Pipeline Fault Diagnosis Using Wavelet Entropy and Ensemble Deep Neural Technique

The maintenance of pipelines is essential for the safe and cost effective transport of important fluids such as water, oil, and gas. The early detection of pipeline faults is vital for avoiding material and economic losses, and more importantly for ensuring the safety of both human life and the environment. This paper proposes a methodology for early fault detection in pipelines using an acoustic emission (AE) based technique. The proposed method incorporates wavelet entropy analysis of the AE signals and ensemble deep neural networks for the effective detection of different types of faults in a pipeline. The proposed method is tested on an experimental testbed, and the results indicate that it can detect various faults in the pipeline with an average accuracy of 96%.