Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network
Rebar corrosion can cause severe deterioration in reinforced concrete structures and needs to be detected using non-destructive testing methods. In particular, eddy-current method utilizing an excitation coil has shown to be a promising solution. However, no research has measured the 3-axis magnetic...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
AIP Publishing LLC
2025-03-01
|
| Series: | AIP Advances |
| Online Access: | http://dx.doi.org/10.1063/9.0000849 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849738950181126144 |
|---|---|
| author | Yuji Ogata Tomonori Yanagida Bunichi Kakinuma Koichiro Kobayashi |
| author_facet | Yuji Ogata Tomonori Yanagida Bunichi Kakinuma Koichiro Kobayashi |
| author_sort | Yuji Ogata |
| collection | DOAJ |
| description | Rebar corrosion can cause severe deterioration in reinforced concrete structures and needs to be detected using non-destructive testing methods. In particular, eddy-current method utilizing an excitation coil has shown to be a promising solution. However, no research has measured the 3-axis magnetic signal over a wide spatial range around the coil. In this study, we compared the magnetic signal of the magnetic sensor directly above the coil with that of a magnetic sensor distant from the coil using a 3-axis 100-channel magnetic sensor array. Moreover, we utilized a convolutional neural network (CNN) to estimate the corrosion level of a rebar with unknown depth. The results showed that when the magnetic sensor was directly above the coil, the magnetic amplitude was not separated into four levels of rebar corrosion. Meanwhile, when the magnetic sensor was 80 mm away from the coil, the amplitude was separated into four levels of rebar corrosion. Therefore, rebar corrosion is easier to detect by measuring the magnetic signal over a wide spatial range. Next, we checked if we could estimate the rebar corrosion level whose depth was not used in the CNN training data. The average accuracy of corrosion level estimation using the three axes data was more than 90% and higher than that using the single axis data, which means that rebar corrosion can be estimated even if the rebar depth is unknown. Overall, our findings indicate that multi-axis and multi-array magnetic measurements are effective in estimating rebar corrosion. |
| format | Article |
| id | doaj-art-d69e21a5db8f478f9f7667a0734fb63f |
| institution | DOAJ |
| issn | 2158-3226 |
| language | English |
| publishDate | 2025-03-01 |
| publisher | AIP Publishing LLC |
| record_format | Article |
| series | AIP Advances |
| spelling | doaj-art-d69e21a5db8f478f9f7667a0734fb63f2025-08-20T03:06:24ZengAIP Publishing LLCAIP Advances2158-32262025-03-01153035106035106-510.1063/9.0000849Estimation of rebar corrosion level using magnetic sensor array with convolutional neural networkYuji Ogata0Tomonori Yanagida1Bunichi Kakinuma2Koichiro Kobayashi3Advantest Laboratories Ltd., 48-4 Matsubara, Kami-Ayashi, Aoba-ku, Sendai, Miyagi 989-3124, JapanAdvantest Laboratories Ltd., 48-4 Matsubara, Kami-Ayashi, Aoba-ku, Sendai, Miyagi 989-3124, JapanAdvantest Laboratories Ltd., 48-4 Matsubara, Kami-Ayashi, Aoba-ku, Sendai, Miyagi 989-3124, JapanGraduate School of Science and Engineering, Iwate University, 4-3-5 Ueda, Morioka, Iwate 020-8551, JapanRebar corrosion can cause severe deterioration in reinforced concrete structures and needs to be detected using non-destructive testing methods. In particular, eddy-current method utilizing an excitation coil has shown to be a promising solution. However, no research has measured the 3-axis magnetic signal over a wide spatial range around the coil. In this study, we compared the magnetic signal of the magnetic sensor directly above the coil with that of a magnetic sensor distant from the coil using a 3-axis 100-channel magnetic sensor array. Moreover, we utilized a convolutional neural network (CNN) to estimate the corrosion level of a rebar with unknown depth. The results showed that when the magnetic sensor was directly above the coil, the magnetic amplitude was not separated into four levels of rebar corrosion. Meanwhile, when the magnetic sensor was 80 mm away from the coil, the amplitude was separated into four levels of rebar corrosion. Therefore, rebar corrosion is easier to detect by measuring the magnetic signal over a wide spatial range. Next, we checked if we could estimate the rebar corrosion level whose depth was not used in the CNN training data. The average accuracy of corrosion level estimation using the three axes data was more than 90% and higher than that using the single axis data, which means that rebar corrosion can be estimated even if the rebar depth is unknown. Overall, our findings indicate that multi-axis and multi-array magnetic measurements are effective in estimating rebar corrosion.http://dx.doi.org/10.1063/9.0000849 |
| spellingShingle | Yuji Ogata Tomonori Yanagida Bunichi Kakinuma Koichiro Kobayashi Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network AIP Advances |
| title | Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network |
| title_full | Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network |
| title_fullStr | Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network |
| title_full_unstemmed | Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network |
| title_short | Estimation of rebar corrosion level using magnetic sensor array with convolutional neural network |
| title_sort | estimation of rebar corrosion level using magnetic sensor array with convolutional neural network |
| url | http://dx.doi.org/10.1063/9.0000849 |
| work_keys_str_mv | AT yujiogata estimationofrebarcorrosionlevelusingmagneticsensorarraywithconvolutionalneuralnetwork AT tomonoriyanagida estimationofrebarcorrosionlevelusingmagneticsensorarraywithconvolutionalneuralnetwork AT bunichikakinuma estimationofrebarcorrosionlevelusingmagneticsensorarraywithconvolutionalneuralnetwork AT koichirokobayashi estimationofrebarcorrosionlevelusingmagneticsensorarraywithconvolutionalneuralnetwork |