Intelligent prediction methods for rock mechanical parameters based on logging while drilling responses

Accurate prediction of rock mechanics parameters during drilling is critical for optimizing resource extraction and ensuring operational safety. This study develops an intelligent model to address this need. First, correlations were established between drilling parameters (conventional and vibration data) and key rock properties. Subsequently, a Particle Swarm Optimization (PSO) algorithm was applied to optimize a BP neural network, constructing a high-performance PSO-BP prediction model. The model was trained and validated using 60 datasets from three lithologies (limestone, sandstone, and coal). Results show that the PSO-BP model achieved an average relative error of 5.252%, substantially lower than that of a standard BP network (10.821%) and empirical formulas, confirming its superior accuracy and stability. This approach offers a promising new direction for automated rock mechanics assessment.