The gray regions in Figure 10e, indicating a lack of
overlap between the ECA and microscopic datasets, result
from the inherent blurriness of the ECA image compared
to the high-resolution microscopic data. This blurriness
occurs because the ECA sensor coils integrate magnetic field
responses over their sensing area, leading to a less sharply
defined defect boundary.
Figure 11 summarizes the validation of ECA results for
all test samples with corrosion damage. The intersections
between post-processed ECA scans and corresponding micro-
scopic images were calculated for randomly selected regions
(D1–D4) in all corroded samples, including the flat Sample A
and the curved sample sets B1–B4 and C1–C4. A comparison
was also made between full scans (single-coil raster scans) and
fast scans (full-array, single-pass along the sample’s long side).
The results showed strong agreement between ECA and
microscopic images across all scans, including those on curved
surfaces, highlighting the robustness of the signal processing
approach used in this study. As shown in Figure 11a, full scans
exhibited a high degree of intersection with microscopic data,
indicating strong correlation. However, in fast scans, some
randomly selected regions (D1–D4) were not effectively covered
due to the single-pass nature of the scan, which followed the
centerline of the sample. Consequently, for these regions,
intersection evaluation was not possible. Instances where the
array probe was out of bounds relative to the microscopic
measurement regions are labeled as “not found” in Figures
11a and 11b.
Conclusions and Future Work
This work presents a robotic eddy current array (ECA) system
developed for detecting corrosion on curved metallic surfaces,
addressing key challenges such as inconsistent probe liftoff, tilt,
and environmental factors. A post-processing workflow, includ-
ing array subtraction and detrending techniques, was imple-
mented to correct these errors and improve defect detection
accuracy. The system showed promising results when tested
on samples with varying curvatures, correlating well with
ground truth data from a digital microscope. While liftoff and
tilt caused some inaccuracies, the system was able to detect
corrosion defects effectively. The integration of computer
vision–based path planning and precise sensor alignment
were essential to optimize measurements, demonstrating the
system’s potential for use in complex geometries and real-
world scenarios.
Future work will focus on improving the robotic ECA
system by addressing liftoff variations and enhancing robot
calibration. More automated and accurate surface reconstruc-
tion methods, such as iterative closest point (ICP) algorithms
or AprilTags (Olson 2011), will replace the current point-pair
picking method. Adaptive path planning will also be explored,
enabling the robot to adjust its scanning path based on real-
time ECA data, which will require further development of
voltage-to-depth calibration techniques. Additionally, the
system will be extended to mobile and aerial platforms,
enabling autonomous inspection of large-scale or remote
structures. These improvements will support NDE 4.0’s goals of
increasing automation, precision, and scalability in real-world
inspection scenarios.
DATA AVAILABILITY
A detailed description of the post-processing procedure used in this study
for the eddy current array image data is available from the corresponding
author upon reasonable request.
REFERENCES
ASTM. 2003. ASTM B117-03: Standard Practice for Operating Salt Spray
(Fog) Apparatus. ASTM International. West Conshohocken, PA. https://
doi.org/10.1520/B0117-03.
Doyle, E. F. 1995. “Application of Laser-Based Profilometry to Tubing in
Power Generating Utilities.” Proceedings of SPIE 2454, Nondestructive Eval-
uation of Aging Utilities. https://doi.org/10.1117/12.209380.
Foster, E. A., G. Bolton, R. Bernard, M. McInnes, S. McKnight, E. Nicolson,
C. Loukas, et al. 2022. “Automated Real-Time Eddy Current Array
Inspection of Nuclear Assets.” Sensors (Basel) 22 (16): 6036. https://doi.
org/10.3390/s22166036.
García-Martín, J., J. Gómez-Gil, and E. Vázquez-Sánchez. 2011.
“Non-Destructive Techniques Based on Eddy Current Testing.” Sensors
(Basel) 11 (3): 2525–65. https://doi.org/10.3390/s110302525.
Hamilton, C. 2024. “Enhanced Corrosion Analysis on Curved Steel Sheets
Using Freeform Robotic ECA.” PhD diss., Michigan State University.
https://doi.org/doi:10.25335/hqqm-v531.
Hamilton, C., O. Karpenko, L. Udpa, M. Haq, and Y. Deng. 2024. “Efficient
Data Acquisition and Reconstruction for Air-Coupled Ultrasonic Robotic
NDE.” Scientific Reports 14 (1): 14021. https://doi.org/10.1038/s41598-024-
60393-z.
Herl, G., J. Hiller, and A. Maier. 2020. “Scanning Trajectory Optimisation
Using a Quantitative Tuy-Based Local Quality Estimation for Robot-Based
X-Ray Computed Tomography.” Nondestructive Testing and Evaluation 35
(3): 287–303. https://doi.org/10.1080/10589759.2020.1774579.
Hespeler, S., H. Nemati, N. Masurkar, F. Alvidrez, H. Marvi, and E.
