Practical Approaches to Generative AI in Manufacturing
By Federico Sciammarella, CTO, MxD
On March 14, 2024, MxD (Manufacturing times Digital), the national digital manufacturing and cybersecurity institute, hosted a workshop on Generative AI for manufacturing. This workshop was a platform to explore the transformative potential of artificial intelligence in the manufacturing sector, with a special focus on generative design, process optimization, predictive maintenance, and quality control. The event brought together industry experts, researchers, and manufacturing professionals to share insights, case studies, and practical applications of Generative AI technologies, enlightening the audience about the benefits of this technology in manufacturing.
GigaOm’s COO Howard Holton delivered the keynote address- Revolutionizing Manufacturing: The Power of Generative AI, which advanced many key insights. He not only emphasized the disruptive potential of Generative AI, particularly in democratizing creativity, but also sparked excitement about the future of this technology. He drew a parallel to the 16th century, when the printing press revolutionized knowledge dissemination, making it accessible through standardized printed books, to illustrate the transformative power of Generative AI.
In his presentation, Holton also advanced how Generative AI can, in the future, enable an unprecedented new level of creativity; however, he cautioned that we are not yet there today. He highlighted that we are still in the experimentation phase and that there are key elements to consider while learning how best to use this tool. Most importantly, we must continue to learn and embrace the technology in the right conditions (personal or professional) as we experiment. Otherwise, you will not know how this can work for you. There are three major criteria to consider when applying AI projects within your organization. The first is to identify a champion in your organization to support the effort (they should establish clear, achievable objectives for the AI initiatives). Secondly, it should solve a wish, not a want. What do you wish you could do that you cannot do today? Finally, Holton emphasized how the organization must budget for Generative AI solutions. Looking ahead, “we envision a future where Generative AI enhances our creative capabilities and revolutionizes how we approach problem-solving and innovation in every sector.”
At the workshop, esteemed experts presented three unique case studies, each demonstrating the current use of Generative AI in manufacturing. The first was from the Director of Field Engineering, Mr. Lauritz Lont, from Cognite. His presentation focused on contextualized industrial data, which refers to data that is enriched with additional information to provide a more comprehensive understanding and its application to their Remote-Operations Control Room (ROCR). By combining Large Language Models (LLMs) with Industrial Knowledge Graphs, it is possible to reap all the benefits of Generative AI and have your operators get the information they need when it is most needed. This is not a trivial task, so the case was presented on providing context to industrial data with knowledge graphs. Creating a pathway and process for digitizing content and codifying it in context allows you to gain new insights into your data using Generative AI. The application of their tool in various scenarios was presented, highlighting how powerful this type of tool can be for manufacturing operations.
The next case study came from Cognizant's AVP & Offering Lead, Mr. Sharath Prasad. This use case focused on how they support an Agri-Products company in their OT maintenance and how they deliver this value to the shop floor. As previously mentioned, data comes from many locations (dozens of inputs within one point of one area in that location). The interface was designed, as in most of their use cases, to allow the operators on the shop floor to ask questions and then provide some responses to which the person can continue in conversation. With their Apex tool, they demonstrated how to identify the independent variables for yield loss and understand the leading indicators for this. They can then correlate the loss to the root cause from real-time data and standard operating procedures.
The CEO and Chair of Digital Manufacturing, Mr. Andre Wegner from Authentise, presented the last case study. With their origins in additive manufacturing (AM), their software tools focus on workflow for most agile engineering and production processes. Given the lack of data in the AM process, they wanted to experiment and understand how they might leverage LLMs for their industry. That resulted in the creation of 3DGPT, which examines 12,000 scientific AM papers to provide answers to questions the public may have in this space. This allowed them to consider using their experience to help augment their current products. They incorporated Generative AI into their “Threads” platform to provide a unique experience for their users. This tool can track a project's data from idea to design and allows for annotation along the way. By adding the Generative AI tool, it can now make recommendations on important actionable items based on information the user provides. This was then applied to a use case for the Department of Defense, as they were asked to develop wind tunnel blades for their NFAC Wind Tunnel in Mountain View, California. With the tools developed, they could contextualize the data meaningfully, allowing them to determine potential risks for the project after analyzing the data.
The final speaker for the workshop was Experience Design & Product Strategy Leader Ms. Anna Garske from Slalom Build, who emphasized keeping the human in the loop. Their key drivers for projects focus on designing Generative AI systems that augment human capabilities, enhance the user experience, and align with the values relevant to the organization and users. One example demonstrated how it is important to start the Generative AI journey by speaking with the people who will use the tool. This will give you a comprehensive understanding of the relationships between users and data. Then, establish a process that will support a curious and scientific mindset and allow the users to ask questions and get comfortable and familiar with the tool. This creates the champions required to make your Generative AI projects successful.
As a crucial part of the workshop, an interactive brainstorming session was held to identify gaps, barriers, and challenges to implementing Generative AI projects, along with the high-value areas and opportunities. This session was a collaborative effort where all the participants shared their insights and ideas.
The output of this session, along with the presentations and associated slide decks, can be found here.
