Generative design is an automated method of creating Computed Aided Design (CAD) geometry for part and assembly models using machine learning. Rather than defining the shape using feature-based or direct modeling methods, a user specifies the design space (like an envelope, including areas to be preserved or excluded), operating environment conditions, materials, and manufacturing constraints. An algorithm then computes one or more potential solutions. The user can then filter down the results to select an optimal choice. Generative design is faster and, in many ways, more reliable than traditional iterative human-driven methods.
Generative design and closely related topology optimization tools have been available for several versions in Creo. Let’s look at four aspects that make generative design different and better than traditional CAD workflows.
You’re building requirements into the design. All products start with requirements. The top-level requirements are decomposed into subsystem requirements, which are then distilled into component requirements. Even though we may know the structural requirements for a part or a subassembly, in the past, those requirements weren’t validated until the design was complete. With generative design, we set up our design by applying load cases to our model. This ensures that the solution meets the criteria from the beginning.
The manufacturing method informs the design. With typical workflows, parts are designed, validated using simulation and analysis tools, and are then handed off to a process engineer to assess if the part can be built using subtractive or additive manufacturing. With generative design, we add manufacturing criteria to the optimization study. These include:
This assures the solution can be manufactured with the appropriate method.
You generate multiple concepts quickly. Product development is always constrained by schedule. Products can never get to market fast enough. Usually during the initial design phase, a handful of potential concepts are created. With the help of machine learning, multiple concepts can be generated in less time than it takes a human to create a single concept.
The resulting model can be modified by the user. The part model created by generative design results in B-rep (boundary representation) geometry. This is the same kind of geometry produced in the subdivisional modeling workflow for surfacing. This geometry can be modified with standard parametric features or direct modeling tools such as the Flexible Modeling Extension.
A study to optimize a model with generative design can be set up in minutes. The general process is as follows:
The speed of this process is a game changer for companies on the leading edge.
Generative design is a proven concept that can help you create lighter parts in less time so you can get to market quicker. Explore more of generative design's potential to help your product development teams here.
Dave Martin is a Creo, Windchill, and PTC Mathcad instructor and consultant. He is the author of the books “Top Down Design in Creo Parametric,” “Design Intent in Creo Parametric,” and “Configuring Creo Parametric,” all available at amazon.com. He can be reached at firstname.lastname@example.org.
Dave currently works as the configuration manager for Elroy Air, which develops autonomous aerial vehicles for middle-mile delivery. Previous employers include Blue Origin, Amazon Prime Air, Amazon Lab126, and PTC. He holds a degree in Mechanical Engineering from MIT and is a former armor officer in the United States Army Reserves.