Abstract: Abstract: Traditional operations of walnut harvest and breaking shell seriously affect the machining quality and efficiency in walnut processing. With continuous exploring areas and increasing output of walnut, developing deep-processing technology is of extremely vital significance to walnut industry. Walnut shell breaking is an important stage of walnut industrialization process. Mechanical force is widely used to obtain a large number of broken walnut kernels in walnut industrialization process. The properties of volume size, shell thickness and texture characteristics of walnuts greatly affect the process of obtaining the kernels. Walnut shell stress and deformation depends on the contents of H2O, volume of size and loading speed during breaking shell. The fruits of Wen 185 sorted by diameter (Divide 4 grades, reference walnut processing standards in Hetian) and H2O content (4%, 6%, 8% and 10%) were compressed by microcomputer-controlled machine using different loading rate (100, 200, 300 and 400 mm/min). Meanwhile, force-deformation curves were analyzed and rupture energy were calculated. It was important to predict the walnut shell rupture energy for improving the design and development of walnut processing equipments. The back-propagating (BP) artificial neural network was an effective prediction model, which highlighted the characteristics of fast, accurate and better adaptability. However, the BP had the deficiencies of insufficient network global search ability, slow convergence and local optimum iteration. The remedy patterns of genetic algorithm that performed global searching would optimize the weights and thresholds in BP network, and thereby improve the accuracy of predictions. For Wen 185 walnut in southern Xinjiang, the H2O content, compression speeds, and transverse diameter were considered as the basic characteristic parameters for BP neural networks models. Genetic algorithm was used to optimize the weights and bias of BP neural work. Optimized BP neural network was applied to predict the rupture energy of walnut shell breaking. The genetic BP prediction neural network model was trained and tested with the experimental data collected from rupture energy. The results showed that the errors between predicted and tested results were small, and there was non-linear relationship between rupture energy and main controlling factors in the model which resulted from the genetic BP network. The correlation coefficient of the network output value between samples and BP network was 0.92488. The optimized BP neural network model had a stronger ability for nonlinear approach, which actually reflected the nonlinear relationship between the rupture energy of walnut shell breaking and main controlling factors. The predicted results from the genetic BP network were better than the back-propagating artificial neural network. Therefore, the genetic BP network is an effective method used for prediction of the rupture energy of walnut shell.