Dehghan-Niri. 2024. “Deep Learning–Based Time-Series Classification
for Robotic Inspection of Pipe Condition Using Non-Contact Ultrasonic
Testing.” Journal of Nondestructive Evaluation, Diagnostics and Prognostics
of Engineering Systems 7 (1): 011002. https://doi.org/10.1115/1.4063694.
Kang, R., P. Guerrero, G. Probst, P. Slaets, and W. Dewulf. 2022.
“Reference-Free Method for Robot CT Imaging Geometry
Estimation.” Precision Engineering 78:102–13. https://doi.org/
10.1016/j.precisioneng.2022.07.005.
Lee, C., and N.-G. Cho. 2012. “Assessment of Surface Profile Data Acquired
by a Stylus Profilometer.” Measurement Science &Technology 23 (10):
105601. https://doi.org/10.1088/0957-0233/23/10/105601.
Mineo, C., D. Cerniglia, and A. Poole. 2022. “Autonomous Robotic Sensing
for Simultaneous Geometric and Volumetric Inspection of Free-Form
Parts.” Journal of Intelligent &Robotic Systems 105 (3): 54. https://doi.
org/10.1007/s10846-022-01673-6.
Morozov, M., S. G. Pierce, C. N. MacLeod, C. Mineo, and R. Summan. 2018.
“Off-Line Scan Path Planning for Robotic NDT.” Measurement 122:284–90.
https://doi.org/10.1016/j.measurement.2018.02.020.
Mukherjee, S., R. Zhang, M. Alzuhiri, V. V. Rao, L. Udpa, and Y. Deng. 2022.
“Inline Pipeline Inspection Using Hybrid Deep Learning Aided Endo-
scopic Laser Profiling.” Journal of Nondestructive Evaluation 41 (3): 56.
https://doi.org/10.1007/s10921-022-00890-1.
Olson, E. 2011. “AprilTag: A Robust and Flexible Visual Fiducial System.”
2011. 2011 IEEE International Conference on Robotics and Automation.
Shanghai, China. https://doi.org/10.1109/ICRA.2011.5979561.
Savin, Adriana. 2024. “Today’s Possible Trail to NDE 4.0.” Catalogul Oficial
al Salonului Cadet INOVA, 35. https://cadetinova.ro/documente/
Supliment_Inova_24.pdf.
ME
|
ROBOTICECA
72
M AT E R I A L S E V A L U AT I O N • A P R I L 2 0 2 5
overlap between the ECA and microscopic datasets, result
from the inherent blurriness of the ECA image compared
to the high-resolution microscopic data. This blurriness
occurs because the ECA sensor coils integrate magnetic field
responses over their sensing area, leading to a less sharply
defined defect boundary.
Figure 11 summarizes the validation of ECA results for
all test samples with corrosion damage. The intersections
between post-processed ECA scans and corresponding micro-
scopic images were calculated for randomly selected regions
(D1–D4) in all corroded samples, including the flat Sample A
and the curved sample sets B1–B4 and C1–C4. A comparison
was also made between full scans (single-coil raster scans) and
fast scans (full-array, single-pass along the sample’s long side).
The results showed strong agreement between ECA and
microscopic images across all scans, including those on curved
surfaces, highlighting the robustness of the signal processing
approach used in this study. As shown in Figure 11a, full scans
exhibited a high degree of intersection with microscopic data,
indicating strong correlation. However, in fast scans, some
randomly selected regions (D1–D4) were not effectively covered
due to the single-pass nature of the scan, which followed the
centerline of the sample. Consequently, for these regions,
intersection evaluation was not possible. Instances where the
array probe was out of bounds relative to the microscopic
measurement regions are labeled as “not found” in Figures
11a and 11b.
Conclusions and Future Work
This work presents a robotic eddy current array (ECA) system
developed for detecting corrosion on curved metallic surfaces,
addressing key challenges such as inconsistent probe liftoff, tilt,
and environmental factors. A post-processing workflow, includ-
ing array subtraction and detrending techniques, was imple-
mented to correct these errors and improve defect detection
accuracy. The system showed promising results when tested
on samples with varying curvatures, correlating well with
ground truth data from a digital microscope. While liftoff and
tilt caused some inaccuracies, the system was able to detect
corrosion defects effectively. The integration of computer
vision–based path planning and precise sensor alignment
were essential to optimize measurements, demonstrating the
system’s potential for use in complex geometries and real-
world scenarios.
Future work will focus on improving the robotic ECA
system by addressing liftoff variations and enhancing robot
calibration. More automated and accurate surface reconstruc-
tion methods, such as iterative closest point (ICP) algorithms
or AprilTags (Olson 2011), will replace the current point-pair
picking method. Adaptive path planning will also be explored,
enabling the robot to adjust its scanning path based on real-
time ECA data, which will require further development of
voltage-to-depth calibration techniques. Additionally, the
system will be extended to mobile and aerial platforms,
enabling autonomous inspection of large-scale or remote
structures. These improvements will support NDE 4.0’s goals of
increasing automation, precision, and scalability in real-world
inspection scenarios.
