Virtual portfolios for earthquake insurance related risk simulations

Seismic Risk, a combination of Hazard, Vulnerability and Loss, is becoming a mainstream topic due to an increase in the losses and damage caused by earthquake related effects worldwide. Therefore, it has become an important aspect for insurer and reinsurers since it is extremely difficult to assess due to the lack of the necessary data compared, for example, to car insurance (chapter 1). As described in chapter 2, where the state of the art of seismic risk assessment and simulation is described alongside a more refined approach to seismic risk, current Vulnerability models are oversimplified and do not provide enough accuracy in predicting the damage. Due to the fact that little information on the behavior of real buildings under earthquakes is known, the buildings are vulnerable, and real datasets of observed damages at structures due to earthquake are rare and vary greatly in terms of quantity and quality. The scope of this thesis, Vulnerability modelling, is presented in more detail in chapter 3. In order to cover the demands of the insurance industry, a tool for generating 3D buildings which are then exposed to several hazard scenarios is needed. This tool has been developed in a general way with clear interfaces between the risk components Hazard, Vulnerability and Loss, and it is able to produce realistic 3D buildings quickly and easily. Its global aim is the generation of a statistically equivalent district of a city or a particular insurance Virtual Portfolio. Therefore, in this thesis, the Sophisticated 3D vulnerability modelling for Seismic Risk Simulation (3D-SRS) concept (chapter 4) has been developed to better capture observed damages. It defines the seismic risk in terms of estimating structural damage areas under an seismic excitation in order to be the base for a more realistic pricing from the point of view of insurers and reinsurers. The 3D-SRS consists of sophisticated numerical 3D Virtual Portfolios, a group of buildings with detailed local models. Although only linear covered by taken in this thesis, the concept is open to non-linear approaches as discussed at the end of that chapter. In terms of seismic risk, this is the first study with such complex local models for Vulnerability. For the purpose of generating Virtual Portfolios, the Simulated Building Portfolio-Tool (SBP-Tool) was developed in cooperation between the University of Kassel and the Munich Re, and it generates hundreds of numerical 3D-FE buildings in an automatic fashion, based on a set of global building parameters. It is also able to model local pre-damages around the foundation, windows or roofs in order to take into account a more realistic building stock. In chapter 5, the SBP-Tool is used to generate the geometry of an Unreinforced masonry building and it is modeled with geometrically linear shell elements and orthotropic material laws. The shaking table test of a two storey URM building from LNEC in Portugal is used to validate the method by comparing observed and estimated damages. The Mechanical Damage Indicator is used in order to link the principal stresses and mechanical damage. As a first assumption, a limit of using linear elastic models is proposed using the linear model approach. Chapter 6 demonstrates the use of 3D-SRS with the clear interfaces. During the evaluation of the estimated damages, a multivariate test of different building parameters was performed by using a Linear discriminant analysis in order to study the effect as well as the influence of different building parameters on the damage. At the end of this study (chapter 7), a comparison of the well-known and widely used Park-Ang Damage Index between local models, those generated with the SBP-Tool and global models (SDOF) is made. This comparison basically reflects the damage distribution. By using local 3D models, real observed damages can be reflected, whereas the damage distribution of the global models shows no correlation to observed damages.