Inversion Analysis of the In Situ Stress Field around Underground Caverns Based on Particle Swarm Optimization Optimized Back Propagation Neural Network

The in situ stress distribution is one of the driving factors for design and construction underground engineering. Numerical analysis methods based on artificial neural networks are most common effective inversion. However, conventional algorithms often have some drawbacks, such as slow convergence, overfitting, local minimum problem, which will directly affect inversion results. An intelligent inverse method optimizing back-propagation (BP) network with particle swarm optimization algorithm (PSO) applied to back stress. PSO used optimize initial parameters BP network, improving stability accuracy numerical simulation utilized calculate field generate training samples. In application Shuangjiangkou Hydropower Station powerhouse, average relative error decreases by about 3.45% using proposed compared method. Subsequently, shows significant tectonic movement surrounding rock, first principal value 20 26 MPa. fault lamprophyre significantly influence stress, 15–30% localized reduction rock mass within 10 m. research results demonstrate reliability improvement provide a reference similar