Research On Predictive Maintenance System For Industrial Robots

With the increasing level of automation in the production process of modern enterprises,most of the manufacturing tasks are completed by smart devices,which requires more and more continuous work of equipment.However,sudden downtime can affect production continuity and reduce production efficiency.Therefore,intelligent manufacturing puts forward higher requirements for the maintenance of equipment.The maintenance of equipment has undergone a process of development from post-maintenance to predictive maintenance.Predictive maintenance(Pd M)is a state-based maintenance,through the system components of the regular(or continuous)status monitoring,to determine the state of the equipment,predict the future development trend of the equipment state,according to the state development trend of the equipment and possible failure modes,advance the development of predictive maintenance plans.Its purpose is to maximize the availability of systems used in modern manufacturing.As a typical representative of industrial production equipment,the application field of industrial robots is more extensive.Therefore,it is of great significance to conduct research on predictive maintenance methods for industrial robots.However,due to the complexity of the design of industrial robots,the number of sensors used is large and the functions are complex.Faced with sensor data of this scale and corresponding complexity,this thesis aims to solve this challenge with a deep learning-based framework.This thesis proposes a model framework for predictive maintenance systems.The proposed prediction algorithm is applied to the simulation data set and the robot arm fault data set of the industrial robot respectively,and the fault and remaining life prediction are effectively carried out in the sensor operation data used to facilitate the maintenance of the equipment.The main tasks are as follows:(1)Aiming at the problem that traditional fault monitoring methods require a lot of data processing time,and cannot better extract deep fault features and complete fault monitoring classification more accurately,a Random Forest(RF)and Temporal Convolutional Network(TCN)are proposed.Firstly,the feature selection is performed by RF algorithm,and the rolling average and rolling standard deviation of the importance features are extracted to reconstruct the feature variables,and the processed feature variables are input to TCN for predictive maintenance of fault state,so as to improve the performance of fault monitoring and other related classification problems.The results show that the F1 Score is increased by up to 10.56%on the C-MAPSS dataset and up to 24.68% on the industrial robot dataset.(2)Aiming at the regression problem of residual life,a method for predicting the remaining useful life of industrial robots based on autoencoder(AE)and long short term memory network(LSTM)is proposed.First of all,use the random forest algorithm to filter out the strongly important features of the fault,realize the dimensionality reduction processing of the data features to eliminate irrelevant feature variables,and send them to the autoencoder as the input feature data;Then use the auto-encoder to extract the nonlinear features of the relevant importance features after dimensionality reduction,and reconstruct them into new feature input variables;Secondly,calculate the statistics of the reconstructed residuals to obtain the squared prediction error(SPE),integrate SPE with the original input feature variables to form a new time series,and finally input the brand new time series data into LSTM to predict the remaining useful life of the equipment.The results showed that the mean absolute percentage error(MAPE)increased by up to 50.56% on the C-MAPSS dataset and up to 9.10% on the industrial robot dataset.(3)On the basis of theoretical and algorithm research,a predictive maintenance system is constructed by using Lab VIEW,which mainly realizes the fault status judgment,remaining useful life prediction and related information query function of the equipment,and visualizes the prediction results,which is of great significance to the stable and reliable operation of the equipment.