Stress distribution analysis in ZrB2-SiC based thermal protection system using machine learning driven approach

Data-to-knowledge has significantly increased over the last decades, and this led to a new paradigm of science where Machine Learning (ML) and big data are utilised to optimize and solve complex problems. This study is the first of its kind to evaluate a non-discrete stress distribution behaviour of ZrB2/SiC composite based thermal protection system (TPS) using ML. Thermo-fluid analysis using FEM at re-entry conditions indicated maximum external temperature of 5000 K with boundary layer pressure up to 205 MPa. The ML driven approach through Keras library with combination of clustering indicated that high thermal stress occurs within approximately on three quarters of boundary layer separation points where flow is highly stochastic, and oxygen ionization takes place. A maximum stress of ~400 MPa obtained from stress distribution supports the ablative behaviour of TPS. The use of deep learning has significantly reduced the computational time needed for simulation and showed improved accuracy.