Delft AI Lab Lunch with MACHINA and SLIMM Lab

About MACHINA Lab

Machine Intelligence Advances for Materials

AI is key to create the next generation of materials because they need to be adaptive, multi-purpose and tunable. Such complex material behaviour requires the consideration of too many design possibilities for the human mind to process, but not for machine learning algorithms. These methods create an opportunity to invert and improve the design process by shifting from experiments to data-driven design with computer simulations, even if the computer models are missing some information. The essential requisite is that ‘enough’ data about the problem of interest is available. Given enough data, AI can provide a treasure map and the scientist can then find the treasure (new material).

The MACHINA Lab focuses on different aspects of the data-driven process, including speeding up material simulations, as well as exploring new machine learning and optimisation methods that facilitate analysis and design. Currently, these new algorithms are being used to design environmentally friendly plastics and metamaterials for biological applications, i.e. materials with exotic properties (such as supercompressibility) that emerge from exploring new geometries and new combinations of constituents.

About SLIMM Lab

AI for smart materials modelling

Materials of the future will rely on nanostructures painstakingly tailored to design constraints and reach unprecedented levels of performance. In order to design these new materials, accurate simulations of material behavior across the scales are needed. Yet, even one of these models currently takes months to run on conventional computers. The goal of the SLIMM lab is to facilitate virtual material testing across the scales by combining Bayesian machine learning and multiscale solid mechanics into new smart material models and learning frameworks. Why virtual testing through smart multiscale models? Because a) speed: Designing new materials without waiting for years for simulations to run; 2) accuracy: Precise nanostructure optimization and deeper knowledge on material behavior; and 3) sustainability: Fewer destructive experiments needed and lower energy use through faster simulations