Comparison of SI-ANN and Extended Kalman Filter-Based Sensorless Speed Controls of a DC Motor
- PDF / 2,933,972 Bytes
- 16 Pages / 595.276 x 790.866 pts Page_size
- 11 Downloads / 302 Views
RESEARCH ARTICLE-ELECTRICAL ENGINEERING
Comparison of SI-ANN and Extended Kalman Filter-Based Sensorless Speed Controls of a DC Motor Ahmet Gundogdu1
· Resat Celikel1
· Omur Aydogmus2
Received: 28 April 2020 / Accepted: 4 October 2020 © King Fahd University of Petroleum & Minerals 2020
Abstract In this study, a speed sensorless algorithm was developed to control a single-link manipulator connected to DC motor. The armature voltage value can be obtained by using duty cycle information generated by the controller without using any sensor. Thus, the proposed system does not require any additional measurement sensor. This paper presents a study for artificial neural network (ANN)-based speed estimation algorithm which has a closed-loop speed control with the first and second inputs generated via support of system identification (SI). The second SI input was obtained as a simple transfer function with discrete time. The performances of the SI-input ANN structure and the conventional extended Kalman filter (EKF) method were compared in the MATLAB/Simulink environment. It was observed that the proposed method revealed better results than the EKF method in the steady and transient states. Thus, it was shown that high-performance sensorless speed control could be performed with SI-ANN structure in applications. Keywords Artificial neural network · System identification · Speed estimation · Sensorless control · Kalman filter
1 Introduction Electric motors are used intensively in hydraulic, pneumatic, robotic and many other industrial applications. The electric motors require speed/position control to perform a process in many of applications. Therefore, speed and/or position control of motors has an important role in industrial systems. Speed/position controls of mechanical systems can be performed at much higher performance by using feedback control systems. Some sensors such as encoder, resolver and taco generator are connected to the motor shaft in order to measure of the motor speed/position. However, these sensors increase the system cost as well as the volume of the motor [1]. In order to reduce system costs and minimize
B
Ahmet Gundogdu [email protected] Resat Celikel [email protected] Omur Aydogmus [email protected]
1
Department of Electrical and Electronic Engineering, Batman University, Batman, Turkey
2
Department of Mechatronics Engineering, Firat University, Elazig, Turkey
faults caused by sensor failures, sensorless control methods have been developed. Sensorless control methods estimate the speed of the motor by using the current and voltage information of the motor. In the literature, many methods such as sliding mode observer (SMO), Kalman filter (KF), EKF, model references adaptive system (MRAS), state observer (SO), deep learningbased approach and artificial neural network (ANN) have been used to perform sensorless speed control of electric motors [2–6]. Sensorless control negatively affects the performance by changing motor parameters due to some effects such as temperature, electr
Data Loading...
