Analysis of concrete creep, shrinkage and fracture by deterministic and probabilistic methods

Part I of the dissertation is devoted to concrete drying creep. The test results confirmed the existence of stress-induced shrinkage and clarified the entire picture of the contributions from four mechanisms: basic creep, drying shrinkage, stress-induced shrinkage, and microcracking effect.;Two theoretical models are developed. One is a diffusion model. The moisture capacity is evaluated by a modified BET equation with the effects of water to cement ratio, age, temperature, and type of cement taken into account. A general equation for the diffusivity in terms of pore humidity and water-to-cement ratio is developed by analyzing diffusion mechanisms. The second model is for drying creep. Basic creep is modeled by a generalized Kelvin chain model combined with solidification theory for aging behavior. The asymptotic transformation method is used to identify the continuous retardation spectra. A linear local constitutive relation for free shrinkage is proposed. An equation for the coefficient of stress-induced shrinkage is derived, based upon the microdiffusion theory. A unified model for drying creep, which represents a serial coupling of the models for each mechanism, is proposed.;Part II deals with the random deviations in structural creep analysis. Three sources of the randomness are considered: material parameter variation, model uncertainty, and environmental fluctuation. All of the composition parameters and model uncertainty parameters are treated as correlated normal random variables, and the density function of multinormal distribution is established, based on 137 test data. The latin hypercube sampling method is generalized to multinormal random variables by the orthogonal matrix transformation. The environmental relative humidity is treated as stochastic processes. The stress responses to the three random sources are analyzed by sampling and spectral methods.;The randomness in crack propagation with monotonic loading condition is studied by Markov chain model. The probability transition matrix is evaluated by (1) a mean curve that is based on R-curve behavior obtained from size effect method, and (2) a linear relation between the deviation of nominal failure stress and the crack extension.