采用BP神经网络预测碳纤维增强树脂基复合材料的钻削力

Prediction of Drilling Force in Drilling Process of CFRP via Back Propagation Neural Network

采用双锋角钻头对碳纤维复合材料进行钻削试验,基于反向传播算法的人工神经网络建立钻削轴向力与主轴转速、进给速度之间的非线性关系模型,对比分析三种不同第二主切削刃与第一主切削刃之比的双锋角钻头在试验加工参数下钻削轴向力变化规律。结果表明:与多元线性回归预测模型对比,在相同试验数据为基础的预测计算下,BP神经网络预测值相对误差明显减小,网络预测值误差均在3%之内,而多元线性回归模型最大误差值达到了12.46%,BP神经网络能建立更精准轴向力预测模型。从降低钻削轴向力的角度分析,应采用第二主切削刃与第一主切削刃之比为1的双锋角钻头进行钻削加工。

The double cone drill was used to drill carbon fiber reinforced plastics （CFRP）. A model for describing the drilling axial force and the spindle speed and feed rate was established by uasing the artificial neural networks with Back-propagation algorithm. Under the processing parameters, the comparative analysis of the change law of drilling axial force among three double cone drill which have variant ratio of the second cutting edge to the principal cutting edge. The results show that comparing with the multivariable linear regression model, the relative error prediction value via BP neural network model is lower than theexperimental, which the prediction errors via BP neural network model were below 3%. The maximum error via multiple linear regression model was 12.46%. BP neural network could be used to establish more accurately axial force prediction model. From the point of view reduce drilling axial force, the double cone drill with the ratio of the second cutting edge to the principal cutting edge has to be equal to 1 should be adopted.