speed drives since it can produce torque-speed characteristic similar to that of a permanent-magnet conventional dc motor while avoiding the problems of failure of brushes and mechanical commutation. The PMBLDC motor is becoming popular in various applications because of its high efficiency, high power factor, high torque, simple control and lower maintenance. BLDC motor is one type of synchronous motor, which can be operated in hazardous atmospheric condition and at high speeds due to the absence of brushes. Pillay et al (1985) and Krishnan et al (1990) have investigated that the PMSM a sinusoidal back emf and requires sinusoidal stator currents to produce constant torque, while the BLDC motor has a trapezoidal back emf and motor requires rectangular stator current to produce constant torque. (Bhim singh et al 2002) have proposed a digital speed controller for BLDC motor and implemented in a digital signal processor (DSP). Later in 1998, the

62 rotor position and the speed of permanent magnet have been estimated using Extended Kalman filter (EKF) by Peter Vas et al (1998). Jadric et al (2001) have proposed that the hall effect sensor are usually needed to sense the rotor position which senses the position signal for every 60 degree electrical. The mathematical model of the BLDC motor has been developed and validated in MATLAB platform with proportional-integral (PI) speed controller (Varatharaju et al 2011). Host of efforts has attempted and solved the problem of non linearity and parameter variations of PMBLDC drive (Lajoie et al 1985, Sebastain et al 1989, Rubai et al 1992, Luk at al 1994 and Radwan et al 2005). The draw-backs of Fuzzy Logic Control (FLC) and Artificial Neural Network (NN) can be over come by the use of Adaptive Neuro-Fuzzy Inference System (ANFIS) (Zhi Rui Huang et al 2006 and Uddin 2007). ANFIS is one of the best tradeoff between neural and fuzzy systems, providing: smooth control, due to the FLC interpolation and adaptability, due to the NN back propagation. Some of advantages of ANFIS are model compactness, require smaller size training set and faster convergence than typical feed forward NN. Since both fuzzy and neural systems are universal function approximators, their combination, the hybrid neuro-fuzzy system is also a universal function approximator. The non-linear mapping in a neurofuzzy network is obtained by using a fuzzy membership function based neural network. Using the developed model of the BLDC motor, a detailed simulation and analysis of a BLDC motor speed servo drive is obtained. Closed loop control of PMBLDC motor drive consisting of PI speed controller and hysteresis current controller is simulated. In addition, the ANFIS controller and SMC also designed.

63 3.2 PI CONTROLLER To accurately control the speed of drive without extensive operator involvement, a speed control system that relies upon a controller is essential. The speed control system accepts the actual motor speed through a speed sensor such as a Tacho generator or Hall sensors, encoders etc as input. It compares the actual speed to the desired control speed otherwise called set point, and provides an output, which adjusts the duty cycle of PWM pulses applied to the IGBT based VSI fed BLDC drive. The controller is one part of the entire control system and the whole system should be analyzed in selecting the proper controller. The following factors should be considered when selecting a controller: (i) Type of input sensor (Hall sensors/encoders/tacho generator/ Revolvers) and speed range (ii) Type of output required (electromechanical relay, solid state relay, analog output or digital output) (iii) Control algorithm needed (On/Off, Proportional, PI, PID) (iv) Type of outputs required for forward/reverse motoring with controlled current mode of operations Only very moderate control of speed can be achieved by causing motor power to be simply switched ON and OFF according to an under or over speed condition respectively. Ultimately, the motor power will be regulated to achieve a desired system speed but refinement can be employed to enhance the control accuracy. Such refinement is available in the form of Proportional (P), Integral (I) and Derivative (D) functions applied to the control loop. These functions, referred to as control terms can be used in combination according

64 to system requirements. The proposed scheme uses combination of P and I controller to achieve precise speed control. To achieve optimum speed control, irrespective of the technique employed, one should ensure: (a) Adequate motor power is available (b) The hall sensor or encoder that acts as a speed sensor, is located in a suitable place so as to measure actual shaft speed such that it will respond for dynamic changes occurring in the motor shaft. (c) Adequate Sensor Measuring Range in the system to optimize the measuring sensitivity under varying load or speed conditions. Proportional Integral controller is most preferable controller in industries for servo drives for closed loop control that requires analytical model and transfer function of the system. Tuning of PI parameters over entire range of control is difficult and time consuming task. The proposed scheme eliminates that PI tuning difficulties. PI controller implementation is done in varies digital controllers. Tuning a PID controller, involves setting the Proportional gain KP, Integral gain KI, and Derivative gain KD values to get the best possible control for a particular process. If the controller does not include an auto tune algorithm, then the unit must be tuned using trial and error. Tuning of the PID settings is quite a subjective procedure, relying heavily on the knowledge and skill of the control or design engineer. Although tuning guidelines are available, the process of controller tuning can still be time consuming with the result that many plant control loops are often poorly tuned and full potential

