基于St-DPMM的带式输送机托辊轴承声信号故障预警方法
Acoustic Signal Fault Warning Method for Belt Conveyor Idler Bearings Based on St-DPMM
孔华永 1高静 1柳跃 1侯继洁2
作者信息
- 1. 国能数智科技开发(北京)有限公司,北京 100011
- 2. 北京化工大学 机电工程学院,北京 100029
- 折叠
摘要
针对带式输送机托辊轴承运行过程中受外部冲击复杂多变,其声信号由多个激励源信号叠加而成,受噪声干扰严重,采用传统的单一特征预警方法难以有效实现故障预警的问题,提出一种基于狄利克雷过程t分布混合模型(St-DPMM)的带式输送机托辊轴承故障预警方法.首先,采集托辊轴承声信号,提取时频域特征构建高维特征空间;其次,训练St-DPMM拟合托辊轴承声信号的统计分布,利用KL散度逼近方法计算基准混合模型与正常状态模型间的距离;最后,基于3σ准则自学习预警阈值,计算基准模型与实时模型间的差异度,并与预警阈值比较实现故障预警.试验结果表明,所提方法预警的准确率、稳定性和时效性较对比方法有明显优势,能够有效实现带式输送机托辊轴承的故障预警.
Abstract
A fault warning method for belt conveyor idler bearings based on Student's t-distribution Dirichlet process mixture model(St-DPMM)is proposed to address the problem of complex and variable external impacts during operation of the bearings.The acoustic signal is superimposed by multiple excitation source signals,which is seriously disturbed by noise.The traditional single feature warning methods are difficult to effectively realize the fault warning.Firstly,the acoustic signal of the bearings is collected,and the time-frequency domain features are extracted to construct a high-dimensional feature space.Secondly,the St-DPMM is trained to fit the statistical distribution of acoustic signal of the bearings,and the distance between benchmark mixture model and normal state model is calculated by using KL divergence approximation method.Finally,based on 3σ criterion,the self-learning warning threshold is used to calculate the difference between benchmark model and real-time model,and the fault warning is realized by comparing with warning threshold.The test results show that the accuracy,stability and timeliness of the proposed method have significant advantages over comparative methods,and can effectively realize the fault warning of the bearings.
关键词
滚动轴承/带式输送机/预警/狄利克雷分布/信号处理Key words
rolling bearing/belt conveyor/early warning/Dirichlet distribution/signal processing引用本文复制引用
出版年
2025
NETL
NSTL
万方数据