Abstract:A new method for the optimum design of slope stabilization is presented in this paper. Firstly,the particular treatment method is determined in term of the detailed conditions of the slope. Using finite element method,combined with limit equilibrium method,the slope stability is analysed and the stresses of support structures are calculated. Upon this basis,a group of design parameters are given,and correspondingly the safety factor of the slope after reinforcement and the cost of the treatment engineering are estimated. Secondly,dozens of design parameters,safety factors and costs are given,which can be used as learning samples for neural networks. Then an evolutionary neural network model whose network structure is optimized by genetic algorithm is made to map the complex nonlinear relationship among the design parameters,safety factors and engineering costs. Thirdly,embedding the above evolutionary neural networks model into the genetic algorithm,taking the cost as fitness function and the safety factor as constraint,an optimum model is made. Thus,the design parameters can be optimized in the whole spectrums,and the most economic result is obtained. Lastly,the above process is programmed in parallel environment,and the searching efficiency and accuracy are both improved. This idea was practiced in Shizibao landslide treatment engineering of the Three Gorges Projects,and an excellent economic result was obtained.