Enhancing development flexibility: a grey-box strategy for adapting to unforeseen adverse conditions

This work introduces an approach to adapt the corridors of the solution space methodology using a grey-box strategy. The need arises when developed components fail to meet the requirements set by the solution space methodology. For example, when additional constraints, such as shape-related conditions, come into play, they render it impossible to find feasible designs. Our two-step approach addresses this challenge: (1) given a parametrized component, identify feasible configurations of the component; (2) change the corridors to find better overall feasible areas. To identify feasible areas, we use Active Learning Reliability to train a classifier to identify feasible configurations. To change the corridors, we add a set of extra conditions to the original solution space formulation. The extra conditions include the desired conditions on a corridor. We test the method in a crashworthiness study, wherein a component failed to fulfill all conditions from the solution space method. Through adaptation and looking into the feasible areas, we identify a set of corridors that all components fulfill. In the end, our methodology improves the solution space method for better support.