Understanding the Role of AI in Generative Engineering Design

Author photo: Dick Slansky
By Dick Slansky

Keywords: Artificial Intelligence, Generative Design, Topology Optimization, Machine Learning, CAD, CAE, Simulation, PDM, Deep Generative Models (DMGs), Artificial Neural Networks (ANNs), Neural-Symbolic AI, ARC Advisory Group.


When a design engineer sets out to design a new part, component, or assembly the intent is to meet the design requirements for fit, form, and function as well as incorporate a certain degree of innovation and elegance to the overall design. Moreover, she does not want to re-invent the wheel by introducing a new design. An important aspect of the design process is design reuse, i.e., finding an existing part design that meets some or many of the new design requirements, with the possibility that there are many similar parts already in existence. 

The concept of design reuse has been a significant part of the design process for decades and became much more refined with the advent of computer aided design (CAD) tools and product data management (PDM) systems. These systems allowed design engineers to search their own internal engineering organizations to find similar design components and details. Later, engineering design search engines expanded the pool of designs well beyond company boundaries and enabled designers to search the Internet in a much larger pool of part designs. In some ways, this process was the forerunner of generative engineering design where the goal was not only to meet basic design requirements based on existing designs, but to produce the most optimal and innovative design that significantly improved on existing parts and components. 

Software driven engineering design optimization tools also emerged that aided the designer in developing the most optimal and efficient design based on functional requirements. 

One such software tool was Topology Optimization. Since its development in the 1990s it has been used to derive an optimal material distribution for a design’s functional usage requirements. Topology optimization is a foundational technology that preceded AI-driven generative design. It is an algorithm that reduces the material in a design object while allowing it to meet design intent and function. 

AI in Generative Engineering

Generative design, a much newer concept, pushes the boundaries of design intent and optimization using AI technologies like Deep Generative Models (DGMs) a form of Machine Learning and Neuro-Symbolic AI, to allow the designer to create innovative designs based on exact engineering requirements. The algorithms will generate many design possibilities that satisfy specified fit, form, and functional requirements including manufacturability.

This Insight will examine the evolution and current use of engineering design tools like topology optimization and AI-driven generative design, how they compare and differ, and how they will allow industrial design enterprises to engineer and design products faster and smarter in the future.

AI-Driven Generative Design Redefines the Engineering Process

While AI can be a powerful tool for solving engineering design problems, it is not going to replace the engineer and designer. AI is certainly helping engineers develop, optimize, and assess design possibilities, but it will not replace human creativity and innovation, at least not yet. What it does is free the designer from repetitive tasks, multiple calculations, searching for optimal designs, and resolves many conflicting design constraints, allowing the engineer to focus on problem-solving and innovation.  

Accepted engineering design techniques are well understood and broadly utilized. The wide use of these traditional approaches has led to and delivered decades of new products and engineering breakthroughs. The engineering design process typically involves several steps:

  • Ideation and conceptual phase – identify the engineering problem and develop a concept.

  • Creation – design (fabricate) a prototype of the concept (usually a CAD model in silico).

  • Redefine and enhance the design.

  • Validate the design – test with CAE.

  • Build – develop optimal production processes for the design.

This process is inherently linear and has significant drawbacks even with current CAD and CAE tools and processes. Extensive subject matter expertise is required at each step. Even with currently used advanced software tools, all aspects of the design requirements, dimensions, features, functions, material, and weight must be exactly defined and tested to produce a practical part that can be manufactured and meet specifications. 

AI-driven generative design addresses all the established engineering methodologies and much more. The constantly improving algorithms not only deal with current design techniques and significantly shorten the product design lifecycle but deliver more design alternatives and possibilities than human designers have time to create or evaluate manually, and, more importantly, concepts that the designer may never have thought possible. 

Generative design software enables the designer to set performance and prioritize parameters such as cost, materials, manufacturing methods, weight, shape, and even aesthetics of the product, and the algorithm generates a menu of alternatives to consider. In terms of the product development lifecycle, generative design is a combination of AI, CAD, simulation and test (CAE), and topology optimization, all working in conjunction. Today, PLM providers are offering various generative design solutions based on these combinations. Evolving AI-driven algorithms will continue to increase the robustness of the generative design process and offer the designers an ever-expanding range of design possibilities. 

Currently, there are specific areas of the design/build process that generative design is having an immediate impact, such as additive manufacturing (AM). Since generative design is an iterative process that generates multiple design outputs the process is made to order for AM. 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. The result are parts that can only be produced by 3D printing and meet very specific functional requirements. 

The AM lifecycle begins with discovering the right material and the application with in-silico materials simulation engineering to find the optimal material compound. Next is function-driven generative design, followed by the manufacturing process definition and production planning of the part. Each phase of this lifecycle process can be driven and enhanced by AI technology. 

ARC Advisory Group clients can view the complete report at the ARC Client Portal,

Please Contact Us if you would like to speak with the author.

Obtain more ARC In-depth Research at Market Analysis

Engage with ARC Advisory Group

Representative End User Clients
Representative Automation Clients
Representative Software Clients