System-Level Performance Prediction for Infrared Systems Based on Energy Redistribution in Infrared Images

The infrared (IR) system has an abroad application in industry, military, medical science, and many aspects of society life. Performance prediction of IR system has great significance to advance maintenance decisions before system severe degradation. This article presents a system-level prognostic method to predict IR system performance. The proposed method treats IR system degradation as energy redistribution between different degradation levels, and considers the system-level degradation as joint effect of degradations from different subsystems. To quantify system degradation level, a health indicator (HI) whose value can be measured from system output IR images is constructed. Then, prediction of the HI is achieved by a modulation transfer function (MTF)-based prognostic model established based on the degradation mechanism of subsystems. Degradation modeling in the proposed MTFbased prognosticmodel is realized in a just-in-time learning framework, in which the model with simple structure is built locally and dynamically using a limited amount of relevant data from historical dataset. Finally, a Gaussian particle filter-based online inference is utilized to update prediction results sequentially as observations of HI become available. The effectiveness and advantages of the proposed method is evaluated and verified through simulations and real IR data collected from flight experiments, with associated comparative analysis.