基于随机森林与主成分分析的刀具磨损评估

发布时间：2018-08-28 13:27

【摘要】：刀具是数控机床的重要零部件,其性能直接影响着生产加工精度。为实现机床刀具磨损程度的在线分级评估,提出一种刀具磨损状态的评估方法,该方法结合随机森林与主成分分析模型,建立不同工况下主轴电机电流传感器信号样本与刀具磨损等级的非线性映射关系。在不同加工条件下进行刀具性能试验,采集主轴电机的电流信号和铣削加工参数。对信号利用小波包分解、时域统计、频域分析提取特征,利用随机森林得到刀具磨损的分级评估结果。该方法可有效解决样本不平衡的问题,与常见的组合分类方法 Ada Boost所得结果相比,该模型能准确地反映刀具的磨损程度,鲁棒性更好。该方法仅利用数控机床内置传感器实现,无需改动机床结构,不影响主轴动态加工性能,可广泛应用于工业数控机床刀具的磨损评估。
[Abstract]:Tool is an important part of NC machine tools, and its performance directly affects the machining accuracy. In order to realize on-line grading evaluation of tool wear degree, an evaluation method of tool wear state is proposed, which combines stochastic forest and principal component analysis model to establish the signal sample of current sensor of spindle motor under different working conditions and to evaluate the wear degree of tool. Nonlinear mapping relation of tool wear grade.Tool performance test is carried out under different machining conditions,current signal of spindle motor and milling parameters are collected.Wavelet packet decomposition,time-domain statistics,frequency-domain analysis and feature extraction are used to extract the signal.Random forest is used to get the result of tool wear grading evaluation.This method can effectively solve the problem. Compared with the results of Ada Boost, this model can accurately reflect the wear degree of cutting tools and has better robustness. This method is realized only by using the built-in sensors of NC machine tools, without changing the structure of the machine tool and without affecting the dynamic machining performance of the spindle. It can be widely used in industrial NC machine tools. Wear assessment.
【作者单位】：上海交通大学机械系统与振动国家重点实验室;上海航天设备制造总厂;
【基金】：国家科技重大专项资助项目(2014ZX04015-021)
【分类号】：TG71

【相似文献】