Integrated system health management-oriented maintenance decision-making for multi-state system based on data mining

To ensure a series of missions can be completed with only finite breaks, many systems are required to guarantee system safety and mission success. Of these, maintenance decision support is vital. One widely used maintenance strategy has been selective maintenance. Most traditional selective maintenance optimisation research has focused on binary state systems, which are subject to distribution deterioration or failure. However, a majority of systems used in aerospace or industrial applications are multi-state systems with more than two states deteriorating at the same time, meaning that real-time state distribution is needed to provide more timely and effective maintenance. This paper presents a novel integrated system health management-oriented maintenance decision support methodology and framework for a multi-state system based on data mining. An aero-engine system numerical example is given to illustrate the methodology, the results of which demonstrate the significant advantages of using data mining to efficiently obtain state distribution information, and the benefits of using a robust optimal model to choose suitable strategies. This methodology, which is applicable to multi-state systems of varying sizes, has the ability to solve maintenance problems when imperfect maintenance quality is considered.