A Model Classifying Four Classes of Defects in Reinforced Concrete Bridge Elements Using Convolutional Neural Networks
Abstract
:1. Introduction
2. Background
- –
- –
- CNNs can learn the normal patterns and visual characteristics of damaged bridge components [20]. By analyzing images or videos of bridges, CNNs can identify anomalies that deviate from the expected patterns, highlighting potential areas of concern that require further inspection or assessment.
- –
- CNNs can perform image segmentation to separate bridge components or regions of interest from the background [21]. This can be useful in identifying specific areas or elements for further analysis, such as detecting cracks or corrosion within a specific part of the bridge. For example, Geng et al. used CNNs for detecting, locating and quantifying corrosion damage in a truss-type bridge through the autocorrelation of vibration signals [22].
- –
- By comparing images or videos captured during different inspection cycles, CNNs can identify changes in the condition of bridge elements over time [23]. This enables the detection of progressive deterioration or changes that might require attention or further investigation.
3. Methodology
3.1. General Description of the Process
- -
- defects that are most often found in various types of reinforced concrete structures [24];
- -
- the ability of computer vision to clearly separate one class of defects from another. The division into classes was motivated by the Guide by the ACI Committee, “Guide for Making a Condition Survey of Concrete in Service” [25]. Its purpose was to standardize the reporting of the condition of the concrete in a structure. The Guide collected a large database on the external indications of damage to concrete, illustrated with photographs.
- -
- “crack”. This class includes all the defects from group A.1 (various type of cracks);
- -
- “spalling”. This class includes two groups of defects—A.2.20 (scaling) and A.2.21 (spalling);
- -
- “popuot”. This class includes defects from group A.2.19 (popout—the breaking away of small portions of a concrete surface that leaves shallow, typically conical, depressions).
- Data collection. It was proven that the presence of a large and high-quality dataset is the key to the success of the CNN modeling process. A large amount is used to mean a dataset that contains from several thousand to several tens of thousands of images [26,27]. The quality of the dataset is based on the fact that the images should be as close to real conditions as possible: images with background noise and different shades of color, including surface roughness, different lighting conditions, background debris, etc. [28,29].
- 2.
- Preliminary data preparation. The next step was to manually crop 256 by 256 pixel defects from the images. This enabled us to obtain a dataset containing 1300 “defect free” images, “crack”, “spalling” and 800 “popout” images. An example of four types of images is shown in Figure 5. At the next stage of the study, “accuracy” metric was applied to evaluate the quality of the model, which is quite sensitive for unbalanced datasets. Therefore, it was decided to equalize the number of images in each class by increasing the number of images in the “popout” set. For this purpose, part of the images was rotated by 90° and saved in files with a new name. The entire dataset was divided into three parts—training, testing and validation in the proportion of 80%/10%/10%, respectively.
- 3.
- Choice of CNN model architecture (described in Section 3.2).
- 4.
- Selection of the main parameters of the CNN model: activation function, loss function, optimization algorithm and quality measure. This process is described in Section 3.3.
- 5.
- Training of the models. The selected models were trained using supervised learning methods. Since one of the research objectives was to determine the training time of several types of models over 20 epochs, the callback function was not used.
- 6.
- Evaluation of the models. Pre-trained models were evaluated on a separate validation dataset to assess their performance and identify any potential problems.
- 7.
- Using of the model. After the evaluation, a model with the highest performance was used to classify the images it “has not seen”.
3.2. CNN Architecture Used in Image Classification Task
3.3. Model Parameters and Evaluation Metric
4. Results
5. Discussion
6. Conclusions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Moshynskyi, V.; Striletskyi, P.; Trach, R. Application of the Building Information Modelling (BIM) for Bridge Structures. Acta Sci. Pol.-Archit. Bud. 2022, 20, 3–9. [Google Scholar] [CrossRef]
- DSTU-N B V. 1.2-18:2016; Guidelines Regarding the Inspection of Building Objects to Determine and Assess Their Technical Condition. Ministry of Regional Development and Construction of Ukraine: Kyiv, Ukraine, 2016.
