Demonstration of using signal feature extraction and deep learning neural networks with ultrasonic data for detecting challenging discontinuities in composite panels

Aldrin, John C.; Forsyth, David

doi:10.1063/1.5099716

Cited by 14 publications

(8 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This method can also be used for defect localisation, material properties' characterisation, and damage characterisation. The future of composites' NDE is headed in the direction of data-driven approaches, opening up newer dimensions of NDE efficiency by involving methods such as deep learning, transfer learning, and physics-informed machine learning, with multiple articles already published [166,167].…”

Section: Future Of Nde Of Structural Compositesmentioning

confidence: 99%

A Review of Sensing Technologies for Non-Destructive Evaluation of Structural Composite Materials

et al. 2021

View full text Add to dashboard Cite

The growing demand and diversity in the application of industrial composites and the current inability of present non-destructive evaluation (NDE) methods to perform detailed inspection of these composites has motivated this comprehensive review of sensing technologies. NDE has the potential to be a versatile tool for maintaining composite structures deployed in hazardous and inaccessible areas, such as offshore wind farms and nuclear power plants. Therefore, the future composite solutions need to take into consideration the niche requirements of these high-value/critical applications. Composite materials are intrinsically complex due to their anisotropic and non-homogeneous characteristics. This presents a significant challenge for evaluation and the associated data analysis for NDEs. For example, the quality assurance, certification of composite structures, and early detection of the failure is complex due to the variability and tolerances involved in the composite manufacturing. Adapting existing NDE methods to detect and locate the defects at multiple length scales in the complex materials represents a significant challenge, resulting in a delayed and incorrect diagnosis of the structural health. This paper presents a comprehensive review of the NDE techniques, that includes a detailed discussion of their working principles, setup, advantages, limitations, and usage level for the structural composites. A comparison between these techniques is also presented, providing an insight into the future trends for composites’ prognostic and health management (PHM). Current research trends show the emergence of the non-contact-type NDE (including digital image correlation, infrared tomography, as well as disruptive frequency-modulated continuous wave techniques) for structural composites, and the reasons for their choice over the most popular contact-type (ultrasonic, acoustic, and piezoelectric testing) NDE methods is also discussed. The analysis of this new sensing modality for composites’ is presented within the context of the state-of-the-art and projected future requirements.

show abstract

Section: Future Of Nde Of Structural Compositesmentioning

confidence: 99%

A Review of Sensing Technologies for Non-Destructive Evaluation of Structural Composite Materials

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Automation of the testing can generate a large amount of data to be processed. Hence machine learning and deep learning methods have been developed to conduct defect detection either on composite [1][2][3] or metallic materials [4][5][6][7][8][9]. One challenge is then to label the data before the learning stage of those methods.…”

Section: State Of the Artmentioning

confidence: 99%

Benefit of Neural Network for the Optimization of Defect Detection on Composite Material Using Ultrasonic Non Destructive Testing

Trouvé-Peloux

Abeloos

Fekih

et al. 2021

2021 48th Annual Review of Progress in Quantitative Nondestructive Evaluation

View full text Add to dashboard Cite

This paper is dedicated to out-of-plane waviness defect detection within composite materials by ultrasonic testing. We present here an in-house experimental database of ultrasonic data built on composite pieces with/without elaborated defects. Using this dataset, we have developed several defect detection methods using the C-scan representation, where the defect is clearly observable. We compare here the defect detection performance of unsupervised, classical machine learning methods and deep learning approaches. In particular, we have investigated the use of semantic segmentation networks that provides a classification of the data at the “pixel level”, hence at each C-scan measure. This technique is used to classify if a defect is detected, and to produce a precise localization of the defect within the material. The results we obtained with the various detection methods are compared, and we discuss the drawbacks and advantages of each method.

show abstract

“…Convolutional Neural Networks [19] are a natural answer to this as they connect only nearby pixels at each layer, vastly reducing the complexity of the network. They have also seen widespread success with natural [20], medical [21]- [23] and NDE [13], [14], [16], [17] images in the past. There are many well-known Convolutional Neural Network (CNN) architectures for image characterization such as LeNet, DenseNet, Inception, AlexNet and ResNet [24].…”

Section: Introductionmentioning

confidence: 99%

Deep Learning for Ultrasonic Crack Characterization in NDE

Pyle

Bevan

Hughes

et al. 2021

IEEE Trans. Ultrason., Ferroelect., Freq. Contr.

View full text Add to dashboard Cite

Machine learning for Non-Destructive 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 paper demonstrates how an efficient, hybrid finite element 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, ±8.6° while the CNN is almost four times more accurate at ±0.29 mm, ±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 mm, ±12° while the CNN has ±0.45 mm, ±3.0°. This demonstrates far superior crack characterization accuracy by using deep learning rather than traditional image-based sizing.

show abstract

Demonstration of using signal feature extraction and deep learning neural networks with ultrasonic data for detecting challenging discontinuities in composite panels

Cited by 14 publications

References 28 publications

A Review of Sensing Technologies for Non-Destructive Evaluation of Structural Composite Materials

A Review of Sensing Technologies for Non-Destructive Evaluation of Structural Composite Materials

Benefit of Neural Network for the Optimization of Defect Detection on Composite Material Using Ultrasonic Non Destructive Testing

Deep Learning for Ultrasonic Crack Characterization in NDE

Contact Info

Product

Resources

About