Management of Earthquake Risks using Condition Indicators

1. Management of Earthquake Risks using Condition Indicators Institute of Structural Engineering (IBK) Yahya Y. Bayraktarli Prof. Dr. Michael Faber

2. The Project Group MERCI • Project started in June 2004. • Interdisciplinary research group. Funded by the Swiss National Fund. • Participating Institutes from Swiss Federal Institute of Technology Zurich • Institute of Structural Engineering • Group Risk and Safety Prof. Dr. Michael H. Faber Dr. Jens Ulfkjaer Yahya Y. Bayraktarli • Group Earthquake Engineering and Structural Dynamics Prof. Dr. Alessandro Dazio Ufuk Yazgan • Institute of Construction Engineering and Management • Institute of Geotechnical Engineering • Dr. Jan Laue • Juliane Buchheister • Institute of Geodesy and Photogrammetry

3. Decision Situations Before During After Optimal allocation of available ressources for risk reduction - retrofitting - rebuilding in regard to possible earthquakes Damage reduction/Control Emergency help and rescue Aftershock hazards Rehabilitation of infrastructure functionality Condition assessment and updating of reliability and risks Optimal allocation of ressources for rebuilding and strengthening

4. Introduction Decision Support Risk assessment Probabilities Consequences Structure Models Lifeline Models CONDITION INDICATORS Photo- grammetry Soil Behaviour Structural Dynamics Socio-econo- mical Modeling Geophysics

5. Uncertainties in the Functional Chain of an Earthquake Site effects Structural response Freefield Structural response Sediment- schichten Festgestein Seismic Wave Input Consequences Damage propagation Earthquake Source Earthquake characteristics Site response

6. Risk Assessment Framework Exposure Vulnerability Risk reduction measures Indicators Robustness

7. Decision Theoretical Framework Variable B Variable A Variable C Consequ- ences Decision The overall theoretical framework is the Bayesian decision theory. Risks will be quantified using Bayesian Probabilistic Networks (BPN). • Causal interrelations of events leading to events of interest are graphically shown in terms of nodes connected by arrows. • A probability structure describing the conditional state probabilities for each node is assigned. • Consequences corresponding to the events in the BPN are assigned.

8. Example: Bauwerksklasse Decision situation, whether to retrofit or not. columns 30x30 16f16 Jacketed columns 50x50 24f16 stirrups f8@18

9. Example: Bayesian Network Soil Type Erdbebenstärke EQ Magnitude EQ Distance Spectral Displacement Periode Damage Structural Class No. of people at risk Costs No. of fatalities Entscheidung Actions

10. Example: Modeling of the Seismic Hazard EQ Magnitude • Gutenberg & Richter (1944) Magnitude recurrence relationship Actions

11. Example: Modeling of Exposure • Attenuation relations Mw=5.5 Mw=6.5 Mw=7.0 Mw=7.5 R=10 km R=20 km R=40 km R=80 km … … … For each spectra 20 time series were simulated.

12. Example: Generation of a Network Soil Type EQ Magnitude EQ Distance Spectral Displacement

13. Example: Generation of a Network Soil Type EQ Magnitude EQ Distance Spectral Displacement Periode Structure Class

14. Example: FE Calculations For each time series the maximum interstory drift ratio (MIDR) is calculated by PreOpenSeesPost. Details will be given in the presentation of Jens-Peder Ulfkjaer.

15. Example: Fragility Functions retrofitted Sd Sd Probability Probability No No Total Total Mod. Mod. Slight Slight Heavy Heavy

16. Example: Generation of a Network Soil Type EQ Magnitude EQ Distance Spectral Displacement Periode Damage Structure Class

17. Example: Modeling of Consequences • Distribution of the No. of people at risk is assumed at the moment by engineering judgement. • No fatalities for the damage classes „No damage“, „slight damage“ und „moderate damage“. • For „heavy damage“ and „totale damage“ a distribution is assumed. 4 5 6 7 8 9 0 1 2 3 10 11 12 No. of people at risk

18. Example: Generation of a Network Soil Type EQ Magnitude Eq Distance Spectral Displacement Periode Damage Structure Class No. of people at risk Costs No. of Fatalities Actions

19. Example: Updating the Network in the „During“ phase Soil Type EQ Magnitude Eq Distance Spectral Displacement Periode Residual Drift Ratio Structure Class Photogram. Measurements Damage No. of people at risk Costs No. of Fatalities Actions

20. Bayesian Network for the Soil Input Soil Type Earthquake Magnitude SPT_CPT Soil Subclass Earthquake Max. Acc. Number of Cycles Seismic Demand Density Liquefaction susceptibility Liquefaction triggering Soil Response Damage Groundwater Level Hollow Cylinder

21. Bayesian Network for the Consequences Structural Damage Number of People at risk Earthquake Time Consequences Number of Fatalities Probability of Escape Number of Injured People Age of People

22. An Example for Extension of the Network EQ Magnitude Soil Type EQ Distance SPT_CPT Soil Subclass EQ Max. Acc. No. of Cycles Seismic Demand Fundamental Period Density Liquefac. Suscep. Liquefac. Triggering Soil Response Structural class Structural Damage GW Level Hollow Cylinder No. of People at risk EQ Time Consequences Number of Fatalities Probability of Escape No. of Injured People Actions Age of People

23. Summary and Conclusions • A systematic approach is suggested by formulating decision problems for three cases in terms of characteristic descriptors (condition indicators), which can be observed and/or measured. • The Bayesian decision theory provides the mathematical framework for the consistent treatment of uncertainties and consequences. • Bayesian Probabilistic Networks are utilized for the consistent consideration of causal dependencies and uncertainties prevailing the identified decision problem. • The modular approach enables the utilisation of different models with varying levels of details.