CHAPTER 1 INTRODUCTION

Transcription

1 1 CHAPTER 1 INTRODUCTION 1.1 GENERAL All industrial, commercial and other units rely on electrical motors for various applications. According to the research by the Electric Power Research Institute (EPRI), motors account for 51% of the world s total energy consumption. The energy consumed by other sectors is comparatively lower. Lighting, for example, accounts for 19%, heating and cooling system 16% and information technology 14%. Clearly, there is necessity for an efficient and robust controller for motor control which will lead to saving in energy. Electric motors influence almost every aspect of modern living. Refrigerators, vacuum cleaners, elevators, air conditioners, Washing machine, fans, computer hard disk drives and industrial processes use electric motors. In fact, motors consume the most of the energy, no matter what the scenario is, residential, industrial or commercial application. The energy efficiency of a motor depends on the type of the motor. Some are built to be more energy efficient while some are not. Also recent rapid proliferation of motor drives in the automobile industry with the new hybrid technology has created a great demand for highly efficient variable speed motor drives. In many adjustable speed drives, the demand is for precise and continuous control of speed with long-term stability, good transient

2 2 performance and better efficiency. Conventional DC motors are highly efficient, simple in construction and have linear torque-speed characteristic. Conventional DC motors satisfy some of these requirements, however due to the presence of commutator and brushes, a DC motor requires periodic maintenance and replacement of brushes. This limits the role of DC motors in commercial applications. In addition, the ratio of delivered torque to the size of the motor is poor, restricting its usage in applications where space and weight are crucial factors- especially in electric vehicles and aerospace applications. Hence, the Brushless DC (BLDC) motors have evolved as a better alternative to the conventional DC motors. 1.2 PERMANENT MAGNET MOTOR DRIVES Improvements in Permanent Magnet materials and power electronic devices have resulted in reliable and cost effective Permanent Magnet motor drives for many applications. There are two types of Permanent magnet motors based on their back emf waveforms: Permanent Magnet Synchronous Motor (PMSM) and Permanent Magnet Brushless DC (PMBLDC) motor type. A PMSM drive requires, continuous and accurate detection of rotor position information, making the system complex. On the contrary, a PMBLDC motor drive requires rotor position information only at six instants, making the commutation process easier and simple. The PMBLDC motor drives are appealing candidates for many high performance applications because of their attractive characteristics. The important features are better speed-torque characteristics, good dynamic response, high power density, reasonable torque to inertia ratio, better efficiency, robustness, long operating life, noise less operation and reliability. From the modelling perspective, a BLDC motor looks exactly similar to a DC motor, possessing a linear relationship between current and

3 3 torque, voltage and speed. It has an electronically controlled commutation system instead of mechanical commutation as in the case of conventional brushed DC motors Details of PMBLDC Motor A PMBLDC motor with the trapezoidal back emf waveform and quasi square wave current is referred as BLDC motor in this thesis. In the BLDC motors, the back emf induced, has trapezoidal waveform and the stator must be supplied with a quasi square wave current waveform, whereas in the sinusoidal motors, the induced emf is sinusoidal with the current fed to the stator being a sinusoidal waveform. Figure 1.1 shows the ideal back emf waveforms of the BLDC motor for the trapezoidal and sinusoidal types. Figure 1.1 Ideal back emf waveforms of the BLDC motor for the trapezoidal and sinusoidal types

4 Operation of BLDC Motors A three phase BLDC motor is normally operated on a two phase on-mode, i.e. the two stator phases that produces the highest torque are energized leaving the third phase open. The commutation instants are determined by the rotor position sensors. Based on the rotor position signal, the stator phase windings are energized. There are two types of control techniques, namely, sensor based and sensorless based control. The BLDC motor control based on sensors has many disadvantages. Mechanical sensors increase the inertia of the rotor shaft and they require additional space. In order to overcome these issues a sensorless technique is preferred. 1.3 ARTIFICIAL INTELLIGENCE TECHNIQUES IN ELECTRIC DRIVE APPLICATIONS Advances in AI techniques like Neural Networks, Fuzzy Logic, Genetic Algorithm, and Particle Swarm optimization techniques have made tremendous impact on the engineering applications. Recently, AI techniques are making a great impact in electrical engineering, particularly in the area of Power Electronics application in motor drives. AI technique is basically a computer simulation of human thinking usually referred as computational intelligence. They are used in many interesting areas of power electronics and motor drives applications. AI techniques are basically classified into four main categories. They are the Expert system, Fuzzy Logic (FL) system, Artificial Neural Network (ANN), and Genetic Algorithm.

