A5.4 - Machine Learning Model Based on Signal Difference Features for Damage Localization on Hydrogen Pressure Vessel Using Ultrasonic Guided Waves
- Event
- 22. GMA/ITG-Fachtagung Sensoren und Messsysteme 2024
2024-06-11 - 2024-06-12
Nürnberg - Band
- Vorträge
- Chapter
- A5 - Sensorik für Wasserstoffwirtschaft
- Author(s)
- H. El Moutaouakil, A. Schütze, T. Schneider - Universität des Saarlandes,Saarbrücken, J. Prager - Bundesanstalt für Materialforschung- und Prüfung,Berlin
- Pages
- 130 - 136
- DOI
- 10.5162/sensoren2024/A5.4
- ISBN
- 978-3-910600-01-0
- Price
- free
Abstract
Hydrogen is already shaping the future of energy resources; therefore, its storage must adhere to the highest safety standards due to its nature as a highly explosive gas. Consequently, it is imperative to ensure accurate and reliable detection of damage in pressure vessels at an early stage. Despite the existence of various machine learning methods, particularly those based on deep learning, they often face challenges related to interpretability and overfitting. This paper introduces an explainable machine learning (ML) model capable of localizing structural damage on Composite Overwrapped Pressure Vessels (COPVs) using ultrasonic guided waves, that overcomes the introduced drawbacks.