Study on the Mechanical Property and the Evolutionary Neural Network Constitutive Model for Limestone under Chemical Corrosive Environments

Abstract

In this paper, the experimental study on the mechanical property of limestone under triaxial compression with different hydrochemical environments is conducted and the non-linear characteristics of complete stress-strain process of limestone subjected to chemical corrosion, are analyzed. The behaviors of deformation and strength of limestone eroded by different chemical solutions are obtained. It is known from the experimental and analytical results that different chemical environments such as chemical composition, pH value etc can affect the mechanical property of rock differently. How to establish a multifactor characteristic constitutive model which can reflect different chemical environments is the key problem to study the coupling corrosion effect of stress and chemistry of rock. The constitutive model of evolutionary neural network for rock under chemical corrosive environments is put forward, and the neural network constitutive relationship under stress-chemistry coupling corrosion is established by applying the theory of evolutionary computation and neural network. The neural network constitutive model established in this presented paper mainly takes into account the following three aspects: chemical environments of rock specimens, content of main mineral compositions that are liable to being eroded by chemical solution, and mechanical environments. The results are in good agreement with the experimental data.