Protection of the health and life of the occupants in case of a hostile fire is the main safety objective of Fire Protection Engineering. On many occasions, open-space buildings, such as atria or multi-functional assembly buildings cannot be realized in accordance with the traditional safety requirements. Within this thesis a probabilistic methodology is developed in order to quantify the safety levels implied by the prescriptive codes. The definition of risk and its relation to reliability and probabilistic design procedures is discussed and structured under various aspects. To achieve the most accurate solutions, various aspects of the current procedural methods of life safety analysis are analyzed, discussed, and improved. Based on the findings, stochastic models are found for the various uncertain parameters. After an introduction to reliability theory, an adaptive response surface method based on interpolating moving least squares (IMLS) is developed, validated and benchmarked. This allows for a fast and efficient calculation of failure probabilities of life safety design solution using state-of-the-art numerical fire protection engineering tools. The method is applied to a typical building and the current reliability levels are (quantitatively) derived considering various possible scenarios. A subsequent inclusion of various fire protection barriers (i.e. sprinklers etc.) shows their quantitative effect and how they can be considered within a probabilistic safety format. Thus a way for a risk-informed and performance-based life safety design is paved.weiterlesen