Research On Health Monitoring And Fault Intelligent Diagnosis Of Complex Equipment

With the rapid development of society,highly integrated and intelligent complex equipment is widely used,such as aerospace,transportation,energy and chemical industry and other fields.These devices usually have a high degree of autonomy,complexity and uncertainty.In its working environment,it is easily affected by various factors,and the equipment may produce various abnormalities and failures,and even serious accidents may occur.However,traditional health monitoring and fault diagnosis methods are usually based on human experience and intuition,which have limitations such as subjectivity,uncertainty,and time-consuming,and cannot meet the efficient,real-time,and accurate needs of a big data environment.Therefore,with the development of computer technology and machine learning,how to use sensor data and operation records to realize health monitoring and fault diagnosis of complex equipment has important practical significance and economic value.This research combines statistical process control and neural network to apply complex equipment health monitoring and fault diagnosis.The main research contents are as follows:(1)Due to the high-dimensionality,nonlinearity and imbalance of complex equipment operation data,the nonlinear method of nuclear principal component analysis is used to reduce the data dimension,introduce T~2 and SPE statistics,and combine the kernel density estimation method to calculate the statistical control limit,establish a health monitoring model,realize the data separation of normal state and abnormal fault state,and provide the basis for the next fault diagnosis;(2)Aiming at the problem that traditional fault diagnosis methods require fault type labels,and the fault types of complex equipment are difficult to be clearly divided,the self-organizing map neural network is used to conduct unsupervised clustering analysis on abnormal fault data,combined with the improved BP neural network to classify faults,and establish fault types library,using the offline model for online diagnosis,and can update and optimize the model through continuous data collection to realize remote real-time diagnosis of faults;(3)In view of the problems of low signal-to-noise ratio and incomplete feature display of two-dimensional data reconstruction methods such as traditional grayscale images,a method of converting one-dimensional signal data into two-dimensional radar chart is proposed,which intuitively shows the relationship between multiple dimensions of information,using residual convolution network feature extraction,and adding convolution block attention mechanism to increase the network’s adaptive learning and selection capabilities,and construct a fault diagnosis model.Applying the above method to an actual case,through calculation and analysis,the results show that the method in this paper can carry out health monitoring and fault intelligent diagnosis on 7 kinds of faults of the track circuit system and 5 kinds of operating states of the high-pressure roller mill,and can realize real-time monitoring and warning of complex equipment,timely detecting equipment abnormalities,rapid troubleshooting of hidden dangers,effectively improving equipment reliability,reducing downtime in the maintenance process,reducing operating costs and maintenance costs,improving operational efficiency and safety,and promoting digital and intelligent transformation of enterprises.It provides support for further in-depth exploration of complex equipment operation rules and optimal design,and has broad application prospects.