5 5 Expert Systems Expert system is basically an intelligent computer program based on Boolean logic that is designed to impart the expertise of a human being in certain domains to solve complex problems. Expert system is referred as hard or precise computing, whereas FL, ANN and GA are referred as soft or approximate computing. The Software for the knowledge base or the rule base is well organized that, it has easy learning, altering and updating capabilities. Expert systems are applied in the tuning of controller parameters, fault diagnostics, and automated testing of drives. Fuzzy Logic Systems Fuzzy logic control is another class of AI technique, but its history is more recent than the evolutionary systems. A FLC is a heuristic approach that embeds the knowledge and key elements of human thinking in the design of nonlinear systems. In the fuzzy set theory, a particular object has a degree of membership in a given set that may be anywhere in the range of 0 to 1. Each fuzzy set is defined by a linguistic variable, which is again defined by a multi valued membership function. The FL controllers are based on fuzzy set theory. Therefore, the boundaries of fuzzy sets are, vague and ambiguous making themselves useful for approximation models. Artificial Neural Networks Artificial neural networks have found wide spread use in function approximation. The ANN techniques do not require a mathematical model. They give an improved performance when tuned properly and requires less tuning effort than conventional controllers. They have a simple design

6 6 procedure which includes the data from a real system even in the absence of necessary expert knowledge. ANN is used in power electronics and motor drive applications to estimate the rotor speed, flux, resistance and position of the rotor. Genetic Algorithm Genetic algorithms are a part of evolutionary computation algorithm, which use a probabilistic method of solving optimization problems and search problems. In GA based solutions, an initial population is assumed and then the optimization is reached after several generations that involve reproduction, crossover and mutation operations. Genetic algorithms are applied in many situations like solving nonlinear system transcendental equations and in the optimization of FLC. Genetic algorithms are derivative free and they are not affected by the local minima. Most of the consumer products such as washing machine, auto focus cameras and air conditioners involve a FL control technique. The application of the Evolutionary system and the GA in the field of Power Electronics and motor drives are limited when compared with FLC and ANN. 1.4 LITERATURE SURVERY This section deals with a brief review of recent developments in the field of BLDC motors, sensorless control techniques and AI techniques for the speed control of BLDC motor drives BLDC Motor Modelling and Sensorless Control Techniques Bimal Bose (2006) has discussed the entire issues in Power Electronics and motor drives, covering almost all the types of control

7 7 techniques applicable to the Induction motor, PMBLDC motor, PMSM and Switched reluctance motor drives through simulation and experimental results. Pillay and Krishnan (1988) modeled a Permanent Magnet motor drive and studied the system responses under steady state and dynamic conditions. A later work by the same authors (1989) was on the exclusive modelling of BLDC motor drives and its analysis. They (1991) have also analyzed the operating characteristics of the PM sinusoidal and PM trapezoidal motor for servo applications. Gencer et al (2006) also have modeled a three phase brushless DC motor using Matlab/ Simulink to predict the accurate system behavior. An advanced simulation model of a sensorless controlled BLDC motor drive using Matlab/Simulink was developed by Byoung et al. (2003). With the developed simulation model they have analyzed the steady state and dynamic characteristics of the motor drive. The advantage of this sensorless model is that, it is free from phase delay circuit and it is independent of the machine parameters. A similar work was also carried out by Somanathan et al (2006) for the PMBLDC motor drives. The sensorless control technique was based on the detection of the back emf zero crossing without using the motor neutral point. Sensorless operation of the BLDC motor drive has been a research topic for more than two decades. The back emf sensing and the detection of the freewheeling diode conduction are the main categories of the sensorless control techniques. Iizuka et al (1985) originally proposed a simple motional emf based scheme for the sensorless operation of trapezoidal machines. The zero crossing events of the motional emf of the open phase were used in sensing

