CoChat

Fatigue Crack Evaluation with the Guided Wave–Convolutional Neural Network Ensemble and Differential Wavelet Spectrogram

Open access

The DOI matches its Crossref record; 52 of the 52 references it cites were found in a registry or an index.

Abstract

On-line fatigue crack evaluation is crucial for ensuring the structural safety and reducing the maintenance costs of safety-critical systems. Among structural health monitoring (SHM), guided wave (GW)-based SHM has been deemed as one of the most promising techniques. However, the traditional damage index-based method and machine learning methods require manual processing and selection of GW features, which depend highly on expert knowledge and are easily affected by complicated uncertainties. Therefore, this paper proposes a fatigue crack evaluation framework with the GW–convolutional neural network (CNN) ensemble and differential wavelet spectrogram. The differential time–frequency spectrogram between the baseline signal and the monitoring signal is processed as the CNN input with the complex Gaussian wavelet transform. Then, an ensemble of CNNs is trained to jointly determine the crack length. Real fatigue tests on complex lap joint structures were carried out to validate the proposed method, in which several structures were tested preliminarily for collecting the training dataset and a new structure was adopted for testing. The root mean square error of the training dataset is 1.4 mm. Besides, the root mean square error of the evaluated crack length in the testing lap joint structure was 1.7 mm, showing the effectiveness of the proposed method.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/s22010307

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Illustration of the GW-based crack monitoring.
Figure 1 Illustration of the GW-based crack monitoring.CC BY · PubMed Central
Figure 2 Architecture of a typical CNN.
Figure 2 Architecture of a typical CNN.CC BY · PubMed Central
Figure 3 Implementation of the GW–CNN ensemble-based fatigue crack evaluation.
Figure 3 Implementation of the GW–CNN ensemble-based fatigue crack evaluation.CC BY · PubMed Central
Figure 4 Specimen dimensions and sensor layout.
Figure 4 Specimen dimensions and sensor layout.CC BY · PubMed Central
Figure 5 Specimen with the PZT smart layer.
Figure 5 Specimen with the PZT smart layer.CC BY · PubMed Central
Figure 6 Setup of the fatigue test.
Figure 6 Setup of the fatigue test.CC BY · PubMed Central
Figure 7 Fatigue cracks of the lap joint structure: (a) Fatigue crack just occurs at No. 4 rivet; (b) Cracks from No. 4 rivet and No. 5 rivet link together.
Figure 7 Fatigue cracks of the lap joint structure: (a) Fatigue crack just occurs at No. 4 rivet; (b) Cracks from No. 4 rivet and No. 5 rivet link together.CC BY · PubMed Central
Figure 8 Total crack length versus fatigue loading cycles.
Figure 8 Total crack length versus fatigue loading cycles.CC BY · PubMed Central
Figure 9 Typical guided wave signals with the increase of the crack lengths: (a) Whole guided wave signal collected during the fatigue test; (b) Magnification of the direct wave packet of the collected GW signal.
Figure 9 Typical guided wave signals with the increase of the crack lengths: (a) Whole guided wave signal collected during the fatigue test; (b) Magnification of the direct wave packet of the collected GW signal.CC BY · PubMed Central
Figure 10 Extracted TFS images with the fatigue crack growth in specimen D1: (a) The TFS image of the GW signal collected when the total crack length is 0 mm obtained by the continuous wavelet transform; (b) The total crack length is 4.2 mm; (c) The total crack length is 9.5 mm; (d) The total crack length is 16 mm; (e) The total crack length is 22 mm; (f) The total crack length is 32 mm.
Figure 10 Extracted TFS images with the fatigue crack growth in specimen D1: (a) The TFS image of the GW signal collected when the total crack length is 0 mm obtained by the continuous wavelet transform; (b) The total crack length is 4.2 mm; (c) The total crack length is 9.5 mm; (d) The total crack length is 16 mm; (e) The total crack length is 22 mm; (f) The total crack length is 32 mm.CC BY · PubMed Central
Figure 11 Extracted differential TFS images with the fatigue crack growth in specimen D1: (a) The total crack length is 0 mm obtained by the continuous wavelet transform; (b) The total crack length is 4.2 mm; (c) The total crack length is 9.5 mm; (d) The total crack length is 16 mm; (e) The total crack length is 22 mm; (f) The total crack length is 32 mm.
Figure 11 Extracted differential TFS images with the fatigue crack growth in specimen D1: (a) The total crack length is 0 mm obtained by the continuous wavelet transform; (b) The total crack length is 4.2 mm; (c) The total crack length is 9.5 mm; (d) The total crack length is 16 mm; (e) The total crack length is 22 mm; (f) The total crack length is 32 mm.CC BY · PubMed Central
Figure 12 Structure of the CNN for fatigue crack evaluation.
Figure 12 Structure of the CNN for fatigue crack evaluation.CC BY · PubMed Central
Figure 13 Loss varying with the training process.
Figure 13 Loss varying with the training process.CC BY · PubMed Central
Figure 14 Fatigue crack evaluation result with the GW–CNN ensemble: (a) Output of the CNNs in the CNN ensemble; (b) Final output of the CNN ensemble.
Figure 14 Fatigue crack evaluation result with the GW–CNN ensemble: (a) Output of the CNNs in the CNN ensemble; (b) Final output of the CNN ensemble.CC BY · PubMed Central
Figure 15 Fatigue cracks of specimen D6 at different stages: (a) Total crack length of 0 mm; (b) Total crack length of 2 mm; (c) Total crack length of 17 mm; (d) Total crack length of 29.5 mm.
Figure 15 Fatigue cracks of specimen D6 at different stages: (a) Total crack length of 0 mm; (b) Total crack length of 2 mm; (c) Total crack length of 17 mm; (d) Total crack length of 29.5 mm.CC BY · PubMed Central
Figure 16 Widely used damage indexes extracted from the GW signal: (a) Root mean square deviation damage index; (b) Drop-in correlation coefficient damage index.
Figure 16 Widely used damage indexes extracted from the GW signal: (a) Root mean square deviation damage index; (b) Drop-in correlation coefficient damage index.CC BY · PubMed Central
Figure 17 Fatigue crack evaluation result compared with the damage index-based method.
Figure 17 Fatigue crack evaluation result compared with the damage index-based method.CC BY · PubMed Central
Figure 18 Fatigue crack evaluation errors: (a) Absolute error; (b) Relative error.
Figure 18 Fatigue crack evaluation errors: (a) Absolute error; (b) Relative error.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 52 references