- World Bank Group. Ukraine Rapid Damage and Needs Assessment: February 2022–February 2023; World Bank Group: Washington, DC, USA, 2023. [Google Scholar]
- Szymanek, S. Construction Production Trends and Industry Optimism in EU Countries after the COVID-19 Pandemic. Acta Sci. Pol. Archit. 2022, 21, 21. [Google Scholar] [CrossRef]
- Bodnar, L.; Koval, P.; Stepanov, S.; Panibratets, L. Operational state of bridges of Ukraine. Avtošljachovyk Ukr. 2019, 2, 57–68. [Google Scholar] [CrossRef]
- Rashidi, M.; Samali, B.; Sharafi, P. A New Model for Bridge Management: Part A: Condition Assessment and Priority Ranking of Bridges. Aust. J. Civ. Eng. 2016, 14, 35–45. [Google Scholar] [CrossRef] [Green Version]
- Frangopol, D.M.; Kong, J.S.; Gharaibeh, E.S. Reliability-Based Life-Cycle Management of Highway Bridges. J. Comput. Civ. Eng. 2001, 15, 27–34. [Google Scholar] [CrossRef]
- Wakjira, T.G.; Nehdi, M.L.; Ebead, U. Fractional Factorial Design Model for Seismic Performance of RC Bridge Piers Retrofitted with Steel-Reinforced Polymer Composites. Eng. Struct. 2020, 221, 111100. [Google Scholar] [CrossRef]
- Rokneddin, K.; Ghosh, J.; Dueñas-Osorio, L.; Padgett, J.E. Bridge Retrofit Prioritisation for Ageing Transportation Networks Subject to Seismic Hazards. Struct. Infrastruct. Eng. 2013, 9, 1050–1066. [Google Scholar] [CrossRef]
- Forcellini, D.; Alzabeebee, S. Seismic Fragility Assessment of Geotechnical Seismic Isolation (GSI) for Bridge Configuration. Bull. Earthq. Eng. 2023, 21, 3969–3990. [Google Scholar] [CrossRef]
- Mackie, K.R.; Lu, J.; Elgamal, A. Performance-Based Earthquake Assessment of Bridge Systems Including Ground-Foundation Interaction. Soil Dyn. Earthq. Eng. 2012, 42, 184–196. [Google Scholar] [CrossRef]
- Trach, R.; Moshynskyi, V.; Chernyshev, D.; Borysyuk, O.; Trach, Y.; Striletskyi, P.; Tyvoniuk, V. Modeling the Quantitative Assessment of the Condition of Bridge Components Made of Reinforced Concrete Using ANN. Sustainability 2022, 14, 15779. [Google Scholar] [CrossRef]
- Trach, R.; Ryzhakova, G.; Trach, Y.; Shpakov, A.; Tyvoniuk, V. Modeling the Cause-and-Effect Relationships between the Causes of Damage and External Indicators of RC Elements Using ML Tools. Sustainability 2023, 15, 5250. [Google Scholar] [CrossRef]
- Assaad, R.; El-adaway, I.H. Bridge Infrastructure Asset Management System: Comparative Computational Machine Learning Approach for Evaluating and Predicting Deck Deterioration Conditions. J. Infrastruct. Syst. 2020, 26, 04020032. [Google Scholar] [CrossRef]
- Yang, D.Y. Deep Reinforcement Learning–Enabled Bridge Management Considering Asset and Network Risks. J. Infrastruct. Syst. 2022, 28, 04022023. [Google Scholar] [CrossRef]
- Pena-Caballero, C.; Kim, D.; Gonzalez, A.; Castellanos, O.; Cantu, A.; Ho, J. Real-Time Road Hazard Information System. Infrastructures 2020, 5, 75. [Google Scholar] [CrossRef]
- Zhang, Y.; Li, Y.; Zhou, X.; Luo, J. cST-ML: Continuous spatial-temporal meta-learning for traffic dynamics prediction. In Proceedings of the 2020 IEEE International Conference on Data Mining (ICDM), Sorrento, Italy, 17–20 November 2020; pp. 1418–1423. [Google Scholar]
- Gomez-Cabrera, A.; Escamilla-Ambrosio, P.J. Review of Machine-Learning Techniques Applied to Structural Health Monitoring Systems for Building and Bridge Structures. Appl. Sci. 2022, 12, 10754. [Google Scholar] [CrossRef]
- Hajializadeh, D. Deep Learning-Based Indirect Bridge Damage Identification System. Struct. Health Monit. 2023, 22, 897–912. [Google Scholar] [CrossRef]
- Hammouch, W.; Chouiekh, C.; Khaissidi, G.; Mrabti, M. Crack Detection and Classification in Moroccan Pavement Using Convolutional Neural Network. Infrastructures 2022, 7, 152. [Google Scholar] [CrossRef]
