A general approach for health index-based remaining useful life prediction using functional principal component analysis

Accurately predicting the remaining useful life (RUL) is crucial for industrial device prognostics and health management. With the advancement of sensor technology, multiple sensor data can be collected during the monitoring process of these devices. This requires analyzing data collected from multiple sensors and extracting useful information to predict the RUL of the product. Existing studies have proposed various methods for fusing multi-sensor data to construct a health index (HI). However, most of these methods rely on the assumption that sensors are independent of each other. In reality, during equipment operation, there is a certain correlation between sensor signals. Ignoring these correlations when constructing the HI may affect the accuracy of RUL predictions. To address this issue, this paper proposes a method based on functional principal component analysis (FPCA) for multi-sensor signal fusion. Compared to other data fusion methods, using FPCA to analyze data can extract orthogonal principal component functions from multiple sensors, effectively modeling the correlations between sensor signals. The proposed method utilizes FPCA to fuse data from multiple sensors, achieving dimensionality reduction while avoiding collinearity. Simulation studies and a case study on the C-MAPSS dataset demonstrate that the proposed method can effectively fuse data from multiple sensor signals and accurately predict RUL.