การพัฒนากระบวนการประมาณค่าแรงบิดสำหรับมอเตอร์ไร้แปรงถ่านด้วยเทคนิคไร้เซนเซอร์

Abstract

In this research, we propose a torque estimation without using a torque sensor for applying to robot operation to reduce production costs and the complexity of robot design. The objective of the research is to compare the efficiency of the torque estimation with machine learning, which focuses on real-time operation. This research has designed and built hardware devices used for experiments based on the operation of the robot arm. The operation of the robot arm uses a brushless de motor to operate. This type of motor is a 3-phase motor that can take back electromotive force to analyse to find information which will lead to the torque estimation. Controlling a brushless dc motor using a vector control method, also known as Field orientation control, is a 3-phase motor control that makes the most of torque performance. The experiments in this research have chosen to use variables that are a part of motor control method because that variable has a direct change to the torque of the motor. The regression model, Support vector machine and Neural network can be used for torque estimation. The experiment witl include both offline and online experiments, including application to the robot arm to, see the performance. In an online experiment, the Neural network model provides the best solution when compared to a torque sensor. However, it takes much more estimation time than the Regression model.