摘要
为实现泵站工程在生产运行过程中有效预测设备潜在故障风险,提升泵站设备运行效率,在数字孪生水利工程数据底板基础上,基于现有硬件设备,以结构故障机理为导向,提出了一种HMM-MLP的泵站设备故障预测方法。先由连续小波包变换处理设备的运行信号,然后通过HMM模型生成设备运行状态序列作为MLP网络的输入预测设备故障,最后通过仿真实验表明,HMM-MLP模型可有效提高泵站设备故障的预测准确率。同时,依托在线监测数据和离线检查与试验数据,建立了设备健康评价指标体系,并开发了泵站监测健康诊断系统,协助运行管理人员充分了解和掌握机组设备的“健康”状态,提升设备管理的信息化水平。结果表明:该研究可为泵站健康系统建设提供实际案例指导与经验启示。
To effectively predict potential equipment failures during the production and operation of pump station projects and enhance the operational efficiency of pump station equipment,this study proposes a HMM-MLP fault prediction method for pump station equipment based on the digital twin hydraulic engineering data platform and existing hardware equipment,and guided by the structural fault mecha⁃nism.First,continuous wavelet packet transform is used to process the equipment operation signals.Then,the HMM model is used to gener⁃ate the state sequences of devices′operation,serving as inputs to the MLP network to predict equipment failures.Finally,simulation experi⁃ments demonstrate that the HMM-MLP model significantly improves the accuracy of pump station equipment fault prediction.Additionally,leveraging online monitoring data and offline inspection and test data,a health assessment index system for equipment is established.And a pump station monitoring health diagnostic system is developed to assist operational managers in fully understanding and monitoring the“health”status of unit equipment,thereby enhancing the informatization level of equipment management.Practical results indicate that this research provides practical guidance and experiential insights for the construction of pump station health systems.
作者
匡正
袁志波
徐振磊
KUANG Zheng;YUAN Zhi-bo;XU Zhen-lei(Jiangsu Jiangdu water conservancy project management office,Yangzhou 225000,Jiangsu Province,China)
出处
《中国农村水利水电》
北大核心
2024年第7期255-261,269,共8页
China Rural Water and Hydropower
基金
江苏省水利科技项目(2023038)。
关键词
智慧泵站
信号处理
机器学习
隐马尔可夫模型
故障预测
intelligent pumping station
signal processing
machine learning
Hidden Markov Model
fault prediction
