Research On Fault Diagnosis Method Of Chemical Process Based On Reinforced Convolutional Neural Network

Intelligent fault diagnosis of chemical process is a key means to accurately identify the health status of chemical equipment and ensure the safety and reliability of the production process.With the characteristics of high latitude,nonlinearity and time sequence in the modern chemical process,the problems caused by the limitations of traditional fault diagnosis methods are becoming more and more prominent.Therefore,it has become an inevitable trend to choose different deep network models to improve data processing capabilities and fault classification capabilities.Considering the problems of insufficient feature extraction ability and slow model training speed in the existing deep network models,two fault diagnosis models based on enhanced convolutional neural networks are proposed for fault diagnosis of complex chemical processes.Through comparative experiments,it has been verified that the two models have very high diagnostic accuracy,which has a huge advantage in dealing with the faults of the chemical process.details as follows:(1)For the existing fault diagnosis methods in the face of the current complex chemical process data,the features are not sufficiently proposed,resulting in the final diagnosis effect is not ideal.A multi-scale zigzag residual dilated convolution method based on model fusion is proposed,which can realize end-to-end fault diagnosis.First,a new MRJDCNN model is designed by integrating the principle of residual learning into the improved sawtooth dilated convolution,and then it is combined with the LSTM network delicately.It can not only greatly improve the extraction rate of nonlinear high-latitude spatial features at different scales while preventing model degradation,but also skillfully extract time-related features,which effectively reduces the loss of necessary features.At the same time,the accuracy of model diagnosis is greatly improved.(2)Most of the current methods are designed based on the depth of the model,the model is relatively complex,and the fault diagnosis cannot meet the real-time requirements.From the perspective of "attention",a new fault diagnosis model integrating attention convolutional neural network and XGBoost algorithm is designed.First,by introducing the attention module into the classic convolutional neural network,a new convolutional neural network model ACNN with attention mechanism is designed,which can enhance features through channel dimension and spatial dimension.The survival rate of useful features is improved,and useless redundant features are reduced.Secondly,the original softmax classifier is replaced by the XGBoost classifier,so as to achieve the purpose of improving the diagnostic accuracy of the model.It is worth mentioning that the model proposed in this paper has the characteristics of light weight,which enables it to complete the diagnosis of fault data in a relatively short time,while ensuring very high diagnostic accuracy.