An improved turbo machinery condition monitoring method using multivariate statistical analysis

Abstract

Industrial practitioners require a well-structured, proactive and precise conditionmonitoring package in order to optimize turbomachinery operation. Typically, conventional condition monitoring uses built-in software to capture faults or degradation processes based on threshold limits recommended by the Original Equipment Manufacturer (OEM). However, because OEM manual concurrent monitoring involves abundant information parameters, it is dependent on human intervention, insensitive to the development of machinery faults and tends to generate error-prone outcomes. This study proposes a simplified and advanced healthmonitoring method for turbomachinery using a multivariate statistical analysis (MSA) technique. By exploiting mathematical relationships between OEM recommended variables, the significance of input parameter is identified based on weighting factor. With a highly-correlated input subset, the revised condition monitoring method delivershigher sensitivity and a more accurate performance in investigating machine assessment mode.