ABSTRACT
Tunnel design is a complex complicated task since it is strongly associated with a great number of load and material uncertainties. Moreover, modelling the soil-structure interaction multiplies the complexity and non-linearity of tunnel engineering problems. Finite Element (FE) stochastic modelling has proven to be a very efficient tool in dealing with such uncertainties. Capturing uncertainties, though, which allows for a more realistic and pragmatic design decision aid, requires probabilistic approaches. The only general tool for probabilistic analysis is represented by Monte Carlo simulation (MC), simulating uncertainties with their complete probability distribution and statistical correlation. The complexity of the FE models renders then such a procedure virtually impossible. The computational burden of MC represents the main obstacle to the approach for time-consuming FE computational models since it is not computationally feasible for industrial applications. Consideration of uncertainties, therefore, remains usually at the level of partial safety factors, the approach which is in most cases on the very conservative side. That is why alternative techniques of so-called safety formats are becoming more and more attractive recently. Methods focused on the Estimation of Coefficient of Variation (ECoV) of structural resistance have been developed and applied recently mainly for concrete structures like bridges. They represent a compromise between the simple and in most cases conservative approach of partial safety factors and highly computationally demanding MC. The paper shows the possibility of application of these methods for tunnels. Selected efficient semi-probabilistic methods based on the ECoV method according to fib Model Code 2010, an improved approach called Eigen ECoV, Taylor Series Expansion and numerical quadrature proposed by Rosenblueth are compared.
