ADVANCED MATERIALS & PROCESSES | NOVEMBER/DECEMBER 2024 31 reshoring data that is currently stored with vendors, adding markup language to legacy PDFs, and implementing software development workflows to improve the infrastructure. As more of the structured data is migrated to graphical representations and more of the unstructured content is enhanced with AI/ML tools, the easier the search and retrieval process will be for ASM members and customers to access the information. The Society has been testing a custom chatbot based on ChatGPT LLM, but with additional training on ASM content sets including the ASM Handbooks Online, ASM Technical Books, and ASM Failure Analysis Database. Saal (Citrine): Central to what Citrine offers its customers is the combination of integrated computational materials engineering (ICME) with AI tools to improve manufacturing processes. Their AI platform can learn from small data sets and product experts, combining that knowledge with AI models that mimic a scientist’s intuition. Ultimately, the platform recommends winning candidates—recipes for ideal experiments to run in the lab—resulting in the desired materials design. It is common that materials datasets are often on the small side, trapped in difficult formats, poorly labeled, and siloed. Companies like Citrine overcome these challenges by developing algorithms that capture small datasets in a graph-based data model that documents a material’s origin and processing history. They also provide consulting expertise to help with an organization’s digital transformation and ensure that data privacy is fundamental. Stuckner (NASA): Informatics has been applied to many NASA materials and manufacturing projects. Automatic microscopy analysis is a key tool for quantifying features such as grains and defects. Its application can improve speed, accuracy, and cost and also establish quantitative processing, structure, and properties relationships to help inform the materials design process. Ultrafast surrogate modeling is used for rapid design tools, larger simulations, and optimization by having AI models trained to emulate physics-based models. The results are delivered with both high speed and high accuracy, bypassing any trade-off. AI/ML is also used to determine the optimal next experiment in a materials development process, saving time and money. In the future, the industry can expect an AI-driven robotics revolution to automate research labs. Materials will be developed faster at a fraction of the cost. Taheri-Mousavi (CMU): When using machine learning to design next-generation structural alloys, the goal is for mechanical performance to be enhanced. To this end, the CMU team uses a hybrid approach for alloy design that combines ML and CALPHAD calculations. In other cases, additional ICME models are also added to the framework. These hybrid approaches can be used to design various alloys produced by various manufacturing techniques. Simulations are employed as they are less expensive and less time consuming than traditional experiments. Several recent projects showcase Taheri-Mousavi and her team’s adeptness with these AI/ML tools. In a DARPA METALS project, they designed gradient additively manufactured alloys using materials informatics. For the Naval Nuclear Laboratory, the CMU team designed next-generation structural alloys in extreme environments, which involved developing a generative AI framework for multi-field, multi-modal, agent-based design of alloys. KEY TAKEAWAYS The panelists’ presentations were followed by audience questions. Take- aways from the panel include: With any materials design or manufacturing process, start by defining the problem to determine where AI can add value; AI can accelerate the design and discovery of materials and optimize manufacturing; cloud computing will become the standard; and most importantly, with any application of artificial intelligence and machine learning in materials and manufacturing, keep the human in the loop. ~AM&P GET ENGAGED, GET INVOLVED, GET CONNECTED The ASM Materials Informatics Committee is a group of ASM members working to drive industry-wide efforts in implementing AI and ML methods to accelerate the design, discovery, and deployment of new materials, processes, and analysis techniques. Projects include developing a new ASM Handbook volume and organizing IMAT conference programming. Members with similar interests are welcome to join. Contact committee chair Joshua Stuckner at josh.stuckner@gmail.com. Automated labs will become more common as AI-driven robotics continues to develop. Courtesy of Lawrence Berkeley National Laboratory.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==