Bayesian probabilistic machine learning analysis of ceramic‐coated ultra‐high‐temperature carbon/carbon composites

Abstract Regulatory agencies and key stakeholders are increasingly promoting the use of probabilistic approaches in design processes for large corporations. This shift is particularly emphasized in analyzing mechanical properties, such as fatigue and failure prediction. Additionally, the use of probabilistic artificial intelligence represents a transformative advancement in material science that leads to enhanced predictive accuracy and robust decision‐making capabilities. These artificial intelligence methods enable more informed decision‐making in the design and evaluation of advanced materials by quantifying uncertainty and offering probabilistic assessments, particularly for applications involving extreme environments. High‐temperature materials, such as carbon/carbon (C/C) composites, are essential for modern technological applications. However, their vulnerability to oxidation poses a significant barrier, indicating the necessity for effective protective coatings. The application of these coatings to C/C composites is complex and has hindered their widespread use in high‐temperature settings. In this study, we utilize finite element analysis (FEA) and machine learning (ML) combined with Bayesian probability to examine the behavior of silicon carbide ceramic‐coated cubic C/C composites. The investigation focuses on how stress and strain evolve under varying thermal conditions and cyclic thermal loading from a probabilistic perspective. This work integrates FEA and Bayesian probabilistic‐based ML to enhance the predictive power for evaluating ultra‐high‐temperature materials.