Material constitutive model is highly nonlinear and multimodal in the large parameter space. A genetic evolution algorithm is thus proposed for its recognition. The nonlinear stress±strain relationship presented by several added multinomials, whose structure is not simpli®ed, is automatically recognized from global response information, e.g., load vs. re¯ection data, obtained from a structure test through genetic evolution in global space. Nonlinear ®nite element analysis is used as a bridge to build a relationship between stress± strain data and load±de¯ection information. The potential of the proposed method is demonstrated by applying it to the macromechanical modeling of nonlinear behavior in advanced composite materials. A nonlinear material model for the ply is recognized using experimental results on a lamina plate AE45 6 s to be a modi®cation of Hahn±Tsai model [H.T. Hahn, S.W. Tsai, J. Compos. Mater. (1973) ( ) 102]. The obtained nonlinear constitutive model is subsequently used to predict nonlinear behaviors of the AE30 6 s and [(0/AE45 4 s plates. The results are satisfying. This modeling procedure can be used as a method to guide to improve analysis of nonlinear behavior and damage of composite materials.