Implementation of Principal Component Analysis (PCA) as Fault Detection Approach for A Three-Tank System: An Education Perspective

Abstract

Precise and efficient fault management is essential in today's industrial environment to avoid downtime and monetary losses. Any faulty events could affect the factory process flow as well as the product quality. Recently, due to the advancement in signal monitoring with cloud data storage, fault detection approach has change from model-based to data-driven approach. The main goal of this study is to provide an initial insight of data-driven fault detection approach focusing towards undergraduate or postgraduate students including practitioner that just new in this area. At first stage, the data generation and collection approach were described with one process plant example, the three-tank system. Later, detail principal component analysis (PCA) and threshold determination was explained using normal dataset. After that, faulty dataset was used to determine the performance of PCA approach. It is found that, PCA approach only able to detect six out of eleven faulty condition which results in 54.4% of the fault detection performance.