多信息融合的城市燃气管道泄漏诊断技术研究

Research on multi-information fusion technique for diagnosis of urban gas pipeline leakage

为提高城市燃气管道泄漏检测能力,预防燃气泄漏事故,将信息融合技术应用于管道泄漏检测领域,并提出一种基于D-S证据理论与支持向量机(SVM)的多传感器信息融合诊断技术。在实验室条件下,采集管道不同位置和不同周期的声发射传感器信号,提取特征参数,构建SVM分类器,该分类器的输出结果经处理后作为基本概率指派(BPA)。然后,利用多层D-S理论,依次融合不同空域与时域传感器信息。最后,根据分类判决门限得出最终诊断结果。研究表明:多传感器信息融合使诊断不确定度明显下降,所属信度值显著提高,最终决策结果与实际一致;与单传感器诊断相比,多传感器信息融合后诊断技术具有更高的准确率和更好的稳定性。

Abstract ： For the sake of improving the ability to detect urban gas pipeline leak and preventing gas leak- age accidents, the information fusion technology was applied to the field of pipeline leak detection. A multi-sensor information fusion diagnosis method was worked out based on D-S evidence theory and SVM. Under laboratory conditions, acoustic emission sensor signals were collected at different positions and dif- ferent periods. Characteristic parameters were extracted. An SVM classifier was built. The processed clas- sifier outputs were taken as the values of BPA. Pieces of information from different sensors were fused ac- cording to multi-layer D-S theory. Finally, recognition results were given based on classification thresh- olds. The test results show that multiple sensor information fusion makes the uncertainty of diagnosis markedly decrease and its reliability increase significantly, that final decision conforms with reality, and that this diagnosis technique has higher accuracy and stability compared with single sensor diagnosis.