Abstract
Concrete displays long-term time-dependent behavior due to its rheological properties. The prediction of long-term behavior of concrete is difficult, even under laboratory conditions, due to the stochastic nature of its rheological phenomena. In concrete structures, long-term prediction is even more challenging due to the presence of uncontrolled conditions, such as variations in temperature, humidity, and loading. Current approaches for prediction of long-term time-dependent behavior at structural scale involve computationally intensive stochastic finite-element methods in which multiple creep and shrinkage models are implemented. However, these models are often calibrated using a database of experiments that are not the most informative for a specific structure. Structural health monitoring can improve prediction accuracy by providing structure-specific in-situ measurements of strain and temperature. Strain sensors, however, measure a multitude of effects simultaneously present in the structure, making it difficult to decouple effects of interest. In this work, a hybrid method employing probabilistic neural networks and engineering code models is proposed for the prediction of long-term behavior in concrete structures. A modular architecture is employed to decouple temperature-dependent environmental strain from long-term time-dependent strain. Generalized creep and shrinkage code models are fitted to the resulting time-dependent strain component data and used for prediction. The method is applied to a concrete pedestrian bridge instrumented with several embedded strain and temperature sensors. Excellent accuracy is achieved in the prediction of structural behavior multiple years beyond the training range. This, in turn, enables the detection of unusual structural behaviors with both gradual and sudden manifestation.
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig1_HTML.jpg)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig2_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig3_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig4_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig5_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig6_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig7_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig8_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig9_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig10_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig11_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig12_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig13_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig14_HTML.png)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs13349-022-00582-4/MediaObjects/13349_2022_582_Fig15_HTML.png)
Similar content being viewed by others
References
Bažant ZP, Jirásek M (2018) Creep and hygrothermal effects in concrete structures. Springer, Dordrecht
Glasser FP, Marchand J, Samson E (2007) Durability of concrete—degradation phenomena involving detrimental chemical reactions. Cem Concr Res 38:226–246
Abdellatef M, Vorel J, Wan-Wendner R, Alnaggar M (2019) Predicting time-dependent behavior of post-tensioned concrete beams: discrete multiscale multiphysics formulation. J Struct Eng 145(7):04019060
Glisic B, Inaudi D, Lau JM, Fong CC (2013) Ten-year monitoring of high-rise building columns using long-gauge fiber optic sensors. Smart Mater Struct 22:055030
Abdel-Jaber H, Glisic B (2019) Monitoring of long-term prestress losses in prestressed concrete structures using fiber optic sensors. Struct Health Monit 18:254–269
Mehta PK, Monteiro PJM (2013) Concrete microstructure, properties, and materials. McGraw-Hill Education, New York
Di Luzio G, Cusatis G (2009) Hygro-thermo-chemical modeling of high performance concrete. I: theory. Cement Concr Compos 31:301–309
Di Luzio G, Cusatis G (2009) Hygro-thermo-chemical modeling of high performance concrete. II: numerical implementation, calibration, and validation. Cement Concr Compos 31:309–324
