基于功率传感器的刀具磨损量预测方法

发布时间：2018-06-13 11:13

本文选题：刀具磨损 + 预测　；参考：《同济大学学报(自然科学版)》2017年03期

【摘要】：使用功率传感器监测机床加工功率,和切削力、声发射等传感器相比,功率传感器具有实用性强、对加工过程无影响等优点.针对采集到的功率信号,在分析信号特征相关性的基础上,提出了一个多目标优化RP-SBL的刀具磨损量预测方法.对信号特征进行后处理(Re-processing,RP)消除电网波动和切削中其他偶然因素的影响,进一步提高特征对刀具磨损敏感性.基于处理后的特征,运用稀疏贝叶斯学习(Sparse Bayesian Learning,SBL)方法建立刀具磨损量预测模型.此外,使用非支配排序遗传算法(Non-dominated Sorting Genetic Algorithm II,NSGA-II)对SBL模型相关参数进行优化以提高预测精度.实验研究表明,该方法能够实现刀具磨损量的准确预测.不同预测方法的对比表明,通过特征后处理提高信号特征对刀具磨损的敏感性,保证了刀具磨损量的准确预测,对SBL模型参数进行优化可进一步提高预测精度,减小预测误差的最大值.
[Abstract]:Compared with cutting force and acoustic emission sensors, the power sensor is more practical and has no influence on machining process. On the basis of analyzing the correlation of signal characteristics, a multi-objective optimized RP-SBL tool wear prediction method is proposed for the collected power signals. The signal features are processed by Re-processing RPs) to eliminate the influence of power grid fluctuation and other accidental factors in cutting, and to further improve the sensitivity of features to tool wear. Based on the processed features, a prediction model of tool wear is established by using sparse Bayesian learning Bayesian learning (Sparse Bayesian learning SBL) method. In addition, non-dominated sorting genetic Algorithm (NSGA-IIA) is used to optimize the parameters of SBLs to improve the prediction accuracy. Experimental results show that the method can accurately predict tool wear. The comparison of different prediction methods shows that the sensitivity of signal features to tool wear is improved by feature post-processing, and the accurate prediction of tool wear is ensured, and the prediction accuracy can be further improved by optimizing the parameters of SBL model. Reduce the maximum prediction error.
【作者单位】：同济大学中德工程学院;同济大学机械与能源工程学院;温州大学机电工程学院;
【基金】：国家自然科学基金(71471139) 国家国际科技合作专项资助(2012DFG72210) 浙江省自然科学基金(LY14E050020)
【分类号】：TG71

【参考文献】