面向发电设备预测性维护的传感数据特征抽取方法  

A Lagged Correlation Oriented Feature Extraction Method for Power Equipment Predictive Maintenance

作　　者：张守利[1,2,3] 苏申 刘晨[2,3] 韩燕波 

机构地区：[1]天津大学计算机科学与技术学院,天津300072 [2]北方工业大学大规模流数据集成与分析技术北京市重点实验室,北京100144 [3]北方工业大学计算机学院,北京100144

出　　处：《太原理工大学学报》2018年第1期79-85,共7页Journal of Taiyuan University of Technology

基　　金：国家自然科学基金面上资助项目(61672042);北京市属高等学校创新团队建设与教师职业发展计划项目(IDHT20130502);北京市自然基金重点项目(4131001)

摘　　要：针对发电厂海量传感器数据,基于相关系数最大化的曲线排齐算法,提出了一种面向多传感序列的特征抽取方法,此方在尽量避免信息丢失的前提下,实现高维传感数据的特征抽取,建立轻量级预测模型。通过实验表明,本文方法能够在保证预测准确度的前提下,减少预测模型的训练时间,降低模型的计算开销。In this paper,a feature extraction method was proposed to solve the problem of lagged correlation between multi-sensor sequences based on the curve registration model of maximizing the correlation coefficient.The dimensionality reduction is realized in the premise of avoiding the loss of information as far as possible and then a lightweight predictive model is built.The experiment results indicate that our method can effectively reduce training time and lower the computational cost of predictive model in the premise of ensuring the accuracy of prediction.

关 键 词：预测性维护 特征提取 延迟相关 轻量级模型 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]