65 of the control system is not achieved. The effect of change is gain of P, I and D controller upon the system response are highlighted in Fig 3.1 to 3.3. Figure 3.1 Effect of changes in K P Figure 3.2 Effect of changes in K I

66 Figure 3.3 Effect of changes in K D PI controller is widely used in industry due to its ease in design and simple structure. The rotor speed r(n) is compared with the reference speed r * (n) and the resulting error is estimated at the nth sampling instant as : (n) (n) (n) (3.1) * e r r The new value of torque reference is given by: T T K ( (n) (n 1) K ( (n) (3.2) (n) (n1) p e e 1 e Where e (n 1) is the speed error of previous interval, and e(n) is the speed error of the working interval. K p and K 1 are the gains of PI speed controller. By using Ziegler Nichols method the K p and K I values are determined. Figure 3.4 show the block diagram of typical closed loop control of BLDC motor drive.

67 Rectifier PWM 1~6 * r r r Speed PI Regulator slip * s r PWM Technique Voltage source 3- phase Inverter r Speed by Hall Sensors 3-Phase BLDC Motor Figure 3.4 Closed Loop arrangement with motor and sensors 3.3 ANFIS CONTROLLER In this section basics of ANFIS and development of ANFIS controller are given. ANFIS uses the neural network s ability to classify data and find patterns. It then develops a fuzzy expert system that is more transparent to the user and also less likely to produce memorization error than a neural network. ANFIS keeps the advantages of a fuzzy expert system, while removing (or at least reducing) the need for an expert. The problem with ANFIS design is that large amounts of training data require developing an accurate system. The ANFIS, first introduced by Jang (1993), is a universal approximator and, as such, is capable of approximating any real continuous function on a compact set to any degree of accuracy (Jain et al 1999, Jang et al 1993 and Jang et al 1997). ANFIS is a method for tuning an existing rule base with a learning algorithm based on a collection of training data. This allows the rule base to adapt.

68 The section presents a methodology for developing adaptive speed controllers in a BLDC motor drive system. A PI controller is employed in order to obtain the controller parameters at each selected load. The resulting data from PI controller are used to train ANFIS that could deduce the controller parameters at any other loading condition within the same region of operation. The ANFIS controller is tested at numerous operating conditions with hysteresis current controlled position determination. The section also provides comparison of PI controller with ANFIS controller. The BLDC motor drive system with PI controller exhibits higher overshoot and settling time when compared to the designed ANFIS controller. 3.3.1 Rules As a simple example, a fuzzy inference system with two inputs x and y and one output z is assumed. The first-order Sugeno fuzzy model, a typical rule set with two fuzzy If Then rules can be expressed as: Rule 1: If x is A 1 and y is B 1, then f 1 =p 1 x+q 1 y+r 1 (3.3) Rule 2: If x is A 2 and y is B 2, then f 2 =p 2 x+q 2 y+r 2 (3.4) The resulting Sugeno fuzzy reasoning system is shown in Figure 3.5. Here, the outputzis the weighted average of the individual rules outputs and is itself a crisp value. 3.3.2 ANFIS Architecture The ANFIS architecture is shown in Figure.3.6. If the firing strengths of the rules arew 1 andw 2, respectively, for the particular values of the inputs A i and integral of B i, then the output is computed as weighted average, w f f w w f w 1 1 2 2 1 2 (3.5)

