Nonlinear, non-stationary image processing technique for eddy current NDE

Yang, Guang; Dib, Gerges; Kim, Jaejoon; Zhang, Lu; Xin, Junjun; Удпа, Лалита

doi:10.1063/1.4716293

Cited by 4 publications

(2 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, Hilbert -Huang transform (HHT), an advanced time -frequency analysis method, has been widely applied in the fields of vibration recognition, fault diagnosis as well as economy and geography data analysis due to its satisfying efficiency and resolution in time domain and frequency domain. [16 -19] In Yang et al [20], a feature extraction method of ECT signal based on HHT was proposed, in which several parameters measuring signal complexities, e.g. root mean square, variance and Shannon entropy, were calculated for each IMF to form feature vector for the flaw detection in steam generator tubes.…”

Section: Introductionmentioning

confidence: 99%

Application of Hilbert–Huang transform for defect recognition in pulsed eddy current testing

Liu

Huang

Hou

et al. 2015

Nondestructive Testing and Evaluation

View full text Add to dashboard Cite

Defect recognition plays an important role in the structure integrity and health monitor of in-service equipment. However, it is difficult to recognise deep-layer defect or smallsize defect in conductive structure during pulsed eddy current (PEC) testing. Aiming at the issue, this article proposes a method based on Hilbert -Huang transform which consists of two modules: data processing and defect recognition. In the data processing module, the PEC response signal is decomposed into a few of intrinsic mode functions (IMFs) using ensemble empirical mode decomposition method. The IMFs whose variance contribution rates are bigger than 1% are chosen to reconstruct signal in order to remove noise. In the defect recognition module, the features based on specific frequency components of marginal spectrum (MS) of the reconstructed signals are extracted to discriminate those defects in surface and subsurface. Furthermore, the normalisation MS energy ratio is proposed to quantify defects which cannot be distinguished using peak value in time domain. Experiments show that the proposed method can achieve better de-noising effect and defect evaluation, which contributes to the recognition of those complicated defects such as deep-layered and small-sized defect.

show abstract

Section: Introductionmentioning

confidence: 99%

Application of Hilbert–Huang transform for defect recognition in pulsed eddy current testing

Liu

Huang

Hou

et al. 2015

Nondestructive Testing and Evaluation

View full text Add to dashboard Cite

show abstract

“…A classifier in NDE is a decision system which classifies a measured signal into one of two possible categories, namely defect or non-defect. Till date, numerous rule-based and pattern recognition based algorithms have been implemented for detection and classification of defects in various NDE inspections such as Fisher Linear Discriminant method for classifying flaws in eddy current signals [1], neural networks [2] [3] for classifying signals from ultrasonic weld inspection, Support Vector Machines [4] for non-stationary image processing of eddy current NDE images and K-means clustering for differential probe signals [5]. Such algortihms have provided good classification results in different NDE applications, yet they focus primarily on minimizing misclassifcation error without throwing much light on the reliability of data anaysis.…”

Section: Introductionmentioning

confidence: 99%

Ensemble of classifiers for confidence-rated classification of NDE signal

Banerjee

Safdarnejad

Удпа³

et al. 2016

AIP Conference Proceedings

Self Cite

View full text Add to dashboard Cite

Abstract. Ensemble of classifiers in general, aims to improve classification accuracy by combining results from multiple weak hypotheses into a single strong classifier through weighted majority voting. Improved versions of ensemble of classifiers generate self-rated confidence scores which estimate the reliability of each of its prediction and boost the classifier using these confidence-rated predictions. However, such a confidence metric is based only on the rate of correct classification. In existing works, although ensemble of classifiers has been widely used in computational intelligence, the effect of all factors of unreliability on the confidence of classification is highly overlooked. With relevance to NDE, classification results are affected by inherent ambiguity of classifica-tion, non-discriminative features, inadequate training samples and noise due to measurement. In this paper, we extend the existing ensemble classification by maximizing confidence of every classification decision in addition to minimizing the classification error. Initial results of the approach on data from eddy current inspection show improvement in classification performance of defect and non-defect indications.

show abstract

Multi-frequency fusion algorithm for detection of defects under fasteners with EC-GMR probe data

Karpenko

Efremov

et al. 2020

NDT & E International

View full text Add to dashboard Cite

Nonlinear, non-stationary image processing technique for eddy current NDE

Cited by 4 publications

References 4 publications

Application of Hilbert–Huang transform for defect recognition in pulsed eddy current testing

Application of Hilbert–Huang transform for defect recognition in pulsed eddy current testing

Ensemble of classifiers for confidence-rated classification of NDE signal

Multi-frequency fusion algorithm for detection of defects under fasteners with EC-GMR probe data

Contact Info

Product

Resources

About