A data-driven approach for real-time prediction of thermal gradient in engineered structures

Predicting thermal gradient of engineered structures with uncertain parameters is crucial for the safety. Many numerical computation methods have been developed in literatre; however, a practical approach is still pursued. To bridge this research gap, a data-driven approach is proposed to realize real-time prediction of thermal gradient in engineered structures with uncertain parameters. First, tetrahedral elements are used to build structures with complex geometry, and the face-based smoothed finite element method (FS-FEM) is employed to calculate the thermal response of the structures. By doing so, a good trade-off between the efficiency and accuracy for modelling the structures can be achieved. Then, the generated datasets are used for training a back propagation neural network (BPNN) model, and real-time prediction of thermal gradient in engineered structures can be performed without knowing the exact distribution functions of stochastic parameters. Lastly, two case studies were investigated to evaluate the performance of the proposed approach. The analysis results demonstrate that the proposed data-driven approach is able to offer accurate predictions of the thermal gradient for the engineered structures with uncertain parameters.