OPERATION DATA ANALYSIS OF DIESEL ENGINE IN INDUSTRY USING PRINCIPAL COMPONENT ANALYSIS (PCA) METHOD CASE STUDY : SYMPTOMS DETECTION - IDENTIFICATION OF KNOCKING AND INJECTOR FAULT

• Main Author: WIRAHADIKUSUMA MUALIM (NIM : 23109026); Tim Pembimbing : Dr. Ir. Arief Hariyanto Prof. Wira, ALFONSUS
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/14806
• Institution: Institut Teknologi Bandung

In large Diesel engine, sensors that are used for measuring operation data have been installed, either analog or digital. Each measured variable has a limit value issued by the Diesel engine manufacturers. Abnormal symptoms of the Diesel engine can occur before a variable passes its limit value and this condition can not be easily detected by a maintenance operator manually. In this study, a multivariate statistical method, Principal Component Analysis (PCA), is applied for operation data analysis in detecting and identifying fault of large Diesel engine. Three PCA algorithms i.e., conventional PCA, RPCA (Recursive PCA) and MWPCA (Moving Window PCA) are used to analyze operation data. Operation data used in the study are obtained from a Diesel engine in Tarakan Diesel Power Plant. The data, consists of 43 operating variables, are measured from analog sensors mounted on Diesel motor. Two cases, knocking and injector fault, are used to evaluate these three algorithms. <br /> <br /> <br /> From applying three PCA algorithm, it can be concluded that the model derived from conventional PCA can not be applied in the Diesel process monitoring because it can not represent the dynamic operating conditions of Diesel engine. Both adaptive PCA methods, MWPCA and RPCA, can detect the two cases mentioned earlier. The Adaptability of MWPCA is better than RPCA because the amount of data used as model in MWPCA can be less than RPCA. Besides, MWPCA can detect fault symptoms earlier than RPCA.