Artificial Intelligence-Based Optimal PID Controller Design for BLDC Motor with Phase Advance

Received Jun 1, 2019 Revised Sep 9, 2019 Accepted Nov 27, 2019 This paper proposes the artificial intelligence (AI)-based optimal PID controller design optimization of brushless direct current (BLDC) motor speed control with phase advance approach. The proposed control system allows the speed adjustment of the BLDC motor by phase advance technique. In this paper, two selected AI algorithms, i.e., the adaptive tabu search (ATS) and the intensified current search (ICS) are conducted as the optimizer for the PID controller design. The proposed control system is simulated by MATLAB/SIMULINK. Results obtained by the ATS and ICS will be compared with those obtained by the Ziegler-Nichols (ZN) tuning rule and the genetic algorithm (GA). It shows that the speed response of the BLDC motor by phase advance with the PID controller optimized by the ICS outperforms better than the ZN, GA and ATS.


INTRODUCTION
The brushless direct current (BLDC) motor is a kind of synchronous motor having permanent magnets on the rotor and stator winding. It has been broadly utilized in automotive, hard disk, robotics, vehicle, aerospace, home appliance and instrumentation. The BLDC motor has many advantages over the induction motor including simple structure, low size, excellent efficiency, less maintenance, low noise and wide speed range [1] - [3]. In the future, motor controllers need to be improved for application targets through simple and economical design.
The BLDC motor can be expanded to operation speed range more than base speed by phase advance [4]. This allows the utilization of BLDC motor in various applications for high speed range. So, it is necessary to have a controller that can improve performance for developing three phases of the BLDC motor drive. The PID controller is extensively used in the BLDC motor drive because of the effectiveness and simple structure [5], [6]. The parameters of PID controller are very essential for the performance of motor control system, particularly in motor device with nonlinearity and large inert [7]. From the previous researches [4], [8]- [11], the BLDC motor system with phase advance is regulated by conventional control methods. It is found that the performance of motor has steady state error and slow speed response. Therefore, the speed response of the BLDC motor system with phase advance can be done effectively by optimizing the values of PID controller parameter using artificial intelligence (AI) algorithms. Furthermore, the applications of AI algorithms for PID controller of the BLDC motor system with phase advance have not been presented in the previous researches [4], [8]- [11]. The AI algorithms have been extensively accepted for the controller design in many industrial applications [12]. For example, designing of BLDC rotor speed by fuzzy logic [13], genetic algorithm (GA) for controller design [14], self-turning the PID controller using GA [15], controller design using adaptive tabu search (ATS) [16], aircraft electric control by ATS [17] and control synthesis using current search (CS) [18],  [19]. Hence, this paper presents the use of AI algorithms for tuning the PID parameters (KP, Ki and Kd) for controlling the BLDC motor with phase advance. In this paper, there are divided six sections. Firstly, the introduction is described. Secondary, the details of the modeling of BLDC motor and phase advance technique are presented. Thirdly, the details of AI algorithms are demonstrated. Fourthly, the application of selected AI algorithms is tuned the PID parameters for the BLDC motor system. Fifthly, the simulation results are illustrated. Finally, the conclusions are described.

MODELING OF BLDC MOTOR AND PHASE ADVANCE TECHNIQUE 2.1. Modeling of BLDC motor
The equivalent circuit model of the BLDC motor is illustrated as shown in Figure 1. The modeling of BLDC motor can be performed in equation (1) [1], [2].

Phase advance
The phase advance technique is a common control of the phase current to lead the phase back emf. It can increase the speed of BLDC motor over the base speed. Figure 2 shows the phase currents shifted by the angle of adv to phase back emf. The phase currents can be varied from 0 o to 60 o , electrical degree [8]- [10]. The phase advance is described in equation (5), where Ls is an indicative of phase inductance, adv is the angle of phase advance, Vdc is an indicative of voltage source, n is rotor speed and Iph denotes the phase current.
In Figure 3, it shows the three phase inverter for BLDC motor drive. In this work, the proposed drive system composes of three phase inverter, switch control signals, Hall effect sensors, BLDC motor and DC voltage source. The switching states of three phase inverter based on Hall effect signals are visualized in Figure  4.   Table 1 shows the switching states for the BLDC motor drive. Each step has an interval of 60 o ,electrical degree. Table 1. Switching states

ARTIFICIAL INTELLIGENCE
Recently, AI algorithms have been broadly utilized to solve combinatorial and engineering problems. In this section, the AI algorithms consisting of the intensified current search (ICS) and the ATS are briefly reviewed.

