计算机集成制造系统 ›› 2018, Vol. 24 ›› Issue (第10): 2502-2513.DOI: 10.13196/j.cims.2018.10.012

• 当期目次 • 上一篇    下一篇

支持少样本的高速滚齿工艺参数优化

曹卫东,阎春平+,吴电建   

  1. 重庆大学机械传动国家重点实验室
  • 出版日期:2018-10-31 发布日期:2018-10-31
  • 基金资助:
    国家自然科学基金资助项目(51575071)。

Optimization of cutting parameters for high-speed gear hobbing based on small sample problem

  • Online:2018-10-31 Published:2018-10-31
  • Supported by:
    Project supported by the National Natural Science Foundation,China(No.51575071).

摘要: 针对少量历史加工案例支撑下的工艺参数优化问题,提出一种融合支持向量回归和多目标蜻蜓算法的高速滚齿工艺参数优化方法。建立高速滚齿加工效果评价模型,设定ε-支持向量回归参数:核函数类型、惩罚因子和松弛变量,利用多目标蜻蜓算法生成支持向量回归参数组,利用支持向量回归生成包括主轴转速、进给量在内的工艺参数组,计算包括加工质量、加工时间、加工成本、环境影响在内的目标值,运用多目标蜻蜓算法中的Pareto最优方法寻找非支配解,循环上述过程,满足终止条件得到支持向量回归参数和工艺参数非支配解集,即优化支持向量回归参数组和工艺参数组。通过与实际滚齿加工比较验证了所提方法的可行性,并通过与单独支持向量回归等算法的对比说明了该方法在少样本的情况下具备一定优势。

关键词: 高速滚齿加工, 工艺参数优化, 支持向量回归, 多目标蜻蜓算法

Abstract: To solve the problem of unsatisfactory processing effect under small number of historical cases,an optimization approach of cutting parameters was proposed based on Support Vector Regression (SVR) and Multi-Objective Dragonfly Algorithm (MODA).Gear hobbing objective function model was established.ε-SVR was chosen and the type of kernel function,penalty factor and slack variable were generated by MODA.The process parameters of spindle speed and feed rate were produced with ε-SVR.The function model was applied to calculate the target value that included machining quality,machining time,machining cost and environmental impact.The non-dominated solutions were obtained by using the Pareto optimal method in MODA.Once the termination condition was satisfied to obtain the ε-SVR parameters and the non-dominated solution sets (optimal process parameter sets).By comparing with the actual hobbing process,the feasibility of the method was verified;by comparing with the separate SVR and other algorithms,the proposed method had some advantages in the case of small samples.

Key words: high-speed gear hobbing, machining parameters optimization, support vector regression, multi-objective dragonfly algorithm

中图分类号: