Predictive Modelling for Concrete Failure at Anchorages Using Machine Learning Techniques
Abstract
:1. Introduction
1.1. Behaviour of Anchors in Tension Subject to Concrete Cone Failure
1.2. Analogies to Shear Concrete Edge Failure
2. Development of the ML-Based Predictive Models
2.1. Processing and Preparation of the Experimental Database
2.2. Machine Learning (ML) Techniques
2.2.1. Gaussian Process Regression (GPR)
2.2.2. Support Vector Regression (SVR)
2.2.3. Other Algorithms Considered
2.3. Implementation of the ML Algorithms
2.4. Performance Evaluation Measures
3. Results and Discussion
3.1. Performance Evaluation of the Developed Models: Comparison of GPR and SVR Model Predictions to Experimental Data
3.2. Influence of Input Variables on Model Performance
3.3. Comparison of the Developed ML Based Predictions to Existing Methods
4. Model Explainability Based on Analogous Rational and Mechanical Phenomena (MEARM)
5. Implementation of the Machine Learning Algorithms as General Probabilistic Models
5.1. Association of Predictive Efficiency to Model Uncertainty Characteristics
5.2. Trends in the Model Uncertainty with Basic Input Variables
6. Conclusions
- The GPR and SVR predictions are in good agreement with the experimentally observed tensile capacities. The results demonstrated that the ML-based models learnt and predicted the experimental data reasonably well.
- The assessment revealed that the GPR model yielded the best result with an R2 value of 0.97 (the closest to the value of one) and MAE values as low as 5.56 (nearest value to zero). It can be said that the GPR model reasonably predicted the experimental data better than the other models investigated.
- The model explainability was also described on the basis of the ML models’ correspondence to existing predictive design equations. Moreover, a very good coincidence has been established between the efficiency of the algorithms predicting the experimental data and the mechanical characteristics of the underlying failure phenomenon. This is ascribed to the Model Explainability based on Analogous Rational and Mechanical phenomena (MEARM), which is for the first time defined and used herein on structural engineering problems.
- Both the SVR and GPR predictive models can technically be used as the basis for establishing probabilistic models GPMs for reliability assessment of fastening design formulations. This is justified by the fact that they deliver low-error predictions, and their predictions are—if at all—only mildly correlated to, and as such biased by the input attributes. These attributes or predictors can be used as stochastic variables. Furthermore, the corresponding model uncertainty characteristics have been quantified.
- The model uncertainty related to the GPR model has a mean value of (closer to the mean value of 1) and the lowest dispersion of all the models investigated with . The model has no major trends with its input parameters; and thus, the most suitable as a GPM for reliability assessment of fastening design formulations. The SVR model has the highest dispersion and the CCD method has some trend with the embedment depth.
- The goodness of fit test indicates that the GPR and CCD model has an underlying normal distribution, whereas the CCD model has a lognormal distribution.
Author Contributions
Funding
Data Availability Statement
Acknowledgments
Conflicts of Interest
Nomenclature
standard deviation of the prediction error | |
mean value of the prediction error | |
edge distance of a single anchor | |
anchor diameter | |
Pearson’s correlation coefficient | |
is the concrete compressive strength. | |
is a product-specific value for non-cracked concrete | |
is the characteristic compressive strength from 150 mm concrete cube specimens | |
