EMCS-PGNN: Physics-guided neural network-based enhanced Monte Carlo simulation for high-dimensional reliability analysis

High input dimensionality, high computational cost, high nonlinearity, and low failure probability are the main challenges in current structural reliability analysis. Developing efficient, accurate, and robust reliability analysis methods is crucial for solving complex engineering problems. In this work, a novel physics-guided neural network-based enhanced Monte Carlo simulation is proposed. Besides, a novel scaling formula and training interval are established. The key contribution lies in constructing the physics-guided loss function to enhance the robustness of the method. The effectiveness of the proposed method is verified through two high dimensional, low failure probability numerical cases and engineering cases. The results show that the proposed method outperforms existing methods and is capable of solving practical engineering problems.