ICS algorithm
The CS is one of the AI algorithms based on the current divider of the electric circuits and network systems [18], [19]. Theoretically, the CS only composed the memory list (ML). For solving the computational problems, the ICS algorithm has been modified version by adding the adaptive radius (AR) and the adaptive neighborhood (AN). In order to find the optimal solution, the ICS is able not only to escape the local entrapment but also to speed up the search process in the search space. The movement of the ICS over search space is illustrated in Figure 6. The pseudo code of the ICS is described in Figure 7 [20], [21]. Figure 6. Movement of the ICS algorithm [20], [21] Step

ATS algorithm
The ATS was firstly launched in 2002 [22], [23] as one of the AI algorithms. It has been modified version of the TS [24], [25] and based on an interactive neighborhood search approach for solving combinatorial and nonlinear problems. The mechanisms of ATS have the adaptive radius (AR) and back tracking (BT). To speed up the search process of ATS, AR mechanism is utilized to intensity. Moreover, the BT mechanism is able to conduct to escape the local entrapments. The movement of ATS over search space can be illustrated in Figure 8. The pseudo code of the ATS is shown in Figure 9 [22], [23].

PID CONTROLLER OF BLDC MOTOR
The PID controller was firstly conducted to industrial applications in 1939 [26]. It has been extensively used in the industrial control systems because of its simple structure and effectiveness. Figure The PID controller model in the s-domain transfer function is stated in equation (7). Therefore, the PID controller can be written in the form of closed loop as shown in equation (8).
) ( ) The AI search techniques are applied to the PID controller design for the BLDC motor system as depicted in Figure 11. The objective function (fobj), sum absolute-error (SAE) between R(s) and C(s) as stated in equation (9) will be fed back to the AI tuning block (ATS or ICS). The fobj will be minimized by ATS or ICS to find the optimal PID parameters (Kp, Ki and Kd) to satisfy the inequality constrainted functions as stated in equation (10), where ess is steady-state error, ess_max is maximum allowance of ess, tr is rise time, tr_max is maximum allowance of tr, Mp is maximum percent overshoot, Mp_max is maximum allowance of Mp, ts is settling time, ts_max is maximum allowance of ts, Kp_min and Kp_max are the search space of the Kp, Ki_min and Ki_max are the search space of the Ki and Kd_min and Kd_max are the search space of the Kd. Flowcharts of ATS and ICS algorithms for the PID controller design optimization for the BLDC motor system are shown in Figure 12 and Figure 13, respectively.
Subject to:  (6) can be written in equation (11 Referring to Figure 11, the AI algorithms were simulated via MATLAB/ SIMULINK run on CPU Core i5, 2.5 GHz and 6 Gbytes DDR-RAM computer. The optimal gains using AI techniques are setup. The design optimization by the ATS and ICS are performed as shown in Table 2 and Table 3, respectively. In addition, results obtained from the ATS and ICS will be compared with Ziegler-Nichols (ZN) tuning rule [28], [29] and GA [30]. The ZN tuning rule is the analytical design method, while the GA is one of the most popular metahueristic optimization techniques. However, both ZN and GA are not new anymore. Then, details of ZN and GA are omitted in this paper. Readers can find their details from the given references. The parameters of the GA are shown in Table 4. For design the PID controller, the search spaces of PID parameters in equation (10) are specified in Table 5.    The convergent rates of the proposed objective function in equation (9) and its corresponding inequality constrained functions in equation (10) proceeded by GA, ATS and ICS are depicted in Figure 14. The PID parameters of the BLDC motor system obtained by ZN tuning rule and optimized by GA, ATS and ICS are expressed in (12), (13), (14) and (15), respectively. The system responses of the BLDC motor without controller and with PID controllers are plotted in Figure 15 and summarized in Table 6.   From to Figure 15 and Table 6, the performance with the PID controller optimized by the GA gives the slowest response. The system response with the PID controller obtained by the ZN tuning rule provides faster response but greater overshoot than that by the GA. The system response with the PID controller optimized by the ATS performs faster response and smaller overshoot than that by the GA and ZN tuning rule. However, the system response with the PID controller optimized by the ICS outperforms that by ZN tuning rule, GA and ATS with the fastest response and smallest overshoot according to inequality constrained functions in equation (10).