mean value of concrete compressive cylinder strength | |
is the mean compressive strength measured at 200 mm concrete cube specimens | |
is the influence length of the anchor loaded in shear | |
is equal to the value of 13.5 for metal expansion anchors and bonded anchors | |
is equal to the value of 1.0 for an undisturbed uniaxial compression stress state (this occurs for equal to approximately 5 anchor diameters) is equal to the value of 0.8 for anchors within compression zones with cracks developing parallel to the compression direction (this occurs for greater than 5 anchor diameters) | |
are calibration factors | |
is the anchor embedment depth (the distance of the failure invitation point to the free surface) | |
mean shear breakout resistance | |
tensile breakout resistance | |
mean tensile breakout resistance | |
characteristic tensile breakout resistance | |
experimentally observed failure load for a single anchor test in tension | |
represents the input variable set which include the embedment depth , anchor and the concrete strength | |
represents mean concrete breakout model prediction for the same anchor test |
References
- Eligehausen, R.; Malleé, R.; Silva, J.F. Anchorages in Concrete Construction; Ernst & Sohn: Berlin, Germany, 2006. [Google Scholar]
- European Committee for Standardization (CEN). En 1992-4. Eurocode 2: Design of Concrete Structures—Part 4 Design of Fastenings for Use in Concrete; CEN/TC 250 2019; European Committee for Standardisation: Brussel, Belgium, 2019. [Google Scholar]
- ACI Committee and International Organization for Standardization. Building Code Requirements for Structural Concrete (ACI 318-14) and Commentary; American Concrete Institute: Farmington Hills, MI, USA, 2014. [Google Scholar]
- BSI—Bristish Standards Institute. BS 8539:2012: Code of Practice for the Selection and Installation of Post-Installed Anchors in Concrete and Masonry; Bristish Standards Institute: London, UK, 2012. [Google Scholar]
- SAISC (Southern African Institute of Steel Construction). Structural Steel Connections: Limit State Design; SAISC: Johannesburg, South Africa, 2012. [Google Scholar]
- Standards Australia. As 5216:2018 Design of Post-Installed and Cast-In Fastenings in Concrete; Standards Australia: Sydney, Australia, 2018. [Google Scholar]
- Comitéeuro-International du Béton. Fastenings to Concrete and Masonry Structures: State of the Art Report (No. 216); Comitéeuro-International du Béton: Thomas Telford, UK, 1994. [Google Scholar]
- Adeli, H. Artificial intelligence in structural engineering. Eng. Anal. 1986, 3, 154–160. [Google Scholar] [CrossRef]
- Fruchter, R.; Gluck, J.; Gold, Y.I. Application of AI programming techniques to the analysis of structures. Comput. Struct. 1988, 30, 747–753. [Google Scholar] [CrossRef]
- Chou, J.S.; Ngo, N.T.; Pham, A.D. Shear strength prediction in reinforced concrete deep beams using nature-inspired metaheuristic support vector regression. J. Comput. Civ. Eng. 2016, 30, 04015002. [Google Scholar] [CrossRef]
- Novák, D.; Lehký, D. ANN inverse analysis based on stochastic small-sample training set simulation. Eng. Appl. Artif. Intell. 2006, 19, 731–740. [Google Scholar] [CrossRef]
- Wang, Z.; Zentner, I.; Pedroni, N.; Zio, E. Adaptive artificial neural networks for seismic fragility analysis. In Proceedings of the 2017 2nd International Conference on System Reliability and Safety (ICSRS), Milan, Italy, 20–22 December 2017; pp. 414–420. [Google Scholar]
- Kawamura, K.; Miyamoto, A.; Frangopol, D.M.; Kimura, R. Performance evaluation of concrete slabs of existing bridges using neural networks. Eng. Struct. 2003, 25, 1455–1477. [Google Scholar] [CrossRef] [Green Version]
