截齿磨损程度的多特征信号融合识别研究  

Research on wear degree recognition of picks based on multi feature signal fusion

作　　者：张强[1,2,3] 刘志恒 王海舰[1] 张赫哲[1] ZHANG Qiang;LIU Zhi-heng;WANG Hai-jian;ZHANG He-zhe(College of Mechanical Engineering,Liaoning Technical University,Fuxin 123000,China;Coal Resource Safety Mining and Clean Utilization Engineering Research Center,Fuxin 123000,China;State Key Laboratory of Mineral Processing Science and Technology,Beijing 100160,China)

机构地区：[1]辽宁工程技术大学机械工程学院,辽宁阜新123000 [2]煤炭资源安全开采与洁净利用工程研究中心,辽宁阜新123000 [3]矿物加工科学与技术国家重点实验室,北京100160

出　　处：《工程设计学报》2018年第3期278-287,共10页Chinese Journal of Engineering Design

基　　金：国家自然科学基金资助项目(51504121);辽宁省自然科学基金资助项目(201601324);煤炭资源安全开采与洁净利用工程研究中心开放课题(LNTU16KF02)

摘　　要：为了解决采煤机开采过程中截齿磨损程度在线监测和状态识别的工程难题,提出一种基于多特征信号融合的截齿磨损程度识别方法。搭建截齿磨损程度监测实验台,分别测试提取不同磨损程度截齿截割过程中的振动加速度信号、声发射信号、红外热像信号和电机电流信号,建立了截齿截割的多传感信号数据样本库;针对数据样本库中两相邻磨损状态截齿特征样本存在数据交集、系统识别精度低的问题,构建最小模糊度优化模型并计算各特征信号的最优模糊隶属度函数,获取特征样本最大隶属度。构建截齿磨损程度的神经网络识别模型,运用多特征数据样本对Back-Propagation(BP)神经网络进行学习和训练。实验结果表明:BP神经网络识别模型的识别结果和试样的实际磨损程度类别相同,此识别模型能够对截齿磨损程度类型进行实时监测和准确识别。研究结果为实际工程中截齿监测和更换提供了解决方案。In order to solve the engineering problems of on-line monitoring and state recognition for the wear degree of picks during mining process , a method to recognize the wear degree of picks based on multi feature signal fusion was proposed . An experimental platform for recognizing the wear degree of picks was set up and the vibration acceleration signal acoustic e-mission signal infrared thermal signal and motor current signal in cutting process were extracted and tested respectively .A sample library of multi-sensor data for picks cutting was established . Aiming at the problems of existing data intersection between two adjacent samples of wear states,which reduced the system recognition accuracy,the minimum fuzzyness optimization model was established to calculate the optimal fuzzy membership function of each characteristic signal and the maximum membership degree of feature samples were obtained .A neural network identification model for different wear degree of picks was constructed . The Back-Propagation （BP） neural network was studied and trained by using multi feature data samples . The experi-mental results showed that the BP network discriminant results of the recognition model were the type of wear degree of picks .The research results provide a solution for monitoring and repla- cing picks in practical engineering .

关 键 词：截齿 磨损程度 振动信号 声发射信号 红外热像信号 电机电流信号 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]