| Prediction of deflection and crack width of reinforced concrete (RC) structures is a critical aspect in the design of concrete members for satisfactory performance during their use. Due to inherent uncertainties in concrete (materials, microstructure, mixing, casting and so on), concrete properties related to the serviceability calculation have been modeled as random (aleatoric ) uncertainty and probability theory has been predominantly used to model those random uncertainties.;However, there exist other types of uncertainties known as epistemic uncertainties that cannot be represented by probability theory. As those types of uncertainties arise from lack of knowledge, they are reducible. Epistemic uncertainties occur in serviceability calculations due to non-specificity, fuzziness and ambiguity of concrete properties.;This dissertation provides the first framework to represent aleatoric and epistemic uncertainties in serviceability of RC structures. Moreover, a method to quantify the propagated uncertainty in final serviceability is described. Theories to model various types of uncertainties are introduced. The propagation of epistemic uncertainties in concrete properties to the final serviceability calculation is examined by introducing the robustness to uncertainties. Non-specificity modeling of concrete cracking strength prediction models is discussed with model looseness. The propagation of uncertainty represented by non-specificity of cracking strength to the final deflection is quantified using interval analysis. Moreover, experimental methods are developed to establish concrete cracking intervals. A new test setup for direct tension test is suggested and verified using finite element models. Application of epistemic uncertainty to structural reliability is presented. |
