Abstract: Engineering practice shows that shear slip is the major failure mode of rock, so to choose the appropriate constitutive model of rock joints is an important part of analysis for failure of rock joints. Current commonly used constitutive models for deformation of rock are exponential model, curve model and power function model, their parameters come from actual experience or laboratory tests, cannot simulate well the actual deformation. Analog values derived from exponential model are too large, derived from curve model are too small, derived from power function model are too small at high stress. Due to that, a pervasive constitutive model of rock surface structure has been proposed in this paper, then its analog values have been identified through improved genetic algorithm, finally constitutive model of rock surface structure composed by ‘equivalent parameters’ has been proposed. Results from verification examples show that the constitutive model identified by improved genetic algorithm can simulate well the actual deformation, can be well adapted to different engineering applications.