Jichao Sun

Joint particle size calculation model (JCM): a neural network approach for efficient rock and soil simulation

Abstract The numerical simulation of rock and soil mass in hydropower engineering will consume a lot of human resources and electric power required for simulation. A rapid and efficient establishment of a rock and soil numerical model is urgent for numerical simulation. The equivalence of the material grading curve is an essential criterion for the accurate simulation of geotechnical material by the discrete element, and particle size calculation is the premise of the grading curve. The discrete element of joint particles can accurately simulate the geotechnical particles in reality, but it is difficult and time-consuming to calculate the particle size of joint particles. This paper presents a neural network calculation model for joint particle size, that is, the joint particle size calculation model (JCM), quickly estimating particle size. The results show that the neural network model with 3–5 circle balls and 11–13 hidden neurons can obtain the gradation curve with a good coincidence degree. The model significantly reduces the calculation time. This paper establishes a joint particle database, which provides standard data for simulation and substantially improves simulation efficiency. The database enhances the comparability and standardization of research and makes research results more universal. The research significantly reduces computers’ human resources and power consumption in numerical simulation.

Doi https://doi.org/10.5200/baltica.2025.1.3

Keywords soil simulation; rock simulation; simulation efficiency; particle simulation; particle database; particle model database; discrete element method

Full text