Research On The Leakage Diagnosis For Heating Network

Pipeline transportation as a common mode of transport,in the oil,gas,heating,water supply and so has a unique advantage.At present,China's rapid urbanization,urban pipe network staggered parallel,old and new pipe fittings,the original heating network is mostly buried directly.Heat supply pipe network is difficult to meet the safety requirements,can not do regular investigation,it is very easy to burst the pipe,resulting in a great waste of energy and resources.On the basis of data mining,signal processing,statistics and other modern tools,this paper studies the leakage diagnosis of heating network.First,design and build an experimental platform on the basis of the characteristics of heating network,it has L2,L3,L4,L5 of 4 leaky pipes,this paper selects node pressure signal as the research object.To solve the problem of pressure signal data cleaning,it is proved that the linear interpolation and particle swarm diagnosis can effectively solve the problem of data loss and data anomaly.Aiming at the noise problem of pressure signal data,the source of signal noise is analyzed,and the discrete wavelet transform(DWT)is used to verify the denoising effect of three thresholds,including Fixed form threshold,Heuristic SURE and Minimax,in this paper,the common wavelet function Haar wavelet,db8 wavelet,Sym8 wavelet,coif5 wavelet and Bior6.8 wavelet are analyzed,and the denoising effect law of the commonly used wavelet functions is obtained.Secondly,aiming at the problem of pressure signal mixing,the EMD method is used to decompose the original pressure signal and study the single basic modal component IMF.It is verified that the EMD method effectively solves the problem of signal mixing.Aiming at the problem of modal aliasing in EMD decomposition,EEMD method is introduced to optimize it.EMD and EEMD are used to simulate the pressure signal.It is proved that EEMD method can effectively solve modal aliasing problem.The sample entropy algorithm is introduced to optimize the sample entropy parameter,like modal dimension,similarity tolerance and data volume,and get the change rule of the sample entropy calculation affected by the parameters,and solve the problem of parameter selection.Using a sample extraction method based on EEMD entropy,sample entropy of each node basic modal component of IMF is calculated,analyzed the internal factors of entropy change,found that the IMF sample entropy mean leak adjacent nodes collection than other nodes,to the maximum value and the maximum time.Finally,taking the entropy mean of the IMF samples of each node as the signal characteristics,we use the typical method of K-Means algorithm to identify the leakage pipes and leak points.in order to make up for the difficulty and high error of diagnosis and identification of leakage diagnosis of pipe network,a leak diagnosis process(Step1~Step6)based on data mining is proposed in this paper.At the same time,40 leakage experiments were carried out for 4 leak points.The results showed that the average accuracy of leak point diagnosis was 75%,and the accuracy of partial leakage diagnosis was more than 80%.In a comprehensive analysis,it is feasible to use the idea of data mining to solve the problem of leakage diagnosis of heating network.