Abstract: In order to determine the optimum scheme of filling ratio, on the basis of experimental, neural network and genetic algorithm based on predicted global optimal filling experimental conditions, optimal experimental conditions sand than 0.2024, curing time day 5.863 and solubility 67.8%, filling the largest compressive strength 0.6777 MPa, and the best ratio of lime sand ratio 1:4, curing days 28 days, solubility of 75%, the maximum compressive strength 5.48 MPa difference is large, the prediction results is not very satisfied. The method has good applicability, and the prediction accuracy of neural network has an effect on the optimization of genetic algorithm.