一种基于证据理论的多源信息融合管道泄漏诊断

摘要

Because of influence by factors of the strong background noise, complex construction conditions and flaws of sensor itself, unitary detection method of pipeline leakage has the problem of low exactitude and uncertainty of recognition. Through incorporating technology of wireless sensor network and data fusion, a novel multi-sensor fusion leakage diagnosis algorithm based on Dempster-Shafer (D-S) evidence theory was proposed. In the method, diagnosis information of multi-period measurement of multi-sensor of multi-node in a cluster was used as independent evidence respectively by relying on the model of clustered network structure. By applying distributed data fusion approach and modified evidence combination rule, the temporal of a single sensor, spatial data of a single node and that of multi-node were aggregated step by step, and finally the fusion result was obtained according to decision rule. The experimental result shows that proposed method improves diagnosis of pipeline leakage exactitude effectively and decreases the recognition uncertainty markedly.%受强背景噪声、复杂工况以及传感器自身缺陷等因素的影响,单传感器管道泄漏检测方法存在诊断精度差和识别不确定性的问题.结合无线传感器网络技术和信息融合技术,提出一种基于D-S证据理论的多传感器数据融合的泄漏诊断算法.该方法依据分簇的网络结构模型,分别将簇内不同节点上多种传感器的多个测量周期的诊断信息作为独立证据体,采用分布式数据融合结构和修正证据合成法则,逐级进行单传感器时域、单节点空域以及多节点空域的融合,最后通过决策法则输出结果.实验结果表明,该方法降低了识别的不确定性,有效提高了管道泄漏诊断的精度.