Modeling

Previous approaches to detailed modeling of fire spread in rail vehicles have required a large number of experiments, extensive optimization, and extensive variant calculations. In this project, two complementary AI approaches, KIM (Material AI System) and KIB (Fire AI System), are being researched and implemented. KIM aims to completely reduce the optimization calculations required for determining material parameters, allowing for the determination of necessary material parameters solely from data obtained from a few tests, such as the Cone Calorimeter. This test is already used for the fire protection assessment of all relevant components of rail vehicles, as required by the European standard (EN 45545-2), so the proposed AI approach (KIM) will eliminate the need for additional effort in design fire simulations in the future. KIB aims to significantly shorten the high computational cost for variation calculations of the entire vehicle and focuses on fast calculation of fire spread variations. The goal is to replace the cost-intensive traditional models used for variant calculations of design fires with the KIB AI system.

The data required for training the AI systems are sourced from both laboratory and real-scale experiments, as well as synthetic data generation. This database will be made publicly available and continuously updated with new data for each use of the AI systems. This enables the AI systems to be continuously trained and improved. Central to this project is the validation, combined with the assessment of uncertainties associated with the AI systems.

The methods developed in this project will be generalized to the extent that they can be applied outside of rail transportation, such as in preventive fire protection or forest fires. These globally unique AI systems will open up new perspectives for fire protection in both application and research, offering innovative solutions in various domains.