Generative Design

The Analysis Agenda - What Next for Engineering Simulation?

Generative Design

Making the Impossible a Reality

 

While engineers and designers can provide the best starting points, design exploration can quickly build on those parameters and generate the most effective and refined alternatives that we may never have thought possible. The engineering simulation community is now embracing this technology more than ever before.

NAFEMS has always been about ensuring no technology is “black box”, and that the physics is still understood. That’s important now more than ever.

 


Generative Design – State your objective, define your constraints, specify the number of design options, press GO and wait for the computer code to provide you with feasible design options. It is easy to see why the Generative Design vision has garnered interest over the last five years (see Figure 1).

We are all becoming used to allowing algorithms to make decisions for us and it would be foolish not to consider how the same assistance can be utilised to improve the engineering design process. So what can generative design offer the simulation community?

Figure 1 - Interest in the Google Search Term “Generative Design” over the last 5 years. The red line represents a moving average based on the last 13-week period of data.

We know that marketing departments are not going to be shy about presenting existing capabilities as part of the generative design tool kit. The cynics amongst us will look at Generative Design as another way of packaging up existing topology optimisation capabilities. These tools have become more relevant over recent years due to the rollout of additive manufacturing capabilities which can readily manufacture the biological designs that often evolve out of these toolkits and a generative design philosophy gives these tools another shot in the arm.

For Generative Design to truly offer something new, it must look at integrating both parametric and topology optimisation approaches when developing a design. But is there more, leading companies have been storing their simulation data for decades and have amassed huge databases of simulation results. As well as providing the “who, why, when, where” audit trail, these datasets can be used to train Machine Learning algorithms. This allows decades of hard wrought experience to be harnessed when developing future designs. Perhaps Generative Design is the glue that can bring big data and mature design optimisation techniques together?

If the Generative Design toolkit proves effective, then the entire design process may need to evolve. The current iterative process may be slow, but it allows the engineer to develop a physical understanding of a design problem. It is this understanding that allows intuitive leaps to be taken. I suspect that if these tools show their value we will be looking at them as a further weapon in the designer’s arsenal and not a replacement for the designer themselves.

Ian Symington, Technical Officer, NAFEMS