Abstract: In light of the problem of describing nonlinear rheology of rock and soil with principle of linear superposition, a neural network model for rheology of rock and soil under step loading has been established with artificial neural network instead of traditional mathematical and mechanical methods. By simulating the experimental creep curves of gypsum breccias, it is shown that the model can effectively describe nonlinear rheology of rock and soil with better prediction. This approach provides a new way for studying rheological properties of rock and soil especially nonlinear rheological properties under step loading.

Key words: rheology, step loading, nonlinearity, neural network, model