8 8 the speed and position of the rotor. However, the back emf zero crossing events did not match with the actual position of the rotor. Therefore, the signals were phase shifted by 90 electrical degrees. Besides, it cannot be used at low speed operation and requires heavy filtering circuit that affects the dynamic characteristics of the drive. Different commutation strategies for brushless DC motors have been studied by Berendsen et al (1993). The influence of torque on the different control strategies were discussed with the simulation and experimental results. A low cost sensorless control technique for brushless DC motor drives using a frequency independent phase shifter was studied by Jung and Ha (2000). The problems of tackling the phase shifting have been discussed. Toliyat H et al (2002) discussed a position sensing technique applicable to surface mounted PM machine with insignificant saliency. They have developed a tapped machine winding to cancel the third harmonic component and the resistive voltage drop. This led to a considerable reduction in errors caused due to the change in resistance because of the heating effect. Kuang et al. (2003) proposed a modified back emf sensing scheme to sense the position of the rotor without using hall sensors. They have introduced a digital phase compensation circuit to reduce the commutation error caused due to low pass filtering and the non ideal effects of back emf waveform. Paul et al (2006) reviewed the sensorless techniques of the BLDC motor. They have discussed the merits, demerits and limitations of the various sensorless techniques available in the literature from 1974 to Shao et al (2003) presented a novel microcontroller based sensorless control technique for the BLDC motor drive. In this method, the

9 9 true back emf signals were extracted directly from each phase without using neutral point. The advantage of this control technique is that it does not require filters and is insensitive to switching noise. The proposed method was successfully tested on an automotive fuel pump. A new phase delay free method to detect the true back emf zero crossing instants for the sensorless operation of spindle motors has been reported by Jiang et al. (2004). They have used digital filters to identify the true zero crossing points. This method is independent of motor parameters and suitable for high speed operation. Cheng et al (2005) discussed the steps involved in the design and implementation of a sensorless controlled BLDC motor drive. The advantage of this method is that it is insensitive to common mode noise, and does not require either phase shifting or neutral point voltage. Jang et al (2005) adopted a bipolar starting and unipolar running method for a hard disk drive spindle motor operated at high speed with large starting torque. A novel inverter circuit was developed such that, high starting torque was achieved with the bipolar starting and high speed was maintained with the unipolar driving. A novel position sensorless control technique based on coordinate transformation was proposed by Yuanyuan et al. (2010). They have used back emf zero crossing technique for the sensorless operation of the BLDC motor drive. The rotor position errors are compensated by controlling the difference in phase angle. Abolfzal et al (2008) proposed a novel cost-effective sensorless control technique for a three phase BLDC motor drive. They have used only four switches instead of six switches to control a three phase BLDC motor

10 10 drive. The sensorless technique used in this paper was based on the detection of zero crossing point (ZCP) derived from the filtered terminal voltages. Alternate methods of position sensing schemes based on observers, inductance and flux linkage variations are also known for more than a decade. A mathematical model of the motor drive and converter were used in the process of estimating the output of the system. The estimated outputs were compared with the measured outputs to generate an error signal. This error signal is fed back to the motor model to correct the estimates. Several authors have highlighted these features in their research work. Kim et al (2008) proposed a new approach to the sensorless control of the BLDC motors to alleviate the drawbacks in the back emf based sensorless techniques during starting. They have modeled the trapezoidal back emf waveform as an input observer to estimate the line to line back emf which is used to extract the position of the rotor. This technique gives better results at low speeds than at high speed range. The sensorless control technique for the BLDC motor drive based on the zero crossing detection of back emf from the line voltage difference was presented by Damodharan et al. (2010). An alternate method of sensorless position detection technique for the BLDC motor drives was reported by Kim et al (2004). The commutation instants of the drive from near zero speed to high speed were estimated by using a speed independent position function. This method does not require an external hardware to sense the terminal voltage and it is independent of the measured back emf. Su et al. (2004) presented a low cost sensorless control technique for BLDC motor drives. They have indirectly extracted the rotor position

