Machine learning Techniques Based Gas Pipeline Leakage Detection

The detection of leaks in natural gas pipelines is of paramount importance for ensuring safety, environmental protection, and economic viability of gas transmission systems. Traditional leak detection methods often rely on periodic inspections which may not provide real-time monitoring or sufficient coverage, particularly in remote or inaccessible areas.Recently, there has been a significant uptick in interest regarding the utilization of acoustic signal analysis as a promising method for detecting and pinpointing pipeline leaks. In this paper we discuss a new methodology for diagnosing the gas pipeline leakage based on statical analysis of acoustic data. For this project we are using GPLA12_v3 dataset containing 24 different classification which are classified into 3 groups based on the pressure values. We have also used MFCC features for fault diagnosis which have produced a result of 66% accuracy for Gaussian Navie Bayes classifier. For statical features we got the best accuracy for Random Forest Classifier with an accuracy of 76%