Multidisciplinary Design Optimization

In its truest essence, MDO means optimizing a design, within a set design space, across multiple disciplines by computer. For those unfamiliar with it, a good example from the automotive industry is a body-in-white (BIW). Many disciplines are involved in the design of a car body, each with their own objectives and requirements. For example, a stylist might suggest a certain shape that looks good but is aerodynamically inefficient. Or the crash safety team could propose a certain material that helps to achieve a 5 star crash rating more easily but might be poor from an NVH perspective. And so on….so the key in MDO is not necessarily optimizing the performance of a particular discipline but rather trading and balancing the competing requirements of individual disciplines to optimize the overall system and find the best compromise. 

In MDO, we can distinguish single-objective (e.g. minimize weight) and multi-objective optimization (e.g. minimize weight and maximize strength). In case of multi-objective optimization, the key is finding the designs that comprise the pareto frontier. This is where MDO frameworks come into play. If the optimization problem is small enough, i.e. one discipline, one objective and only a handful of independent design variables, an experienced engineer may be able to ‘manually’ optimize the design. Most real world applications however are more complex, thus necessitating smart algorithms to explore the design space and quickly narrow down to the region of interest, and ultimately finding the (set of) optimal design(s).