69 Let the membership functions of fuzzy sets A i and B i, are µ Ai and µ Bi. Layer 1: Each neuron i in layer 1 is adaptive with a parametric activation function. Its output is the grade of membership function; an example is the generalized bell shape function. (x) 1 x c 1 a 2b (3.6) Where [a, b, c] is the parameter set. As the values of the parameters change, the shape of the bell-shape function varies. Layer 2:.Every node in layer 2 is a fixed node, whose output is the product of all incoming signals. w i = µ Ai (x) µ Bi (y), i=1,2 (3.7) Layer 3: This layer normalizes each input with respect to the others (The i th node output is the i th input divided the sum of all the other inputs). w i wi w w 1 2 (3.8) Layer 4: This layer si th node output is a linear function of the third layer si th node output and the ANFIS input signals. wifi w i(pix qiy r i ) (3.9) Layer 5: This layer sums all the incoming signals. f w f w f (3.10) 1 1 2 2

70 A 1 B 1 Main (or) Pr oduct w 1 f1 px 1 qy 1 r1 X Y B A 2 2 w 2 f2 px 2 q2y r weighted average x X y Y f wf w f w w 1 1 2 2 1 2 Figure 3.5 Two-input first-order Sugeno fuzzy model with two rules Layer 1 Layer 2 Layer 3 Layer 4 x y A 1 A 2 1 N 1 1f 1 Layer 5 f B 1 N 2f 2 B 2 2 2 x y Figure 3.6 Equivalent ANFIS architecture 3.4 SLIDING MODE CONTROLLER Sliding mode control is an important robust control approach. For the class of systems to which it applies, sliding mode controller design provides a systematic approach to the problem of maintaining stability and consistent performance in the face of modelling imprecision.

71 This section investigates variable structure control (VSC) as a highspeed switched feedback control resulting in sliding mode. For example, the gains in each feedback path switch between two values according to a rule that depends on the value of the state at each instant. The purpose of the switching control law is to drive the nonlinear plant s state trajectory onto a prespecified (user-chosen) surface in the state space and to maintain the plant s state trajectory on this surface for subsequent time. The surface is called a switching surface. When the plant state trajectory is above the surface, a feedback path has one gain and a different gain if the trajectory drops below the surface. This surface defines the rule for proper switching. This surface is also called a sliding surface (sliding manifold). Ideally, once intercepted, the switched control maintains the plant s state trajectory on the surface for all subsequent time and the plant s state trajectory slides along this surface. The most important task is to design a switched control that will drive the plant state to the switching surface and maintain it on the surface upon interception. A Lyapunov approach is uesd to characterize this task. Lyapunov method is usually used to determine the stability properties of an equlibrium point without solving the state equation. Let V(x) be a continuously differentiable scalar function defined in a domain D that contains the origin. A function V(x) is said to be positive definite if V(0)=0 and V(x)>0 for x. It is said to be negative definite if V(0)=0 and V(x)>0 for x. Lyapunov method is to assure that the function is positive definite when it is negative and function is negative definite if it is positive. In that way the stability is assured. A generalized Lyapunov function, that characterizes the motion of the state trajectory to the sliding surface, is defined in terms of the surface. For each chosen switched control structure, one chooses the gains so that

72 the derivative of this Lyapunov function is negative definite, thus guaranteeing motion of the state trajectory to the surface. After proper design of the surface, a switched controller is constructed so that the tangent vectors of the state trajectory point towards the surface such that the state is driven to and maintained on the sliding surface as shown in Figure 3.7. Such controllers result in discontinuous closed-loop systems. Let a single input nonlinear system be defined as (n) x f (x, t) b(x, t)u(t) (3.11) Here, x (t) is the state vector, u(t) is the control input (in our case braking torque or pressure on the pedal) and x is the output state of the interest (in our case, wheel slip). The other states in the state vector are the higher order derivatives of x up to the (n-1) th order. The superscript n on x(t) shows the order of differentiation. f(x,t) and b(x,t) are generally nonlinear functions of time and states. The function f(x) is not exactly known, but the extent of the imprecision on f(x) is upper bounded by a known, continuous function of x; similarly, the control gain b(x) is not exactly known, but is of known sign and is bounded by known, continuous functions of x. The control problem is to get the state x to track a specific time-varying state x d in the presence of model imprecision on f(x) and b(x). A time varying surface s(t) is defined in the state space R (n) by equating the variables(x;t), defined below, to zero. d (3.12) dt n1 s(x,t) ( ) x(t) Here, is a strict positive constant, taken to be the bandwidth of the system, and x(t) x(t) x (t) is the error in the output state where x d (t) is the desired d state. The problem of tracking the n-dimensional vector x d (t ) can be replaced by a first-order stabilization problem in s. s(x; t ) verifying (3.12) is referred

