基于异源信号特征融合的刀具磨损状态识别  

Tool Wear State Recognition Based on Heterogeneous Signals Features Fusion

作　　者：潘贺丰 朱坚民[1] 

机构地区：[1]上海理工大学机械工程学院,上海200093

出　　处：《电子科技》2017年第5期105-109,共5页Electronic Science and Technology

基　　金：国家自然科学基金(50975179);上海市教委科研创新基金(11ZZ136);上海市科委科研计划基金(13160502500);沪江基金(D14005)

摘　　要：针对单一刀具切削信号的局限性和磨损特征存在较强的非线性,提出一种基于异源信号特征融合的刀具磨损状态识别方法。同时采集加工过程中的振动信号和声发射信号,分别提取它们小波包频带的相关时域特征,采用局部保留投影算法对原始特征进行降维,并将两信号的低维特征进行融合,对融合特征通过灰色接近关联模型进行磨损状态识别。通过3把车刀的切削磨损实验,结果表明该方法可充分利用异源信号的互补信息和特征数据之间的非线性关系,更全面地表征刀具磨损状态,相比单信号法具有更高的识别精度。A tool wear state recognition method based on heterogeneous signals features fusion is proposed to ad-dress the limitation of single tool turning signal and strong nonlinear relation of wear features. The vibration signal and acoustic emission signal are acquired simultaneously during the processing, and the wavelet packet frequency band relevant time -domain features are extracted respectively. Local preserving projection algorithm （LPP） is used to re-duce the original features dimension. The low dimension features of the two signals are fused and sent to the grey nearness incidence model for wear state recognition. Experiment of the three turning tools show that the proposed method can make full use of complementary information from data, and characterize tool wear state more comprehensively. different signals and nonlinear relation among features The proposed method can obtain higher recognition accuracy than single signal method.

关 键 词：异源信号 局部保留投影 特征融合 刀具磨损 灰色接近关联模型 

分 类 号：TN911.7[电子电信—通信与信息系统] TB93[电子电信—信息与通信工程]