Degradation Modeling and Performance Monitoring of Electro-optical Detection System via Dynamic Bayesian Network

Abstract

This paper proposes a degradation modeling approach for Electro-Optical detection system based on dynamic Bayesian network. Modulation Transfer Function (MTF) is firstly used in each subsystem for degradation description from the perspective of energy domain, which helps the whole degradation modeling keep away from complicated description of interactions between subsystems. To enable uncertainty and time-variant description of degradation process, a dynamic Bayesian network (DBN) constructed from MTF model is developed. Considering that parameters in DBN cannot be full recognized, Gaussian particle filtering (GPF) is applied with kernel smoothing as inference, combined with which DBN is capable of self-Learning unknown model parameters based on observation data and tracking the dynamic degradation process of time-dependent variables. A case study based on simulation data is presented to show the effectiveness of the proposed method in degradation modeling and performance monitoring for Electro-Optical detection systems.