Details
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the 11th European Workshop on Structural Health Monitoring |
| Subtitle of host publication | EWSHM 2024 |
| Pages | 11 |
| Publication status | Published - 1 Jul 2024 |
| Event | 11th European Workshop on Structural Health Monitoring, EWSHM 2024 - Potsdam, Germany Duration: 10 Jun 2024 → 13 Jun 2024 |
Abstract
In non-destructive testing, acoustic emission testing is an established technique with numerous applications in academia and industry. In nearly all applications, the detection of relevant acoustic emission signals in continuous structure-borne sound recordings is a crucial step that forms the basis for subsequent analyses and should therefore neither miss any relevant acoustic emissions nor detect too many irrelevant events. That is because, in its most prevalent form, acoustic emission testing utilizes ultrasound signals that require large amount of storage, which might be an economically limiting factor especially for structural health monitoring of large civil infrastructures. In these monitoring applications, the signal-to-noise ratio is also expected to be worse compared to more controlled laboratory conditions, as they can be often found, for example, in materials research. We therefore propose the use of linear prediction, a time series forecasting technique, to extract relevant acoustic emissions from continuous recordings. The proposed methodology utilizes the residual between the signal predicted from a linear combination of previous time steps and the actual measurement to detect any anomalous events. Since the linear predictor can be initialized using the environmental noise only, no specific knowledge of the acoustic emission signals of interest is required beforehand and hence can automatically adapt to each measurement environment. We compare our approach with a simple amplitude-based detection as it is commonly implemented in commercially available acoustic emission systems. Especially for low signal-to-noise ratios, an increase in the area under the receiver operating characteristic curve of up to 0.6 compared to the amplitude-based detection is found.
Keywords
- Acoustic Emission, Damage Detection, Non-Destructive Testing, Structural Health Monitoring, Time Series Forecasting, Wind Energy
ASJC Scopus subject areas
- Computer Science(all)
- Computer Science Applications
- Health Professions(all)
- Health Information Management
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
Proceedings of the 11th European Workshop on Structural Health Monitoring: EWSHM 2024. 2024. p. 11.
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Acoustic Emission Detection in Noisy Environments using Linear Prediction
AU - Lange, Alexander
AU - Xu, Ronghua
AU - Kaeding, Max
AU - Marx, Steffen
AU - Ostermann, Joern
N1 - Publisher Copyright: © 2024 11th European Workshop on Structural Health Monitoring, EWSHM 2024. All rights reserved.
PY - 2024/7/1
Y1 - 2024/7/1
N2 - In non-destructive testing, acoustic emission testing is an established technique with numerous applications in academia and industry. In nearly all applications, the detection of relevant acoustic emission signals in continuous structure-borne sound recordings is a crucial step that forms the basis for subsequent analyses and should therefore neither miss any relevant acoustic emissions nor detect too many irrelevant events. That is because, in its most prevalent form, acoustic emission testing utilizes ultrasound signals that require large amount of storage, which might be an economically limiting factor especially for structural health monitoring of large civil infrastructures. In these monitoring applications, the signal-to-noise ratio is also expected to be worse compared to more controlled laboratory conditions, as they can be often found, for example, in materials research. We therefore propose the use of linear prediction, a time series forecasting technique, to extract relevant acoustic emissions from continuous recordings. The proposed methodology utilizes the residual between the signal predicted from a linear combination of previous time steps and the actual measurement to detect any anomalous events. Since the linear predictor can be initialized using the environmental noise only, no specific knowledge of the acoustic emission signals of interest is required beforehand and hence can automatically adapt to each measurement environment. We compare our approach with a simple amplitude-based detection as it is commonly implemented in commercially available acoustic emission systems. Especially for low signal-to-noise ratios, an increase in the area under the receiver operating characteristic curve of up to 0.6 compared to the amplitude-based detection is found.
AB - In non-destructive testing, acoustic emission testing is an established technique with numerous applications in academia and industry. In nearly all applications, the detection of relevant acoustic emission signals in continuous structure-borne sound recordings is a crucial step that forms the basis for subsequent analyses and should therefore neither miss any relevant acoustic emissions nor detect too many irrelevant events. That is because, in its most prevalent form, acoustic emission testing utilizes ultrasound signals that require large amount of storage, which might be an economically limiting factor especially for structural health monitoring of large civil infrastructures. In these monitoring applications, the signal-to-noise ratio is also expected to be worse compared to more controlled laboratory conditions, as they can be often found, for example, in materials research. We therefore propose the use of linear prediction, a time series forecasting technique, to extract relevant acoustic emissions from continuous recordings. The proposed methodology utilizes the residual between the signal predicted from a linear combination of previous time steps and the actual measurement to detect any anomalous events. Since the linear predictor can be initialized using the environmental noise only, no specific knowledge of the acoustic emission signals of interest is required beforehand and hence can automatically adapt to each measurement environment. We compare our approach with a simple amplitude-based detection as it is commonly implemented in commercially available acoustic emission systems. Especially for low signal-to-noise ratios, an increase in the area under the receiver operating characteristic curve of up to 0.6 compared to the amplitude-based detection is found.
KW - Acoustic Emission
KW - Damage Detection
KW - Non-Destructive Testing
KW - Structural Health Monitoring
KW - Time Series Forecasting
KW - Wind Energy
UR - http://www.scopus.com/inward/record.url?scp=85202571630&partnerID=8YFLogxK
U2 - 10.58286/29837
DO - 10.58286/29837
M3 - Conference contribution
AN - SCOPUS:85202571630
SP - 11
BT - Proceedings of the 11th European Workshop on Structural Health Monitoring
T2 - 11th European Workshop on Structural Health Monitoring, EWSHM 2024
Y2 - 10 June 2024 through 13 June 2024
ER -