Research On Leak Detection Method Of Residential Gas System Based On Improved Negative Pressure Wave

Once the residential gas pipeline system leaks,the gas is more likely to accumulate in the building,causing a series of serious secondary disasters such as combustion,explosion and structural collapse.The existing method for detecting residential gas leakage mainly rely on combustible gas alarms,which has prominent problems such as high failure probability,short effective life,and passive monitoring methods.Indoor gas accidents occur frequently.This article proposes an improved negative pressure wave method and conducts interpretability verification to address the problem of single and outdated indoor gas leak detection methods.The theory of weak pressure feature estimation under complex strong noise background and the framework of deep auto-encoder leak detection based on semi-supervised anomaly diagnosis are established.The specific research is as follows:Firstly,based on the theoretical analysis and simulation of hydraulics,the improved theory of negative pressure wave method is discussed.Steady-state calculation models for leakage and gas usage of low-pressure gas pipelines are established,and the difference in steady-state characteristics mainly related to the orifice diameter and orifice flow coefficient.Based on the hydraulic transient theory,the transient models of pipeline flow and negative pressure wave of orifice outlet flow are established.The pressure transient processes under different conditions are simulated based on Flowmaster.The simulation results show that the transient fluctuation characteristics of the negative pressure wave of gas usage and leakage have certain differences under the same flow rate.Based on the potential difference between the steady-state and transient characteristics of the negative pressure wave of gas usage and leakage,the definition and criteria of the improved negative pressure wave method are proposed,and the research focus of the improved negative pressure wave method is clarified.Secondly,based on the key issues and verification requirements of the improved negative pressure wave method,the indoor gas system leakage simulation experimental system and the actual measurement system are built.The experimental system takes air as the gas source,simulates the residential gas pipeline system through multistage pressure regulation,and obtains the experimental data with high signal-to-noise ratio and balanced sample distribution.The actual measurement system selects a real residential gas user for building pressure regulation and obtains the measured data with low signal-to-noise ratio and uneven sample distribution.Through the analysis of the experimental and measured results,the fluctuation characteristics of the negative pressure wave and the differences between the samples under different conditions are explored to provide a direct basis for the verification research and application research of the improved negative pressure wave method.Thirdly,an interpretability verification of the improved negative pressure wave method is carried out.A negative pressure wave preprocessing method based on multi-scale linear fitting is proposed,and a static index system is established based on the fluctuation characteristics of different stages of negative pressure wave.Based on the principle of acoustic sensor,the negative pressure wave signal is converted into a dynamic sound pressure signal to realize the enhancement of pressure transient oscillation,and the dynamic pressure enhancement amplitude index is proposed.Based on the interpretable two-dimensional feature index system,the supervised leak detection model of limit learning machine is established.Based on the experimental data,the leakage detection success rate can reach 91.33%,which verifies the feasibility of the improved negative pressure wave method.The differences in the fluctuation characteristics of negative pressure waves between gas usgae and leakage can support the identification of leakage to a certain extent.Then,based on the measured negative pressure wave signals,the noise components,sources,characteristics and classifications of the residential gas system are sorted out.The filtering and evaluation methods of baseline noise of pipeline network are discussed,and a robust detection method of low-frequency noise period in complex background based on the black top-hat autocorrelation average amplitude difference transform is established.Combined with the simulation signal,the superiority of the robust Kalman filter in estimating the objective fluctuation state of the measured negative pressure wave is demonstrated.Based on the discrete scale theory,an improved variational modal decomposition method is proposed,and the adaptive calculation method of modal parameters and empirical calculation formula of quadratic penalty factor are established,which effectively solves the problem of complex noise modal aliasing.The adaptive reconstruction method based on improved Fourier series is discussed,and the accurate estimation of the real state of negative pressure wave is realized.The reconstruction result based on the simulated signal shows that the Spearman rank correlation coefficient between the reconstructed waveform and the real waveform is 0.9834.Finally,a deep auto-encoder fluctuation feature learning framework based on semi-supervised anomaly detection is proposed,which only needs the participation of the gas usage samples to avoid the dependence of the improved negative pressure wave method on the leakage samples.Considering the high-dimensional and dense characteristics of negative pressure waves,a 7-layer sequence-to-sequence deep autoencoder model is established based on long short-term memory neural network and Bahdanau attention mechanism.The number of iterations required for the reconstruction of the model to achieve stable convergence is about 20,and the reconstruction error is as low as 5.1563 10^-6.Based on the measured gas usage data,the superiority of depth auto-encoder model to reconstruct negative pressure wave time series data is demonstrated.By analyzing the distribution characteristics of the error data reconstructed by the negative pressure wave of gas usage,a self-driven anomaly diagnosis method based on Tukey rule data is established,and experts are also allowed to make decisions based on the improved Grubbs test method.The success rate of comprehensive leakage detection can reach 73.3%,realizing semi-supervised intelligent detection of leakage conditions.The research has enriched the leak detection methods for residential gas systems,provided useful supplements to existing leak detection methods,opened up new ideas for negative pressure wave leak detection,and provided theoretical references for the application of artificial intelligence and software leak detection methods in residential gas systems.