Fault Detection and Identification with Kernel Principal Component Analysis and Long Short-Term Memory Artificial Neural Network Combined Method

A new fault detection and identification approach is proposed. The kernel principal component analysis (KPCA) first applied to the data for reducing dimensionality, occurrence of faults determined by means two statistical indices, T2 Q. K-means clustering algorithm then adopted analyze perform clustering, according type fault. Finally, using a long short-term memory (LSTM) neural network. performance proposed technique compared with (PCA) method in early detecting malfunctions on continuous stirred tank reactor (CSTR) system. Up 10 sensor other system degradation conditions are considered. LSTM network three machine learning techniques, namely support vector (SVM), K-nearest neighbors (KNN) algorithm, decision trees, determining results indicate superior suggested methodology both identification.