Design Of Online Monitoring System For Rolling Bearing Based On Edge Computing

Industrial big data is taking massive strides forward,but the data processing efficiency and real-time performance of the online rolling bearing inspection system on the cloud platform are poor,which are restricted by the computing power and data transmission bandwidth of the cloud center.An online monitoring system for rolling bearings based on edge computing is proposed.The system adopts a hierarchical progressive mode.Firstly,big data was trained and tested in a continuous hidden Markov model which is arranged on the edge layer.It extracts time-frequency features of vibration signals of the rolling bearing,evaluates the importance of these features by random forest algorithm,establishes the sensitive feature sets and inputs them into the continuous hidden Markov model.All of this is to finish condition monitoring and preliminary fault diagnosis on the edge layer.Then,the primary diagnosis is uploaded to the cloud platform,and the final judgment and maintenance arrangement are made by using envelope spectrum analysis.According to the analysis the experiment signals of rolling bearings,it is be verified that the system has high stability and recognition accuracy to meet the real-time requirements with high monitoring efficiency.The main work of this paper as fellows:(1)Theoretical analysis of the feasibility of an online monitoring system for rolling bearings based on edge computing.Through theoretical modeling of intelligent fault diagnosis methods,edge computing,hidden Markov model,etc.of rolling bearings,they are organically combined to build an online monitoring system for rolling bearings based on edge computing.The system performs invalid data rejection,bearing status monitoring and fault status classification at the edge layer.(2)Study the influence of different feature evaluation methods on the fault classification effect of the CHMM model.The influence of the sensitive features extracted by three methods: random selection,random forest algorithm,and compensation distance evaluation technology on the fault classification effect of the CHMM model is studied.Comprehensive analysis of the three aspects of convergence speed,diagnostic accuracy,and classification dispersion,finally choose to establish a random forest-based CHMM diagnostic model.(3)The online monitoring system for rolling bearings based on edge computing is realized by Lab VIEWThe online monitoring system for rolling bearings based on edge computing was implemented in Lab VIEW,and various functional modules corresponding to the terminal layer,edge layer,and cloud layer were built,and each module was verified with the measured signals of the rolling bearing.The results showed that the The system can realize the state monitoring and fault location judgment of the rolling bearing stably and efficiently.Through theoretical analysis and research,an online monitoring system for rolling bearings based on edge computing is established.This system can realize the state monitoring and fault diagnosis of the bearing,and has high real-time performance,high stability,fast recognition speed,and good application prospects.