The Detection On Water Supply Pipeline Leak Based On Hilbert-Huang Transformation

Water is the source of life. It’s the precious resource which is endowed by the nature, but it is not inexhaustible. With the development of industry and the growth of population, a great deal of water resource suffers from serious pollution and destruction. However, with the development of social economy, human’s demand of water resources is increasing. Meanwhile, as the water supply enterprise developing, the problem of underground water pipeline leakage is very serious, which not only increases the water supply in vain, but also causes large economic loss and severe waste of resources. Therefore, strengthening the research of water pipeline leak detection technology and taking advanced detection methods to improve the management level of pipeline monitoring are of great significance.Firstly, this paper discusses and does research on the theory of water supply pipeline leak detection. We analyze the main leaking sound source and its generating mechanism, and then focus on the leak sound caused by pipeline vibration on the subsequent analysis work. We use power spectrum density (PSD) and Hilbert-Huang transformation (HHT) to analyze the frequency characteristics of vibration signal and extract its typical frequency characteristics, then applying them to leak detection. Moreover, we extract the normalized energy of IMF components, the singular values of Hilbert spectrum, and combine them with BP neural network for recognition of leakage types.In this paper, we present a set of software and hardware combination scheme of water supply pipeline leak detection and alarm bases on MATLAB GUI, vibration sensor, acquisition card etc. peripheral equipments, In this scheme, the collection, display, processing of leakage source signal are integrating in GUI monitoring platform, which makes the realization of pipeline leak monitoring and detecting easy and convenient. On this basis, we put the sorting technique combining BP neural network with the normalized energy of intrinsic mode function (IMF) components into the presented scheme to recognize the type of leakage.We collect a number of signals under different leakage types by experiment, then we use them to validate and analyze the presented scheme. It’s confirmed in this paper that the detecting technique based on signal frequency characteristics has high leak detection accuracy and a certain practical application value. Besides, the classification algorithm combining BP neural network with the normalized energy of IMF components has higher classification accuracy and can be used for classification recognition of leakage types.