### Filtern

#### Dokumenttyp

- Beitrag zu einem Tagungsband (4)
- Vortrag (2)
- Posterpräsentation (2)

#### Schlagworte

- Fire (3)
- Risk (3)
- Tunnel (3)
- Analysis (2)
- Road (2)
- CFD (1)
- Comparative study (1)
- Evacuation (1)
- Metamodel (1)
- Risk assessment (1)

#### Organisationseinheit der BAM

Methodologies on fire risk analysis in road tunnels consider numerous factors affecting risks (risk indicators) and express the results by risk measures. But only few comprehensive studies on effects of risk indicators on risk measures are available. For this reason, this study quantifies the effects and highlights the most important risk indicators with the aim to Support further developments in risk analysis. Therefore, a system model of a road tunnel was developed to determine the risk measures.
The system model can be divided into three parts: the fire part connected to the fire model Fire Dynamics Simulator (FDS); the evacuation part connected to the evacuation model FDS+Evac; and the frequency part connected to a model to calculate the frequency of fires. This study shows that the parts of the system model (and their most important risk indicators) affect the risk measures in the following order: first, fire part (maximum heat release rate); second, evacuation part (maximum preevacuation time); and, third, frequency part (specific frequency of fire). The plausibility of These results is discussed with view to experiences from experimental studies and past fire incidents.
Conclusively, further research can focus on these most important risk indicators with the aim to optimise risk analysis.

Filter criteria in the Frame of Fire PSA identify compartments in a first qualitative analysis for which the contribution to the overall core damage frequency of the NPP is negligible. The aim of the filter criteria is to reduce the number of compartments to be analysed precisely in Fire PSA. One example for filter criteria is the 'fire load criterion'. By the fire load criterion compartments with a fire load density of less than 90 MJ/m² are 'screened out' which means to exclude them from a precise analysis in Fire PSA. Neither the justification of the particular value of 90 MJ/m² is well documented nor does this criterion take into account varying compartment configurations such as ventilation conditions, physical and chemical properties of the fire load as well as compartment characteristics.
A probabilistic set of filter criteria was developed to overcome the restrictions of the fire load criterion. In line with the 'fire load criterion', the probabilistic set of filter criteria assumes that a compartment can be screened out if a fire is not able to cause any damage to other components within the compartment. Therefore, the electrical failure of an electrical cable conservatively represents the damages of all components. It is assumed that the electrical cable failure occurs when the maximum cable temperature exceeds an experimentally determined failure temperature. The maximum cable temperature that can occur in a compartment fire is mainly influenced by the four significant factors: 1. inlet air stream of the mechanical ventilation, 2. the fire growth rate, 3. the compartment floor area and 4. the compartment height. A parameter study revealed how the significant factors affect the maximum cable temperature in fictitious compartment fires. The results of the parameter study are transferred on true Nuclear Power Plant compartments. However, it is not possible to determine precisely the occurrence of an electrical cable failure because of uncertainties in the maximum cable temperature and the failure temperature. The probabilistic set of filter criteria considers these uncertainties and determines the probability of cable failure for true compartments to be screened in Fire PSA. Finally, a compartment can be screened out in Fire PSA if the failure probability exceeds a predefined accepted threshold value for the failure probability. The theoretical application of the methodology is shown at the end of the paper.

Safety measures like tunnel emergency Ventilation Systems cause high financial costs. Hence, safety measures have to be chosen with the focus on the expected reduction of the consequences like fatalities or damage on structures and in conjunction with the investments. Since 2004, the European directive EU 2004/54/EC proposes therefore the application of risk assessments. Because the EU directive provides only few legal requirements on risk assessments, the methodologies developed on this basis have large differences. After one decade of intensive research, the comparative study now highlights common aspects and differences of several methodologies.

Fires in road tunnels constitute complex scenarios with interactions between the fire, tunnel users and safety measures. More and more methodologies for risk analysis quantify the consequences of these scenarios with complex models. Examples for complex models are the computational fluid dynamics model Fire Dynamics Simulator (FDS) and the microscopic evacuation model FDS+Evac. However, the high computational effort of complex models often limits the number of scenarios in practice. To balance this drawback, the scenarios are often simplified. Accordingly, there is a challenge to consider complex scenarios in risk analysis.
To face this challenge, we improved the metamodel used in the methodology for risk analysis presented on ISTSS 2016. In general, a metamodel quickly interpolates the consequences of few scenarios simulated with the complex models to a large number of arbitrary scenarios used in risk analysis. Now, our metamodel consists of the projection array-based design, the moving least squares method, and the prediction interval to quantify the metamodel uncertainty. Additionally, we adapted the projection array-based design in two ways: the focus of the sequential refinement on regions with high metamodel uncertainties; and the combination of two experimental designs for FDS and FDS+Evac.
To scrutinise the metamodel, we analysed the effects of three sequential refinement steps on the metamodel itself and on the results of risk analysis. We observed convergence in both after the second step (ten scenarios in FDS, 192 scenarios in FDS+Evac). In comparison to ISTSS 2016, we then ran 20 scenarios in FDS and 800 scenarios in FDS+Evac. Thus, we reduced the number of scenarios remarkably with the improved metamodel. In conclusion, we can now efficiently integrate complex scenarios in risk analysis. We further emphasise that the metamodel is broadly applicable on various experimental or modelling issues in fire safety engineering.