Development of Reliability-Based Load and Resistance Factor Design (LRFD) Methods for Piping
6 Modeling Uncertainty
This chapter analyzes the modeling uncertainty associated with prediction models used to obtain nominal values in the design of pipes compared to experimental results. The objective herein is to evaluate the bias, which is the ratio of the true (mean) value of a variable based on experimental results to its nominal value calculated according to a model. First the bias (Experimental Value ∕ Nominal Value) of models used for the prediction of the hoop stress and in consequence of the pipe's thickness is presented and then the bias in the prediction of pipes bending moments. This bias can be used to compare the conservatism (bias>1) or non conservatism (bias<1) of different models. The COV of the bias should be small and combined with the COV that reflects the basic, “inherent”, statistical variability of the load and resistance variables.