Research And Application Of Adaptive Fault Diagnosis Method Based On DPCA

With the development of industrial automation,there is a higher demand for the accuracy and efficiency of fault detection technique in the industrial process.In the complex industrial processes such as oil drilling process,traditional fault detection methods have many problems,such as low detection efficiency,poor real-time performance and weak robustness to noise.At present,there is little research on the adaptive fault detection method for oil drilling process.To solve this problem,this paper focuses on the study of the adaptive fault detection methods,and puts forward an adaptive fault detection method applied to the oil drilling process.The main achievements of this paper include:(1)For the disadvantage of the sensitivity of principal component analysis(PCA)to noise,this paper studies several de-noising algorithms in detail and proposes an off-line fault detection method based on PCA.This method utilizes the advantages of PCA in projecting the data in high-dimensional space into low-dimensional space,and combines the variational mode decomposition method with Hilbert-Huang transform(HHT)to denoise principal component in low-dimensional space.The effectiveness of the proposed algorithm is verified by the fault detection experiment of Tennessee Eastman(TE)process.(2)In view of the fact that the fault detection model based on traditional PCA method is a static model and cannot effectively monitor the process data with time-varying and dynamic characteristics in the industrial process,improved methods are proposed based on moving window PCA(MWPCA)method and dynamic PCA method in this paper.In consider of the correlation between time-varying process data and model will decay with time,an adaptive fault detection method based on HHT-MWPCA is proposed,which combines with Hamming window.(3)Considering the strong dynamic characteristics of drilling process data,an adaptive fault detection method based on HHT-MWDPCA is proposed by introducing the augmented matrix of dynamic PCA(DPCA)method.The fault detection experiments of TE process and drilling process prove that the improved algorithms can self-adaptively monitor the time-varying process with dynamic characteristics,reduce the false alarm rate and missed alarm rate,and enhance the robustness of the model to faults.