Research On Fault Detection And Fault Prediction Method For Ethercat Master-Slave Station Equipment

With the advent of Industry 4.0 era,in the field of industrial technology,various industrial equipment is developing in the direction of intelligence.Among them,the motion control device changed from the traditional communication method based on pulse signal to the communication method based on real-time industrial Ethernet frame.With the complexity of equipment,the possibility of equipment failure also increases,and failure has an important impact on the quality of products and equipment reliability.Traditional devices use sensors to monitor the health of the device and use limit-based judgment methods,which can only indicate the two states of fault and health,and insufficient representation of the internal state of the device.This thes describes the health status of the system by obtaining control signals and feedback signals,which can not only determine whether the entire device is faulty,but also obtain relevant data values to reflect the trend of the health status of the device.This thesis is based on the project of "Advanced Manufacturing Intelligent Service" of Sichuan Province's major science and technology project in 2019.Based on the design project of the Ethercat master-slave station of an automation equipment manufacturing company in Chengdu,this thesis proposes a research method for fault detection and fault prediction by analyzing the communication data of Ethercat real-time industrial Ethernet.This thesis focuses on the following two aspects:First,it addresses the problems of large amount of data per unit time,unlabeled data,and no explicit correspondence between data and fault characteristics in real-time industrial Ethernet communication data.A fault detection method based on clustering analysis and association rule mining is proposed.After preprocessing the data,the method clusters the samples by fuzzy c clustering algorithm based on subtraction clustering,and the clustering result As the key parameter of adaptive density clustering,it is used to perform cluster analysis on real-time data,and analyze the clustering cluster formed by density clustering using association rule mining algorithm to obtain data-based fault characteristics.Finally,simulation experiments prove the feasibility of the method.Secondly,according to the real-time industrial Ethernet internal communication data,the amount of data per unit time is large and it cannot be directly predicted by time series.An ESN-FESN two-step prediction model is proposed.Based on the echo state network and fuzzy model,the ESN algorithm is used to predict the temporal attributes of the clusters obtained by fault detection,and the prediction results are used It is used to activate the fuzzy rules of FESN,so as to predict the multi-dimensional angle of the cluster cluster eigenvalue attribute,and judge the prediction result through the fault threshold.Finally,simulation experiments prove the feasibility of the method.