基于Cotes求积法和神经网络的稳定域判断及铣削参数优化新方法  

A Novel Method of Stability Judgment and Milling Parameter Optimization Based on Cotes Integration Method and Neural Network

作　　者：秦国华[1,2] 娄维达 林锋[1] 徐勇 QIN Guohua;LOU Weida;LIN Feng;XU Yong(School of Aeronautical Manufacturing Engineering,Nanchang Hangkong University,Nanchang 330063,Jiangxi,China;School of Mechanical Engineering,Northwestern Polytechnical University,Xi'an 710072,Shaanxi,China)

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

出　　处：《兵工学报》2024年第2期516-526,共11页Acta Armamentarii

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

摘　　要：再生效应引起的铣削颤振,是制约加工效率和工件质量的主要因素之一。准确高效地识别铣削颤振的稳定域,是抑制颤振、提高生产效率的关键步骤。为此,根据铣削振动微分方程,利用柯特斯积分法(Cotes Integration Method,CIM)提出一种新的铣削稳定域预测方法。利用CIM获得主轴转速-轴向切深平面二维稳定性叶瓣图(Stability Lobe Diagram,SLD),与半离散法(Semi-discretization Method,SDM)和全离散法(Full-discretization Method,FDM)等方法进行比较,结果表明CIM的收敛性更优。考虑径向切深的影响,建立三维SLD的迭代计算方法,通过等效地离散三维SLD曲面为节点集合,随机以90%的节点作为训练集,构建轴向切深与主轴转速、径向切深之间的神经网络预测模型。10%验证集的预测结果表明神经网络的预测误差不超过6%。以材料去除率为效率目标,刀具寿命为成本目标,建立稳定铣削多目标优化模型并应用基于分解的多目标进化算法(Multiobjective Evolutionary Algorithm Based on Decomposition,MOEA/D)求解,为获得高效率、低成本的最优铣削参数提供基础理论和技术支撑。Milling chatter caused by regeneration effect is one of the main factors restricting machining efficiency and workpiece quality.The accurate and efficient identification of milling chatter stability is crucial to suppress the chatter and improve the production efficiency.Therefore,a predication method for milling chatter stability is investigated using the Cotes integration method(CIM)on basis of the milling vibration differential equation.The 2D stability lobe diagram(SLD)is obtained by CIM,which is compared with 1st semi-discretization method(SDM),1st semi-discretization method(FDM)and 2nd FDM.The results show that CIM is of better convergence.Considering the effect of radial depth of cut on SLD,an iterative calculation method of 3D SLD is established.After equivalently discretizing 3D SLD surfaces as a set of nodes,a neural network prediction model between axial depth of cut and spindle speed and radial depth of cut is constructed by randomly using 90%of the nodes as the training set and 10%of the nodes as the validation set.The predicted results show that the predicted error of neural network is no more than 6%.Finally,a multi-objective optimization model of stable milling is suggested with the objective of the material removal rate and tool life in addition to the corresponding MOEA/D.It provides basic theory and technical support for obtaining the optimal milling parameters with high efficiency and low cost.

关 键 词：铣削 颤振 Cotes积分 神经网络 遗传算法 

分 类 号：TH161[机械工程—机械制造及自动化]