Research works printed but not linked: 1

  1. Kingma, D.P., and Ba, J.L. (2017). Adam: A Method for Stochastic Optimization. arXiv, Available online: https://arxiv.org/abs/1412.6980. · arXiv

Cited by

See all 29

Metadata

Authors
Chen, J., Wu, W., Ren, Y., & Yuan, S.
Affiliations
State Key Laboratory of Mechanical Structure Mechanics and Control (CN)
Corresponding author
Yuan, Shenfang
Venue
Sensors
Publisher
MDPI AG
Published
2021-12-31
Volume
22
Issue
1
Pages
307
ISSN
1424-8220 (ISSN-L)
OpenAlex
W4205829692
Cited by
29
References
51
Citations by year
2026: 5 · 2025: 9 · 2024: 6 · 2023: 8 · 2022: 1
FWCI
1.83
Type
journal article
Language
en
Open access
Locations
  • Sensors (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Foundation for Innovative Research Groups of the National Natural Science Foundation of China (51921003)
  • Key Program of National Natural Science Foundation of China (51635008)
  • Jiangsu Provincial Key Research and Development Program of China (BE2018123)
  • Research Fund of State Key Laboratory of Mechanics and Control of Mechanical Structures (Nan-jing University of Aeronautics and astronautics) (MCMS-I-0521K01)
  • Priority Academic Program Development of Jiangsu Higher Education Institutions of China (none)
Topics
  • Physical Sciences › Engineering › Mechanics of Materials › Ultrasonics and Acoustic Wave Propagation
  • Physical Sciences › Engineering › Civil and Structural Engineering › Structural Health Monitoring Techniques
  • Physical Sciences › Engineering › Mechanical Engineering › Non-Destructive Testing Techniques
Keywords
guided wave inspection, ensemble convolutional neural networks, crack sizing, structural health monitoring, lap joints, time-frequency analysis
MeSH terms
Machine Learning, Normal Distribution, Neural Networks, Computer, Wavelet Analysis

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.