Application of DPCA to oil data pH model building and comparison of optimal CBM policies.

摘要

This thesis builds Proportional Hazards Model (PHM) for Condition Based Maintenance (CBM) using covariates obtained from Dynamic Principal Component Analysis (DPCA) of oil data. DPCA covariates, derived from DPCA of the oil data, represent most of the variability in the original data with reduced dimension and little cross-correlation. This makes DPCA covariates ideal variables in maintenance modeling and decision making. The analysis is performed in two steps. Firstly, analyze the correlation in the original data and apply DPCA to the multivariate time series fitted to the data. Then, test the correlation of selected DPCA covariates, define their states as Markov process, and build the PHM. A case study using oil data obtained from monitoring heavy hauler trucks is developed. The maintenance policy using DPCA covariates appears superior to the one using original variables.