Abstract
Pipeline infrastructure systems in service are aging and continue to deteriorate with the passage of time. For public safety, it is extremely important to accurately detect harmful defects in these pipelines and replace the corresponding pipe sectors before they lead to leakages. However, due to regular usages, the inner pipe surfaces can be rather non-smooth and inundated with several scratches and tiny cavities. Their minor defects within the pipeline do-not need immediate repair. As such, it will be very expensive if pipe sections just containing minor defects are replaced. In this paper, we have developed a novel method for accurate identification of large cavities and potentially harmful obtrusive defects using magnetic flux leakage (MFL) based nondestructive evaluation (NDE) technique. A substantial challenge in our set-up is the detection of possible harmful defects in the presence of several minor and tiny cavities. This translates into defect recognition under extremely noisy conditions as the MFL's intensity-based signals is heavily influenced by the presence of multiple minor defects and cavities. Based on MFL data from a wide range of feasible scenarios, we develop a robust detection algorithm that is sensitive in the detection of harmful large defects and is simultaneously also cost effective by not classifying most of the harmless cavities as harmful defects. Our detection analysis is based on nearest neighbor-based divergence measure in Wavelet transformed domain of the flux signals. We study the performance of our procedure across different regimes and obtain encouraging results.
| Original language | English |
|---|---|
| Title of host publication | Proceedings - 2019 Prognostics and System Health Management Conference, PHM-Paris 2019 |
| Editors | Chuan Li, Jose Valente de Oliveira, Ping Ding, Ping Ding, Diego Cabrera |
| Publisher | Institute of Electrical and Electronics Engineers Inc. |
| Pages | 297-302 |
| Number of pages | 6 |
| ISBN (Electronic) | 9781728103297 |
| DOIs | |
| State | Published - May 2019 |
| Externally published | Yes |
| Event | 2019 Prognostics and System Health Management Conference, PHM-Paris 2019 - Paris, France Duration: May 2 2019 → May 5 2019 |
Publication series
| Name | Proceedings - 2019 Prognostics and System Health Management Conference, PHM-Paris 2019 |
|---|
Conference
| Conference | 2019 Prognostics and System Health Management Conference, PHM-Paris 2019 |
|---|---|
| Country/Territory | France |
| City | Paris |
| Period | 05/2/19 → 05/5/19 |
Funding
V. CONCLUSION AND FUTURE WORK We present a wavelet based nearest neighbor algorithm that can detect large harmful defects in pipelines with very high accuracy. Our proposed methodology can also simultaneously distinguish minor harmless defects and cavities. Our proposed method will save expenditure over for competing NDE methods that can not distinguish between defect types and would recommend these minor defects for further scrutiny along with the large defects It will be interesting to extend the methodology developed here to study pipeline inspection systems based on 3-dimensional probes and NDE based techniques. Another interesting future direction would be to provide a holistic analytical and data gathering system that can collect and integrate data from multiple sensors VI. ACKNOWLEDGMENT This work is partially supported by the U.S. Department of Transportation Grant: Improvements to Pipeline Assessment Methods and Models to Reduce Variance (Award No.693JK1810001).
Keywords
- detection with noise and uncertainty
- false alarms
- K-NN
- Magnetic flux leakage
- Nondestructive Evaluation
- robust
- sensitivity
- Wavelets