Probabilistic seismic risk is affected by several sources of uncertainty. Investigating their influence on loss analysis results is essential to obtain reliable quantitative estimations of seismic performance. Within the framework developed by the Pacific Earthquake Engineering Research (PEER) Center for probabilistic seismic loss analysis, this study incorporates the effect of seismic demand model class uncertainty on seismic risk performance metric estimation. The extended formulation is illustrated with an application example code-designed reinforced concrete moment resisting frame building with unreinforced masonry (URM) infill walls, where model class uncertainty related to URM infill walls modeling is propagated to the estimation of seismic financial losses. Model class uncertainty accounts for the variability arising from the use of different modeling solutions, such as the ones associated with the adoption of three equivalent strut macro-models and their modeling parameters. Probabilistic distributions are assigned to selected infill strut model parameters, and a large set of finite element models (FEMs) are generated for each infill strut model class by sampling the model parameter distributions through Latin Hypercube Sampling (LHS). Expected values of repair cost and life-cycle annualized loss are evaluated and compared for two sets of three building performance models. The first set considers the original (median) values of infill strut backbone parameters, while the second set includes model parameter uncertainty. The uncertainty propagation from structural response results to repair costs is also reported for the six performance models. Finally, the contribution of different structural and non-structural element categories to the financial losses is presented.

概率地震风险受多种不确定性因素影响。研究其对损失分析结果的影响对于获得可靠的地震性能定量评估至关重要。在太平洋地震工程研究（PEER）中心开发的用于概率地震损失分析的框架内，本研究纳入了地震需求模型类别不确定性对地震风险绩效度量估计的影响。通过应用示例代码设计的具有未增强砌体（URM）填充墙的钢筋混凝土抗弯框架建筑，说明了扩展的公式，其中，与URM填充墙建模有关的模型类别不确定性被传播到地震财务损失的估计中。模型类的不确定性是由于使用不同的建模解决方案（例如与采用三个等效的支撑宏模型及其建模参数相关联的建模解决方案）所引起的可变性。将概率分布分配给选定的填充支撑模型参数，并通过拉丁超立方体采样（LHS）对模型参数分布进行采样，从而为每个填充支撑模型类生成大量的有限元模型（FEM）。评估并比较了两组三个建筑性能模型的预期维修成本和生命周期年度损失。第一组考虑填充支柱主干参数的原始（中间值），而第二组包括模型参数不确定性。还报告了六个性能模型从结构响应结果到维修成本的不确定性传播。最后，介绍了不同结构和非结构要素类别对财务损失的贡献。