Considering the importance of monitoring piping systems, the aim of this work is to develop a technique to detect gas leaks in pipes, based on a acoustic method, in order to determining the magnitude of leaks through the use of neural artificial networks. Audible noise, generated by leaks, was detected in a pipe 100 m long. The experimental data, obtained through a microphone installed inside the pressure vessel and connected to a data acquisition system, are decomposed into sounds of different frequencies. The audible noise was analyzed generated by the occurrence of gas leaks in a flexible pipe. This hose has a diameter of 0.5 in and 100 m of length. Using a continuous feed of gas, the hose was fed through a pressure vessel. Compressed air was used as working fluid. The operating pressure was monitored by a manometer installed at the inlet pipe. The maximum pressure used in the line is 7 kgf/cm2. The leaks were triggered manually through a open/close quick valve installed in the hose. The magnitude or size of the leakages was controlled by the installation of the holes at 0 m and 50 m from the beginning of the pipe, with diameters ranging between 1.0 and 3.0 mm. The monitoring of leaks was performed through a microphone installed inside the pressure vessel, connected to a microcomputer. These noises were processed in a filter bank, resulting in three signals of voltage with frequency bands chosen from 1 kHz, 5 kHz and 9 kHz.The dynamics of these noises in time is used as input to the neural model to determine the occurrence and magnitude of the leaks. The experimental results showed that it is possible to detect leaks in gas pipelines based on the acoustic method, since there was a sharp increase in the amplitude of noise in the presence of the leak. It was also possible to analyse the occurrence and determine the magnitude of the leak through neural networks, because the models successfully predicted the leaks of diameters 1 mm, 2 mm and 3 mm for a data set unseen in the training procedure. Absolute errors of less than 10 % in size of leakage were observed, what was considered acceptable. Different operating pressures are under study for broadening the model application. The tests using different leak positions at the hose are being conducted as well for the development of a secondary neural model to act in the location of the leak. Preliminary results indicate the potential of this new model, which would be activated only when there was a leak detected by the neural model already developed. A general subroutine for the neural model calculations is under development in C language, in order to provide online monitoring for anysize-leakages in gas pipelines, warning the operator on the occurrence of leakages.


    Access

    Access via TIB

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    A neural model to detect and determine the magnitude of leaks in gas pipelines


    Additional title:

    Ein neuronales Modell zum Entdecken und Bestimmen der Größe von Lecks in Gasppelines




    Publication date :

    2011


    Size :

    6 Seiten, 10 Bilder, 11 Quellen



    Type of media :

    Conference paper


    Type of material :

    Print


    Language :

    English