11 11 information from the back emf waveform. But, it requires a low pass or a band pass filters to retrieve the position information from the measured back emf. Matsui et al (1992, 1996) proposed a closed loop observer based position sensing scheme for a PM machine with significant rotor saliency. Two alternative models namely, the voltage and current models were used in the position estimation technique. The limitation of this method is that it has not addressed the problems associated with the sensorless starting and it also requires a separate starting arrangement. Ertugrul et al (1994) and French et al (1996) proposed a control strategy for estimating the position of the rotor. They have used the varying flux linkage to estimate the position of the rotor. A novel speed control technique for the PM sine wave, and square wave motors was presented by Chan et al (1995). The advantages of the speed control technique are: it can be applied directly to any brushless dc motor drive without any co-ordinate transformation and it can be used for a wide range of speed control. An alternate method of sensing the position of the rotor by monitoring the rate of change of winding current was proposed by Nakashima et al (2000). The presented technique is applicable only to the surface mounted PMSM drives. Bhim Singh et al (2003) presented a gain scheduling scheme for the PI speed controller used in the PMBLDC motor drives. The PI gains were allowed to vary within a predetermined range to eliminate the problems associated with the PI controller.

12 12 Sensorless operation of the BLDC motor drive using the back emf information has been widely used in low cost applications like hard disk drives. But, when the motor is at stand still or at low speed, it is very difficult to get the position information from the back emf. Therefore, a special starting arrangement was required to overcome this difficulty. A Sensorless rotor position estimation technique implemented to the interior PM motor from its initial states was discussed by Ha JI et al (2003). In this paper, authors have discussed the various control techniques regarding the torque, speed and position control method at stand still and at low speed for the interior permanent magnet motor drive system. In order to estimate the initial position of the rotor, a high frequency signal was injected into the estimation circuit to detect the initial position of the rotor. Tursini et al (2003) reported on the application of inductance methods to address the problem of sensorless starting. However, the sensed position was valid only for a period of 180 degree electrical, whereas, it is problematic for 360 degree electrical. Ying et al (2003) proposed novel DSP-based indirect rotor position estimation for the permanent magnet AC motors without using the rotor saliency. The problems associated with the flux integrator drift have been addressed by considering the relationship between incremental changes in flux linkage with rotor position to improve the response of speed. Cho et al (2004) presented a model based approach to sense the back emf in the surface mounted PM machine used for a direct drive washing machine. In this method, the temperature of the washing machine was estimated using the stator resistance when the rotor speed was zero, and an appropriate correction was carried out in the model.

13 13 Jang et al (2004) conducted a detailed experimental and finite element analysis to investigate the impact of injecting a high frequency signal. A new rotor position detection method for the sensorless control of the spindle motors in the hard disk drive was presented by Quang J et al (2005). In this method, the digital filters are used to identify the true and false zero crossing points of the back emf. Wook et al (2006) proposed a novel method of detecting the rotor position at standstill in the BLDC motors used in the Hard Disk Drive. The back emf based sensorless technique was used in the analysis. They have proposed a startup method to accelerate the rotor up to a certain speed till it acquires sufficient back emf. It is demonstrated experimentally that the proposed method showed a stable operation during the entire range of operation even in the presence of several mechanical disturbances. Kim et al (2007) reviewed the position sensorless control techniques of the BLDC motor and generators. The authors covered a wide range of topics related to the control of BLDC motor, their limitations, advantages and its future scope. Rodriguez and Emadi (2007) presented a novel digital control for brushless DC motor drives using conduction angle control and current mode control. The experimental demonstration was carried out using a real time interfacing system. Dong et al (2008) analyzed the characteristics of terminal voltage which is used for detecting the rotor position. A simulation model of the interior PMBLDC motor was developed to analyze the advancement in the ZCP due to the variation of inductance. They have obtained a relationship between the abnormal currents and the position errors. This relationship has been included in the simulation model to minimize the position errors.

