Fault Diagnosis Of Industrial Robot Joint Based On Digital Twin

Industrial robots are the core equipment in the industrial field,which can effectively improve production efficiency.As industrial level improves by leaps and bounds,industrial robots are moving towards high precision,high reliability and high automation,and their structure is more precise and more complex.Ensuring the reliability,accuracy and safety in the operation process is the key to adapt industrial robots to the high automation requirements of intelligent manufacturing,and also an significance guarantee for the good implementation of intelligent manufacturing.State monitoring and fault diagnosis during the operation of industrial robots can effectively ensure the stability and reliability of equipment operation,real-time monitoring of industrial robot equipment operation,and prevent the impact of motion deviation caused by faults on the production process.Performing real-time fault diagnosis is an important means to improve the fault-free operation time of industrial robots and to improve the accuracy of end position and end attitude during industrial robot motion.At present,there are mainly data-driven methods,model methods,and expert knowledge methods,all of which have limitations in their single methods,such as low efficiency of data utilization in data-driven methods,poor simulation of model construction in model methods,and low timeliness of traditional knowledge methods.Based on digital twin technology and neural network algorithm,real-time virtual-real mapping of devices,real-time training output of motion data,and multi-dimensional and multi-level virtual-real fusion provide ideas to solve the above problems.Based on digital twin theory,the key problems of industrial robot condition monitoring and fault diagnosis are studied as follows:(1)A digital twin five-dimensional model for industrial robot is proposed for industrial robot joint fault diagnosis,and its overall framework and internal components are designed in detail,including entity,digital twin,data communication,information space and knowledge service.(2)The construction method of the digital twin model and platform for industrial robots is proposed.A Unity3D-based digital twin model and platform construction method for industrial robots is proposed to determine the optimal digital twin model construction software scheme.The kinematic modeling of the UR10 is completed,and the process of building the UR10 robot digital twin model and its digital twin platform is designed.(3)A neural network-driven network model for industrial robot joint fault diagnosis is established.The process of data training using neural network in joint fault diagnosis is given through the multiple mapping model of industrial robot motion reliability,and the end pose fault data set caused by the joint angle error is established.The optimal sampling frequency of end position for joint fault diagnosis of industrial robots is given,and the results show that when the sampling frequency reaches 250 Hz,the comprehensive accuracy of fault diagnosis is 99.17% for 0.5° joint angle error and 96.52% for 0.1° joint angle error.By comparing the data-driven method with the inverse solution method,it is demonstrated that the data processing algorithm using neural network as fault diagnosis has higher accuracy and adaptability.(4)Validation of industrial robot fault diagnosis scheme.A digital twin test platform for industrial robot joint fault diagnosis are build,realize real-time virtual mapping between the digital twin platform and the industrial robot entity,conduct offline virtual machine simulation test and online robot entity test,verify the dynamic interaction performance of the digital twin platform,conduct joint fault diagnosis test data collection and analysis,and verify the correctness of the proposed digital twin-based industrial robot joint fault diagnosis scheme is verified to be correct.