New trends in safety systems design optimization

Design and maintenance-strategy optimization of safety systems for potentially hazardous facilities have become a first-order social demand in the last decade. Not only because the nocive effects of contamination and industrial spillages on the environment are less tolerated, but also because the great amount of materials dangerous to human beings that are manipulated in the cities, or at their outskirts, turn some spots of our cities into very dangerous zones for their inhabitants. As a result, in the last decade safety system modeling methods have been developed, especially the so-called Fault Tree Analysis, in order to allow the introduction of new design alternatives as well as different maintenance strategies. The increasing complexity of the resulting optimization problems has led to the application of global optimization evolutive methods, especially genetic algorithms, in a single-objective environment as well as in a multiobjective one. Despite recent progress in this subject, there persist important limitations still not solved that limit the applicability in multiple real problems, and therefore current methodologies are not part yet of the tool set used by most engineering design groups. Among them, the computational cost of fault tree quantitative evaluation, and the loss of efficiency of the genetic algorithms due to the different nature of the variables to optimize. In this paper, a review of relevant contributions in safety system design, from a conceptual point of view as well as from an algorithmic one, is presented. Emphasizing the advantages and drawbacks of the use of genetic algorithms. Non-solved problems will be analyzed and described using current methodologies and the most recent contributions to solving such problems will be presented, including the usage of the new global optimization evolutive method, known as flexible evolution, which shows excellent properties for design optimization. The role that Monte Carlo simulation methods may play when they are used for approximated evaluation of the objective function will be also analyzed, as well as the necessary modifications performed in the genetic algorithm in order to keep its robustness. Several safety systems design examples will be presented as practical application of the methodological revision. Refs. 4 (author)