고장예지 및 건전성관리에서의 AI 기법

Abstract

Purpose: Today, prognostics and health management (PHM) is being focused upon in the manufacturing, aerospace, and defense industries. This study aimed to identify the characteristics of diagnosis and PHM, and review the artificial intelligence (AI) techniques used for their implementation. Methods: PHM involves both diagnosis and prognostics, and AI is used for the diagnosis and PHM of a component or a system. There are two approaches to PHM: model-based and databased approaches. The former involves a system of differential equations, an expert system, a finite state machine, and qualitative reasoning, while the latter involves conventional numerical methods such as linear regression and Kalman filtering, and machine learning methods such as those involving a neural network, a decision tree, or a support vector machine. We review some of these approaches in this paper. Results: The historical maintenance paradigm is presented along with the evolution of the production paradigm. Diagnosis and PHM are essential for the efficient maintenance of a system, and methods for each of them are briefly described. Conclusion: A proper understanding of diagnosis and PHM and their appropriate implementation are essential for the efficient maintenance of a system. Model-based and data-based approaches to PHM are briefly described. Both types of approaches have their merits and demerits and they should be tried out.