14 14 Lee et al (2008) proposed a novel position sensorless starting method for the surface mounted permanent magnet BLDC motors used in the low cost applications. In order to have current lesser than the demagnetization current and also to limit the motor current during the starting period and the commutation was carried out based on the measured phase current. The BLDC motor drives with four switches have an advantage that there is a considerable amount of reduction in conduction losses and reduction in switches and freewheeling diode current. Hence, they are used in the low cost applications. Lee J H et al (2000) and Lee K (2003) have developed a BLDC motor drive using a four switch, three phase inverter topology. They have used position transducers to sense the commutation instants of the drive. But, the sensorless technique based on the back emf zero cross detection requires an unexcited phase to detect the zero crossing instants. To overcome this drawback, Lin et al (2008) proposed a novel FPGA based position sensorless control for a three phase BLDC motor drive using four switches. In this scheme, a novel asymmetrical PWM technique with six commutation models was used. The main drawback of this four switch topology is that it produces undesirable torque ripple. The sensorless techniques based on the back emf detection have a drawback that, either it can be used in the low speed range or in the high speed range. Yen Shen et al (2008) proposed a novel technique for the BLDC motor drive which can be used for both low duty-ratio and high duty-ratio control. An FPGA based novel digital PWM control scheme for the BLDC motor was proposed by Anand et al (2009). Bhim Singh et al (2009) studied the different sensorless control techniques for the BLDC motor drives.

15 15 An advanced double loop control strategy based on the four switch topology applied to the three phase BLDC motor was proposed by Xia et al (2009). A single neuron was used in the adaptive Proportional Integral speed control algorithm to improve the performance. A novel starting method of sensorless control technique for the BLDC motor used in the electric vehicle was presented by Mei Ying et al (2010). The principle applied in the estimation of the position signal was based on the variation of current response caused by the magnetic saturation of the stator core, when the current flows along the magnetic axis. A sensorless control technique suitable for the BLDC motor drives used in the electric propulsion of small ships and underwater vehicles was presented by Min- Fu et al. (2010). In this technique, a low pass filter with two cut off frequencies were used to regulate the speed from a low range to a high range. Yen et al (2011) presented a unified approach for detecting the back emf zero crossing points in the sensorless operation of BLDC motor. They have also investigated the influence of various PWM techniques for the estimation of position in the sensorless approach Artificial Intelligence Techniques in Electric Drives and Control In recent times, AI techniques are making a great impact in the motor drives and control industry. Computers can be made to think like humans with the help of AI which has the capability to mimic human intelligence. This capability of AI makes it suitable for many applications in the area of power electronics and motor drives applications. Lee et al (1990) carried out a study on FLC schemes. The general methodology for constructing a FL controller, assessment of its performance and its future scope are discussed in this paper. Kim et al (1994) proposed a

16 16 Fuzzy logic based pre-compensation scheme to complement the PID controllers. In their research work, an FLC was integrated with the PID controller to improve its performance. The proposed technique was also verified by implementing it on a DC servo motor drive. Homaifar et al (1995) examined the applicability of the GA to the simultaneous design of membership functions and rule sets of the FL controller. Sousa et al (1995) discussed the applications of FL controller and reviewed the FL Control theory in power electronics and drives. A discussion on the design and implementation aspects of the FL controller was presented, and the interaction of neural networks and FL control techniques are also considered in their research work. Tipuswan et al (1999) proposed an alternative method to implement the FL speed controller to a DC motor using a microcontroller. The controller has the advantage of high performance with compact size and low cost. Modelling and controlling of non linear system based on the evolutionary design procedure was presented by Kang et al (2000). They have explained the design procedure with four different numerical examples. Ahmed et al (2002) adopted an online training algorithm to the BLDC motor drive using FL and NN. The proposed controller integrates the ideas of the FLC and NN structure into an intelligent control system. An ANN-based structure was introduced for the FL control system. The hardware implementation of the microprocessor based fuzzy logic controller is carried out by Rubaai et al (2002). The results discussed indicate that it possess excellent tracking performance for both speed and position trajectories. After studying the problems in the design and tuning of fuzzy controller, Changliang (2004) proposed an auto tuning method based on

17 17 GA. The GA based fuzzy controller was applied to the speed control of the BLDC motor. Kim et al (2005) investigated the nature of torque in a sensorless controlled BLDC motor as a function of commutation delay. A commutation unbalance compensator and maximum torque controller were introduced in the drive for advancing the commutation and maximizing the torque generation. Timothy J Ross (2005) published a text book titled Fuzzy logic with engineering applications. The entire concepts of the FLC with a few industrial applications were discussed in this book. Nasir Uddin et al (2005) presented a novel speed control scheme using a GA based FL controller (GFLC) for the interior PMSM drive. This controller was designed and implemented in the real time using a digital signal processor board. The parameters of the FL controller were tuned by GA. To improve the performance of conventional sensorless control technique, Park et al (2006), proposed a robust fuzzy back emf observer for the continuous estimation of speed. The robustness of the proposed algorithm was proved through the simulation studies and is compared with other sensorless control methods. A PID type FL based on NN controller was designed by Muammer et al (2007). In this paper, the steady state and dynamic behavior of the motor drive were analyzed. Ahmed et al (2008) discussed the findings of the design and experimental verifications carried out on a hybrid FL control system used in the BLDC motor drive. The laboratory implementation of this controller was based on a linguistic fuzzy controller.