- Papadrakakis, M.; Lagaros, N.D. Reliability-based structural optimisation using neural networks and Monte Carlo simulation. Comput. Methods Appl. Mech. Eng. 2002, 191, 3491–3507. [Google Scholar] [CrossRef]
- Golafshani, E.M.; Rahai, A.; Sebt, M.H.; Akbarpour, H. Prediction of bond strength of spliced steel bars in concrete using artificial neural network and fuzzy logic. Constr. Build. Mater. 2012, 36, 411–418. [Google Scholar] [CrossRef]
- Sakla, S.S.; Ashour, A.F. Prediction of tensile capacity of single adhesive anchors using neural networks. Comput. Struct. 2005, 83, 1792–1803. [Google Scholar] [CrossRef]
- Eligehausen, R.; Fuchs, W.; Lotze, D.; Reuter, M. Befestigungen in der Betonzugzone. Beton Stahlbetonbau 1989, 84. [Google Scholar] [CrossRef]
- Ashour, A.F.; Alqedra, M.A. Concrete breakout strength of single anchors in tension using neural networks. Adv. Eng. Softw. 2005, 36, 87–97. [Google Scholar] [CrossRef]
- Alqedra, M.A.; Ashour, A.F. Prediction of shear capacity of single anchors located near a concrete edge using neural networks. Comput. Struct. 2005, 83, 2495–2502. [Google Scholar] [CrossRef]
- Fuchs, W.; Eligehausen, R.; Breen, J.E. Concrete capacity design (CCD) approach for fastening to concrete. Struct. J. 1995, 92, 73–94. [Google Scholar]
- Gesoğlu, M.; Güneyisi, E. Prediction of load-carrying capacity of adhesive anchors by soft computing techniques. Mater. Struct. 2007, 40, 939–951. [Google Scholar] [CrossRef]
- Güneyisi, E.M.; Gesoğlu, M.; Güneyisi, E.; Mermerdaş, K. Assessment of shear capacity of adhesive anchors for structures using neural network-based model. Mater. Struct. 2016, 49, 1065–1077. [Google Scholar] [CrossRef]
- Samui, P. Support vector machine applied to settlement of shallow foundations on cohesionless soils. Comput. Geotech. 2008, 35, 419–427. [Google Scholar] [CrossRef]
- Hu, B.; Su, G.S.; Jiang, J.; Xiao, Y. Gaussian Process-Based Response Surface Method for Slope Reliability Analysis. Adv. Civ. Eng. 2019, 2019, 9185756. [Google Scholar] [CrossRef]
- Guo, M.; Hesthaven, J.S. Reduced order modeling for nonlinear structural analysis using gaussian process regression. Comput. Methods Appl. Mech. Eng. 2018, 341, 807–826. [Google Scholar] [CrossRef]
- Hoang, N.D.; Pham, A.D.; Nguyen, Q.L.; Pham, Q.N. Estimating compressive strength of high performance concrete with Gaussian process regression model. Adv. Civ. Eng. 2016, 2016, 2861380. [Google Scholar] [CrossRef] [Green Version]
- Yan, K.; Xu, H.; Shen, G.; Liu, P. Prediction of splitting tensile strength from cylinder compressive strength of concrete by support vector machine. Adv. Mater. Sci. Eng. 2013, 2013, 597257. [Google Scholar] [CrossRef] [Green Version]
- Cheng, M.Y.; Chou, J.S.; Roy, A.F.; Wu, Y.W. High-performance concrete compressive strength prediction using time-weighted evolutionary fuzzy support vector machines inference model. Autom. Constr. 2012, 28, 106–115. [Google Scholar] [CrossRef]
- Olalusi, B.; Spyridis, P. Machine learning-based models for the concrete breakout capacity prediction of single anchors in shear. Adv. Eng. Softw. 2020, 147. [Google Scholar] [CrossRef]
- Cheng, M.Y.; Cao, M.T. Evolutionary multivariate adaptive regression splines for estimating shear strength in reinforced-concrete deep beams. Eng. Appl. Artif. Intell. 2014, 28, 86–96. [Google Scholar] [CrossRef]
- Grosser, P.R. Load-Bearing Behavior and Design of Anchorages Subjected to Shear And Torsion Loading in Uncracked Concrete. Ph.D. Thesis, University of Stuttgart, Stuart, German, 2012. [Google Scholar]
- Hofmann, J. Tragverhalten und Bemessung von Befestigungen unter beliebiger Querbelastung in ungerissenem Beton (Load—Bearing Behavior and Design of Fastenings Under Arbitrary Shear Loading in Uncracked Concrete). Ph.D. Thesis, University of Stuttgart, Stuart, Germany, 2004. (In German). [Google Scholar]