73 to as a sliding surface, and the system s behaviour once on the surface is called sliding mode or sliding regime. From (3.12) the expression of s contains (n 1) x, differentiating s once will make the input u to appear. Furthermore, bounds on s can be directly translated into bounds on the tracking error vector x, and therefore the scalar s represents a true measure of tracking performance. The corresponding transformations of performance measures assuming x (0) 0 is: 0, ( ) 0, i i t st t x ( t) (2 ) (3.13) where / n1. In this way, an nth-order tracking problem can be replaced by a 1st-order stabilization problem. 1 d s 2 s 2 dt (3.14) The simplified, 1st-order problem of keeping the scalar s at zero can now be achieved by choosing the control law u of (3.11) such that outside of S(t) where is a strictly positive constant. Condition (3.14) states that the squared distance to the surface, as measured by s 2, decreases along all system trajectories. Thus, it constrains trajectories to point towards the surface s(t). In particular, once on the surface, the system trajectories remain on the surface. In other words, satisfying the sliding condition makes the surface an invariant set (a set for which any trajectory starting from an initial condition within the set remains in the set for all future and past times). Furthermore, equation (3.14) also implies that some disturbances or dynamic uncertainties can be tolerated while still keeping the surface an invariant set.

74 x finite-time reaching phase Sliding mode exponential convergence x (t) d s 0 x Figure 3.7 Graphical interpretation of equations (3.12) and (3.14) (n=2) Finally, satisfying (3.12) guarantees that if x(t=0) is actually off x d (t=0), the surface S(t) will be reached in a finite time smaller than s(t=0) /. Assume for instance that s(t=0)>0, and let t reach be the time required to hit the surface s=0. Integrating (2.14) between t=0 and t reach leads to 0-s(t=0)=s(t=t reach )-s(t=0) - (t reach -0) Which implies that t reach s(t=0)/ Starting from any initial condition, the state trajectory reaches the time-varying surface in a finite time smaller than s(t=0) /, and then slides along the surface towards x d (t) exponentially, with a time-constant equal to 1/. In summary, the idea is to use a well-behaved function of the tracking error, s, according to (3.12), and then select the feedback control law u in (3.11) such that s 2 remains characteristic of a closed-loop system, despite the presence of model imprecision and of disturbances.

75 3.5 SIMULATION RESULTS 3.5.1 PI Controller The set of equations representing the model of the drive system developed in chapter 2 and simulated with PI Speed controller. The results are observed for the 3 phase, 2.0 hp, 4- pole 1500 rpm, 4 A motor (Refer table 2.2). Figure 3.8 and 3.9 show the simulated results for the steady state and transient responses respectively. Figure 3.8 Torque and Speed Waveforms when moment of inertia = 0.013 kg-m 2 Figure 3.9 Torque and speed waveforms for Step Change in Moment of Inertia at 0.5 sec

76 In Figure 3.8 shows the Torque and Speed variations for moment of inertia 0.013kg-m 2. It reaches the steady state torque and speed suddenly at time 0.03seconds. From these figures it is inferred that increasing the moment of inertia ploys an important role in settling time. 3.5.2 ANFIS Controllers Fuzzy membership functions can take many forms, but simple straight-line functions are often preferred. Triangular membership functions are often selected for practical applications and different membership functions are tried for the minimum mean root square errors (MRSE). A set of modified membership functions are derived through training the ANFIS by using data obtained from PI controller as illustrated in Figure.3.10 and Figure.3.11. ANFIS controller is designed with two inputs (speed error and change in speed error) and one output and shown in Figure.3.12. Figure3.13 shows the stator currents while Figure.3.14 and Figure.3.15 show the torque and speed responses respectively for ANSFIS controller. Figure 3.10 Membership Functions Obtained after Training for Speed Error

77 Figure 3.11 Membership Functions Obtained after Training for Change in Speed Error Figure 3.12 Architecture of ANFIS

78 Figure 3.13 Stator Current ANFIS controller Figure 3.14 Torque Response -ANFIS Controller Figure 3.15 Speed Response ANFIS Controller 3.5.3 SMC The performance comparison of SMC and the PI controller is provided in this section. Both load and line variations are obtained. The load torque is varied at time 2S from 11 N-m to 15 N-m. Figure 3.16 show the speed response for both the controllers. Figure 3.17 and 3.18 show the stator current, torque and back EMF respectively for PI controller and SMC. Speed response results are illustrated for reduction of load torque from 11N-m to 5N-m in Figure 3.19.