Farrar CR, Worden K (2013) Structural health monitoring: a machine learning perspective. Wiley, West Sussex
Huston D (2011) Structural sensing, health monitoring, and performance evaluation. CRC Press, Boca Raton
Alnaggar M, Cusatis G, Di Luzio G (2013) Lattice discrete particle modeling (LDPM) of alkali silica reaction (ASR) deterioration of concrete structures. Cement Concr Compos 41:45–59
Alnaggar M, Di Luziom G, Cusatis G (2017) Modeling time-dependent behavior of concrete affected by alkali silica reaction in variable environmental conditions. Materials 10(5):471
Bažant Z, Hubler M, Yu Q (2011) Pervasiveness of excessive segmental bridge deflections: wake-up call for creep. ACI Struct J 108(6):766–774
Bažant ZP, Yu Q, Li G-H (2012) Excessive long-time deflections of prestressed box girders. I: record-span bridge in Palau and other paradigms. J Struct Eng 138(6):676–686
American Concrete Institute Committee 209 (2008) Guide for modeling and calculating shrinkage and creep in hardened concrete. Farmington Hills
CEB-FIP (1993) CEB-FIP model code 1990. Committee Euro-International du Béton
Bažant ZP, Baweja S (1995) Creep and shrinkage prediction model for analysis and design of concrete structures—model B3. Matériaux et Constructions 28:357–365
Hubler MH, Wendner R, Bažant ZP (2015) Statistical justification of model B4 for drying and autogenous shrinkage of concrete and comparisons to other models. Mater Struct 48:797–814
Pham A-D, Ngo N-T, Nguyen T-K (2020) Machine learning for predicting long-term deflections in reinforce concrete flexural structures. J Comput Des Eng 7(1):95–106
Ghasemzadeh F, Manafpour A, Sajedi S, Shekarchi M, Hatami M (2016) Predicting long-term compressive creep of concrete using inverse analysis method. Constr Build Mater 124:496–507
Han B, **ang T-Y, **e H-B (2017) A Bayesian inference framework for predicting the long-term deflection of concrete structures caused by creep and shrinkage. Eng Struct 142:46–55
Sousa H, Santos LO, Chryssanthopoulous M (2019) Quantifying monitoring requirements for predicting creep deformations through Bayesian updating methods. Struct Saf 76:40–50
Strauss A, Wan-Wendner R, Vidovic A, Zambon I, Yu Q, Frangopol DM, Bergmeister K (2017) Gamma prediction models for long-term creep deformations of prestressed concrete bridges. J Civ Eng Manag 23(6):681–698. https://doi.org/10.3846/13923730.2017.1335652
Abdel-Jaber H, Glisic B (2016) Systematic method for the validation of long-term temperature measurements. Smart Mater Struct 25:125025
Che Z, Purushotham S, Cho K et al (2018) Recurrent neural networks for multivariate time series with missing values. Sci Rep 8:6085. https://doi.org/10.1038/s41598-018-24271-9
CheZ, Purushotham S, Li G, Jiang B, Liu Y (2018) Hierarchical deep generative models for multi-rate multivariate time series. In: Dy J, Krause A (eds) Proceedings of the 35th international conference on machine learning, ser. proceedings of machine learning research, Stockholm, Sweden: PMLR, 10–15 Jul 2018, vol 80, pp 784–793. http://proceedings.mlr.press/v80/che18a.html
Bal L, Buyle-Bodin F (2013) Artificial neural network for predicting drying shrinkage of concrete. Constr Build Mater 38:248–254
Hauge M (2019) Machine learning for predictions of strains due to long-term effects and temperature in concrete structures. Master’s thesis. Norwegian University of Science and Technology.
Hu W-H, Cunha Á, Caetano E, Rohrmann RG, Said S, Teng J (2017) Comparison of different statistical approaches for removing environmental/operational effects for massive data continuously collected from footbridges. Struct Control Health Monit 24:e1955. https://doi.org/10.1002/stc.1955
Cross EJ, Worden K, Chen Q (2011) Cointegration: a novel approach for the removal of environmental trends in structural health monitoring data. Proc R Soc A Math Phys Eng Sci 467:2712–2732. https://doi.org/10.1098/rspa.2011.0023
