Tuning of Servo Drive Controller Based on Boosted Tree Model and Particle Swarm Optimization

This paper presents a gain tuning method for servo drives that combines machine learning model (LightGBM) and optimization algorithm (PSO). The LightGBM model can predict the response characteristics of the servo system under different gain parameters of the servo drive. Then, the PSO will tuning the gain parameters to obtain the best performance. A new performance criterion for evaluating the servo drive is proposed. It consists of several time-domain position response characteristics, such as percentage of overshoot and settling time etc., and the servo drive performance can be optimized by utilizing those characteristics. The experimental result shows that the appropriate gain parameters can be obtained through this proposed AI tuning method.

I. Introduction

Automation equipment is usually composed of a servo motor connected to various mechanisms. The performance of the equipment is therefore closely related to the response of the servo motor. Servo motors are controlled by servo drives, which contain many adjustable gain parameters, which will significantly affect the servo response. In order to achieve the optimal equipment performance, users must tune the gain parameters appropriately. Previous researches related to gain tuning can be divided into two types, one is the model-based method [1]–[3], which need to establish a linear approximation model of the equipment, and calculate the suitable gain parameters by numerical methods, like GPM, IMC, optimization method. The other is the model-free method [4]–[7], which need to input test signals to the equipment and then directly adjusting the parameters based on the behavior of system response, like Ziegler-Nichols method.