- Dzięcioł, J.; Sas, W. Perspective on the Application of Machine Learning Algorithms for Flow Parameter Estimation in Recycled Concrete Aggregate. Materials 2023, 16, 1500. [Google Scholar] [CrossRef]
- Nguyen, T.; Nguyen, T.; Sidorov, D.N.; Dreglea, A. Machine Learning Algorithms Application to Road Defects Classification. Intell. Decis. Technol. 2018, 12, 59–66. [Google Scholar] [CrossRef]
- Duan, Y.; Chen, Q.; Zhang, H.; Yun, C.B.; Wu, S.; Zhu, Q. CNN-Based Damage Identification Method of Tied-Arch Bridge Using Spatial-Spectral Information. Smart Struct. Syst. 2019, 23, 507–520. [Google Scholar]
- Geng, Q.; Zhou, Z.; Cao, X. Survey of Recent Progress in Semantic Image Segmentation with CNNs. Sci. China Inf. Sci. 2018, 61, 051101. [Google Scholar] [CrossRef] [Green Version]
- Yanez-Borjas, J.J.; Valtierra-Rodriguez, M.; Machorro-Lopez, J.M.; Camarena-Martinez, D.; Amezquita-Sanchez, J.P. Convolutional Neural Network-Based Methodology for Detecting, Locating and Quantifying Corrosion Damage in a Truss-Type Bridge Through the Autocorrelation of Vibration Signals. Arab. J. Sci. Eng. 2023, 48, 1119–1141. [Google Scholar] [CrossRef]
- Yessoufou, F.; Zhu, J. One-Class Convolutional Neural Network (OC-CNN) Model for Rapid Bridge Damage Detection Using Bridge Response Data. KSCE J. Civ. Eng. 2023, 27, 1640–1660. [Google Scholar] [CrossRef]
- Żółtowski, B.; Castañeda, L.F.; Żółtowski, M.; Wierzbicki, T. Research Condition of Complex Technical Objects. Acta Sci. Pol. Archit. 2022, 21, 21. [Google Scholar] [CrossRef]
- ACI 201.1 R-92; ACI Committee 201 Guide for Making a Condition Survey of Concrete in Service. American Concrete Institute: Farmington Hills, MI, USA, 1997.
- Fujita, Y.; Hamamoto, Y. A Robust Automatic Crack Detection Method from Noisy Concrete Surfaces. Mach. Vis. Appl. 2011, 22, 245–254. [Google Scholar] [CrossRef]
- Nishikawa, T.; Yoshida, J.; Sugiyama, T.; Fujino, Y. Concrete Crack Detection by Multiple Sequential Image Filtering: Concrete Crack Detection by Image Processing. Comput.-Aided Civ. Infrastruct. Eng. 2012, 27, 29–47. [Google Scholar] [CrossRef]
- Luo, H.; Xiong, C.; Fang, W.; Love, P.E.D.; Zhang, B.; Ouyang, X. Convolutional Neural Networks: Computer Vision-Based Workforce Activity Assessment in Construction. Autom. Constr. 2018, 94, 282–289. [Google Scholar] [CrossRef]
- Zhang, K.; Zhang, Y.; Cheng, H.-D. CrackGAN: Pavement Crack Detection Using Partially Accurate Ground Truths Based on Generative Adversarial Learning. IEEE Trans. Intell. Transp. Syst. 2021, 22, 1306–1319. [Google Scholar] [CrossRef]
- Simonyan, K.; Zisserman, A. Very Deep Convolutional Networks for Large-Scale Image Recognition. arXiv 2014. [Google Scholar] [CrossRef]
- Szegedy, C.; Liu, W.; Jia, Y.; Sermanet, P.; Reed, S.; Anguelov, D.; Erhan, D.; Vanhoucke, V.; Rabinovich, A. Going deeper with convolutions. In Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, 7–12 June 2015; pp. 1–9. [Google Scholar] [CrossRef] [Green Version]
- He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, 27–30 June 2016; pp. 770–778. [Google Scholar]
- Chollet, F. Xception: Deep Learning with Depthwise Separable Convolutions. arXiv 2016. [Google Scholar] [CrossRef]
- Huang, G.; Liu, Z.; van der Maaten, L.; Weinberger, K.Q. Densely Connected Convolutional Networks. arXiv 2016. [Google Scholar] [CrossRef]
- Michele, A.; Colin, V.; Santika, D.D. MobileNet Convolutional Neural Networks and Support Vector Machines for Palmprint Recognition. Procedia Comput. Sci. 2019, 157, 110–117. [Google Scholar] [CrossRef]
