基于数据挖掘的故障诊断算法

项目名称： 基于数据挖掘的故障诊断算法

项目编号： No.U1232115

项目类型： 联合基金项目

立项/批准年度： 2013

项目学科： 物理学II

项目作者： 敖新宇

作者单位： 中国科学院上海应用物理研究所

项目金额： 60万元

中文摘要： 重大系统的安全运营对国民经济建设意义巨大，而故障诊断是建立安全系统的核心支撑技术。而现在的系统日趋复杂，与之相关的数据也以指数的趋势在增长，为了对复杂系统进行全面而有效的监测，需要发展能够有效处理海量的传感器数据的故障诊断算法。在本项目中，试图对光源运行的物理数据库中通过数据挖掘算法发现和运维相关的数据模式，结合运行的实际经验，建立统计模型，提取发生故障的状态空间子集，确立一种故障原因的发现算法进而建立多层次的监控模型，并能够预测可能潜在的故障；并根据上海光源的运维实际需要，建立界面友好的基于Web的故障分析与预测综合平台来保证上海光源的更为顺畅的运行。

中文关键词： 时间序列；数据挖掘；经常模式联结；知识系统；

英文摘要： Important Systems'' security is vital to ecnomic development, and the fault detection and diagnosis (FDD) method is the basement for safety of systems. The structure of system tends to become more and more complex, also, the storeage of accumulated sensory data keeps increasing with exponential rate. To achieve an effective and fully monitoring to such a system, Huge amounts of data needed to be processed. In this project, we try to get useful patterns for fault detection and diagonsis through related data mining algorithm, which is related with the operation process; constructing their statistical model with the consideration of maintaining experiences. Then, the fault subspace is collected, and build relationship between the subspace and sensory input with the learning algorithm. Based on that, the fault detection algorithm is supposed to build to monitoring the possible incoming fault and isolating the fault which already happen. Also, by endowing the semantic meaning to the undering pattern, a user-friendly interface could be built, which will facilitate the better maintance and smoothing future operations.

英文关键词： Time series sequence；data mining；frequent pattern association；knowledge base；