Generative Design - a new Mindset

The world of engineering has always been in flux and will continue to change. Engineers are constantly striving to find new solutions to improve machines, products and processes. And not only for the result of their work, but also for their daily work itself. This is why another change from manual drawing to digital drawing (CAD) to automated design is expected. The usage of CAE tools will increase to more and more workstations and will be simplified to be used by the normal engineer instead of highly specialized experts.

Thus, conventional CAE processes need to be rethought. While conventional software is based on a deep mathematical basis for finite element computation and requiring detailed knowledge of model construction and result interpretation, it must be transformed into a more user-oriented product. The technology of “topology optimisation” is one way to generate optimal designs for structural parts. Therefore, such an FE model is solved iteratively, while the number of elements and thus the material is reduced in all areas where low relevance can be found. These iterations are performed until certain boundary conditions (displacement, mass, stress) are not exceeded and as little material as possible is consumed. Unfortunately, this requires a high effort to create and solve highly accurate FE-models with perfect meshes, while the result is far from a final solution. Often unnecessary accuracy is used to determine a rough shape that is never used in this way but has to be redesigned manually with CAD software. The final geometry therefore often shows a lack of optimisation quality and shows semi-optimal results, even though a lot of time and expertise has been spent on.

This is where Generative Design kicks in. One of “the 7 habits of highly effective generative design” (see our White paper: https://www.mscsoftware.com/MSC-Apex-Generative-Design-Whitepaper ) is “Design for usability”, which exactly refers to a smart, lean and fast process of generating optimal designs with low effort. The above graphic shows how the long and exhausting process in existing FE-Software is replaced by a much shorter and highly automated process of MSC Apex Generative Design.

Thus, we expect the user to think more about what he wants and not what the mathematic basis requires. This is a major shift in the CAE business and will allow a much broader usage from more engineers for more tasks in a shorter time and with better results.

Better results are gained because of a combination of “Design for Exploration” and “Design for productivity”, two more habits of a highly effective generative design. We do see two major key differentiators of Generative Design towards Topology optimisation, although it is based on the well-known technology. On the one hand this is the generation of manufacture ready designs without the need of manual reworking and the other one is to generate a bunch of slightly different designs meeting the same requirements but differing in terms of number of struts or overall complexity. Delivering this in a reasonable time frame, the aim is over lunchtime, requires high performance computing using the latest capabilities and a perfect balance between speed and accuracy. Thus, MSC Apex Generative Design does not rely on outdated FE-code with a useless dedication to accuracy, but on an automated mesh generation with careful resolution changes during the simulation and an overall reasonable simulation accuracy. Every design is always geometrically and mechanically feasible, can be used and trusted. Nevertheless, for critical components and minimum safety factors, a verification with the reliable, tested and verificated FE-solvers such as MSC Nastran is always recommended.

Therefore, Generative Design is never a straight line from setting up a model and taking the outcome. It is more a creative process where variety is a key for successful designs. You can think about a funnel where hundreds (or software internally even thousands!) of different designs are thrown in and then thinned out by the optimisation goals such as stress, weight, design or manufacturing constraints. In the future our solution will be developed further and combined with other solutions from the MSC portfolio such as Simufact Additive, Digimat AM and MSC Nastran to go through all of these decisions with a completely automated workflow.

In the end, the idea is to combine known data points (mainly geometrical ones, like connection points, bearings, fixations etc.) with information (loads, constraints, materials) with knowledge (FEM, load flux etc.) to gain creativity. And this creativity often looks very different from what was expected and consequently creates new solutions to known problems to enhance products and processes.