Fault diagnosis of Tennessee Eastman process with multi-scale PCA and ANFIS

Monitoring of chemical industrial processes using integrated complex network theory with PCA

“…From Eq. (12), it can be seen that the network edge is simultaneously changed, which means that the number of edges connected to a vertex is also changed. The local structure information of network vertices should thus be studied.…”

“…For this fault, KPCA-SPE and KPCA-T 2 statistics fail to detect the fault, and there are many samples below the control limit from sample 161, which induces a high miss detection rate. In contrast to the KPCA method, the proposed method is more sensitive in detecting the small Reactor cooling water inlet temperature step IDV (5) Condenser cooling water inlet temperature step IDV (6) A feed loss step IDV (7) C header pressure loss-reduced availability step IDV (8) A, B, and C feed composition Random variation IDV (9) D feed temperature Random variation IDV (10) C feed temperature Random variation IDV (11) Reactor cooling water inlet temperature Random variation IDV (12) Condenser cooling water inlet temperature Random variation IDV (13) Reaction kinetics Slow drift IDV (14) Reactor cooling water valve Sticking IDV (15) Condenser cooling water valve Sticking IDV (16)(17)(18)(19)(20) Unknown Unknown IDV (21) The valve fixed at steady state position Constant position changes that occurred in the measured variables. Fig.…”

“…This shortcoming thus limits the use of such methods. In contrast to the model-based methods, the process history-based methods (data-driven methods) [6], such as PCA [7,8], DPCA [9,10], DPCA-DR [11], multi-scale PCA [12], ICA [13], DICA [14], KDICA [15], the GPLV model [16], and DHKPLS [17], do not need a priori knowledge and only rely on process knowledge. These methods eliminate the use of detailed models, which are expensive and difficult to develop for quantitative or qualitative model-based methods; thus, the data-driven methods have been extensively studied and developed over the past few decades.…”

Variable selection method for fault isolation using least absolute shrinkage and selection operator (LASSO)

“…To give some examples, Lee et al conducted root cause fault diagnosis of this process based on system decomposition and dynamic partial least squares [34], Zhao and Gao tested the monitoring performance of the fault-relevant PCA also using this process [35], Lau et al proposed to use multi-scale PCA and adaptive neuro-fuzzy inference system to detect and diagnose the faults contained in the TE data [36], Chiang et al used this process to demonstrate the utilization of causal analysis [37], and Rato and Reis conducted fault detection in the TE process using dynamic principal components analysis based on decorrelated residuals [38]. In the survey papers, Kano et al [39] and Yin et al [40] studied the performances of various statistical process monitoring methods through their applications to the TE process.…”

Designing supervised local neural network classifiers based on EM clustering for fault diagnosis of Tennessee Eastman process

“…Various supervised learning models such as support vector machines (SVM) [17], the fisher discriminant analysis (FDA) [18], artificial neural networks (ANN) [18], the Bayesian network classifier [19,20] and multi-scale PCA and ANFIS [21] have been applied for fault classification of industrial processes. Among the mentioned methods, the artificial neural network of multilayer perceptron (MLP) type has received considerable attention due to its simplicity and high efficiency as non-linear classifier.…”