ABSTRACT
Machine learning for nondestructive evaluation (NDE) has the potential to bring significant improvements in defect characterization accuracy due to its effectiveness in pattern recognition problems. However, the application of modern machine learning methods to NDE has been obstructed by the scarcity of real defect data to train on. This article demonstrates how an efficient, hybrid finite element (FE) and ray-based simulation can be used to train a convolutional neural network (CNN) to characterize real defects. To demonstrate this methodology, an inline pipe inspection application is considered. This uses four plane wave images from two arrays and is applied to the characterization of cracks of length 1-5 mm and inclined at angles of up to 20° from the vertical. A standard image-based sizing technique, the 6-dB drop method, is used as a comparison point. For the 6-dB drop method, the average absolute error in length and angle prediction is ±1.1 mm and ±8.6°, respectively, while the CNN is almost four times more accurate at ±0.29 mm and ±2.9°. To demonstrate the adaptability of the deep learning approach, an error in sound speed estimation is included in the training and test set. With a maximum error of 10% in shear and longitudinal sound speed, the 6-dB drop method has an average error of ±1.5 mmm and ±12°, while the CNN has ±0.45 mm and ±3.0°. This demonstrates far superior crack characterization accuracy by using deep learning rather than traditional image-based sizing.
ABSTRACT
Plane wave imaging (PWI) is an ultrasonic array imaging technique used in nondestructive testing, which has been shown to yield high resolution with few transmissions. Only a few published examples are available of PWI of components with nonplanar surfaces in immersion. In these cases, inspections were performed by adapting the transmission delays in order to produce a plane wave inside the component. This adaptation requires prior knowledge of the component geometry and position relative to the array. This article proposes a new implementation, termed PWI adapted in postprocessing (PWAPP), which has no such requirement. In PWAPP, the array emits a plane wave as in conventional PWI. The captured data are input into two postprocessing stages. The first reconstructs the surface of the component; the latter images inside of it by adapting the delays to the distortion of the plane waves upon refraction at the reconstructed surface. Simulation and experimental data are produced from an immersed sample with a concave front surface and artificial defects. These are processed with conventional and surface corrected PWI. Both algorithms involving surface adaptation produced nearly equivalent results from the simulated data, and both outperform the nonadapted one. Experimentally, all defects are imaged with a signal-to-noise ratio (SNR) of at least 31.8 and 33.5 dB for, respectively, PWAPP and PWI adapted in transmission but only 20.5 dB for conventional PWI. In the cases considered, reducing the number of transmissions below the number of array elements shows that PWAPP maintains its high SNR performance down to the number of firings equivalent to a quarter of the array elements. Finally, experimental data from a more complex surface specimen are processed with PWAPP resulting in detection of all scatterers and producing SNR comparable to that of the total focusing method.