- Eligehausen, R.; Ožbolt, J. Design of fastenings based on the fracture mechanics. In Proceedings of the 11th International Conference on Fracture (ICF11), Turin, Italy, 20–25 March 2005; pp. 1405–1410. [Google Scholar]
- Yang, K.H.; Ashour, A.F. Mechanism Analysis for Concrete Breakout Capacity of Single Anchors in Tension. ACI Struct. J. 2008, 105, 609–616. [Google Scholar]
- Appl, J.J. Tragverhalten von Verbunddübeln unter Zugbelastung (Load-Bearing Behavior of Bonded Anchors under Tension Laod). Ph.D. Thesis, University of Stuttgart, Stuttgart, Germany, 2009. (In German). [Google Scholar]
- Piccinin, R.; Ballarini, R.; Cattaneo, S. Linear elastic fracture mechanics pullout analyses of headed anchors in stressed concrete. J. Eng. Mech. 2010, 136, 761–768. [Google Scholar] [CrossRef] [Green Version]
- Marcon, M.; Vorel, J.; Ninčević, K.; Wan-Wendner, R. Modeling adhesive anchors in a discrete element framework. Materials 2017, 10, 917. [Google Scholar] [CrossRef] [Green Version]
- Ninčević, K.; Boumakis, I.; Marcon, M.; Wan-Wendner, R. Aggregate effect on concrete cone capacity. Eng. Struct. 2019, 191, 358–369. [Google Scholar] [CrossRef]
- Guggenberger, T. Einfluss der Zementart auf das Tragverhalten von Befestigungselementen in Beton (Influence of the Cement Type on the Load-Bearing Behaviour of Fasteners in Concrete). Ph.D. Thesis, University of Natural Resources and Life Sciences Vienna, Vienna, Austria, 2009. [Google Scholar]
- Su, G.; Yu, B.; Xiao, Y.; Yan, L. Gaussian process machine-learning method for structural reliability analysis. Adv. Struct. Eng. 2014, 17, 1257–1270. [Google Scholar] [CrossRef]
- Su, G.; Peng, L.; Hu, L. A Gaussian process-based dynamic surrogate model for complex engineering structural reliability analysis. Struct. Saf. 2017, 68, 97–109. [Google Scholar] [CrossRef]
- Rasmussen, C.E.; Williams, C. Gaussian Processes for Machine Learning; Series Adaptive Computation and Machine Learning; MIT Press: Cambridge, MA, USA, 2006; Volume 38, pp. 715–719. [Google Scholar]
- Olalusi, O.B.; Awoyera, P.O. Shear capacity prediction of slender reinforced concrete structures with steel fibers using machine learning. Eng. Struct. 2020, 227, 111470. [Google Scholar] [CrossRef]
- Bentsen, F.H. Model Construction with Support Vector Machines and Gaussian Processes through Kernel Search. Master’s Thesis, The University of Bergen, Bergen, Norway, 2019. [Google Scholar]
- Dibike, Y.B.; Velickov, S.; Solomatine, D.; Abbott, M.B. Model induction with support vector machines: Introduction and applications. J. Comput. Civ. Eng. 2001, 15, 208–216. [Google Scholar] [CrossRef]
- Vapnik, V.; Golowich, S.E.; Smola, A.J. Support vector method for function approximation, regression estimation and signal processing. In Advances in Neural Information Processing Systems; MIT Press: Cambridge, MA, USA, 1997; pp. 281–287. [Google Scholar]
- Mozumder, R.A.; Roy, B.; Laskar, A.I. Support vector regression approach to predict the strength of FRP confined concrete. Arab. J. Sci. Eng. 2017, 42, 1129–1146. [Google Scholar] [CrossRef]
- Du, K.L.; Swamy, M.N. Neural Networks and Statistical Learning; Springer: Berlin, Germany, 2013. [Google Scholar]
- Kuhn, M.; Johnson, K. Applied Predictive Modeling; Springer: New York, NY, USA, 2013. [Google Scholar]
- Olalusi, O.B.; Spyridis, P. Uncertainty modelling and analysis of the concrete edge breakout resistance of single anchors in shear. Eng. Struct. 2020, 222, 111112. [Google Scholar] [CrossRef]
- Olalusi, O.B.; Viljoen, C. Model uncertainties and bias in SHEAR strength predictions of slender stirrup reinforced concrete beams. Struct. Concr. 2020, 21, 316–332. [Google Scholar] [CrossRef]
