基于变分自编码器的液压泵腰槽特征加工参数评估模型  

Cutting Parameter Evaluation Method Based on Variational Autoencoder

作　　者：李晶晶 蔡跃波 李佳怡 文嘉怡 陈姝元 LI Jing-jing;CAI Yue-bo;LI Jia-yi;WEN Jia-yi;CHEN Shu-yuan(AVIC Jincheng Nanjing Mechatronic Hydraulic Engineering Research Center,Nanjing 211106;College ofMechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016)

机构地区：[1]中国航空工业集团公司金城南京机电液压工程研究中心,江苏南京211106 [2]南京航空航天大学机电学院,江苏南京210016

出　　处：《制造业自动化》2025年第3期62-70,共9页Manufacturing Automation

基　　金：航空制造企业技术攻关课题(ZY23027);国家自然科学基金(52175467)。

摘　　要：数控加工工艺参数决策高度依赖工艺人员的经验,导致工艺准备周期长,工艺质量稳定性差。分油盘、分油盖和壳体零件是液压泵系统中非常重要的组成部分,其腰槽特征存在加工效率低、成本高等缺点,是生产企业降本增效任务的重点研究对象。因此建立铣削工艺参数与工艺指标之间的关联关系,是评估优化加工参数,进而提高加工效率、保证加工质量的重要基础。由于加工中涉及复杂的热力耦合,加工参数与工艺指标之间是复杂的非线性关系,传统拟合方法与数据驱动方法难以有效建模。针对以上问题,提出了一种基于变分自编码器的切削参数评估模型,建立加工参数与反映切削特性的关键物理量之间的预测模型,为企业实现加工参数优化提供基础。验证结果表明,所建立的模型预测误差为1.2%,满足加工参数优化的要求。The decision-making process of CNC machining parameters relies heavily on the engineers'experience,leading to long preparation and unstable machining quality.The oil distributor cap,oil distributor and shell parts are critical components of the hydraulic pump system.Their internal slot features often suffer from low machining efficiency and high costs,which are the key targets for production enterprises'cost reduction and efficiency improvement.Therefore,establishing the relationship between milling process parameters and process indicators is essential for evaluating and optimizing machining parameters,thus improving machining efficiency and ensuring process quality.Due to the complex thermo-mechanical coupling involved in the machining process,the relationship between machining parameters and process indicators is highly nonlinear,making it difficult to model effectively using traditional fitting or data-driven methods.To address this issue,this paper proposes a cutting parameter evaluation model based on a Variational Autoencoder(VAE),which establishes a predictive model between machining parameters and key physical quantities that reflect cutting characteristics.This model serves as a foundation for optimizing machining parameters in production enterprises.The validation results show that the prediction error of the proposed model is merely 1.2%,fulfilling the requirements for machining parameter optimization.

关 键 词：智能制造 变分自编码器 腰槽特征 加工参数评估 

分 类 号：TG547[金属学及工艺—金属切削加工及机床] TP18[自动化与计算机技术—控制理论与控制工程]