A feature extraction and visualization method for fault detection of marine diesel engines

High efficiency fault-detection and fault-tolerant control approach in Tennessee Eastman process via fuzzy-based neural network representation

“…The proposed structure to detect and identify a fault in fault-tolerant control in the TE process has been realized. The faults in this research are taken as idv (1), idv (4), idv (8), idv (12), simult aneously, wit h idv (15). The faults idv (1) and idv (8) have the direct impact on products concentration that are G and H to be shown.…”

“…All classifiers include methods based on the neural networks such as MLP, 1 RBF 2 and statistical methods such as Bayesian, KNN, 3 and Parzen Windows and also SVM. 4 In Fig. 2, the diagram of fusion classifier methods has been presented.…”

“…In all, matrix confusion, class one represents the normal state. Classes 2-5 expresses faults idv (1), idv (4), idv (8), idv(12) simultaneously with idv (15) . In one such case, the results are illustrated in Figs.…”

“…Therefore, after occurring these faults, output loop, which has to control G material concentration, have to be adjusted. idv (4) that is related to reactor temperature. In R structure, this variable is controlled by a local PI controller by input cold water to reactor cooler.…”

“…The TE process is unstable in open-loop mode and the first limitation that causes this instability is exceeding from reactor allowed pressure range. Realized control structure should be able to save process in operational conditions and can put process variables in its proposed constraints range [4][5][6][7][8][9][10][11][12][13][14][15][16][17]. Regarding the proposed research novelty and its innovation, it can be expressed in three lines.…”