Rubanova Y, Chen RTQ, Duvenaud DK (2019) Latent ordinary differential equations for irregularly-sampled time series. http://papers.nips.cc/paper/8773-latent-ordinary-differential-equations-for-irregularly-sampled-time-series.pdf
Oh BK, Park HS, Glisic B (2021) Prediction of long-term strain in concrete structure using convolutional neural networks, air temperature and time stamp of measurements. Autom Constr 126:103665
Sigurdardottir DH, Glisic B (2015) On-site validation of fiber-optic methods for structural health monitoring: Streicker bridge. J Civ Struct Heal Monit 5:529–549
Sigurdardottir DH, Glisic B (2013) Neutral axis as damage sensitive feature. Smart Mater Struct 22:075030 (p 18)
ACI 423.10R-16 (2016) Guide to estimating prestress losses
Abadi M, Agarwal A, Barham P, Brevdo E, Chen Z, Citro C, Corrado GS, Davis A, Dean J, Devin M, Ghemawat S, Goodfellow I, Harp A, Irving G, Isard M, Jia Y, Jozefowicz R, Kaiser L, Kudlur M, Levenberg J, Man ́e D, Monga R, Moore S, Murray D, Olah C, Schuster M, Shlens J, Steiner B, Sutskever I, Talwar K, Tucker P, Vanhoucke V, Vasudevan V, Vi ́egas F, Vinyals O, Warden P, Wattenberg M, Wicke M, Yu Y, Zheng X (2015) Software available from tensorflow.org [Online]. https://www.tensorflow.org/
Kingma DP, Ba J (2015) Adam: a method for stochastic optimization. In: 3rd international conference for learning representations, San Diego
Karniadakis GE, Levrelodos IG, Lu L, Perdikaris P, Wang S, Yang L (2021) Physics-informed machine learning. Nat Rev Phys 3:422–440
Acknowledgements
The authors would like to thank the support of Princeton University for the financial support, and Hiba Abdel-Jaber and Vivek Kumar for the help with data preparation.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
At certain positions in Streicker bridge, a long-term time-dependent strain relaxation is observed (e.g., see Fig. 13a). This relaxation implies that at some time \(t\), the rate of the long-term time-dependent effective strain \({\varepsilon }_{\rm EF}^{t}\) becomes positive. We show, under the simplified model outlined in the Methods section and using the CEB MC90-99 code, that the rate of \({\varepsilon }_{\rm EF}^{t}\) should remain negative, characterizing strain relaxation as anomalous structural behavior.
Taking the time derivative of Eq. (23) gives the effective strain rate
The goal is to determine if the strain rate shown in Eq. (67) can be positive for some time \(t\). That is nontrivial because of concrete aging and the recovery of creep strain due to prestress loss.
The time derivative of Eq. (17) is
Using the CEB MC90-99 model for the aging of concrete,
in Eq. (14), gives
where \(s>0\) is a constant that depends on the type of concrete [15].
In view of Eq. (70), Eq. (68) can be expanded as
For Streicker bridge, \(s=0.2,\) \({\beta }^{28}\approx 36\), and the expression above can be numerically evaluated for any \(t>0.\) Notice that in general \(\frac{{\rm d}{\chi }^{t}}{{\rm d}t}<0\) and, furthermore, because \({\varepsilon }_{R}^{t}<0\), the first term on the RHS of Eq. (67) is negative. Hence, by neglecting the negative contribution of the first term, the following inequality holds:
We would like to show that a conservative upper bound on \(\frac{{\rm d}{\varepsilon }_{\rm EF}^{t}}{{\rm d}t}\) is still negative, such that the effective strain rate must necessarily be negative. Since \(\left(1-{\chi }^{t}\right)>0\), towards more conservative upper bounds, further considerations should seek to make the rheological strain rate \(\frac{{\rm d}{\varepsilon }_{R}^{t}}{{\rm d}t}\) a larger number.
The rheological strain is given by
where \({\varepsilon }_{\rm Cr}^{t}\) is the creep strain, as given in Eq. (12), and \({\varepsilon }_{Sh}^{t}\) is the shrinkage strain. Introducing Eq. (12) in Eq. (73) gives
The first two terms contribute with negative values, while the integral term contributes with positive values, corresponding to the creep recovery associated with prestress loss.
