阻力墙参数对转向节锻件充填影响的神经网络分析

为研究阻力墙结构参数对转向节锻件长叉充填的定量影响,设计了广义回归( GR)人工神经网络模型。用“舍一法”训练了模型,并采用3个样本对模型进行预测检验,散点表明预测值和实验值拟合较好。统计学指标为：均方误差M1为0. 898 0,相对均方误差M2为0.167 0%,拟合分值V为1.973 9,说明人工神经网络具有较高的预测精度。最后用神经网络分析阻力墙关键参数对锻件长叉充填的定量影响,结果表明：长叉侧边桥部宽度和阻力墙斜度的增加对长叉充填作用不明显;阻力墙间隙的加大不利于长叉充填;阻力墙宽度对长叉充填的影响呈抛物线关系,先增大后减小,存在一个极大值。GR人工神经网络模型能够定量预测各阻力墙参数对长叉充填的影响。

Influence of Resistance Wall Parameters on Filling of Steering Knuckle Based on GR Neural Network

ln order to study the influence of the structural parameters of the resistance wall on the long cross filling of the steering knuckle, a generalized regression ( GR) artificial neural network model was designed." Leave-one out" method was used to train the ANN model and 3 samples were used to test the model, and the scatter plots showed that thepredicted values and the experimental values fitted well. Statistical indicators were M1= 0.898 0, M2 =0.167 0% , v =1.973 9, which showed that the prediction results of ANN model had high prediction accuracy. The quantitative effectsof resistance wall key parameters on the long cross filling were analyzed by artificial neural network model. The results showed that increase of the bridge width and resistance wall slope had less influence on the long cross filling, and the increase of the resistance wall clearance was not conducive to long cross filling, the influence of resistance wall width to the long cross filling was parabola, first increased and then decreased, and there was a maximum value. GR artificial neural network model could be used to predict the impact of the parameters of the resistance wall quantitatively on the long cross filling.