- Olalusi, O.B. Reliability Assessment of Shear Design Provisions for Reinforced Concrete Beams with Stirrups. Ph.D. Thesis, Stellenbosch University, Stellenbosch, South Africa, 2018. [Google Scholar]
- McBean, E.A.; Rovers, F.A. Statistical Procedures for Analysis of Environmental Monitoring Data and Risk Assessment; Prentice-Hall: Upper Saddle River, NJ, USA, 1998. [Google Scholar]
Parameters | Symbol | Range | ||||
---|---|---|---|---|---|---|
Minimum | First Quartile (P25) | Median (P50) | Third Quartile (P75) | Maximum | ||
Concrete cylinder compressive strength (MPa) | 7.5 | 20.6 | 24.1 | 30.8 | 63.9 | |
Embedment depth (mm) | 17.6 | 46 | 66.5 | 90.3 | 185 | |
Diameter of anchor (mm) | 6 | 10 | 15.9 | 19.5 | 32 | |
Tensile capacity of anchors (kN) | 3.3 | 22.2 | 40.9 | 71.8 | 273.8 |
Statistical Parameter | Datasets | (mm) | |||
---|---|---|---|---|---|
Range | Training | 6 to 32 | 17.6 to 185 | 7.5 to 63.9 | 5.6 to 273.8 |
Testing | 6 to 32 | 17.6 to 185 | 7.5 to 61.0 | 3.3 to 271.0 | |
Mean | Training | 15.7 | 73.6 | 26.5 | 54.5 |
Testing | 15.2 | 76.4 | 26.1 | 58.3 | |
St. deviation | Training | 5.8 | 32.8 | 9.6 | 43.8 |
Testing | 5.6 | 45.3 | 7.5 | 57.1 |
(a) | |||||||
Model | Training Dataset | Testing Dataset | |||||
RMSE (kN) | R2 | MAE (kN) | RMSE (kN) | R2 | MAE (kN) | ||
GPR | 8.15 | 0.97 | 5.56 | 10.26 | 0.97 | 6.18 | |
SVR | 11.73 | 0.93 | 7.98 | 13.12 | 0.95 | 8.61 | |
Random Forest (Boosted Trees) | 12.1 | 0.92 | 7.55 | 13.33 | 0.94 | 8.9 | |
Decision Tree (Fine) | 13.4 | 0.91 | 7.7 | 15.5 | 0.94 | 9.4 | |
Neural Network | 19.2 | 0.91 | 14.5 | 12.5 | 0.97 | 8.6 | |
(b) | |||||||
Model | Input | Training Dataset | Testing Dataset | ||||
RMSE (kN) | R2 | MAE (kN) | RMSE (kN) | R2 | MAE (kN) | ||
GPR | , , | 8.15 | 0.97 | 5.56 | 10.26 | 0.97 | 6.18 |
, | 10.7 | 0.94 | 6.3 | 11.7 | 0.96 | 7.04 | |
, | 16.8 | 0.86 | 10.9 | 19.3 | 0.89 | 11.2 | |
, | 24.3 | 0.69 | 14.3 | 31.6 | 0.7 | 18.4 | |
SVR | , , | 11.73 | 0.93 | 7.98 | 13.12 | 0.95 | 8.61 |
, | 12.5 | 0.92 | 8.2 | 11.3 | 0.96 | 7.9 | |
, | 17.5 | 0.84 | 11.6 | 19.9 | 0.88 | 12.3 | |
, | 31.9 | 0.47 | 18.5 | 45.1 | 0.53 | 25.7 |
No. | Parameters | GPR | SVR | CCD |
---|---|---|---|---|
1 | R2 | 0.97 | 0.95 | 0.96 |
2 | RMSE | 10.26 | 13.12 | 13.3 |
3 | MAE | 6.18 | 8.61 | 8.79 |
4 | Mean | 0.97 | 0.99 | 0.97 |
5 | Standard deviation | 0.14 | 0.22 | 0.19 |
6 | Skewness | 0.08 | 0.23 | −0.21 |
7 | Minimum () | 0.59 | 0.39 | 0.39 |
8 | Maximum | 1.46 | 1.38 | 1.67 |
9 | Ranking for | 2 | 1 | 2 |
10 | Ranking for | 1 | 3 | 2 |
11 | Ranking for / | 1 | 2 | 3 |
12 | Overall ranking | 1.3 | 2.0 | 2.3 |
Parameters | GPR | SVR | CCD |
---|---|---|---|
Concrete compressive strength (MPa) | −0.13 | −0.22 | 0.08 |
Embedment depth (mm) | −0.001 | 0.11 | 0.25 |
Diameter of anchor (mm) | −0.01 | 0.08 | - |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2020 by the authors. 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 (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Spyridis, P.; Olalusi, O.B. Predictive Modelling for Concrete Failure at Anchorages Using Machine Learning Techniques. Materials 2021, 14, 62. https://doi.org/10.3390/ma14010062
Spyridis P, Olalusi OB. Predictive Modelling for Concrete Failure at Anchorages Using Machine Learning Techniques. Materials. 2021; 14(1):62. https://doi.org/10.3390/ma14010062
Chicago/Turabian StyleSpyridis, Panagiotis, and Oladimeji B. Olalusi. 2021. "Predictive Modelling for Concrete Failure at Anchorages Using Machine Learning Techniques" Materials 14, no. 1: 62. https://doi.org/10.3390/ma14010062
APA StyleSpyridis, P., & Olalusi, O. B. (2021). Predictive Modelling for Concrete Failure at Anchorages Using Machine Learning Techniques. Materials, 14(1), 62. https://doi.org/10.3390/ma14010062