摘要
针对水轮发电机碳刷温度传感器因长期运行引发的温漂故障诊断难题,提出一种基于DBSCAN密度聚类与多传感器协同校验的故障诊断方法;通过对比K-means、层次聚类、孤立森林与DBSCAN算法在三维特征空间(转速、电流密度、温度)中的误判率、计算效率以及算法参数和鲁棒性,验证了DBSCAN算法在处理三维数据的适用性(误判率低于18.7%、计算效率达0.41 s/‰);采用小波变换对历史工况数据降噪处理,结合动态阈值预警机制与邻近传感器协同校验,实现了对传感器缓变偏移的精准诊断;实验模拟0.5℃/h温度偏移及±10℃噪声干扰条件下,故障簇均值差异达12.23℃,超出阈值触发预警,并通过停机标定验证诊断准确性;该方法解决了传统方法对缓变故障敏感度不足的缺陷,经实际测试满足水轮发电机复杂工况下的实时监测需求,为工业设备智能化状态监测提供了潜在可行的技术方案。
To address the challenge of diagnosing temperature drift faults caused by a long-term operation in hydro-generator brush temperature sensors,a fault diagnosis method based on the density-based spatial clustering of applications with noise(DBSCAN)density clustering and multi-sensor collaborative verification is proposed.Compared with the misjudgment rates,computational efficiency,algorithm parameters,and robustness of K-means,hierarchical clustering,isolation forest,and DBSCAN algorithms in a three-dimensional feature space(rotational speed,current density,temperature),the DBSCAN algorithm verifies the applicability of processing the 3D data(the misjudgment rate below 18.7%,the computational efficiency up to 0.41 s/‰).the wavelet transform is used to denoise the historical operational data.A dynamic threshold early-warning mechanism is combined with a collaborative verification from adjacent sensors to achieve the precise diagnosis of gradual sensor offset faults.Under the experimental simulation conditions of the temperature drift of 0.5℃/h and noise interference of±10℃,the mean difference between fault clusters reaches up to 12.23℃,exceeding the threshold to trigger a warning.The accuracy of the diagnosis is verified through shutdown calibration.This method overcomes the insufficient sensitivity of the traditional approach to gradual faults,meets the real-time monitoring requirement of hydro-generators under complex operating conditions by on-site testing,and provides a potential feasible technical solution for the intelligent detection of industrial equipment.
作者
李若松
王金鹏
周军长
LI Ruosong;WANG Jinpeng;ZHOU Junzhang(Eastern Electric Group Dongfang Electric Co.,Ltd.,Deyang 618000,China)
出处
《计算机测量与控制》
2025年第8期102-111,共10页
Computer Measurement & Control
基金
企业科技项目(P0202302230106)。
关键词
碳刷温度
数据偏移故障
DBSCAN密度聚类算法
多传感器协同校验
设备智能化
brush temperature
data drift fault
DBSCAN density clustering algorithm
multi-sensor collaborative verification
equipment intelligence
