A Health Assessment Method Based on Directed Acyclic Graph and Improved Mahalanobis- Taguchi System

Health status assessment is a key procedure for fault prediction and health management for complex systems. To accurately evaluate the health status of complex systems, this paper proposes a health assessment method based on a directed acyclic graph and an improved Mahalanobis- Taguchi system. Firstly, the Mahalanobis distance between samples of different health statuses is calculated by using the Mahalanobis distance system, and the discrimination degree of each sample combination is analyzed and compared. Then, the multi-classification process of equipment health status is constructed by the directed acyclic graph. Next, different feature combinations are formed by the features selected in the orthogonal experiment, and the Mahalanobis- Taguchi system is constructed separately for each combination. After voting, an improved Mahalanobis- Taguchi system used to divide the final health status category is obtained. Finally, the PHM08 data provided by the National Aeronautics and Space Administration of the United States is applied to evaluate the health status of the aero-engine.