基于多传感器信息决策级融合的刀具磨损在线监测

Tool Wear Online Monitoring Based on Multi-sensor Information Decision-making Level Fusion

针对无法精确掌控机械加工过程中刀具磨损状态的现状,提出一种基于多传感器信息决策级融合的刀具磨损在线动态监测模型。该模型对采集的振动、力、声发射传感器信号进行特征提取后,将特征按传感器类型划分为独立样本。划分后的独立样本分别对同一个刀具磨损量进行回归预测,进而对每一个独立样本预测得到的刀具磨损量进行加权综合决策,最终决策出刀具磨损量。实验结果表明:刀具磨损在线动态监测模型能够有效地提高刀具磨损动态预测精度,平均预测准确率可达97.9%;与现有研究方法相比,预测准确率至少提升4%以上,预测时间仅为0.016 s,具有较大优势。

An online dynamic monitoring model of tool wear based on multi-sensor information decision-making level fusion is proposed for accurately controlling the tool wear state in the machining process.After extracting the time,frequency and time-frequency features from the collected vibration,force and acoustic emission sensor signals,the monitoring model divides the sensor signal features into independent samples according to the sensor type.The same tool wear extent is regressively predicted using the independent samples,respectively.Then the tool wear extent predicted from the signal characteristics of each sensor is comprehensively determined.Finally,the tool wear extent is determined.The experimental results show that the on-line dynamic monitoring model of tool wear can effectively improve the accuracy of tool wear dynamic prediction,and the average prediction accuracy is 97.9%.Compared with existing research methods,the proposed method is used to increase the prediction accuracy rate by at least 4%,and the prediction time is only 0.016 s.