Research On Condition Monitoring And Performance Prediction Method Of Harmonic Reducer

Harmonic reducer has the advantages of small size,high transmission efficiency,light weight and strong carrying capacity,so it is increasingly widely used in the mechanical field.In industrial robots,as an important transmission component,the harmonic reducer mainly plays the role of deceleration and torque increase,which determines the positioning accuracy,carrying capacity and service life of the robot end.In the long-term and highintensity work,the performance of the harmonic reducer gradually declines and the reliability decreases,which will directly affect the overall safety and production efficiency of the robot.A typical maintenance plan is to replace a reducer that has been running for a set number of hours.However,due to complex conditions such as working conditions,the reducer may suddenly fail without proper condition monitoring methods.These may lead to a high level of risk and affect the overall service life of the robot.In order to reduce the risk caused by the sudden failure of the harmonic reducer and optimize the use of the reducer,it is necessary to conduct comprehensive state monitoring and performance prediction of the robot harmonic reducer.The main research contents are as follows:(1)In terms of physical failure mechanism,the different types of failure modes of harmonic reducers are listed,and the main failure modes of industrial robot harmonic reducers are found: flexible wheel fatigue fracture,flexible bearing failure,and flexible wheel fatigue fracture is the most common Form of failure.Finally,based on the analysis of the failure physical model,the degradation functions of the failure modes of the flexspline and the flexible bearing are established respectively.(2)In terms of the state monitoring of the harmonic reducer,as the use time of the harmonic reducer increases,the accumulated internal wear degree gradually increases,and the state of health evolves into a state of performance decline.A state monitoring method based on Multivariate State Estimation Technique(MSET)is proposed,a memory matrix of health state data is constructed,the residuals are compared with actual observations,and the fault warning threshold is set.It is verified through accelerated life test data of harmonic reducer.The proposed method can inform the failure point18 minutes before the sample life is 5.7 hours,and realize the effect of failure warning.(3)In the aspect of transmission performance of harmonic reducer,the performance parameters of harmonic reducer are predicted based on feature dimension reduction and long short term memory(LSTM)network.The manifold learning dimensionality reduction method is introduced into the feature extraction of reducer signal,and a degradation model based on exponential function is proposed.The LSTM method is used to predict the performance degradation.The prediction method is evaluated by statistical indicators.The prediction error of the proposed method is small and has strong ability of data trend prediction.(4)Based on the above research content,a complete set of robot harmonic reducer monitoring and management system including data acquisition,storage,analysis and display has been designed and built.At the same time,the system can be extended for different industrial robots and other key sub-components.Provide visual technical support for industrial robot condition monitoring and performance prediction,and realize longterm intelligent operation and maintenance management.