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AdaBoost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。
赵帅
,
黄亦翔
,
王浩任
,
刘成良
,
刘晓
,
梁鑫光
. 基于随机森林与主成分分析的刀具磨损评估[J]. 机械工程学报, 2017
, 53(21)
: 181
-189
.
DOI: 10.3901/JME.2017.21.181
The cutting tool is a critical part of the CNC machine. Its performance directly affects the machining accuracy. A method to assess the wear status of the cutting tools is proposed based on the combination of random forest analysis and PCA to establish a nonlinear mapping relationship between the features of the spindle current signals and tool wear. The degrees of the tool wear are divided into several classifications. Experiments have been conducted by testing the tools of different machining conditions. Wavelet packet decomposition, time domain statistics and frequency domain analysis are performed on the signals for feature extraction. Then the random forest method is applied to evaluate and classify different tool status. Compared with the results from AdaBoost which is a common boost classification method, results show that the proposed model is more accurate and robust. The method can avoid the problem of imbalance samples. In addition, it can be realized on the build-in sensors of the industrial CNC machines so that there is no need to change the original structure design to avoid the potential interferences of the spindle's dynamic processing performance, which enables a wide industrial applications of the tool wear assessment for the CNC machines.
[1] LI Lihong, AN Qingbin. An in-depth study of tool wear monitoring technique based on image segmentation and texture analysis[J]. Measurement, 2016, 79(2):44-52.
[2] SAMIK D, SURJYA K P, RANJAN S. Progressive tool flank wear monitoring by applying discrete wavelet transform on turned surface images[J]. Measurement, 2016, 77(1):388-401.
[3] MAURICIO E N, PAULO R A, HILDO G J, et al. Evaluation of neural models applied to the estimation of tool wear in the grinding of advanced ceramics[J]. Expert Systems with Applications, 2015, 42(20):7026-7035.
[4] FARBOD A N, DURUL U, LAINE M. In-process tool flank wear estimation in machining gamma-prime strengthened alloys using Kalman filter[J]. Procedia Manufacturing, 2015, 1(1):696-707.
[5] MEHDI N, BARRY K F, BETH L Z, et al. Real-time tool wear monitoring in milling using a cutting condition independent method[J]. International Journal of Machine Tools and Manufacture, 2015, 89(2):1-13.
[6] AZMI A I. Monitoring of tool wear using measured machining forces and neuro-fuzzy modelling approaches during machining of GFRP composites[J]. Advances in Engineering Software, 2015, 82(4):53-64.
[7] WANG Guofeng, YANG Yinwei, XIE Qinglu, et al. Force based tool wear monitoring system for milling process based on relevance vector machine[J]. Advances in Engineering Software, 2014, 71(5):46-51.
[8] 陈保家,陈雪峰,李兵,等. Logistic回归模型在机床刀具可靠性评估中的应用[J]. 机械工程学报, 2011, 47(18):158-164. CHEN Baojia, CHEN Xuefeng, LI Bing, et al. Reliability estimation for cutting tool based on logistic regression model[J]. Journal of Mechanical Engineering, 2011, 47(18):158-164.
[9] KADU R S, AWARI G K, SAKHALE C N, et al. Formulation of mathematical model for the investigation of tool wears in boring machining operation on cast iron using carbide and CBN tools[J]. Procedia Materials Science, 2014, 6(1):1710-1724.
[10] YANG C Y, WU T Y. Diagnostics of gear deterioration using EEMD approach and PCA process[J]. Measurement, 2015, 61(2):75-87.
[11] LEO B. Random forests[J]. Machine Learning, 2001, 45(1):5-32.
