Development Of Intelligent Fault Diagnosis System For Diesel Engines Based On Vibration Analysis

As the power core of many industries such as vehicles,ships,and electric power,diesel engines play an important role in national production and life.However,factors such as harsh working environment and complex mechanical structure make diesel engine failures occur frequently.Once the diesel engine fails,it will reduce work efficiency and cause a certain degree of economic loss,and even cause damage to the unit and even threaten personal safety.Therefore,it is of great significance to study diesel engine fault diagnosis technology and develop a high-accuracy and real-time fault diagnosis system to ensure the good and stable operation of diesel engines and the smooth progress of production and life.Because the signal acquisition is convenient and the anti-interference ability is strong,this paper adopts the vibration analysis method to carry out the diesel engine fault diagnosis research.Taking a certain type of six-cylinder diesel engine as the research object,a fault simulation experiment was carried out for the common misfire faults of diesel engines.And system development laid the data foundation.The traditional diesel engine fault diagnosis algorithm includes three relatively independent links: signal decomposition,feature extraction and pattern recognition,and the links are incoherent,the recognition accuracy is low,and the calculation speed is slow.the existing diesel engine fault diagnosis algorithm based on convolutional neural network often has too many model parameters,and it is easy to overtake during training.fit.In response to the above problems,this paper proposes an improved convolutional neural network(Robust Scaler and Spatial Pyramid Pooling based Convolutional Neural Network,RSCNN)based on robust scaling and pyramid pooling.RSCNN preprocesses the original vibration signal through robust scaling,which reduces the impact of abnormal peaks on diagnostic accuracy;uses scaling exponential linear unit instead of traditional activation function to avoid information loss in the activation process,and uses pyramid pooling module to extract multi-scale features of the signal,which improves the feature extraction ability of the network;uses separable convolution instead of ordinary convolution layer,and global average pooling instead of fully connected layer to reduce model parameters and reduce the risk of overfitting.Through the above improvements,the number of trainable parameters of RSCNN is only 11887,and the average diagnostic accuracy of the 15-type diesel engine misfire fault data set is as high as 99.18%,so it is used as the built-in fault localization algorithm of the fault diagnosis system.On the basis of the above research,the diesel engine condition monitoring and fault diagnosis system is developed by using Lab VIEW.The system mainly includes four modules: user login,data acquisition,data storage,and fault diagnosis.System user rights are divided into administrators and operators.Only administrators can adjust the status monitoring threshold;the data acquisition and storage module can realize multi-channel high-speed parallel acquisition and classified data storage;the fault diagnosis module can realize real-time status monitoring and fault diagnosis.Two major functions are located.The type of the system state monitoring value is determined as the form factor by performing common time-domain feature statistics on the signals of each working condition in the data set.The RSCNN algorithm is deployed in the system by calling the python node,and the "end-to-end" End" quick fault cylinder location.After debugging and verification,each module of the system works well,and can achieve high-accuracy,real-time and intelligent monitoring and diagnosis,and no manual intervention is required in the whole process of operation.