Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges

A leak detection method is developed for leaks typically encountered in industrial production. Leaks of 1 mm diameter and less are considered at operating pressures up to 10 bar. The system uses two separate datasets—one for the leak noises and the other for the background noises—both are linked usi...

Full description

Saved in:
Bibliographic Details
Main Authors: Deniz Quick, Jens Denecke, Jürgen Schmidt
Format: Article
Language:English
Published: MDPI AG 2025-06-01
Series:AI
Subjects:
Online Access:https://www.mdpi.com/2673-2688/6/7/136
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849409136167485440
author Deniz Quick
Jens Denecke
Jürgen Schmidt
author_facet Deniz Quick
Jens Denecke
Jürgen Schmidt
author_sort Deniz Quick
collection DOAJ
description A leak detection method is developed for leaks typically encountered in industrial production. Leaks of 1 mm diameter and less are considered at operating pressures up to 10 bar. The system uses two separate datasets—one for the leak noises and the other for the background noises—both are linked using a developed mixup technique and thus simulate leaks trained in background noises. A specific frequency window between 11 and 20 kHz is utilized to generate a quadratic input for image recognition. With this method, detection accuracies of over 95% with a false alarm rate under 2% can be achieved on a test dataset under the background noises of hydraulic machines in laboratory conditions.
format Article
id doaj-art-469ae9d4d8074fa5bde4ec1c4bf5bd5a
institution Kabale University
issn 2673-2688
language English
publishDate 2025-06-01
publisher MDPI AG
record_format Article
series AI
spelling doaj-art-469ae9d4d8074fa5bde4ec1c4bf5bd5a2025-08-20T03:35:36ZengMDPI AGAI2673-26882025-06-016713610.3390/ai6070136Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise ChallengesDeniz Quick0Jens Denecke1Jürgen Schmidt2CSE Center of Safety Excellence (CSE-Institut), 76327 Pfinztal, GermanyCSE Center of Safety Excellence (CSE-Institut), 76327 Pfinztal, GermanyCSE Center of Safety Excellence (CSE-Institut), 76327 Pfinztal, GermanyA leak detection method is developed for leaks typically encountered in industrial production. Leaks of 1 mm diameter and less are considered at operating pressures up to 10 bar. The system uses two separate datasets—one for the leak noises and the other for the background noises—both are linked using a developed mixup technique and thus simulate leaks trained in background noises. A specific frequency window between 11 and 20 kHz is utilized to generate a quadratic input for image recognition. With this method, detection accuracies of over 95% with a false alarm rate under 2% can be achieved on a test dataset under the background noises of hydraulic machines in laboratory conditions.https://www.mdpi.com/2673-2688/6/7/136convolutional neural networksacoustic leak detectionacoustic mixupbackground noiseshort-time Fourier transformationchemical plants
spellingShingle Deniz Quick
Jens Denecke
Jürgen Schmidt
Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges
AI
convolutional neural networks
acoustic leak detection
acoustic mixup
background noise
short-time Fourier transformation
chemical plants
title Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges
title_full Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges
title_fullStr Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges
title_full_unstemmed Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges
title_short Enhancing Acoustic Leak Detection with Data Augmentation: Overcoming Background Noise Challenges
title_sort enhancing acoustic leak detection with data augmentation overcoming background noise challenges
topic convolutional neural networks
acoustic leak detection
acoustic mixup
background noise
short-time Fourier transformation
chemical plants
url https://www.mdpi.com/2673-2688/6/7/136
work_keys_str_mv AT denizquick enhancingacousticleakdetectionwithdataaugmentationovercomingbackgroundnoisechallenges
AT jensdenecke enhancingacousticleakdetectionwithdataaugmentationovercomingbackgroundnoisechallenges
AT jurgenschmidt enhancingacousticleakdetectionwithdataaugmentationovercomingbackgroundnoisechallenges