基于D-S的决策级信息融合的故障诊断.docVIP

 基于 D-S 的决策级信息融合的故障诊断 王璇1，周丰2* （1. 嘉兴学院南湖学院，浙江 嘉兴 314000； 5 10 2. 嘉兴学院机电工程学院，浙江 嘉兴 314000） 摘要：为保证石油钻井过程安全可靠运行，对其故障诊断方法的研究具有重大意义。本文在 国内外故障诊断技术的基础上，提出了基于传感器间相互支持的系数矩阵的 D-S 证据理论的 决策级融合诊断方法。传统证据理论的故障诊断方法未考虑证据的权重不同，会导致诊断精 度下降甚至出现误诊，针对这种现象，引入了传感器间相互支持的系数矩阵。采用基于正态 分布的算法代替典型样本算法来获得基本概率赋值，以解决随目标模式增多融合后的特征结 果快速弱化问题。再与传统 D-S 证据理论相结合。实验室数据仿真表明，该方法可以对石油 钻井的故障进行准确的诊断，而且计算速度快，实时性强，具有实际应用价值。 关键词：石油钻井；故障诊断；D-S 证据；信息融合 中图分类号：TP206.3 15 Based on D-S decision level information fusion and its application in the fault diagnosis WANG Xuan1, ZHOU Feng2 (1. College of NanHu,Jiaxing University, ZheJiang JiaXing 314000; 20 25 30 35 40 2. College of Mechanical and Electrical ,Jiaxing University, ZheJiang JiaXing 314000) Abstract: In order to ensure the safe and reliable operation of the petroleum drilling process, the method of fault diagnosis is of great significance. In this paper, the fault diagnosis technology at home and abroad on the basis of, is put forward based on the sensor support each of the coefficient matrix D-S evidence theory decision level fusion diagnosis method. Traditional evidence theory fault diagnosis method does not take into account the weight of evidence is different, can lead to diagnostic accuracy decreased or even misdiagnosis, according to this phenomenon, introduced the sensor mutual support of the coefficient matrix. Based on the normal distribution of the algorithms instead of the typical sample algorithm to obtain basic probability assignment, to solve the target mode increase with the fused features results in rapid weakening. With traditional D-S theory of evidence combination. Laboratory data simulation show that, this method can of oil drilling fault for accurate diagnosis, and the calculation speed is fast, real-time, has practical value. Keywords: Oil drilling; Fault diagnosis; D-S evidence; information fusion 0 引言 石油钻井是一种连续的作业过程，一旦设备发生故障，将造成巨大的经济损失，而故障 诊断技术已经发展成为应用工程的一门综合学科，因此对石油钻井故障诊断方法的研究具有 重大的现实意义，以保证钻井过程安全可靠运行。 1 证据理论的基本概念 D-S 证据理论有 3 个重要的函