18 18 A DSP based laboratory implementation of hybrid fuzzy PID controller using the GA based optimization for the high performance motor drives was reported by Ahmed et al (2008). In this work, authors have made a real-time implementation of the GA based hybrid Fuzzy-PID controller for the industrial motor drives. The design of Fuzzy-PID controller and its integration with the conventional PID for the design of the hybrid controller have been demonstrated. The transient and steady state behavior of the BLDC motor drive using a multi input FL controller was studied by Bharathi et al (2008). Mirtalei et al (2008) presented a novel sensorless control strategy for the BLDC motor drives using the Neural Network Observer. They have replaced the FL based observers with the two neural networks to make the process easier. This method gave an accurate position information for a wide range of speed. Amit et al (2009) adopted a novel switching function for the speed control of PM synchronous motor using a hybrid (Fuzzy + PI) controller. The switching functions were calculated based on the weights of the controller output. Jiao et al (2009) discussed the various problem associated with the conventional PID controller and proposed a FL controller based on the sliding mode control for the position sensorless operation of electric vehicle. The controller developed had a facility to adjust the switching gains of the FLC to avoid the whippings. Patil et al (2009) discussed the steps involved in the design and real time implementation of integrated FL controller for the high speed PMDC motor. They have analyzed the performance of the proposed controller by testing it with various test signals such as square, triangular, sinusoidal and step signals.

19 19 Mahesh Kumar et al (2009) proposed a simple speed control technique for the BLDC motor drives using FL controller. The FL was used as an observer to estimate the speed of the drive by measuring the stator line voltage and current. The limitations of the fixed gain and its effect on the performance of the BLDC motor drive were discussed by Srinivas et al (2009). To overcome the limitations of the fixed gain, a FL based gain scheduling scheme was implemented in the Digital Signal Processor. A literature survey on the various AI techniques for the BLDC motor drives were presented by Gupta et al (2010). This literature survey acts as guidelines and quick reference for the researchers and practicing engineers who are working in the area of the PMBLDC motor drives. Varatharaju et al (2010) made a Matlab/Simulink model of the PMBLDC motor drive to study the performance of the drive for the change in load torque and inertia. The simulation model was not validated either with analytical model or with the experimental setup. The simulation model was not subjected to different load perturbation. Therefore, this model cannot be used to predict the actual performance of the drive. Malhotra et al (2010) presented a review on the technological innovations in soft computing techniques to the new level of applications. The performance of different process controllers has been analyzed and compared with the FL controller. Shun-Chung et al. (2011) proposed a modified PI like FLC with self tuning mechanism. The proposed technique was based on the redevelopment of control rule base to reduce the number of rules and membership functions, thereby reducing the complexity of the controller. Rajani and Nikhil (2011) proposed a robust self tuning scheme for FLC. The output scaling factor was adjusted online by fuzzy rules according to the

20 20 current trend of the controlled process. Authors have proved that, the complexity in the control and cost in implementing the hardware were considerably reduced. Varatharaju et al (2011) overcame the drawbacks of the FLC and ANN based controller by using an Adaptive Neuro-Fuzzy Inference System (ANFIS). A Comparison between FL controller and Adaptive FLC in the speed control of the BLDC motor was presented by Alizadeh (2011). Nagaraj (2011) discussed the different methods of tuning a PID controller. The methodology, efficiency and the performance of the AI speed controller are compared with the traditional PID controllers. Subbarami Reddy et al (2011) compared the performance of the PID and the Hybrid controller by considering the various performance measures such as the rise time, settling time, steady state error, peak overshoot, the Integral of the Absolute value of the Error and the Integral of the Time-weighted Squared Error. Siong et al (2011) developed a fuzzy logic model to carry out the simulation and experimental studies on a PMBLDC motor drives. The dynamic characteristics of the drive such as, speed, torque, current and voltage of the inverter components are observed and analyzed using the developed MATLAB model. Nikhil et al (2011) presented the steps involved in the implementation of DSP based FLC scheme for a BLDC motor. A three phase four switch topology was used in the analysis. Kandiban et al (2012) proposed an adaptive Fuzzy-PID controller for the speed control of BLDC motor dive. The performance comparison of the PID controller, Fuzzy Logic-PID controller and an adaptive Fuzzy PID controller were discussed in their research work. The difficulty in tuning the

