Research On Fault Diagnosis Technology Of Industrial Robot Based On Extreme Learning Machine

With the rapid development of computer science and technology,the application fields of industrial robots are also expanding and deepening,and industrial robots have become a high-tech industry.However,the field of fault diagnosis and monitoring of industrial robots is not so perfect.In this paper,the fault diagnosis of the transmission system of RV reducer of six-axis industrial robot is studied from the perspective of data-driven with the help of low-precision attitude sensor.In this study,the attitude sensor is installed at the end of six axes,and the fault type is set for data acquisition.Finally,an on-line monitoring system is built by using the optimal ultimate learning machine model to observe the running state of industrial robots in real time and remotely.In terms of fault diagnosis,the extreme learning machine(ELM)has a high operating efficiency.However,the recognition accuracy of ELM algorithm is actually affected by the activation function.In addition,most of the test data came from high-precision and expensive sensors.In this paper,a low-cost attitude sensor is used to collect the original data,which is installed at the end of the six-axis industrial robot.A doublet activation function is proposed to improve the performance of ELM,named doublet ELM(DELM).The method is evaluated with the experimental data collected from the robot,and its superiority is verified by comparing with other activation functions such as sin,sigmoid,tribas,radbas and hardlim.The experimental results show that the method can be in low accuracy of fault diagnosis precision is higher than other activation function,and can fast convergence,in the face of different fault category is at the same time,improve the curve of the activation function almost no too big fluctuations,sin performance is superior to the sigmoid activation function in the fault diagnosis,and tribas,radbas,hardlim,also explains the sin activation function in the industrial robot the superiority of the fault diagnosis.According to the basic theoretical research of ELM,it is found that Extreme learning machine(ELM)has a good performance in learning speed than traditional gradient descent algorithms.However,the random inputs weights and hidden biases are influential factors for the accuracy and generalization performance of ELM.Therefore,this paper proposes an extreme learning machine based on level-based learning swarm optimizer(LLSO-ELM)is proposed to diagnose faults of industrial robots.LLSO has strong competitiveness in solving quality and computational efficiency of large-scale optimization problems,and can be used to solve the optimal configuration of ELM weight and deviation.The proposed model is experimented by using the attitude data of industrial robots working in different fault modes.Experimental results show that this method has good generalization performance and stability.Finally,the python development platform was built,and the sensor data communication was realized by using LLSO-ELM algorithm and combining Modbus communication protocol and pyserial serial port communication module.Py Qt5 was used to build the human-computer interaction interface,Threading module was used to complete multi-threading tasks,and the pickle module was used to transform the algorithm into a local module,finally achieving online fault detection and diagnosis.