基于FCM和CNN-BiLSTM-MHA的矿用带式输送机健康状态评估  

Mining Belt Conveyor Health State Assessment Based on FCM and CNN-BiLSTM-MHA

作　　者：孙琪雅 袁逸萍[1] 张润泽 陈彩凤 SUN Qiya;YUAN Yiping;ZHANG Runze;CHEN Caifeng(Intelligent Manufacturing Modern Industry College,Xinjiang University,Urumqi Xinjiang 830017,China)

机构地区：[1]新疆大学智能制造现代产业学院,新疆乌鲁木齐830017

出　　处：《机床与液压》2025年第7期201-206,共6页Machine Tool & Hydraulics

基　　金：国家自然科学基金项目(72361032)。

摘　　要：受频繁启停机、负载突变等影响,带式输送机监测数据存在大量噪声、异常值和空值等,从而无法准确表征其运行状态。提出一种基于FCM聚类算法和CNN-BiLSTM-MHA模型的健康状态评估方法。对采集到的多传感器数据,利用动态时间规整进行预处理,采取自适应特征融合方法将降维后的健康指标进行融合;利用FCM聚类分析设备全生命周期退化数据,划分其健康状态;将划分好健康状态的数据输入CNN-BiLSTM-MHA模型进行训练,得到最终的健康状态评估结果。实验结果表明:与CNN和CNN-BiLSTM模型相比,CNN-BiLSTM-MHA模型在准确率、精确率、召回率和F1分数这4个评价指标上表现更优。Due to the effects of frequent startups and shutdowns,sudden load changes,etc.,the monitoring data of the belt conveyor contains a large amount of noise,outliers,and null values,which makes it difficult to accurately characterize its operating state.A health state assessment method based on FCM clustering algorithm and CNN-BiLSTM-MHA model was proposed.The collected multi-sensor data were preprocessed using dynamic time warping(DTW),and the adaptive feature fusion method was adopted to fuse the dimensionality-reduced health indicators.FCM clustering was used to analyze the whole life cycle degradation data of the equipment to classify its health state.Finally,the data with the classified health state were input into the CNN-BiLSTM-MHA model for training,and the final health state assessment results were obtained.The experimental results show that the CNN-BiLSTM-MHA model performs better in the four evaluation metrics of accuracy,precision,recall,and F1 score compared with CNN and CNN-BiLSTM models.

关 键 词：关矿用带式输送机 健康状态评估 多传感器融合 模糊C均值聚类 CNN-BiLSTM-MHA 

分 类 号：TD614[矿业工程—矿山机电]