基于多目标灰狼算法的H13模具钢铣削工艺优化研究

Research on Optimization of H13 Die Steel Milling Process Based on Multi-Objective Grey Wolf Algorithm

摘要: 模具产品在结构上多为复杂型腔，具有高精度、多特征、难加工等特点，而在模具加工有着显著优势的高速铣削加工，最优的铣削工艺参数选择影响着工件的生产效率、表面质量。首先采用了正交试验设计方法，开展了H13模具钢铣削试验，对结果进行了极差分析，并采用GA-BP神经网络构建了表面粗糙度预测模型；其次，以建立的GA-BP表面粗糙度预测模型和材料去除率为目标函数，相关工艺参数为主要优化变量，基于多目标灰狼优化算法，建立多目标铣削参数优化模型。结果表明，铣削工艺参数对表面粗糙度的影响敏感性排序为：主轴转速 > 切宽 > 进给速度 > 切削深度；所建立的GA-BP神经网络预测模型能够有效预测加工表面粗糙度，平均相对误差为8.72%；并通过实例验证了优化的工艺参数，对比优化前，材料去除率提高了40.1%，表面粗糙度降低了36.4%，加工时间减少了29.3%。

Abstract: Die products are in the structure of complex cavity, with high precision, multi-features, difficult to process and other characteristics, and have a significant advantage of high-speed milling processing in the die processing, the optimal milling process parameters affect the production efficiency of the workpiece, surface quality. Firstly, the orthogonal experimental design method was used to carry out the milling experiment of H13 die steel, the range analysis of the results was carried out, and the GA-BP neural network was used to construct the surface roughness prediction model. Secondly, with the established GA-BP surface roughness prediction model and material removal rate as the objective function and related process parameters as the main optimization variables, a multi-objective milling parameter optimization model was established based on the multi-objective grey wolf optimization algorithm. The results show that the sensitivity of milling parameters to surface roughness is in the order of spindle speed > cutting width > feed speed > cutting depth. The GA-BP neural network prediction model can effectively predict the machining surface roughness, and the average relative error is 8.72%. The optimized process parameters are verified by an example. Compared with before optimization, the material removal rate is increased by 40.1%, the surface roughness is reduced by 36.4%, and the processing time is reduced by 29.3%.