Deep Learning based Finite Element Analysis

During the design process of an autonomous underwater vehicle (AUV), the pressure vessel has a critical role. The objective of this study is to develop a deep learning-based surrogate model that can effectively replace traditional Finite Element Analysis (FEA) simulations during the pressure vessel design process. The methodology involves training a deep learning model to predict stress effects based on sparse datasets, as generating dense data for FEA simulations is computationally expensive. The results indicate that the deep learning-based surrogate outperforms classical regression models such as random forest and gradient boost in handling sparse data scenarios. By implementing this surrogate model, the prediction speed for design evaluations is significantly enhanced compared to direct FEA simulations. The findings highlight the potential of deep learning as a promising approach for surrogate modeling in engineering design tasks, particularly in applications involving complex relationships between design parameters.