In addition, the BLDC motor speed control system with phase advance technique and the PID controllers obtained by the ZN tuning rule and optimized by the GA, ATS and ICS is simulated by MATLAB/SIMULINK as depicted in Figure 16. The BLDC motor in the simulink block diagram is a three phase motor rated 3,000 rpm. The effectiveness of the obtained PID controllers is analyzed under different operating conditions and compared with the conventional control of phase advance. Figure 17(a) depicts the responses of BLDC motor when speed changed from 0 to 2,800 rpm with no load condition without phase advance. From the Figure 17(a), the responses of speed control are very satisfactory. Figure 17(b) shows the relation between the phase current (ia) and the back emf (ea). It was found that ia and ea are in phase.  Figure 17. Responses of BLDC motor speed control system without phase advance (no load) Figure 18(a) depicts the responses of BLDC motor speed control system when speed changed from 2,800 to 3,100 rpm with no load condition with phase advance. From the figure18(a), the BLDC motor system can extend the speed range over the 3,000 rpm. The speed response with the PID controller optimized by the ICS outperforms other controllers. Figure 18(b) depicts the relation between ia and ea, it was found that ia leads ea at 0.5 s. (a) Step change from 2,800 to 3,100 rpm (b) Comparison between ia and ea Figure 18. Responses of BLDC motor speed control system with phase advance (no load) Figure 19 depicts the responses of BLDC motor speed control system when speed changed from 3,100 to 2,800 rpm at 0.5 s with no load in Figure 19(a) and with load of 2 N.m in Figure 19(b). From Figure 19, the speed response with the PID controller optimized by the ICS also provides better response than other controllers with the least settling time of both no load and load.  Figure 20(b) depicts the speed responses of BLDC motor speed control system with phase advance when speed changed from 3,400 to 3,100 rpm. From Figure 20, the speed response with the PID controller optimized by the ICS can yield better response than other controllers with the least settling time. Figure 21 depicts the regulating responses of BLDC motor speed control system with phase advance at constant speed of 3,100 rpm with load of 2 N.m at 0.4 s and no load at 0.7 s. From Figure 21, it was found that the speed responses once load applied and load released can be successfully controlled by the optimal PID controllers for load regulation purpose. For overall system performance, the PID controller optimized by the ICS for BLDC motor speed control system with phase advance outperforms other controllers. Figure 21. Responses of load regulation of BLDC speed control system

CONCLUSION
The AI-based optimal PID controller design optimization for BLDC motor speed control with phase advance has been presented in this paper. Two selected AI algorithms, i.e., the ATS and ICS, have been applied as the optimized to the PID controller design. In this paper, the mathematical model of the BLDC motor and the phase advance technique have been provided. Algorithms of the ATS and ICS have been clearly reviewed. The proposed design approach has been conducted by using MATLAB/SIMULINK. The PID controllers optimized by the ATS and ICS have been compared with the ZN tuning rule and the GA. As results, the ATS and ICS could successfully give optimal PID controllers. The simulation results shown that the responses of the BLDC motor speed control by phase advance with the PID controller designed by the ICS outperform that obtained by the ZN, GA and ATS, respectively. Moreover, the proposed ICS algorithm for PID controller design can be improved the system response effectively.