In civil aviation, but also in many other disciplines, safety-driven probabilistic requirements are imposed on systems. In conventional development approaches, a required functionality is developed in a non-probabilistic, conservative way with independent performance budgets for individual subfunctions, thereby sacrificing possible system performance. As opposed to this, this paper presents a novel, model-based probabilistic development approach. This approach utilizes the knowledge of desired system dynamics, known uncertainties, and disturbances to gradually develop implementation-independent simulation models that are used for derivation of lower-level requirements. These models are also used to validate the requirements before actual implementation, which reduces development risk. The actual implementation is supported and optimized by using executable requirements that allow for automatic proof of compliance during simulation. This minimizes risk that the actual implementation violates safety-critical requirements and hence safety can be increased. Verification of implementation against probabilistic requirements is accomplished by stochastic simulations, using sophisticated algorithms that allow for an efficient evaluation of small probabilities related to safety-critical events. By dynamic allocation of admissible performance budgets to individual subfunctions during operation, the availability of the implemented functions is increased. This paper gives an introduction to this total capability approach for development of safety-critical functions. Examples for the individual steps of the specific development process are given, which show promising results.     «

In civil aviation, but also in many other disciplines, safety-driven probabilistic requirements are imposed on systems. In conventional development approaches, a required functionality is developed in a non-probabilistic, conservative way with independent performance budgets for individual subfunctions, thereby sacrificing possible system performance. As opposed to this, this paper presents a novel, model-based probabilistic development approach. This approach utilizes the knowledge of desired s...     »