Abstract: In order to understand a large amount of fuzzy and uncertain factors of red slate rock materials in geotechnical engineering, an implicit constitutive model is established on the basic of ANN(artificial neural network) method and experiment on geotechnical materials. In the paper, the ANN s learning and training are done directly on the experimental results, the knowledge of mechanical behaviors of the materials can be obtained and stored in the weight structure of a multilayer feed forward back-propagation neural network, and hence we obtain an implicit constitutive model expressing stress-strain relationships of geotechnical materials.