Product Design Optimisation Methodology
Market demand is continuously adding more and more constraints onto industries, such as safety standards, environmental standards, cost reductions, and shorter product development cycles. To validate the choice among several design options, companies need to run a large number of numerical simulations requiring more and more computing power. In this context, industries must implement a more efficient design methodology. Eurodecision proposes such a methodology, that enables the user to efficiently explore the design space while minimising the number of simulations to run.
Eurodecision offers services for carrying out this type of study using its Test Management Toobox (OGE). Their purpose is to optimise the product during the design phase by factoring in numerous criteria and finding the best possible compromise within the allotted time for the study. This provides invaluable decision support for designers. We have already carried out a large number of such studies for Renault.
Design problem definition
Designers receive specifications setting objectives for services such as pollution emission levels; mechanical, thermal, or aerodynamic performance; or the product's mass or volume. They must list the influential design parameters—such as part thickness, materials, and shapes; whether or not a given part is present; an angle, etc.—as well as their level of variation, while factoring in both technical and manufacturing constraints (removal from mould, weldability, etc.). In the methodology, parameters called "factors" correspond to the design decision variables, and "responses" are criteria based upon the specifications, mainly the output data for digital simulations or real tests.
Design optimisation methodology
Once the factors and responses have been defined, the actual study phase may begin. The figure below illustrates this process:
Typical steps of the design optimisation methodology
An initial design of experiments is produced, aiming at getting the maximum amount of information from the minimum number of experiments (simulations to run). According to the specific structure of the selected design of experiments, it is possible, for instance, to approximate the influence of the factors on each response, and to determine their interactions. Responses obtained via simulation of these first experiments allow the user to build statistical models (surface response models). Such models provide an approximation of the responses for any new experiment. They are used to generate new experiments obtained with multi-objective optimisation methods. The new configurations are validated using simulations and are then added to the experiment database. This makes it possible both to reassess the models in order to improve their quality, and to repeat the process iteratively, until satisfactory solutions are obtained.
This methodology generates solutions which are on the Pareto frontier presenting all the best compromises based on the many criteria of the study. Designers therefore obtain detailed results that are useful for decision-making.
Related topics:
The Test Management Toolbox (OGE)
Example of design optimisation studies