About the author
Federico Sciammarella is the CTO for MxD. In this role, he is responsible for developing the technology roadmaps of the Institute and its ecosystem in digital and cyber for manufacturing. In 2020, he initiated a program targeted at emerging technologies that connect academics to the industry to bring up early-stage technology through to later stages of technology readiness. He also supports MxD Cyber as the national center for cybersecurity in manufacturing to ensure its continued growth and impact within the American supply chain. Before joining the Institute, he served as the Interim Chair for the Mechanical Engineering Department at Northern Illinois University’s College of Engineering & Engineering Technology. As the Director of the Advanced Research of Materials and Manufacturing (ARMM) Laboratory, his research efforts brought in millions of dollars and produced many journal articles on laser-based manufacturing focusing on metal 3D printing.
By Federico Sciammarella, CTO, MxD
On March 14, 2024, MxD (Manufacturing times Digital), the national digital manufacturing and cybersecurity institute, hosted a workshop on Generative AI for manufacturing. This workshop was a platform to explore the transformative potential of artificial intelligence in the manufacturing sector, with a special focus on generative design, process optimization, predictive maintenance, and quality control. The event brought together industry experts, researchers, and manufacturing professionals to share insights, case studies, and practical applications of Generative AI technologies, enlightening the audience about the benefits of this technology in manufacturing.
GigaOm’s COO Howard Holton delivered the keynote address- Revolutionizing Manufacturing: The Power of Generative AI, which advanced many key insights. He not only emphasized the disruptive potential of Generative AI, particularly in democratizing creativity, but also sparked excitement about the future of this technology. He drew a parallel to the 16th century, when the printing press revolutionized knowledge dissemination, making it accessible through standardized printed books, to illustrate the transformative power of Generative AI.
In his presentation, Holton also advanced how Generative AI can, in the future, enable an unprecedented new level of creativity; however, he cautioned that we are not yet there today. He highlighted that we are still in the experimentation phase and that there are key elements to consider while learning how best to use this tool. Most importantly, we must continue to learn and embrace the technology in the right conditions (personal or professional) as we experiment. Otherwise, you will not know how this can work for you. There are three major criteria to consider when applying AI projects within your organization. The first is to identify a champion in your organization to support the effort (they should establish clear, achievable objectives for the AI initiatives). Secondly, it should solve a wish, not a want. What do you wish you could do that you cannot do today? Finally, Holton emphasized how the organization must budget for Generative AI solutions. Looking ahead, “we envision a future where Generative AI enhances our creative capabilities and revolutionizes how we approach problem-solving and innovation in every sector.”
At the workshop, esteemed experts presented three unique case studies, each demonstrating the current use of Generative AI in manufacturing. The first was from the Director of Field Engineering, Mr. Lauritz Lont, from Cognite. His presentation focused on contextualized industrial data, which refers to data that is enriched with additional information to provide a more comprehensive understanding and its application to their Remote-Operations Control Room (ROCR). By combining Large Language Models (LLMs) with Industrial Knowledge Graphs, it is possible to reap all the benefits of Generative AI and have your operators get the information they need when it is most needed. This is not a trivial task, so the case was presented on providing context to industrial data with knowledge graphs. Creating a pathway and process for digitizing content and codifying it in context allows you to gain new insights into your data using Generative AI. The application of their tool in various scenarios was presented, highlighting how powerful this type of tool can be for manufacturing operations.
The next case study came from Cognizant's AVP & Offering Lead, Mr. Sharath Prasad. This use case focused on how they support an Agri-Products company in their OT maintenance and how they deliver this value to the shop floor. As previously mentioned, data comes from many locations (dozens of inputs within one point of one area in that location). The interface was designed, as in most of their use cases, to allow the operators on the shop floor to ask questions and then provide some responses to which the person can continue in conversation. With their Apex tool, they demonstrated how to identify the independent variables for yield loss and understand the leading indicators for this. They can then correlate the loss to the root cause from real-time data and standard operating procedures.
The CEO and Chair of Digital Manufacturing, Mr. Andre Wegner from Authentise, presented the last case study. With their origins in additive manufacturing (AM), their software tools focus on workflow for most agile engineering and production processes. Given the lack of data in the AM process, they wanted to experiment and understand how they might leverage LLMs for their industry. That resulted in the creation of 3DGPT, which examines 12,000 scientific AM papers to provide answers to questions the public may have in this space. This allowed them to consider using their experience to help augment their current products. They incorporated Generative AI into their “Threads” platform to provide a unique experience for their users. This tool can track a project's data from idea to design and allows for annotation along the way. By adding the Generative AI tool, it can now make recommendations on important actionable items based on information the user provides. This was then applied to a use case for the Department of Defense, as they were asked to develop wind tunnel blades for their NFAC Wind Tunnel in Mountain View, California. With the tools developed, they could contextualize the data meaningfully, allowing them to determine potential risks for the project after analyzing the data.
The final speaker for the workshop was Experience Design & Product Strategy Leader Ms. Anna Garske from Slalom Build, who emphasized keeping the human in the loop. Their key drivers for projects focus on designing Generative AI systems that augment human capabilities, enhance the user experience, and align with the values relevant to the organization and users. One example demonstrated how it is important to start the Generative AI journey by speaking with the people who will use the tool. This will give you a comprehensive understanding of the relationships between users and data. Then, establish a process that will support a curious and scientific mindset and allow the users to ask questions and get comfortable and familiar with the tool. This creates the champions required to make your Generative AI projects successful.
As a crucial part of the workshop, an interactive brainstorming session was held to identify gaps, barriers, and challenges to implementing Generative AI projects, along with the high-value areas and opportunities. This session was a collaborative effort where all the participants shared their insights and ideas.
The output of this session, along with the presentations and associated slide decks, can be found here.