摘要
针对实际工程系统故障建模困难、现有故障预报方法实时性差的问题,从一类挖掘的角度,设计了一种基于一类支持向量机的时间序列暂态挖掘算法,提出了一种既不需要系统近似模型也不需要故障训练数据和先验知识的直接故障预报方法。在系统运行的同时实现学习和预报,提高了实时性。同时该方法简单易用,克服了传统方法在预报故障前需要预测系统未来状态的缺点。具有很强的应用意义。以釜式反应器为对象进行的仿真实验证明了方法的有效性。
Aiming at the difficulty of fault modeling for complex engineering systems and the poor real time capability of existing method for fault prediction, an one class mining algorithm for temporal pattern in time series is designed using one class support vector machines. Based on this mining algorithm, a new kind of realtime fauh prediction method is presented, which needs neither the model to approximate the true system nor the fault training data and prior knowledge. It can learn and predict while system＇s running, so that it can improve the real time capability. Besides. this method is simple and universalizable, and it can overcome the disadvan rage of the conventional method to predict the future state of systems before forecasting fault, so it is provided with strong practical significance. The results of simulation on continuous stirred tank reactor （CSTR）prove the effectiveness of the proposed approach.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2008年第10期2023-2027,共5页
Systems Engineering and Electronics
基金
江苏省博士后基金资助课题(072008C)
关键词
故障预报
数据挖掘
支持向量机
暂态:非线性
时间序列
fault prediction
data mining
support vector machine
temporal pattern
nonlinear
time series
