Abstract:Water supply pipeline leakage not only wastes resources but also causes dangerous accidents. Therefore, detecting the state of pipelines is a critical task. With the expansion of the scale of the water supply pipeline, the amount of data collected by the leak detection system is gradually increasing. Moreover, there is an imbalance of sample in the data. This makes the detection performance of traditional leakage detection methods deteriorate. To solve the above issues, this paper proposes a pipeline leakage intelligent detection method based on a support vector weighted twin-bound support vector machine (SV-WTBSVM). Noise in the data negatively affects the performance of the classifier. To eliminate the effect of noise, a hybrid denoising algorithm based on improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) is used for denoising to filter out the noise in the data. Twin bound support vector machine (TBSVM) is a classical classification algorithm that has been widely used in the detection of pipeline leakage. To solve the decrease in classification accuracy caused by sample imbalance, the SV-WTBSVM algorithm oversamples the minority class samples based on the distance density and integrally undersamples the majority class samples to obtain a balanced sample. Since pipelines often have multiple working conditions, the SV-WTBSVM used for binary classification cannot meet this requirement, and this paper combines the SV-WTBSVM with the "one-to-one" strategy to address the multi-classification problem. Finally, experiments have verified that the SV-WTBSVM algorithm not only retains the advantages of fast training speed and simple operation of the TBSVM but also improves the classification accuracy and generalization ability of the algorithm when dealing with imbalanced data.

Leak Location of Pipelines Based on IPSO-SVM

Leak location of pipelines based on transient model and PSO-SVM

In-Situ Leak Detection in Urban Water Supply Pipelines Using Flexible Vibration Sensors

Sparse Variational Bayesian Inference for Water Pipeline Systems with Parameter Uncertainties

Leak Location of Pipelines Based on Characteristic Entropy

Compressive Sensing-Based Multiple-Leak Identification for Smart Water Supply Systems

Application of support vector machine learning to leak detection and location in pipelines

An improved variational mode decomposition method based on particle swarm optimization for leak detection of liquid pipelines

An Intelligent Water Supply Pipeline Leakage Detection Method Based on SV-WTBSVM

An SAE-based resampling SVM ensemble learning paradigm for pipeline leakage detection

Leak Identification Method of Water Supply Pipeline Based on Compressed Sensing and Least Squares Twin Support Vector Machine

Gas Pipeline Leakage Detection Method Based on IUPLCD and GS-TBSVM

Research on Multipoint Leak Location of Gas Pipeline Based on Variational Mode Decomposition and Relative Entropy

A New Leak Detection Method for Subsea Pipelines

A Calculation Model for Leak Detection and Location of Single Pipeline with Two Leaks

The Method for Leakage Detection of Urban Natural Gas Pipeline Based on the Improved ITA and ALO

Novel Leakage Detection Method by Improved Adaptive Filtering and Pattern Recognition Based on Acoustic Waves

Pipeline Leak Detection Combining Machine Learning, Data Assimilation Approaches and Pipeline Fluid Flow Physics Models

A Study of Multi-Method Based Subsea Pipeline Leak Detection System

An Inverse Transient Nonmetallic Pipeline Leakage Diagnosis Method Based on Markov Quantitative Judgment

An improved pipeline leak detection and localization method based on compressed sensing and event-triggered particle filter