Research On Fault Diagnosis Of TE Chemical Process Based On CNN-SVM

The advent of the big data era is driving the modern chemical industry toward automation and intelligence.However,due to the continuous complexity of process technology,many types of failures occur in process industrial systems,and the failure of each unit has an impact on the safety and reliability of the entire production process.Therefore,early detection and identification of faults is very important to prevent system failures.In order to ensure the safe and stable operation of equipment,chemical process fault detection and diagnosis technology has been widely concerned by researchers and chemical companies.With the improvement of computing and processing analysis capabilities,data-driven machine learning-based diagnostic technology has become a popular technique and has been widely used in various chemical process industries.Therefore,this paper proposes a method that combines Convolutional Neural Network(CNN)and Support Vector Machines(SVM),and optimizes the convolutional neural network(CNN)to realize fault diagnosis in the chemical process.The following research work was mainly conducted.(1)SVM fault diagnosis of chemical process based on feature selection.An adaptive wavelet threshold denoising method is proposed to reduce noise in data.In order to solve the dimensional disaster of chemical process data,a feature selection method based on RFECV-RF was proposed.A new feature subset was selected as the input of the subsequent model through comprehensive analysis,and the fault diagnosis effect of several different kernel functions of SVM was compared.(2)Fault diagnosis model based on CNN-SVM.Aiming at the problems of low accuracy and insufficient nonlinear ability of fault diagnosis caused by traditional data-driven modeling methods,a CNN-SVM model combining CNN and SVM was proposed to enhance the nonlinear fitting ability of the model and improve the accuracy of fault diagnosis of chemical process.Compared with the single SVM model and CNN model,the fusion model improved the fault identification accuracy by 5.6% and 3.2%.(3)Fault diagnosis model based on improved CNN-SVM.Due to the complexity of chemical process,there are great similarities between variables.In order to further improve the accuracy of chemical process fault identification,grid search is used to optimize SVM parameters,CNN structure is improved,attention mechanism is added to the CNN network model,and an improved CNN-SVM model is proposed to realize the fault diagnosis of chemical process.Experiments show that the average accuracy,recall rate and F1 value of the fusion model based on the improved CNN-SVM are all above 90%.