Intelligent condition-based maintenance using Plant Health Index

The concept of utilizing empirical models for plant health monitoring has been quite attractive due to its several advantages over the traditional physical model based approach, such as flexible and customized modeling. Major power plant industries in Korea are adopting this methodology to guide their maintenance practices and to optimize their operational costs. These concepts are expected to become essential in nuclear industry in near future to improve the competitiveness with other electric sources, where unnecessary shut downs for maintenance are not welcomed. Korea South-East Power Company (KOSEP), BNF technology, and Kyung Hee University have developed the Plant Health Index (PHI) system, which is an online condition-monitoring system to provide a viable solution for maintenance management. The PHI system models normal operation patterns from historical data by adopting statistical learning techniques and provides an overall health status of a complex system in advance. The success of PHI lies in its independent modules for a residual generator, margin analyzer, and a health index calculator using success tree techniques, which helps to recognize an abnormal component in the pin-point manner and to implement efficient maintenance strategies. In particular the residual generator calculates the differences between the actually measured values and the model-predicted values. The margin analyzer computes the process margin in terms of plant safety and efficiency. The health index calculator merges these two outputs in functional success trees to calculate the plant-level as well as the system-level health index. The core algorithm of the PHI system is being evolved in terms of computational memory and efficiency, accuracy of model prediction, user-friendly interface, that need to be resolved before their wide acceptance in the industry. In collaboration, we are searching for optimum solutions to tackle these problems. In this paper, we are going to present an overview of the PHI system and its academic outcomes of our R&D activities, particularly on data pre-processing and health index calculation.