A smart manufacturing environment is centered on the use of intelligent, connected production equipment and devices that allow for data-driven decision-making to optimize processes and productivity across the product lifecycle. In many initial smart manufacturing applications, manufacturers used advanced analytics to progress from condition-based monitoring to predictive and prescriptive analytics. The idea was to create a pathway from connected and intelligent systems to AI-enabled self-aware equipment and, ultimately, to autonomous self-healing systems.
These efforts showed that manufacturing data is a good fit for AI (artificial intelligence) and especially ML (machine learning). Manufacturing is a natural source of analytical data which is easier for machines to analyze. Hundreds of variables impact the production process, and while these are very hard to analyze for humans, ML models can easily predict the impact of individual variables in complex situations.
As AI is applied to manufacturing in digital twin applications, it is making much more accurate manufacturing process design decisions, including potential problem diagnosis and resolution when inconsistencies crop up in the fabrication process. While half of a digital twin is a virtual representation of the physical part as designed, it is much more than just a CAD model because it draws from the physical operations of the equipment it represents to make optimal decisions on the production process.
Manufacturers implementing any digital twin project should begin by capturing and managing the actual physical configuration of the asset. Additionally, due to the many use cases for a digital twin across the product lifecycle, implementers would be well served to employ digital twin technology that can integrate a flexible/dynamic data model. ML training algorithms can be developed and implemented much more efficiently and faster with real-time production data.
Generative design
The emergence of AI-powered generative design—a process in which a design engineer enters a set of requirements and parameters for product design that considers fit, form and function—can generate multiple iterations of possible design solutions, considering factors like materials, manufacturing processes, structural integrity, cost and performance.
However, it's important to note that not all generative design processes are exclusively powered by AI. Some generative design approaches may use traditional algorithms, search engines, mathematical models, or expert knowledge to create design alternatives. While AI technologies have enhanced the capabilities of generative design, the term itself refers to the broader concept of generating multiple design solutions through systematic processes, which may or may not involve AI techniques depending on the specific implementation.
Whereas traditional design optimization techniques took a more generalized approach to part optimization, generative design can be much more specific, focusing on individual features and requirements such as manufacturability, mechanical properties based on alternative materials and operational constraints of the part. While many designs are idealized and conceptual, manufacturing processes take place in real-world conditions that might not be constant. An effective generative design algorithm can include a high level of understanding of the design/build process and enable a manufacturing engineer to better solve the challenge of manufacturability.
It should also be noted here that the recent emergence of Large Language Models (LLM) would not be suitable for the generation of engineering design models. These LLM based algorithms could possibly be helpful in generating informational content for manufacturing instructions and processes but would be unable to generate design models based on engineering criteria.
Additive manufacturing
The manufacturing environment where the greatest opportunity for AI to add value is in additive manufacturing (AM). Additive processes are primary targets because their products are generally more expensive and smaller in volume. The combined additive design/build process lends itself to generative design. Engineers can focus on a variety of constraints such as light-weighting, optimal strength-to-weight ratio, fit and any number of functional requirements that best meet the design requirements.
Currently, generative design/build applications allow the manufacturing engineer to design tooling and fixtures along with the specific printing capabilities and material options offered by the printing machine. This integrated AM design/build capability is enabled by AI-powered generative design.
Dick Skansky is a senior analyst at ARC Advisory Group.
