Deep learning-driven predictive tools for damage prediction and optimization in composite hydrogen storage tanks

Abstract

This research presents a comprehensive framework for predicting the damage in lightweight composite high-pressure hydrogen storage tanks and optimizes their design to prevent failure. By integrating advanced analytical methods, numerical analysis, and deep learning techniques, we introduced a novel approach to enhance design optimization and damage prediction. The framework employs the 3D elasticity anisotropy theory to predict mechanical performance, incorporating failure criteria for accurate damage analysis. A parametric study was conducted to examine the effects of thickness, pressure, and diameter on the behavior of the tank. The analytical results were compared against finite element analysis using the WoundSim software, underscoring the significance of the modeling assumptions. Furthermore, we developed a new framework that combines deep neural networks with differential evolution optimization (DNN-DEO) to predict stress and damage in composite pressure vessels while identifying the optimal design parameters (pressure, radius, and thickness) to minimize failure risks and maintain high performance. A graphical user interface (GUI) was also designed to automate calculations and predictions, providing an intuitive tool for users. This integrated approach offers a powerful solution for optimizing the design and operation of lightweight composite hydrogen storage tanks, ensuring the reliability and efficiency of hydrogen storage systems. • Improved computational efficiency using 3D elasticity anisotropy theory. • Detailed parametric study on pressure, layer thickness, and tank radius effects. • High accuracy DNN models for predicting stress components and failure points. • User-friendly GUI for real-time assessment of stress and failure criteria. • Advanced design integration for safer and more efficient hydrogen storage tanks.