将BP神经网络应用于典型的曲面推土板-模型铲刀推土阻力的预测。在模型铲刀推土试验研究结果的基础上,以模型铲刀的切削角、前翻角和切削速度为输入,以模型铲刀推土阻力的水平分力与垂直分力为输出,建立了BP神经网络数值模拟模型。研究结果表明,该BP神经网络有效地预测了模型铲刀的推土阻力,其准确率在94%以上。
The surfaces of soil - engaging components is the combination of curve blade and plane blade commonly, considering the different parameters and conditions involved in the soil - tool interaction system, it became a difficult work to study the interaction between soil and curve blade variously, and it was essential to use numerical simulation instead of physical test to solve this problem. The BP neural network was used to "learn" results of the numerical simulation to the draft resistance of model blade, it had a structure of 3 ×10 × 2, input facts of the BP neural networks were cutting angle, front reversal angle and forward speed, output facts were horizontal and vertical components of draft resistance. The correctness of the output facts is more than 94% ; it means the performance equation of the adhesion interface system has a high forecasting accuracy and favorable generalization ability.