Study And Uncertainty Analysis On Seismic Performance Of Concrete Structures

Uncertainty analysis is an important and comprehensive aspect in civil engineering. Uncertainty analysis theorems and methods are not only the mathematic languages for describing the uncertain realistic engineering problems, but also the computational logic to handle the uncertain realistic engineering problems. Uncertainty theorems, including Bayesian theory, interval analysis theory etc, have made the uncertainty structural analysis based on more broader theory foundations and extent the application domains of uncertainty analysis used to be based on classic probabilistic theorem. Around structural seismic analysis, primary research works of this dissertation conclude as following:(1) Generic prototype of expert system based on Bayesian network is built for structure assessment and diagnosis. In this system, domain knowledge is expressed as discrete Bayesian network, the inference system is based on junction tree algorithm. Modular design is adopted in construction of the expert system. The whole system can be divided into user interface module, Bayesian network edit module and inference module according to the function. The user interface module accepts the user input, including edit of the Bayesian network, evidence input, and outputs results; inference module compiles the Bayesian network to junction tree, inference uncertainties with given evidences and marginalizes the result joint distribution, Bayesian network edit module converts the activity of user interface to the construction of Bayesian network. Take the example of durability diagnosis of reinforced concrete to illuminate the inference process of junction tree algorithm for discrete Bayesian network; combined with seismic performance evaluation of reinforced concret frame structures, demonstrates how the reasoning can be performed in the situation where the input information is uncertain and incomplete that frequently encountered in civil engineering.(2) Nonlinear time history analysis of SDOF with various yield strength coefficient and period are performed based on1918earthquake ground motion records; simplified formula of ductility demand is acquired by curve fitting. A new parameter that quantifies the earthquake spectrum character effect on the ductility demand is constructed by regression analysis. Based on above, a ten-node continuous Bayesian network is established for seismic ductility demand probability analysis. The Bayesian network can acquire the posterior distribution which is more accurate updated by observed data. The posterior probability distribution involved in Bayesian network updating is a complicated multi-dimensional integral, Markov Chain Monte Carlo method, more specifically, Metropolis-Hastings sampling algorithm and convergence diagnosis algorithm, is introduced to update the ductility demand posterior probability distribution with earthquake intensity observations. Case study is carried out to analyze how the earthquake intensity parameters and given observations effect on the calculation results.(3) Based on71time history acceleration records of12near-fault earthquakes occurred in recent years, responses of three typical frame structures are analyzed and two new intensity measures are proposed. First intensity measure, based on the fuzzy structure vibration period, acquiring the corresponding squared spectrum velocity fuzzy set according to extension principle, is determined as representative value by taking the center of gravity of fuzzy set. For the frame structures with medium and long period, the new measure exhibits more efficiency and sufficiency than other9present available intensity measures. Another intensity measure, based on fuzzy valued force shape vector used in pushover analysis, obtaining the fuzzy set of displacement ductility by extension principle and pushover analysis, is also determined as representative value. This new measure is efficiency and sufficiency for the frame structures with medium and short period. In general, new fuzzy representative valued IMs are more efficiency and sufficiency than other available IMs for consideration of vague uncertainty of structure under complicated realistic circumstance.(4) In the seismic design of reinforced concrete structures,"strong shear weak bending" is an important design conception to guarantee the ductibility of the structure. Interval variable is introduced to express the epistemic uncertainty and failure probability interval of strong shear weak bending of reinforced concrete column is analyzed. The interval-valued probabilistic reliability model for "strong shear weak bending" is formulated according to the inclusion relationship of the element events and the failure event. For the calculation of the resistance capability involving interval-valued parameters, Taylor model is introduced in computation to reduce the error induced by interval inflation. The simulated annealing genetic algorithm considering "infeasible degree" is applied to determine the approximate design interval of the "strong shear weak bending". A specific sampling function constructed by such design interval is adopted to obtain the interval-valued probabilistic reliability index; error analysis indicates that the precision of the method is acceptable. Case study is carried out to analyze the different design parameters affecting the reliability and corresponding design suggestions are proposed.