摘要
钻头工况的实时自动监测有助于提高钻削加工过程的可靠性。针对钻头磨损在线监测,提出基于钻头工作声音信号累积功率谱的失效监测法。采用驻极体声电转换器采集声音信号,根据钻头磨损的慢变性,提出基于累积功率谱提取能量特征集的方案,并使用BIF特征选择结合Fisher准则筛选最优特征集,解决特征数量较多的问题。最后,利用二分类逻辑回归实现特征集与磨损量之间的数学建模,以h函数值作为失效判断的依据。结果表明:系统在钻头磨损严重并接近失效时,计算失效概率值>0.7,近似等于真实值,能为钻头更换决策提供可靠依据。
Real-time and automatic detection of drill＇s working conditions contribute to improve the reliability of drilling process. An approach for online drill wear monitoring was proposed according to the cumulative power spectrum of drill acoustic information. The acoustic information was acquired by an electret microphone. The optimal feature set was screened by BIF feature selection and Fisher criterion to minimize the number of features. Furthermore, a mathematical model for feature set and wear amount was created by binary logistic regression and the function h was used as the criterion for failure determination. The study has indicated that the failure probability is higher than 0.7 and approximately equals to the true value when the drill is worn heavily and almost out of service. This approach mentioned above can provide a reliable basis for drill replacement.
出处
《中国测试》
CAS
北大核心
2016年第2期111-114,共4页
China Measurement & Test
基金
桂林市科技攻关项目(LD14042E)
广西重点学科重点实验室项目(LD12047B)
