基于数据驱动的矿井主排水设备寿命预测方法

发布时间：2018-03-21 00:04

本文选题：离心泵　切入点：寿命预测　出处：《工矿自动化》2017年11期 　论文类型：期刊论文

【摘要】：以矿井排水系统的离心泵为研究对象,介绍了包括机器学习方法、多元统计分析方法、特征量提取方法和信息融合方法在内的4种基于数据驱动的矿井主排水设备寿命预测方法的基本原理、相关案例、优缺点、尚未解决的问题及其在离心泵寿命预测中的应用;指出了离心泵寿命预测的发展趋势:寿命衰退指标应多样化,只有离心泵的各类指标正常,才能表明离心泵运行正常,多变量综合考虑使预测可靠性更高;决策层信息应高度融合,振动信号、动态摩擦力矩、扬程等因素都会随着寿命的衰退发生一定的变化,将这些信息融合用于寿命预测,效果会更好;融合特征层信息,将多种预测模型进行融合,或者建立一个集更多优点于一体的混合模型,才能更好地满足工业要求。
[Abstract]:Taking the centrifugal pump of mine drainage system as the research object, this paper introduces the methods of machine learning and multivariate statistical analysis. The basic principle, relevant cases, advantages and disadvantages of four data-driven life prediction methods for mine main drainage equipment, including feature extraction method and information fusion method, as well as the unsolved problems and their application in the life prediction of centrifugal pumps; The development trend of life prediction of centrifugal pump is pointed out: the index of life decline should be diversified, only if all kinds of indexes of centrifugal pump are normal, can the operation of centrifugal pump be normal, and the reliability of prediction can be improved by multivariable comprehensive consideration. Decision level information should be highly fused, vibration signal, dynamic friction torque, lift and other factors will change with the decline of life. Only by merging multiple prediction models or establishing a hybrid model with more advantages can we better meet the requirements of industry.
【作者单位】：太原理工大学电气与动力工程学院;煤矿电气设备与智能控制山西省重点实验室;矿用智能电器技术国家地方联合工程实验室;
【基金】：山西省科技重大专项项目(20131101029)
【分类号】：TD442

【相似文献】