From Eq. (24)
Introducing Eq. (75) into Eq. (74)
In general, the integral terms cannot be solved analytically, so it is typical to consider a discretization of the integral terms [1], yielding
where \(N\) is the number of time steps \(\Delta t\), which can be taken arbitrarily small. Taking the time derivative of Eq. (77)
The first two terms are negative rates associated with the creep due to the initial prestressing and shrinkage, respectively, while remaining terms are associated with creep recovery due to prestress loss. However, notice that in the summation terms
the rheological strain \({\varepsilon }_{R}^{{u}_{i}}<0\) contributes towards a more negative strain rate \(\frac{{\rm d}{\varepsilon }_{R}^{t}}{{\rm d}t}\), since \(\frac{{\rm d}\varphi \left(t,{u}_{i}\right)}{{\rm d}t}>0\) and \(\frac{{\rm d}{\chi }^{{u}_{i}}}{{\rm d}t}<0\). Thus, the following inequality, obtained by disregarding this negative contribution, holds:
Expanding the rheological strain rate \(\frac{{d\varepsilon_{R}^{{u_{i} }} }}{{\rm d}t}\) in terms of its creep and shrinkage components gives
Taking the time derivative of Eq. (12) gives the creep strain rate
where the first term contributes negatively to the creep strain rate and is associated with prestressing at time \(t^{\prime}\), while the second term contributes positively to the creep strain rate, as it corresponds to the rate of creep recovery associated with prestress loss. Thus, the following inequality holds:
Then, in view of inequalities (81) and (83), the following inequality obtained by introducing a more positive contribution on the RHS holds:
The CEB MC90-99 creep coefficient is of the form
where \({\varphi }_{0}>0\) is the notional creep coefficient, and the shrinkage strain is of the form
where \({\varepsilon }_{cs0}<0\) is the notional shrinkage coefficient. The time-dependent coefficients \({\beta }_{\mathrm{\rm Cr}}\left(t,{t}^{{\prime}}\right)\) and \({\beta }_{\mathrm{Sh}}\left(t,{t}^{{\prime}}\right)\) are monotonically increasing functions.
Introducing Eq. (85) and (86) in inequality (84) gives
where
and
are, respectively, the creep and shrinkage strain rate upper bound coefficients. Since \({\varepsilon }_{e,P}^{{\prime}}{\varphi }_{0}<0\) and \({\varepsilon }_{cs0}<0,\) if \({c}_{\rm Cr}\left(t\right)>0\) and \({c}_{Sh}\left(t\right)>0\), the rheological strain rate must remain negative per inequality (87). Using the CEB MC90-99 creep and shrinkage models with Streicker Bridge properties gives a notional creep coefficient of \({\varphi }_{0}=2.08\), and the time-dependent coefficient for creep
and shrinkage
The derivatives of the time-dependent coefficients of creep and shrinkage are, respectively
Thus, Eqs. (88) and (89) can be numerically evaluated for any time \(t\). Figures
16 and
17 show, respectively, \({c}_{\rm Cr}(t)\) and \({c}_{Sh}\left(t\right)\) for \({t}^{{\prime}}<t\le 3\times 365\) days, with \(\Delta t=1 min.\) The figures show that both \({c}_{\rm Cr}\left(t\right)\) and \({c}_{Sh}(t)\) remain positive over this period, and, therefore, the rheological strain rate remains negative.
Then, considering inequality (72), and that the shrinkage component \({\varepsilon }_{cs0}{c}_{Sh}\left(t\right)<0,\) the following inequality, obtained by disregarding the negative shrinkage contribution, holds,
where
is the effective strain rate upper bound coefficient. Since \({\varepsilon }_{e,P}^{{\prime}}<0\), if \({c}_{\rm EF}\left(t\right)>0\), the time-dependent effective strain rate must remain negative in view of inequality (94). Figure
18 shows that the resulting values of \({c}_{\rm EF}(t)\) computed numerically for \({t}^{{\prime}}<t\le 3\times 365\) days, with \(\Delta t=1 min,\) remain positive, implying that
over the period of interest here and, thus, characterizing the positive rates, observed between 730 and 1000 days (e.g., see Fig. 13a), at some locations in Streicker Bridge as anomalous structural behavior.
Rights and permissions
About this article
Cite this article
Pereira, M., Glisic, B. A hybrid approach for prediction of long-term behavior of concrete structures. J Civil Struct Health Monit 12, 891–911 (2022). https://doi.org/10.1007/s13349-022-00582-4
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s13349-022-00582-4