Calibration of A 6-DOF IR Based on PSO-BP Neural Network and FEA

Abstract

This paper presents a method for calibrating the position and posture of IR (industrial robots) in industrial sites based on PSO-BP neural network and FEA. Besides the neural network is used to identify the nonlinear relationship between the input and output of the industrial robot, the neural network model of the industrial robot is established. It is concluded that the HSR-JR612A industrial robot is calibrated by the neural network model, and its position mean square error is reduced from 2.15mm to 0.23mm, the mean square error of the industrial robot's posture drops from 0.985° to 0.163°. Consequently, the error of the HSR-JR612A industrial robot is effectively compensated, and the performance index of the industrial robot can be greatly improved by the proposed method.