Physics-Informed Deep Monte Carlo Quantile Regression method for Interval Multilevel Bayesian Network-based Satellite Heat Reliability Analysis
This addresses satellite heat reliability analysis for engineering applications, but it is incremental as it combines existing techniques like DCNNs with physics knowledge and Monte Carlo methods.
The paper tackles the problem of reconstructing satellite temperature fields from noisy unlabeled data by proposing an unsupervised physics-informed deep Monte Carlo quantile regression method, which achieves accurate reconstruction and quantifies aleatoric uncertainty, validated through two case studies.
Temperature field reconstruction is essential for analyzing satellite heat reliability. As a representative machine learning model, the deep convolutional neural network (DCNN) is a powerful tool for reconstructing the satellite temperature field. However, DCNN needs a lot of labeled data to learn its parameters, which is contrary to the fact that actual satellite engineering can only acquire noisy unlabeled data. To solve the above problem, this paper proposes an unsupervised method, i.e., the physics-informed deep Monte Carlo quantile regression method, for reconstructing temperature field and quantifying the aleatoric uncertainty caused by data noise. For one thing, the proposed method combines a deep convolutional neural network with the known physics knowledge to reconstruct an accurate temperature field using only monitoring point temperatures. For another thing, the proposed method can quantify the aleatoric uncertainty by the Monte Carlo quantile regression. Based on the reconstructed temperature field and the quantified aleatoric uncertainty, this paper models an interval multilevel Bayesian Network to analyze satellite heat reliability. Two case studies are used to validate the proposed method.