21 21 parameters of the PID controller was overcome with the adaptive fuzzy logic controller. The exhaustive literature review conducted on the BLDC motor drive has revealed that, the research work carried out on the BLDC motor drive shows the influence of the technical developments in Power electronics, microelectronics and control system on the application of AI. Most of the developments in these fields were in the direction to reduce the hardware by introducing high-speed digital signal processors and to reduce the sensing requirements for the drive system. In the literature, there are several simulation models available for BLDC motor drives. These models employ state-space equations, Fourier series or d-q axis model. Even though these models have made a great contribution in the BLDC motor drives, there is no comprehensive model for the analysis of motor used in sensorless operation. The motor models available in the literature discuss the mathematical analysis of the BLDC motor and a computer simulation of such models. The verification and validation of such models were not discussed in detail. Moreover, the assumptions made during the mathematical modelling of the motor control system limits the feasibility of such models in the prototype development. In the model based design process, these affect the performance of the motor in terms of accuracy, stability and robustness. Hence, there is a need for a simplified and real time model which can be validated, corrected and can be extended directly to the prototype development. This thesis, presents a review of the most commonly used sensorless control techniques with a comparison to determine which one is the best candidate for the industrial applications. The literature survey also revealed that, the sensorless method based on back emf estimation has been

22 22 found to be more promising. Hence, the back emf based sensorless technique is further investigated in detail with the AI controller. Literature survey also reveals that even though AI controllers were analyzed and investigated over a decade, the successful implementation of this controller to the sensorless controlled BLDC motor drives with trapezoidal back emf is very limited. Even though neural network and FL controllers were used as observers for sensorless position estimators, the real time/experimental validation of these techniques is very limited. In this thesis, a study on the performance enhancement of BLDC motor drives operated in a sensorless mode is considered. To show the suitability of the proposed technique on a sensorless environment, Fuzzy logic controller and Genetic Algorithm were analyzed on a BLDC motor without sensors. 1.5 REAL TIME VALIDATION AND DSPACE The recent development in the software technology has made the design and development process simple, cost effective and less time consuming. The new technology has reduced the complexity in the verification, validation and testing process, making the entire process simpler. When changes occur in the system, instead of redesigning the whole system, corrections can be made then and there, therefore reducing the time taken for computation. Even though there are several simulation models in the literature, the direct extension of such models in the prototype development is very limited. The dynamics assumed during the modelling will not be the same as observed in the real time system.

23 23 A simulation model in a Matlab/Simulink will automatically generate the source code with the help of Real Time Workspace (RTW). The generated code is tested and compared with the behavior of the mathematical model and it is validated with a real time simulator. The generated source code is uploaded to the appropriate hardware in loop simulator with the necessary peripherals. The generated code is linked with the real system through Real Time Interface (RTI) and Real Time Embedded coder. This embedded coder again generates a source code which is optimized to mimic the real system Real Time Interface (RTI) The Real Time Interface is the link between DSPACE real time hardware and the Matlab/Simulink development software. The RTI automatically runs through all the steps necessary to build the application such as generating C code and invoking the compiler. It also extends the C code generated from the real time workspace to the simulation models which can be implemented easily on the DSPACE real time environment DSPACE DSPACE and Matlab RTW have created a rapid control prototype environment which has facilitated the design engineers to focus only on the design rather than on programming. DSPACE approach has the advantage of directly extending the design skills of engineers from Matlab/ Simulink environment to the hardware implementation. It is a simple user interface, which requires only a basic knowledge of Matlab/Simulink for designing the controller while the experiment is running. The parameters of the controller and the reference signal can be changed easily with the help of DSPACE.

