Model uncertainty in the prediction of crack widths in reinforced concrete structures and reliability implications

[摘要] ENGLISH ABSTRACT: This research concerns the assessment of model uncertainty of serviceability limit state (SLS) crack models in reinforced concrete structures, where SLS cracking governs design. SLS are treated nominally in design standards without the extensive probabilistic calibration utilised for the ultimate limit state (ULS). Such nominal treatment is not necessarily appropriate when SLS is the governing limit state. A probabilistic approach is therefore taken in assessing SLS cracking in reinforced concrete structures such as liquid retaining structures (LRS), as a precursor to developing a safe but economical design crack width formulation for application in South Africa. In a reliability study of the current Eurocode crack model undertaken by McLeod (2013), model uncertainty displayed a significant influence on reliability. However, it was found that there is little information on the model uncertainty statistical parameters, providing motivation for this research to quantify the model uncertainty of crack models.A probabilistic approach to SLS cracking is employed. The reliability crack model or General Probabilistic Model (GPM) may be described in this context as the best probabilistic description of expected crack widths. This research therefore included the assessment of the crack models of BS 8007 (1987), BS EN 1992-1 (2004), fib MC 2010 (2013) and the proposed amended EN 1992-1 model (Perez Caldentey, 2017) to establish the GPM.Model uncertainty is defined as the ratio of the maximum experimental to predicted crack widths, wexp/ wpredict. Experimental data on load-induced cracking is assembled into a database to quantify model uncertainty, including the statistical quantification of the bias and uncertainty of the prediction model under consideration. The interaction between selected model parameters and model uncertainty is investigated by means of Pearson's correlations and linear regression analyses. The crack width model for the GPM is chosen by using the model uncertainty quantification results. Reliability analyses using the First Order Reliability Method (FORM) are done to assess the importance of the respective random variables, including model uncertainty. The study confirms that model uncertainty is the dominant influence on the crack model. Factors such as the relationship between the target reliability and the reliability of the crack width formulations, and the influence of long-term shrinkage strain on reliability are also investigated. Recommendations for future research are also made.

[发布机构] Stellenbosch University