浆体管道输送高压隔膜泵关键部件故障特征提取及诊断方法研究

Reciprocating high-pressure diaphragm pump is the core power of slurry pipeline transportation equipment. The state monitoring, fault diagnosis and production efficiency are closely related to safety. This project is to study the key components of large-scale reciprocating high-pressure diaphragm pump, and the relative critical technologies such as multi-source information fusion, feature extraction and fault diagnosis are developed. Firstly, count and analyze the historical damage types and damage state of the key components of the high pressure diaphragm pump, and study the "inner-class and between-class" multi-source information fusion structure and fusion method in order to provide reliable data sources for further analysis. The adaptive lifting wavelet and two-dimensional local wave decomposition method are introduced to solve the feature extraction problem of nonlinear and unsteady time-varying complex signals under complex operating conditions. Combined with multi-core learning and cost-sensitive evaluation mechanism to establish multi-kernel cost sensitive fault diagnosis model. Introduce the D-S evidence theory to realize the organic fusion and reasoning of the multi-fault diagnosis model, and further improve the reliability and generalization ability of the fault diagnosis model. And finally get the key components of high-pressure diaphragm pump fault diagnosis of new technologies and new devices, which provide reliable theoretical guarantee and technical support for status monitoring and fault diagnosis of large-scale reciprocating diaphragm pump. At the same time, it has important economic significance and theoretical research value.

往复式高压隔膜泵是浆体管道输送的核心动力设备，其状态监控、故障诊断与生产效率和安全息息相关。本项目以大型往复式高压隔膜泵关键部件为研究对象，开展多源信息融合、特征提取及故障诊断等相关关键技术的研究。首先对高压隔膜泵关键部件的历史损伤类型及损伤状态表征等进行统计与分析，并研究“类内与类间”多源信息融合结构及融合方法，为后续分析提供可靠数据源；引入自适应提升小波和二维局域波分解方法，解决复杂运行工况下非线性、非稳态时变复杂信号的特征提取难题；结合多核学习和代价敏感评估机制建立多核代价敏感故障诊断模型，并引入D-S证据理论实现多故障诊断模型的有机融合与推理，提高故障诊断模型的可靠性和泛化能力。研究成果将形成一套完整的高压隔膜泵关键部件故障诊断新理论与新技术，为大型往复式隔膜泵的状态监测和故障诊断提供理论及技术支持，具有重要经济意义及理论研究价值。

往复式高压隔膜泵是浆体管道输送的核心动力设备，其状态监控、故障诊断与生产效率和安全息息相关。本项目以大型往复式高压隔膜泵关键部件为研究对象，开展多源信息融合、特征提取及故障诊断等相关关键技术的研究。构建了集压力、振动、声音等于一体的高压隔膜泵关键部件数据采集系统，研发了多源信息融合自诊断装置，实现了多传感器数据多方面、多层次和多级别的分析与处理，为后续分析提供可靠数据源。提出了改进的局部均值分解、变分模态分解、多点最优最小熵反褶积等系列信号预处理方法，完成了改进的精细复合多尺度散布熵、循环相干谱、小波能量熵等故障信号敏感特征的有效提取，提高了对复杂运行工况下非线性、非平稳时变复杂信号的表征能力。根据所选不同状态表征量，引入动态核函数和代价敏感学习机制，建立了微分几何代价敏感诊断模型，实现了高压隔膜泵关键部件的故障诊断。同时引入D-S证据理论和弱学习器的集成思想，构建了基于BP_AdaBoost的故障诊断集成模型，实现了多故障诊断模型的有机融合与推理，提高了故障诊断模型的可靠性和泛化能力。基于所提取的敏感特征，进一步构建了能表征关键部件退化过程的健康指标，建立了融合广义回归神经网络的剩余寿命预测模型，实现了轴承等关键部件的状态监测及预测性维护。研究成果形成了一套完整的高压隔膜泵关键部件故障诊断与剩余寿命预测新理论与新技术，为大型往复式隔膜泵的状态监测、故障诊断和预测维护提供理论及技术支持，具有重要经济意义及理论研究价值。