To begin with, a series of reliability analyses of all cases in<br>Sample C are performed to study the effect of the COV component<br>of the model uncertainty. Four scenarios of model uncertainty,<br>each with c1=1.0 and a different COV 0.0, 0.1, 0.2, or<br>0.3, are studied. For each scenario of model uncertainty, 3 values<br>are calculated for all 64 case histories. Similar to the scenario<br>of no model uncertainty, a Bayesian mapping function can be<br>obtained for each model uncertainty scenario. While not shown<br>herein, the results indicate that within the range of 0–0.3, the<br>COV component of the model uncertainty has little effect on the<br>calculated probability.<br>Another series of reliability analyses of all cases in Sample C<br>are performed with four scenarios of model uncertainty, each with<br>COV=0.1 and a different c1 1.0, 1.1, 1.2, or 1.3. A set of<br>Bayesian mapping functions are obtained and shown in Fig. 3.<br>For the scenario of c1=1.0, which represents the case with Eq.<br>1 as the CRR model, PL=0.23 at =0, indicating that the CRR<br>model as expressed in Eq. 1 is quite conservative. As c1 increases,<br>the mapping function is shifted from left to right, and the<br>probability corresponding to =0 increases. An interpolated c1<br>value of 1.24 yields PL=0.5 at =0, which is believed to be an<br>accurate estimate of model uncertainty.
การแปล กรุณารอสักครู่..
