Fusion and Visualization of Bridge Deck Nondestructive Evaluation Data via Machine Learning

Mohamadi, Sara; Lattanzi, David; Azari, Hoda

doi:10.3389/fmats.2020.576918

Cited by 14 publications

(4 citation statements)

References 75 publications

(80 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The method achieved an optimal recognition accuracy of 88%, and experimental results demonstrated that the CNN model outperformed the traditional peak frequency method in imaging internal defects in concrete slabs. Mohamadi et al 10 utilized support vector machine (SVM) to randomly select salient features of the signal, which were then used to identify unlabeled signals. While these methods significantly improve the accuracy of defect recognition, the designed network structures need a large amount of labeled training data, which are not necessarily available in practice.…”

Section: Introductionmentioning

confidence: 99%

Visualization of concrete internal defects based on unsupervised domain adaptation algorithm for transfer learning of experiment-simulation hybrid dataset of impact echo signals

Shang,

Chen

2023

Structural Health Monitoring

View full text Add to dashboard Cite

Detecting concrete internal defects through deep learning analysis of impact echo signals faces two challenges: (1) the traditional signal processing method such as wavelet transform (WT) fails to reflect data-sensitive damage characteristics due to the uncertainty principle and (2) the limited labeled data acquired from real structures impedes network training. To address the first challenge, this paper proposes the WT-based synchrosqueezing transform (WT-SST) for the conversion of time-series data to the time-frequency spectrogram, which can provide effective features for the network in time and frequency domains simultaneously. To overcome the second challenge, numerical simulation data are supplemented for the augment of labeled data. To minimize the effect of data variance between experiments and simulations, this paper uses an unsupervised domain adaptation (DA) network for the transfer training of labeled simulation data (original domain) and unlabeled experimental data (target domain). The DA network extracts domain-invariant features by maximizing the domain recognition error and minimizing the probability distribution distance. The damage probability is calculated by the trained model to produce a 2D defect contour image of concrete specimens, and the three-dimensional visualization of internal defects by estimating the defect depth based on the defect area of contour image. Finally, the recognition precision, recall, F1-score, and accuracy of the model of unsupervised DA network trained by a hybrid dataset reaches 89.4%, 88.4%, 88.9%, and 94.7%, respectively.

show abstract

Section: Introductionmentioning

confidence: 99%

Visualization of concrete internal defects based on unsupervised domain adaptation algorithm for transfer learning of experiment-simulation hybrid dataset of impact echo signals

Shang,

Chen

2023

Structural Health Monitoring

View full text Add to dashboard Cite

show abstract

“…ML has been adopted in NDT over the past decade, and progress has been achieved and demonstrated in particular studies [17], [18], [19] The applications target the holistic assessment of structures in terms of data fusion concepts [20] and the analysis of specific inspection techniques [21] With a wide range of studies conducted, there is still an urgent need for development regarding applying ML to NDT of concrete structures and, more specifically, NDT methods using elastic waves. With powerful algorithms available, ML models can be tailored to specific applications.…”

Section: Introductionmentioning

confidence: 99%

Machine Learning Applications in Nondestructive Testing of Concrete Structures

Algernon,

Münch,

Muller

et al. 2023

ce papers

View full text Add to dashboard Cite

Machine Learning bears great potential for data‐driven solutions in the field of nondestructive testing (NDT) of concrete structures. The analysis of the data collected with NDT methods, such as ultrasonics, impact‐echo and ground penetrating radar, can be complex and requires experience. The expected benefit of Machine Learning applications in this context goes beyond the increase of efficiency obtained by automating the analysis process. While traditional analysis approaches are usually solely based on key features according to the basic principles, Machine Learning algorithms can consider the full data content and reveal hidden correlations. For an organized approach to Machine Learning on NDT data, an analysis tool has been developed. The tool provides a graphical user interface to manage and label training/test data and interactively define the Deep Neural Network architecture. In particular, Convolutional Neural Networks, as proven successful in various image recognition tasks, are implemented. The Machine Learning concepts are demonstrated in show cases, comprising ultrasonic as well as impact‐echo applications. In particular, the relevance of a targeted preprocessing is addressed, comparing the effectiveness of time‐, frequency‐ and joint‐time‐frequency‐representations.

show abstract

“…The results presented in [38] demonstrate the possibility to have a better design of the SHM system for the damage localization where the benefits of different damage indices and evaluation methodologies are explored by means of fusion of the identified results. In [39] a framework is presented to support fusion of multiple nondestructive evaluation techniques Fig. 1 The different levels of data fusion [33]: data-level, feature-level, decision-level.…”

Section: Introductionmentioning

confidence: 99%

A Bayesian-based inspection-monitoring data fusion approach for historical buildings and its post-earthquake application to a monumental masonry palace

Ierimonti

Cavalagli

Venanzi

et al. 2022

Bull Earthquake Eng

View full text Add to dashboard Cite

A Bayesian-based inspection-monitoring data fusion approach model updating and visual inspections.In a preliminary phase, potential damage-sensitive regions in the structure are identified through FE-based numerical analysis and engineering judgment. Then, the solution of the inverse problem aimed at deriving the Bayesian posterior statistics of the uncertain parameters is entrusted to a computational-effective surrogate model (SM). Finally, the Bayesian-based updated parameters are adjusted considering the different allowable sources of information to achieve a final assessment. The effectiveness of the proposed approach is demonstrated by using the recorded data acquired in the Consoli Palace, an historical building located in Umbria, central Italy, which has been continuously monitored since 2017 using dynamic, static and environmental sensors and which has been hit by low-intensity earthquakes in 2021.

show abstract

Fusion and Visualization of Bridge Deck Nondestructive Evaluation Data via Machine Learning

Cited by 14 publications

References 75 publications

Visualization of concrete internal defects based on unsupervised domain adaptation algorithm for transfer learning of experiment-simulation hybrid dataset of impact echo signals

Visualization of concrete internal defects based on unsupervised domain adaptation algorithm for transfer learning of experiment-simulation hybrid dataset of impact echo signals

Machine Learning Applications in Nondestructive Testing of Concrete Structures

A Bayesian-based inspection-monitoring data fusion approach for historical buildings and its post-earthquake application to a monumental masonry palace

Contact Info

Product

Resources

About