DATA AVAILABILITY
A detailed description of the post-processing procedure used in this study
for the eddy current array image data is available from the corresponding
author upon reasonable request.
REFERENCES
ASTM. 2003. ASTM B117-03: Standard Practice for Operating Salt Spray
(Fog) Apparatus. ASTM International. West Conshohocken, PA. https://
doi.org/10.1520/B0117-03.
Doyle, E. F. 1995. “Application of Laser-Based Profilometry to Tubing in
Power Generating Utilities.” Proceedings of SPIE 2454, Nondestructive Eval-
uation of Aging Utilities. https://doi.org/10.1117/12.209380.
Foster, E. A., G. Bolton, R. Bernard, M. McInnes, S. McKnight, E. Nicolson,
C. Loukas, et al. 2022. “Automated Real-Time Eddy Current Array
Inspection of Nuclear Assets.” Sensors (Basel) 22 (16): 6036. https://doi.
org/10.3390/s22166036.
García-Martín, J., J. Gómez-Gil, and E. Vázquez-Sánchez. 2011.
“Non-Destructive Techniques Based on Eddy Current Testing.” Sensors
(Basel) 11 (3): 2525–65. https://doi.org/10.3390/s110302525.
Hamilton, C. 2024. “Enhanced Corrosion Analysis on Curved Steel Sheets
Using Freeform Robotic ECA.” PhD diss., Michigan State University.
https://doi.org/doi:10.25335/hqqm-v531.
Hamilton, C., O. Karpenko, L. Udpa, M. Haq, and Y. Deng. 2024. “Efficient
Data Acquisition and Reconstruction for Air-Coupled Ultrasonic Robotic
NDE.” Scientific Reports 14 (1): 14021. https://doi.org/10.1038/s41598-024-
60393-z.
Herl, G., J. Hiller, and A. Maier. 2020. “Scanning Trajectory Optimisation
Using a Quantitative Tuy-Based Local Quality Estimation for Robot-Based
X-Ray Computed Tomography.” Nondestructive Testing and Evaluation 35
(3): 287–303. https://doi.org/10.1080/10589759.2020.1774579.
Hespeler, S., H. Nemati, N. Masurkar, F. Alvidrez, H. Marvi, and E.
Dehghan-Niri. 2024. “Deep Learning–Based Time-Series Classification
for Robotic Inspection of Pipe Condition Using Non-Contact Ultrasonic
Testing.” Journal of Nondestructive Evaluation, Diagnostics and Prognostics
of Engineering Systems 7 (1): 011002. https://doi.org/10.1115/1.4063694.
Kang, R., P. Guerrero, G. Probst, P. Slaets, and W. Dewulf. 2022.
“Reference-Free Method for Robot CT Imaging Geometry
Estimation.” Precision Engineering 78:102–13. https://doi.org/
10.1016/j.precisioneng.2022.07.005.
Lee, C., and N.-G. Cho. 2012. “Assessment of Surface Profile Data Acquired
by a Stylus Profilometer.” Measurement Science &Technology 23 (10):
105601. https://doi.org/10.1088/0957-0233/23/10/105601.
Mineo, C., D. Cerniglia, and A. Poole. 2022. “Autonomous Robotic Sensing
for Simultaneous Geometric and Volumetric Inspection of Free-Form
Parts.” Journal of Intelligent &Robotic Systems 105 (3): 54. https://doi.
org/10.1007/s10846-022-01673-6.
Morozov, M., S. G. Pierce, C. N. MacLeod, C. Mineo, and R. Summan. 2018.
“Off-Line Scan Path Planning for Robotic NDT.” Measurement 122:284–90.
https://doi.org/10.1016/j.measurement.2018.02.020.
Mukherjee, S., R. Zhang, M. Alzuhiri, V. V. Rao, L. Udpa, and Y. Deng. 2022.
“Inline Pipeline Inspection Using Hybrid Deep Learning Aided Endo-
scopic Laser Profiling.” Journal of Nondestructive Evaluation 41 (3): 56.
https://doi.org/10.1007/s10921-022-00890-1.
Olson, E. 2011. “AprilTag: A Robust and Flexible Visual Fiducial System.”
2011. 2011 IEEE International Conference on Robotics and Automation.
Shanghai, China. https://doi.org/10.1109/ICRA.2011.5979561.
Savin, Adriana. 2024. “Today’s Possible Trail to NDE 4.0.” Catalogul Oficial
al Salonului Cadet INOVA, 35. https://cadetinova.ro/documente/
Supliment_Inova_24.pdf.
ME
|
ROBOTICECA
72
M AT E R I A L S E V A L U AT I O N • A P R I L 2 0 2 5
