AMP_06_September_2021

A D V A N C E D M A T E R I A L S & P R O C E S S E S | S E P T E M B E R 2 0 2 1 2 2 Indeed, it is currently being used suc- cessfully in multiple ways at materials and chemicals companies. NEW PRODUCT DEVELOPMENT AI-guided sequential learning re- duces the number of experiments needed to find materials with target properties [1] . This speeds up develop- ment times and enables companies to get new products to market faster. Be- ing there first, they can charge more un- til their competitors catch up. It also enables the exploration of larger, high-dimensional design spaces. Running a matrix of experiments over a large number of compositions and pro- cessing parameters is expensive. How- ever, first running an AI model to work out the likelihood of achieving target properties with materials candidates means that experiments are more like- ly to succeed. Rapidly exploring a broader design space doesn’t guarantee finding novel, patentable, high-performing materials, but it does help scientists quickly deter- mine if such materials exist within their current experimental capabilities. Of course, better materials lead to higher margins for producers. For example, consider the first 3D-printable aluminum alloy. 3D print- ing with standard aerospace-grade alu- minumalloys results in amicrostructure a bit like a flaky biscuit, which is not ac- ceptable for aerospace applications. HRL Laboratories spotted a market and worked with Citrine Informatics and re- searchers at the University of California, Santa Barbara to explore how nanopar- ticles could seed grain nucleation and refine the resulting microstructure. Citrine was given rules on lattice spacing, thermodynamic stability, and density, and applied them to algorith- mically search through 11.5 million combinations of powder and particles. The next step involved testing 100 can- didate couplings. As a result, HRL Lab- oratories was able to register the first aluminum alloy powder feedstock, shown in Fig. 2, for off-the-shelf addi- tive manufacturing machines with the Aluminum Association [2] . The first com- mercial customer for this new prod- uct is the NASA Marshall Space Flight Center. AGILITY, RESILIENCE AND CUSTOMER RESPONSIVENESS Once AI is part of the application engineering and development work- flows, reusable digital assets are ac- cumulated and can be deployed to respond to changing requests. Inputs can change, as is the case when ingre- dients/raw materials in existing prod- ucts become unavailable due to supply chain issues or regulation. In addition, new customer requests and other fac- tors can result in changes for target out- puts. In both cases, a company can go back to its AI models to make quick ad- justments and rerun the model. In order to have more options and secure its supply chain, a global lead- er in thermoplastics with operations in the U.S. and Europe worked with Citrine to evaluate how U.S.-based ingredients performed in European products. As il- lustrated in Fig. 3, the team developed an AI model to predict the properties of plastics based on ingredients and Fig. 1 — Snapshot of part of a material and process history in open-source graphical expression of materials data (GEMD) format. Fig. 2 — After 100 candidate combinations were put forward and tested, HRL Laboratories registered the first aluminum alloy powder feedstock for off-the-shelf additive manufacturing machines with the Aluminum Association. Courtesy of HRL Laboratories LLC.