Research on Data-Driven Mechanical System Network Construction and Anomaly Monitoring Methods
Aiming at the difficulty of fault diagnosis and anomaly monitoring caused by the increasing complexity of mechanical system,combined with the integral characteristics of the system,complex network technology and system operation process data are used to realize the network model construction and anomaly monitoring of mechanical system.Firstly,the improved cosine similarity algorithm is used to calculate the correlation level between nodes,and the network model is constructed accordingly.Secondly,the distance between the graph sequences is calculated in the time domain to analyze the relative abnormal situation of the system at a certain time,and then judge whether the system produces abnormal behavior.Finally,with the help of Auto Regressive Moving Average(ARMA)model,the system state is predicted.The proposed method is verified by the running data of the rolling bearing.The results show that the model can accurately monitor the time node of the abnormal system and predict the subsequent running state of the system.
NETL
NSTL
万方数据
