Reframing Generative Design

Reframing Generative Design

Design isn't about tools, it's about people.
6 Min Read

What's the point of CAD tools?

The purpose of computer-aided design (CAD) tools is to give time back to product designers so they can focus on the creative tasks that have the most impact. Useful CAD tools dissolve the interface between humans and machines to enable frictionless expression of design intent. The overall trend in the evolution in CAD tools in the past 50 years is pretty clear — the creative emphasis of design work is moving away from repetitive, tedious, and manual design tasks and towards the creative framing of constraints. Thought of another way, designers get to spend more time thinking about what to create and spend less time converting designs into a 3D format.

CAD tools have come a long way in the past 50 years. They still have a long way to go!

Clarifying the status quo of generative design tools

There are so many design terms and tools available to use these days, it's hard to keep track of what tools are useful for what. In general, the term 'generative design' has been used to describe a sub-category of computational design methods, whereby software generates and analyzes thousands of design permutations to evolve a design that satisfies defined constraints. To clarify some 'generative design' buzzwords — and to put our new tool BLANK into historical context — we thought it would be useful to quickly categorize some computational design methods often associated with generative design and their place in the progression of design tools. Specifically, we'll quickly review three terms:

  • Topology optimization: a constraint-driven method for basic geometric optimization
  • Generative design optimization: a constraint-driven method for complex geometric optimization
  • Generative design exploration: a data-driven method for design exploration that represents a new category of generative design tools, provided by our tool BLANK

Topology Optimization (1980's)

Topology optimization (TO) is a constraint-driven process for producing a mass-optimized structure for a given load requirement. TO tools are used to optimize the material distribution within a single 3D mechanical component. These tools are most applicable to high-performance structural design applications in aerospace, automotive, and civil engineering.

Topology optimization was the first step towards the generative design tools of today.

TO Design Workflow (simplified):

  1. Define geometric constraints for your component, the mechanical loads it must endure, and the materials it will be made with.
  2. Run the algorithm.
  3. Voila! A structurally optimized shape that looks like a tree root!

Pros of TO:

  • Structural efficiency — minimize the weight of a part that can meet your mechanical requirements
  • Great for optimizing the use of additive manufacturing — where complex geometries add little additional production cost

Cons:

  • Generated designs are often difficult to manufacture with conventional, economical methods (machining or molding); creates a barrier to practical application.

Example tools: 

Generative Design Optimization (2010's)

Generative Design (GD) is a constraint-driven process for producing multiple, optimized design options that satisfy a defined set of multiple, precise, and functional design constraints for a complex system. GD tools generate optimal designs for a given set of system requirements and differ from TO tools in that designers can specify many different classes of design constraints depending on the industry focus of the application. For this reason, a GD tool can be used for creating an optimized design for a component, a building, or even an entire city! For example, in generative mechanical design tools (e.g. Fusion 360, Paramatters) designers specify design constraints like geometric limits, preferred materials, ideal manufacturing methods, etc.

Generative design optimization tools are excellent for optimimizing complex, multivariate design problems. Like complex urban layouts or intricate mechanical assemblies.

Alternatively, with generative design tools for architecture and infrastructure — e.g. Revit (buildings), Spaceshaper (urban design), Trax (infrastructure) — designers can specify a diverse set of parameters related to the performance of their buildings or infrastructure e.g. energy efficiency, environmental impact, daylight, floor-plan layout, etc. Through digital modeling and algorithmic exploration of the many directions a complex design take, designers can quickly make intelligent, system-level design decisions before they've committed too many resources to a specific design direction. Succinctly put, traditional GD tools enable designers to reduce the technical and project risks associated with complex system design challenges.

GD Design Workflow (simplified):

  1. Define all system constraints.
  2. Run the algorithm.
  3. Manipulate defined constraints and parameters to explore all generated options.
  4. Select and refine one of the optimized designs.

Pros of GD optimization:

  • Great for balancing tradeoffs between many competing constraints related to the design goals of your system e.g. mechanical, environmental, energy, capital performance.
  • Great for simplifying complex structural assemblies through part consolidation and optimizing structural efficiency and maximizing the utility of additive manufacturing processes (just like TO)
  • Accelerates convergence on an optimized solution that satisfies multi-variate constraints.

Cons of GD optimization:

  • Can be computationally intensive
  • Decision fatigue — GD systems generate huge volumes of design solutions. This is a double-edged sword in the sense that it creates a selection burden for the designer. How do you choose the 'best' solution when you have 100 options that are better than anything you could have optimized yourself?
  • Quality vs quantity — algorithms can produce lots of ideas but they can't tell good ideas from bad

Example tools: 

Generative Design Exploration (today)

Generative design exploration is a data-driven process wherein a designer teaches an AI to 'understand' design data and then generate new designs using this learned representation to generate new design possibilities. By training a bespoke instance of BLANK, a designer teaches our AI to uncover hidden or emergent relationships that exist in information embedded in their training data (e.g. 3D designs and associated metadata).

Through manipulating intuitive semantic controls the designer can explore and blend an algorithmic representation of thousands of designs in an instant!

This generative design exploration process is fundamentally different from other generative design processes in that there is no optimization objective — it's purely about exploration. The designer can near-simultaneously take many different types of data about a certain type of design into account to easily create something new. With existing design tools, an expert user spends a lot of time searching through previous designs and related design information to find new inspiration and ideas — all while working with fragmented tools.

With BLANK, we're leveraging leading-edge AI algorithms to create an entirely new design interaction that inspires divergent concept ideation — all in one tool.

GD Exploration Workflow (simplified):

  1. Collect data.
  2. Train your BLANK AI model.
  3. Generate multiple ideas, blend them together.
  4. Cross-breed between different product categories, e.g. cars & airplanes, sneakers & cars
  5. Add texture or color style variations for more lively/ realistic model creations
  6. Export to other tools for expanding the design.

Pros of GD Exploration:

  • By creating statistical representations of 3D designs, BLANK's AI can learn and parameterize previously intangible and semantically defined design parameters — 'strong', ‘aggressive".
  • Computationally light
  • Rapid concept generation and editing

Cons of GD Exploration:

  • Requires access to design data

Example tools: 

  • BLANK

The ultimate goal of generative design is process automation.

The promise of generative design — and all computational design techniques — is to reduce the cognitive demand on designers. It's much more valuable for the outcome of your design (be it a car, a house, a city) to spend more time understanding the user and their design-relevant challenges than for you to spend your time editing thousands of sketches and splines on a computer screen.

Product design is not about understanding tools, it's about understanding people, problems, and emotions.

A designer's work is a balance between research, creative ideation, and practical communication of ideas. Our goal with BLANK is to make the 3D design process more accessible, fun, and creative. We aim to help dissolve the human-machine interface and to reduce the creative strain on designers. Our ultimate goal is to transfer the creative burden on designers away from the mastery of complex software interfaces, or high-skill manual techniques, and towards the framing of product design constraints and expanded customer empathy. We want to help designers focus more of their creative energy on human understanding and less of their energy on tools. We want designers to be empowered to focus on the 'what?' and 'why?' of design — to focus on the human aspects of their designs.

Generative design tools exist to liberate time for research, critical thinking, and exploration. They're not intended to replace designers — they're intended to remove repetitive and wasteful sub-processes from design work.

That said, to achieve quality product designs you always require a designer with common sense, aesthetics, and technical skills to define goals and made decisions. Thankfully, machines are far from capable of this. A designer’s work is never done! Interested to learn more? Reach out for a product demo!

Book a quick demo today.

Learn how AI can accelerate your vehicle development.