Pipeline leaks acoustic detection using mechanically amplified fiber Bragg grating sensors and artificial intelligence

In this work, an acoustic sensor based on FBGs is presented for pipeline leak detection. To enhance its sensitivity, a mechanical amplification structure incorporating a hanging mass was designed and fabricated. Additionally, to improve versatility and detection performance, an Artificial Intelligence algorithm based on an autoencoders was implemented to perform a real-time leak detection regardless of the acoustic frequency at which the leak occurs. The sensors and detection capabilities using autoencoders were successfully tested on a real pipeline with a weak leak. The amplified FBG demonstrated a 25 dB gain in detection performance using the autoencoder’s approach.