79 Figure 3.16 Comparison of speed response PI and SMC Figure 3.17 Current, torque and back EMF PI Controller

80 Figure 3.18 Current, torque and back EMF SMC Controller Figure 3.19 Comparison between PI controller and SMC-load torque step change

81 The study of line voltage variation is done by changing the dc bus voltage from 400V to 600V and again to 400V. The results are provided from Figure 3.20 to 3.23. Figure 3.20 Comparison between PI controller and SMC-Line variation Figure 3.21 Current, torque and back EMF -Line variation (PI)

82 Figure 3.22 Current, torque and back EMF -Line variation (SMC) Figure 3.23 Comparison of speed response -Line variation (PI and SMC)

83 3.6 HARDWARE IMPLEMENTATION When system engineers implement BLDC speed control with a digital signal processor, coding the control algorithm will be the first step. The DSP must provide sufficient resources, particularly enough computing bandwidth to run the control algorithm, interface to sensors, and drive the power switches that supply current to the motor s windings. A system engineer implementing a design that uses a BLDC motor must choose the best combination of control algorithm and DSP, given the application s performance requirements and cost budget. This can be a challenge. The information presented in this thesis is intended to aid the process of finding optimum solutions. Carefully studied BLDC applications and applicable speed control algorithms in the light of MCU resource requirements, particularly the analysis also covered DSP resource needs, such as analog-to-digital converters (ADCs), PWM outputs, special timers, pin counts, and on-chip RAM and flash ROM. This accumulated knowledge is applicable to a wide range of BLDC applications. DSP chips characterized by the execution of most instructions in one instruction cycle, complicated control algorithms can be executed with fast speed, making very high sampling rate possible for digitally-controlled inverters than microprocessors. Shih-Liang Jung et al (1998) proposed a DSP TMS32OC14 based repetitive control scheme to obtain high power factor at low switching frequency. Their control scheme consists of two parts: ac current loop and dc voltage loop. Current loop employs a repetitive controller so that the inherent periodic error caused by low switching frequency can be suppressed. As to the voltage loop, a digital PI controller with load current compensation is designed to regulate the dc-link voltage such that the ripple can be minimized. Recently, direct torque control of multi-phase induction motor using

84 TMS320F2407 DSP was developed (Mythili and Thyagarajah 2005). By using DSP different controllers are designed in the areas of active filter, power factor correction methods etc. A model layout of DSP control in motor control is highlighted in Figure 3.24. The thesis utilizes DSP TMS320LF2407. Figure 3.24 BLDC Motor Control in application 3.6.1 Motor Control Features High-Performance 16-bit CPU core with hardware multiplier Up to three 16-bit timers with 40MHz clock 3.3us ADC conversion with timer-trigger start option

85 Advanced Motor-tuned Timers Up to 6-ch complimentary PWM signals with independent compare registers Programmable Dead-Time Control (16-bit) Selectable buffer operation for fast timer reload PWM signal shut-off using external trigger Safety Mechanism for UL Compliance Watchdog with dedicated on-chip oscillator Power-on Reset, Voltage Detection Circuits Outstanding Development Support Multiple motor control algorithms Free Compiler for up to 64KB code and peripheral code generator 3.6.2 Description of DSP Resources In this configuration, the DSP has six output pins, of which a minimum of three pins must have PWM output capability. The remaining pins are in the high/low or on/off state. It s worth noting that even though only three PWM output pins are required, in many cases six PWM output pins are preferred. A timer or set of timers is required to create the necessary PWM outputs for each phase. The timer creates the PWM output waveforms with a selected carrier frequency and appropriate PWM duty ratio. The PWM duty cycle is determined in various ways. For some applications, the duty cycle is kept constant and Hall sensor signals are used only for commutation switching. In such cases, neither speed nor current

