基于复化Cotes积分和神经网络的稳定铣削工艺参数多目标优化方法  

Multi-objective Optimization Method of Stable Milling Process Parameters Based on BP Neural Network and Composite Cotes Integration

作　　者：娄维达 秦国华[1,2] 王华敏[2] 吴竹溪[2] LOU Weida;QIN Guohua;WANG Huamin;WU Zhuxi(School of Mechanical Engineering,Northwestern Polytechnical University,Xi'an 710072;School of Aeronautical Manufacturing Engineering,Nanchang Hangkong University,Nanchang 330063)

机构地区：[1]西北工业大学机电学院,西安710072 [2]南昌航空大学航空制造工程学院,南昌330063

出　　处：《机械工程学报》2023年第17期279-290,共12页Journal of Mechanical Engineering

基　　金：国家自然科学基金(51765047);江西省自然科学基金重点(20232ACB204019);江西省主要学科学术和技术带头人培养计划(20172BCB22013);江西省重点研发计划(20203BBE53049)资助项目。

摘　　要：铣削过程再生颤振严重影响工件表面质量和生产效率,准确高效地识别铣削颤振稳定域并合理选择工艺参数是抑制颤振、提高生产效率的关键步骤。当前对考虑稳定性约束下的多目标铣削工艺参数优化研究仍然缺乏系统且完整的解决方案。为此,基于复化Cotes积分和神经网络并结合NSGA-Ⅱ遗传算法建立一种稳定铣削工艺参数多目标优化方法。其中基于复化Cotes积分法提出了一种新的铣削稳定域预测方法来获得二维稳定性叶瓣图(Stability lobe diagram,SLD),收敛性分析表明新方法具有更快的收敛速度。考虑径向切深不确定性得到由离散三维SLD曲面构建的铣削稳定域神经网络预测模型。最后,分别以材料去除率、刀具寿命作为效率、成本目标,采用NSGA-Ⅱ遗传算法建立稳定铣削工艺参数的优化模型。与经验方法对比显示,优化后的铣削方案可以提高8.4%的材料去除率和16.3%的刀具寿命。结果表明不仅可以通过更加准确地判断铣削稳定性来保证加工质量,而且能够为高效率低成本的铣削确定工艺参数提供科学理论指导。The regenerative chatter in the milling process seriously affects the surface quality and production efficiency of the workpiece.Accurately and efficiently identifying the stable region of the milling chatter and selecting the process parameters reasonably are the key steps to suppress the chatter and improve the production efficiency.At present,there is still a lack of systematic and complete solutions for the optimization of multi-objective milling process parameters considering stability constraints.Therefore,a multi-objective optimization method of stable milling process parameters is established based on composite Cotes integral and neural network combined with NSGA-II genetic algorithm.Among them,a new milling stability region prediction method is proposed based on the composite Cotes integration method to obtain a two-dimensional stability lobe diagram(SLD).Convergence analysis shows that the new approach has a faster convergence rate.Considering the uncertainty of radial depth of cut,a neural network prediction model of milling stability region constructed by discrete three-dimensional SLD surfaces is obtained.Finally,taking the material removal rate and tool life as the efficiency and cost targets respectively,the NSGA-II genetic algorithm is used to establish the optimization model of stable miing process parameters.Compared with the empirical method,the optimized milling scheme can improve the material removal rate by 8.4%and the tool life by 16.3%.The results show that not only can the machining quality be ensured by judging the milling stability more accurately,but also scientific theoretical guidance can be provided for high-efficiency and low-cost milling to determine the process parameters.

关 键 词：铣削颤振 多目标参数优化 叶瓣图 神经网络 遗传算法 

分 类 号：TH161[机械工程—机械制造及自动化]