Classification of the underground pipeline leakage signals by means of the artificial neural networks

Authors

  • Віктор Олександрович Строганов Sevastopol National Technical University Universitetskaya str, 33, Sevastopol, 99053, Ukraine

DOI:

https://doi.org/10.15587/1729-4061.2012.5680

Keywords:

Artificial neural network, classification, pipeline leaks

Abstract

One of the most promising approaches to finding leaks of underground pipelines is to analyze the acoustic signals generated by the pipeline with a leak. This signal can be recorded on the earth surface above the pipeline. The problem of finding a leak can be viewed as the classical problem of signal classification. Such a problem can be sloved with the artificial neural networks.

In this article we investigated the possibility of classifying the pipeline signals using artificial neural network. Feedforward network with one hidden layer was used. During the investigation we used experimentally obtained signals for pipeline with and without a leak. As a feature vectors we used Fourier spectrums of the signals. The effectivenes of two trainig methods was compared: Levenberg-Marquardt algorithm and "resilient backpropagation" (RPROP). We came to the conclusion that pipeline leakage signals can be classified with the  artificial neural networks. RPROP training method works better than  Levenberg-Marquardt algorithm. The main problem to be solved in the future works is proper network initialization which will improve the quality of network training

Author Biography

Віктор Олександрович Строганов, Sevastopol National Technical University Universitetskaya str, 33, Sevastopol, 99053

Department of Information systems

References

  1. Строганов В. А. Экспериментальное исследование сигналов утечек подземных трубопроводов/В.А. Строганов, В. Н. Хоролич// Вестник СевНТУ. Сер. Информатика, электроника, связь: Сб. науч. тр. – Севастополь, 2010. – Вып.101. – С.29-32.
  2. Hagan M. T. Training feedforward networks with the Marquardt algorithm/M. Т.Hagan, M. Menhaj// IEEE Transactions on Neural Networks. – 1994. – Vol. 5. – No. 6. – P. 989-993.
  3. Riedmiller M. A direct method for faster backpropagation learning/M. Riedmiller// Proceedings of the 1993 IEEE International Conference on Neural Networks (ICNN ’93). – San Francisco, 1993. – Vol. 1. – P. 586-591.

Published

2012-12-14

How to Cite

Строганов, В. О. (2012). Classification of the underground pipeline leakage signals by means of the artificial neural networks. Eastern-European Journal of Enterprise Technologies, 6(4(60), 33–36. https://doi.org/10.15587/1729-4061.2012.5680

Issue

Section

Mathematics and Cybernetics - applied aspects