86 measurement is required. In other applications, a timer is used to measure the time between two Hall sensor signals. The time measurement is then converted to a speed measurement and a proportional integration (PI) speed loop is executed to determine the PWM duty cycle. Speed loop execution is asynchronous to the carrier frequency, however, which makes tuning the gain of the PI loop difficult and reduces performance. Furthermore, the frequency of the speed loop is directly related to the speed of the motor. At lower speeds, the loop rate is lower resulting in poor performance, while at faster speeds, the loop rate is higher, resulting in comparatively better performance. Nevertheless, overall performance is adequate only for some applications. This control method generally does not handle what is known as torque control. For applications requiring torque control, current must be measured. Torque is directly proportional to current. Thus, current measurement allows a system engineer to decipher the torque-speed point of the motor. Based on this point, the load can be estimated and the motor operated at a proper speed for high efficiency. Current measurements have to be synchronous to the carrier frequency, and the speed loop must be synchronous to the current loop. The output of the speed loop is a reference current that becomes an input to the current loop. The output of the current loop is usually the voltage necessary to drive the motor at that instant of time. This voltage is transferred into the PWM duty cycle using binary arithmetic. Thus, three increasingly complex scenarios are possible using Hall-sensor-based trapezoidal control: the PWM duty cycle can be constant; a speed loop alone can be used to determine the PWM duty cycle; or a combination of a speed loop and current loop can be used to determine the PWM duty cycle. 3.6.3 Closed Loop Speed control of BLDC Motor In Closed speed loop, a speed sensor is added and a speed loop used to adjust the frequency and voltage of the applied sine wave. The reason

87 that a sensor is used to measure speed instead of a back-emf signal is simple: When all three phases are used to drive the motor, there is no free phase through which a back-emf signal can be detected. Even if such a signal could be detected within a very short dead time, it would be inadequate for determining rotor position. Thus, some type of sensor is needed on the motor to give either speed information or rotor position data. Figure 3.25 shows the schematic diagram of the hardware system, where the system is controlled via a DSP controller TMS320LF2407A. DSP commands are isolated and amplified via HCPL A316J gate drivers and the phase currents are measured current transducers. Determination of the voltage functions and making the virtual position signals are carried out via hardware. Therefore, it is possible to employ the capabilities of the capture unit in the event manager module of DSP. Figure 3.25 Sensor-controlled BLDC Motor drive based on TMS320LF2407A

88 Hall sensors can provide accurate rotor position information and are suitable for this motor control method. They can be used to determine rotor angle and to synchronize the angle estimation for correct sine computations. By using Hall sensors and a simple timer, a system engineer can determine the motor speed, just as was done in the 6-step trapezoidal method. Both sensor and timer functions can be added easily to the code. A speed loop is executed at the proper time to increase or decrease the frequency and voltage level of the sine wave. Because speed measurement is asynchronous, frequency adjustments will also be asynchronous. Hall sensors are expensive and require special mounting procedures. To reduce cost, it is possible to use only one Hall sensor to measure speed instead of the usual three sensors. The fabricated experiential setup is shown in Figure 3.26 while representative gate pulses are given in Figure 3.27. The line and phase current waveforms are shown in Figure 3.28 and 3.29. Figure 3.26 Experimental setup

89 (a) (b) (c) (d) Figure 3.27 Gate Pulses for closed loop Six Switch Inverter control of BLDC Motor (a) Line (b) Phase Figure 3.28 Current Waveforms using Digital scope Figure 3.29 Phase Current using CRO

90 Table 3.1 Comparison of PI, ANFIS AND SMC Controllers Controller PI ANFIS SMC Peak overshoot 1518 rpm 1506rpm 1501rpm Settling time 1.4 ms 0.7ms 0.2 ms Steady state error 0.5 rpm 0.3rpm 0.1rpm 3.7 SUMMARY The designed trapezoidal BLDC motor based drive system has the closed-loop speed control, in which PI algorithm is adopted and the position determination is done through hysteresis current control. The nonlinear simulation model of the BLDC motors drive system with PI control based is simulated in the MATLAB/Simulink platform. The simulated results in electromagnetic torque and rotor speed are given for inertia change, load variation and line variation. ANFIS served as a basis for constructing a set of fuzzy if-then rules with appropriate membership functions to generate the stipulated input-output pairs. The performance of the developed MATLAB based speed controller of the drive has revealed that the algorithms developed to analyze the behavior of the PMBLDC motor drive system work satisfactorily in software implementation. Using neuro-fuzzy controller error can be reduced and train the membership functions to get the improved speed characteristics. It is found that the ANFIS controller shows reduced overshoot and settling time in both start-up and loaded change conditions and hence robust response. Detailed steps involving implementing the PI controller based drive system is discussed.