基于高斯过程回归的复杂曲面加工切削力高效预测

Cutting Forces Prediction for Complex Surface Machining Based on Gaussian Process Regression

为了准确高效地监测和预报复杂曲面多轴加工的切削力,本文建立了一种基于高斯过程回归(GPR)算法的切削力预测模型。提出根据刀位文件(CLS)提取刀具–工件啮合区刀轴矢量倾角、切宽等特征参数的方法,构建以特征参数为输入的GPR模型,实现复杂曲面加工切削力预测。开发了复杂曲面加工切削力仿真软件,与传统机械力模型利用布尔运算计算刀具–工件啮合区域并预测切削力的方法进行对比,切削力预测误差小于10%,预测结果评价系数维持在0.98以上。设计叶轮流道加工试验,验证了本文提出的GPR模型对复杂曲面加工切削力预测的准确性。在整体叶轮流道加工同一条刀路的预测时间上,基于布尔运算预测切削力的方法耗时161 s,而本文方法耗时仅为1.63 s,实现了复杂曲面加工切削力高效准确预测。

In order to accurately and efficiently forecast the cutting force of complex surface multi-axis machining,a cutting force model based on Gaussian process regression(GPR)algorithm is developed in this paper.Feature parameters of tool–workpiece engagement,such as tool-axis inclination angles and cutting width,which serve as input characteristic parameters of GPR model for cutting force prediction in complex surface machining,are extracted based on cutter location file(CLS).The training set of the GPR model are obtained using the mechanical force model where the tool–workpiece engagement is calculated by Boolean operations.Cutting force simulation software for complex surface machining is developed and the efficiency of the proposed GPR model is verified by comparing with the traditional force prediction model which adopts the Boolean operations to calculate the tool–workpiece engagement.The error of cutting force prediction is less than 10%and the evaluation coefficient of prediction results is maintained above 0.98.An impeller runner machining experiment was designed to verify the accuracy of the GPR model proposed in this paper for predicting cutting force in machining complex curved surface.In force prediction based on the same CLS file for an impeller passage processing,the method using Boolean operation takes 161 s,while the time elapsed of the proposed model is only 1.63 s.The results indicate that the proposed model is efficient and accurate for cutting force prediction in complex surface machining.