Research On Modeling Of Robot Arm System Based On Signal Feature Analysis

The space manipulator plays a vital role in spacecraft.With the remarkable development of my country’s aerospace industry,the requirements for its stability and accuracy are becoming more and more stringent.Due to the particularity of the working environment,the robotic arm will be affected by different nonlinear factors in the process of performing tasks,resulting in vibration of the robotic arm and poor movement performance.Therefore,establishing an accurate dynamic model and analyzing the influence of nonlinear factors on the motion performance of the manipulator is an important prerequisite for improving the control accuracy.This thesis mainly takes the flexible space manipulator as the research object,establishes an accurate dynamic model,and analyzes the vibration signal problem generated by different nonlinear factors and the parameter identification of the friction force in the manipulator.Firstly,this thesis adopts the Lagrange method to establish and derive the dynamic models of the space manipulator with flexible joints and flexible links respectively.The influence of nonlinear factors such as joint flexibility,friction,time delay,and link flexibility on the dynamic characteristics of the manipulator is analyzed.Secondly,the vibration signals caused by different nonlinear factors in the manipulator are studied.The state signal obtained based on the dynamic model is nonlinear and non-stationary.Therefore,the methods of Empirical Mode Decomposition(EMD)combined with Sample Entropy(SE)are used for feature extraction in the time-frequency domain.The data relationship between different nonlinear factors and corresponding entropy values is obtained,and then the different nonlinear factors existing in the manipulator were classified by LSO-KELM.The experimental results are compared with SSA-KELM,GA-KELM,PSO-KELM and KELM,and the obtained results show that the average accuracy of the LSO-KELM algorithm reaches 98.85%,and the effect is best.Finally,the problem of friction parameter identification in the manipulator is further analyzed.A parameter identification method of friction force in manipulator based on EMD-SE-CNN-GRU is proposed.The Coulomb-viscous friction force is considered in the manipulator,and the relationship between the friction force and the end state signal is analyzed.Obtain the corresponding data through experiments.The data is fed into a network model built using a combination of Convolutional Neural Network(CNN)and Gated Recurrent Unit(GRU)for identification.The experimental results are compared with CNN,BiLSTM,GRU and other networks.From the experimental results,it can be seen that the identification accuracy of CNN-GRU for friction force has reached more than 99.9%,which has a good effect on the problem of identifying multiple parameters at the same time.