- Zoph, B.; Vasudevan, V.; Shlens, J.; Le, Q.V. Learning Transferable Architectures for Scalable Image Recognition. arXiv 2017. [Google Scholar] [CrossRef]
- Tan, M.; Le, Q.V. EfficientNetV2: Smaller Models and Faster Training. arXiv 2021. [Google Scholar] [CrossRef]
- Kouretas, I.; Paliouras, V. Simplified hardware implementation of the softmax activation function. In Proceedings of the 2019 8th International Conference on Modern Circuits and Systems Technologies (MOCAST), Thessaloniki, Greece, 13–15 May 2019; pp. 1–4. [Google Scholar]
- Kowalski, J.; Połoński, M.; Lendo-Siwicka, M.; Trach, R.; Wrzesiński, G. Method of Assessing the Risk of Implementing Railway Investments in Terms of the Cost of Their Implementation. Sustainability 2021, 13, 13085. [Google Scholar] [CrossRef]
- Rusiecki, A. Trimmed Categorical Cross-entropy for Deep Learning with Label Noise. Electron. Lett. 2019, 55, 319–320. [Google Scholar] [CrossRef]
- Alay, N.; Al-Baity, H.H. Deep Learning Approach for Multimodal Biometric Recognition System Based on Fusion of Iris, Face, and Finger Vein Traits. Sensors 2020, 20, 5523. [Google Scholar] [CrossRef]
- Trach, Y.; Melnychuk, V.; Michel, M.M.; Reczek, L.; Siwiec, T.; Trach, R. The Characterization of Ukrainian Volcanic Tuffs from the Khmelnytsky Region with the Theoretical Analysis of Their Application in Construction and Environmental Technologies. Materials 2021, 14, 7723. [Google Scholar] [CrossRef]
- Trach, R.; Lendo-Siwicka, M.; Pawluk, K.; Bilous, N. Assessment of the Effect of Integration Realisation in Construction Projects. Teh. Glas. 2019, 13, 254–259. [Google Scholar] [CrossRef]
- Gunawardana, A.; Shani, G. A Survey of Accuracy Evaluation Metrics of Recommendation Tasks. J. Mach. Learn. Res. 2009, 10, 2935–2962. [Google Scholar]
- Li, S.; Zhao, X. Automatic Crack Detection and Measurement of Concrete Structure Using Convolutional Encoder-Decoder Network. IEEE Access 2020, 8, 134602–134618. [Google Scholar] [CrossRef]
- Yang, C.; Chen, J.; Li, Z.; Huang, Y. Structural Crack Detection and Recognition Based on Deep Learning. Appl. Sci. 2021, 11, 2868. [Google Scholar] [CrossRef]
DenseNet 201 | EfficientNet V2 | Inception V3 | Mobile Net | NASNet Large | ResNet 50 | VGG16 | Xception | |
---|---|---|---|---|---|---|---|---|
Total times, s | 2405 | 5949 | 1252 | 1250 | 4996 | 1588 | 2026 | 1522 |
Mean times, s | 120 | 297 | 63 | 63 | 250 | 79 | 101 | 76 |
Categorical cross-entropy | 0.078 | 0.2343 | 0.1414 | 0.0264 | 0.065 | 0.0422 | 0.0525 | 0.0313 |
Accuracy, % | 91.82 | 86.21 | 89.54 | 94.61 | 92.28 | 93.1 | 92.73 | 93.49 |
Defect Free (Class 0) | Crack (Class 1) | Spalling (Class 2) | Popout (Class 3) | Predicted Class | |
---|---|---|---|---|---|
9.994 × 10−1 | 2.474 × 10−8 | 5.386 × 10−4 | 9.499 × 10−7 | (Class 0) | |
7.336 × 10−6 | 9.999 × 10−1 | 8.023 × 10−8 | 2.186 × 10−16 | (Class 1) | |
8.645 × 10−3 | 9.883 × 10−7 | 9.913 × 10−1 | 2.673 × 10−11 | (Class 2) | |
9.499 × 10−11 | 1.737 × 10−7 | 3.734 × 10−1 | 6.265 × 10−1 | (Class 3) |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Trach, R. A Model Classifying Four Classes of Defects in Reinforced Concrete Bridge Elements Using Convolutional Neural Networks. Infrastructures 2023, 8, 123. https://doi.org/10.3390/infrastructures8080123
Trach R. A Model Classifying Four Classes of Defects in Reinforced Concrete Bridge Elements Using Convolutional Neural Networks. Infrastructures. 2023; 8(8):123. https://doi.org/10.3390/infrastructures8080123
Chicago/Turabian StyleTrach, Roman. 2023. "A Model Classifying Four Classes of Defects in Reinforced Concrete Bridge Elements Using Convolutional Neural Networks" Infrastructures 8, no. 8: 123. https://doi.org/10.3390/infrastructures8080123