Sensitivity analysis frameworks for mechanistic-empirical pavement design of continuously reinforced concrete pavements

Abstract

The new AASHTO Mechanistic-Empirical Pavement Design Guide (MEPDG) performance predictions for the anticipated climatic and traffic conditions depend on the values of the numerous input parameters that characterize the pavement materials, layers, design features, and condition. This paper proposes comprehensive local sensitivity analyses (LSA) and global sensitivity analyses (GSA) methodologies to evaluate continuously reinforced concrete pavement (CRCP) performance predictions with MEPDG inputs under various climatic and traffic conditions. A design limit normalized sensitivity index (NSI) was implemented in both LSA and GSA to capture quantitative as well as qualitative sensitivity information. The GSA varied all inputs simultaneously across the entire problem domain while the LSA varied each input independently in turn. Correlations among MEPDG inputs were considered where appropriate in GSA. Two response surface modeling (RSM) approaches, multivariate linear regressions (MVLR) and artificial neural networks (ANN or NN), were developed to model the GSA results for evaluation of MEPDG CRCP input sensitivities across the entire problem domain. The ANN-based RSMs not only provide robust and accurate representations of the complex relationships between MEPDG inputs and distress outputs but also capture the variation of sensitivities across the problem domain. The NSI proposed in LSA and GSA provides practical interpretation of sensitivity relating a given percentage change in a MEPDG input to the corresponding percentage change in predicted distress relative to its design limit value. The "mean plus/minus two standard deviations (μ + 2σ)" GSA-NSI metric (GSA-NSIμ ±2σ) derived from ANN RSM statistics is the best and most robust design input ranking measure since it incorporates both the mean sensitivity and the variability of sensitivity across the problem domain. Keywords AASHTO, concrete, design, pavement, sensitivity analyses Disciplines Civil and Environmental Engineering | Construction Engineering and ManagementComments NOTICE: This is the author's version of a work that was accepted for publication in Construction and Building Materials. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Construction and Building Materials 73 (2014) Reference to this paper should be made as follows: Ceylan, H., Kim, S., Gopalakrishnan, K., Schwartz, C. W., and Reference to this paper should be made as follows: Ceylan, H., Kim, S., Gopalakrishnan, K., Schwartz, C. W., and Li, R.. "Sensitivity Analysis Frameworks for Mechanistic-Empirical Pavement Design of Continuously Reinforced Concrete Pavements," Construction and Building Materials. Vol. 73, pp. 498-508. Sensitivity Analysis Frameworks for Mechanistic-Empirical (2014). "Sensitivity Analysis Fra...

Effects of extended short-term aging duration on asphalt binder behaviour at high temperatures

“…Kavussi et al (2014) utilized RSM to evaluate the effects of aggregate gradation, hydrated lime and Sasobit content on the indirect tensile strength of warm mix asphalt. Ceylan et al (2014) employed RSM to model the Global Sensitivity Analyses (GSA) results for evaluation of Mechanistic-Empirical Pavement Design Guide (MEPDG) input sensitivities across the entire problem domain for continuously reinforced concrete pavement (CRCP). Hamzah et al (2015b) used RSM to determine the effects of aging on the rheological properties of asphalt binders.…”

Alternative Source of Climate Data for Mechanistic–Empirical Pavement Performance Prediction

“…ated the sensitivity of MEPDG performance predictions to traffic, material, and other design input parameters, but few comprehensive sensitivity studies of the effects of climate on performance predictions have been performed (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14). Nearly all sensitivity analyses of MEPDG performance predictions to climate inputs to date have simply compared results using one weather station to results using another (15,16).…”

Predictive modelling to support sensitivity analysis for robust design in aerospace engineering

“…In another study, Taflanidis et al (2011) constructed response surface model with moving least squares method to perform sensitivity analysis to determine the influence of input variables on the outcome variable for the robust design of offshore energy conversion devices . Recently, Ceylan et al (2014) constructed response surface models by multivariate linear regressions and artificial neural networks (ANN) to perform sensitivity analysis for robust design of concrete pavement design, and the authors concluded that ANN is a robust and accurate method to capture the nonlinearities between variables. However, for our datasets, ANN did not perform well in our initial investigations; hence, we exclude the ANN method from this paper.…”

“…Furthermore, the time is limited for design engineers to make decisions about design variables and design concepts. One way to solve this problem is to use response surface models-they are also known as surrogate models and metamodels-for sensitivity analysis (Ceylan et al 2014;Chen et al 2006;Ma et al 2015;Sathyanarayanamurthy and Chinnam 2009;Taflanidis et al 2011). The sensitivity analysis-based response surface models allow for efficient studies of how uncertainties in input variables affect system performance.…”

“…Since simulations are expensive to conduct, the datasets are small in real-world context. From the literature studies, we have learned that popular methods such as polynomial models, Kriging, Radial basis functions and multivariate adaptive regression were studied to construct surrogate models for design problems, and their advantages and disadvantages have been highlighted in Ceylan et al (2014); Huang et al (2011); Sathyanarayanamurthy and Chinnam (2009); Shan and Wang (2009); and Wang and Shan (2007). However, as complexity of design problems increases, the challenges such as curse-of-dimensionality and dealing with highly non-linear design spaces remain the same.…”