Research On Joint Motion State Prediction Of Four-axis Industrial Robot Driven By Current Data

With the progress of modern technology,industrial robots are widely used in the field of industrial automation production.Robot joints are the basic structure that controls the motion of the machine body,and the joint motion state directly affects the working performance of the robot.Therefore,research on predicting the joint motion state of industrial robots is of great practical significance.This article conducts research on the prediction of joint motion states of SCARA four axis industrial robots,aiming to achieve accurate prediction of speed and force based on motor current data driven joint motion states.When the joint torque information of the robot is difficult to obtain,the joint motor current can be used to reflect the joint torque.This article simulates the scene of collision or obstruction during joint movement by applying external resistance to different joint positions under different joint motion speeds,changing the joint load torque state,and achieving the collection of joint motor current data under different motion states.Finally,a Bayesian optimized deep learning neural network prediction model is used to analyze the collected joint motor current data under different motion states,which can accurately distinguish the motion state of the joint.The main research content of this article is as follows.(1)We studied the structure of the industrial robot joint system,the mathematical model of permanent magnet synchronous motor,and vector control technology.Based on the parameters of the SCARA four axis industrial robot joint motor,we conducted simulation verification analysis using MATLAB and found that there is a positive correlation between stator current and load torque.(2)We have designed an industrial robot joint motor current acquisition device and completed the hardware and software design of the experimental platform.Based on the PLC design system as the core component,combined with the use of Hall current sensors and computers,the non-contact current collection task of the joint motor of the four axis industrial robot was completed,and a current dataset for predicting the motion status of the SCARA four axis robot was generated.(3)Based on the theory of deep learning,predictive models for motor current of industrial robot joints and classification models for robot joint motion states were designed based on RNN,LSTM,and Bi LSTM,respectively.Bayesian optimization models were used to optimize the RNN,LSTM,and Bi LSTM models.Model validation was conducted on the constructed dataset,and experimental results showed that compared with the manually adjusted model,The performance of the Bayesian optimized model has been further improved,with the BO-LSTM model showing the best prediction performance in joint motion state classification prediction.