Flexible Acoustic Emission Sensor Signal Classification Using Convolutional Neural Networks for Pipeline Leak Detection

This paper presents a method for the real-time detection of pipeline leaks using flexible Acoustic Emission (AE) sensors. The signals gathered from the AE sensor are transformed into RGB images through the application of Mel-spectrogram and color coding. These converted images serve as input for a Convolutional Neural Network (CNN) based on ResNet18. With this approach, both the presence and intensity of leaks in a pipeline can be identified using the AE sensor. The effectiveness of the proposed method was validated through data collected from a testbed featuring a galvanized pipe.