24 24 The design and implementation phases are carried out using Matlab/Simulink and the commercial real time hardware DSPACE DS2202/DS1104/DS2205 and other series of DSP board. The controller block uses Matlab RTW routine that automatically generates the C code from the Simulink block. The generated code is used by the DSPACE DSP board for the real time control. The Simulink and DSPACE together help to validate the model on a laboratory platform, at a click of mouse button. 1.6 OBJECTIVE OF THE THESIS This research work deals with various aspects of applying AI techniques to enhance the performance of the BLDC motor drive. In the presented work, AI technique is used in the speed control of the BLDC motor drive operated in the sensorless mode. The specific objectives are as follows To develop a simplified simulation model of the BLDC motor drive with hall sensors and to validate the model with real time/experimental setup, and extend the simplified simulation model for the sensorless control of the BLDC motor drive. To optimally tune the parameters of PI controller using GA To design, develop and investigate the influence of fuzzy logic and the genetic algorithm on the BLDC motor drive. To design an improved GA based optimal tuning of the parameters of FLC and to compare the performance with a modified self tuned FLC.

25 25 The scope of the presented work is given in Figure.1.2. Performance enhancement of BLDC motor drives using AI techniques Simplified modeling and sensorless speed control of BLDC motor drive Design of AI controller for the BLDC motor drive FL based speed controller GA based speed controller Simple FL controller Modified STFL controller GA tuned PI controller Improved GA based FLC Compare the speed response of the BLDC motor drive with AI controllers Identify the suitable controller for the enhancement Figure 1.2 of the Scope BLDC of motor the present drive work Figure 1.2 Scope of the presented work

26 ORGANIZATION OF THE THESIS The work reported in the thesis is organized in to seven chapters, Chapter one presents the problem under investigation, the objective of the present work, scope, organization of the thesis, a brief review of literature on the modelling of the BLDC motor drive, sensorless control techniques of the BLDC motor drive, and AI control techniques for electrical motor drives. The description of a simplified simulation model of the BLDC motor drive with hall sensors is presented in Chapter two. The simulation model consists of several sub blocks like the hall sensor block, the voltage and current equations block, which also includes the inverter and switching logic control. The responses obtained from the BLDC motor drive during different operating conditions were compared with the ideal waveforms and also with the waveforms available in the literature. The results of the simplified simulation model were compared with those obtained from the developed hardware setup. The simulation and hardware results matched closely, and hence, this model was further extended to the sensorless control technique and the design of the AI controller. The sensorless control, based on the back emf zero cross detection is a superior and widely used technique by the motor drives industry. Since this work is focused only on the study of AI controller techniques for the BLDC motor drives, the sensorless control scheme based on the back emf zero crossing was used to study the proposed controller. The simplified simulation model developed for the BLDC motor drive with hall sensors has been extended, to obtain a sensorless speed control technique in Chapter three. The developed sensorless model of the BLDC

27 27 motor drive with the PI speed controller was compared with the responses available in the literature and also with the experimental results. A detailed analysis carried out on the sensorless controlled BLDC motor under different operating conditions showed, that the sensorless model can be further used in the study of the proposed AI controller for BLDC motor drives. The GA based tuning of the parameters of the PI controller is discussed in Chapter four. The steps involved in the optimal tuning of the parameters of the PI controller, the problem formulation and the results obtained, are also included in this chapter. A simple Fuzzy logic controller (SFLC) was designed to study the direct impact of the fuzzy logic in the BLDC motor drive. The results obtained from the SFLC clearly reveal that it requires an optimal tuning to improve the performance of the drive. A modified self tuned FLC (MSTFLC) was proposed in Chapter five to improve the performance of the drive. The MSTFLC gave a good improvement in the performance of the drive when compared with the SFLC, but the initial gains of the controller were obtained by trial and error method. Hence, an Improved GA based optimal tuning of the parameters of the FLC (IGAFLC) was proposed in Chapter six, to further improve the performance of the FL controller. The results obtained from the proposed GA based optimal tuning of parameters of the FL controller shows an improvement in the performance of the BLDC motor drive. Chapter seven gives a summary of the conclusions obtained from the research work, and suggestions for future work. The references are listed at the end.