APPLICATION OF PRINCIPAL COMPONENT ANALYSIS (PCA) FOR DATA RECONCILIATION ON MULTISTAGE COMPRESSOR

• Main Author: Yudha NIM: 23004301, Wielianto
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/9608
• Institution: Institut Teknologi Bandung

Abstract: <br /> <br /> <br /> <br /> <br /> Measurement data always contain errors. These errors could be in the form of gross error, systematic error, and random error. Gross error detection in measurement data need to be done first such that it could be distinguished whether the disturbance in the process is really due to the process itself or is caused by malfunction in the measurement system. <br /> <br /> <br /> <br /> <br /> Data reconciliation activity is performed to treat contaminated data into consistent information. Meaningful data adjustments can be obtained if and only if there is no gross error in the data. The purpose of data reconciliation is to resolve the contradictions between the measurements and their constraints. <br /> <br /> <br /> <br /> <br /> The latest development of gross error detection has been focused on the use of multivariate statistical techniques. These techniques give more accurate and correct detection in determining the variables suspected to contain gross error. These techniques are reliable to solve the problem of correlation present in measurement data. As one of multivariate statistical technique, PCA can be used to detect gross error in the measurement. <br /> <br /> <br /> <br /> <br /> The objectives of this research are to apply PCA for detection and identification of gross error in measurement data, to apply data reconciliation technique to obtain reconciled data, and to investigate application of PCA leading to its use in the dynamic data reconciliation. In this research, the operation of multistage compressor in urea fertilizer plant is chosen as the studied case. <br /> <br /> <br /> <br /> <br /> The results of this research prove that PCA can be applied for detection and identification of gross error in measurement data of mass flow, pressure, and temperature. PCA application on steady state data reconciliation can be used for detection the location of gross error in the measurement data. Gross error in the data can be eliminated and corrected exactly. Moreover, PCA can be used to detect and isolate fault quickly leading to its use in the dynamic data reconciliation. Nevertheless, process knowledge base must be collected first before the application of PCA. If the process knowledge is more complete, the confidence of the inference becomes higher.