A look at Generative Design

A series of progressively optimised designs for a critical aircraft bracket, each using less material and lighter than the previous [Photo: courtesy Frustum]
A series of progressively optimised designs for a critical aircraft bracket, each using less material and lighter than the previous [Photo: courtesy Frustum]
Generative design promises to a very useful tool in a designer’s toolbox. Harnessing the massive computing power of the cloud, designers should be able to make a computer churn out hundreds of designs, speeding up the design process and saving time and money.

­­­The designer has to tell the computer what his requirements are and the computer would generate thousands of designs that meet the requirements. For example, a designer can tell a computer that he needs a chair that can take the weight of a person and must have 4 legs. The computer will generate thousands of designs, far more than what he could have thought of.

Small selection of the thousands of chair designs generated by Autodesk’s internal testing software [Photo: courtesy Autodesk]

Each design solution would meet the requirements but each with tradeoffs. The designer has to select the ones he feels meets the design requirements and can carry on refining them. Besides generating multiple designs, the computer can be used to optimise a design. For example, minimising weight of a component’s form while maximising its structural strength. This often leads to organic looking, highly complex structures.

Side by Side comparison of the manually designed steel node and the optimised design [Photo: courtesy Arup]

The Civil engineering firm Arup recently did some experiments to develop light weight steel nodes for a complex lighting installation. A comparison of the traditionally designed node and the design optimised node is shown here. Due to the irregular shape of the structure proposed, all the nodes had to be slightly different. With traditional designing largely centred on extruding and rotating, the designs take considerable amount of time to design by hand and are not well optimised.

Leaving the computer to do the hard work, Arup suggests that there could be weight savings of up to 75% for individual nodes and 40% reduction in weights of the whole lighting structure.

The Inner Workings

Generative design algorithms generally tend to mimic how organisms evolve in the real world. A starting design and requirements has to be fed into the computer at the beginning. The algorithm makes small changes to the design, testing if it meets the requirements with each change. Designs that meet the requirements are kept while those that don’t are culled. Eventually, a small number of highly optimised designs are left, far fewer than the nearly infinite possible design solutions.




The final rendering of the aircraft partition [Photo: courtesy Autodesk]

There is a great deal of potential in areas where high performance design is key.  Minimising weight is critical in automotive and aircraft design and generative design algorithms are being used to remove every last gram of material from the design while still maintaining its strength and capabilities.

Numerous possible designs for an aircraft cabin partition. They are arranged by the stress it sustains and its weight. A project by Autodesk for Airbus A320 aircraft. 45% weight reduction over current designs have been reported. [Photo: courtesy Autodesk]

Highly Customised Solutions

3D printed sole of shoe, designed using generative design. [Photo: courtesy New Balance]
Generative design opens the door to highly customised designs. Foot wear companies Adidas and New Balance have recently explored the possibility of custom designing the soles of their shoes to suit individual runner’s pace. Using an under-foot sensor, designers at the Massachusetts design studio Nervous System measured the forces across the feet to customise the best fitting sole for the user.

The stylish braces by New York based designer Francis Bitonti [Photo: courtesy Studio Bitonti]
3D Printing pioneer Francis Bitonti has designed braces for scoliosis sufferers, offering a customised brace for users. They offer a more “fashionable” option over existing solutions with curvy, futuristic looking appearances. This is particularly important for the self-conscious 12-15 year old typical users who would have to wear it for at least 18 hours a day to get most benefit.

A good number of possible layouts for floor space at autodesk’s new offices [Photo: courtesy Autodesk]
Generative algorithms have also been adapted to be able to calculate and map out floor plans with optimum natural light, efficient circulation and layouts based on the different needs of the building’s users. Autodesk used these algorithms to design the layout of their new Toronto offices, generating 10,000 design options and finally narrowing down to  8 designs to present to their management.

However, current generated designs have some draw backs. These complicated structures generated are very difficult to manufacture using traditional mass manufacturing techniques. Currently the best way to generate these physical forms is Additive Manufacturing. Restricting manufacturing to Additive Manufacturing brings its own set of limitations such as small batch sizes and high costs.

Will generative design put designers out of work? Far from that. Instead, it dramatically speeds the design process up, saving time and money.

Human creativity is not replaceable. The computer is merely a slave, automating mundane tasks so that we, human beings can focus on more difficult tasks such as finding out what people need, how they feel about the design. No longer does the designer have to spend endless time designing variations on the design that meets the mechanical criteria (is it strong enough? light enough?  Etc.) or other numerical parameters.

Instead, the designer becomes the “problem framer and curator of the design solution set”. He can focus on human factors and other areas of the design process. Understanding the semiotics, aesthetics and semantics of the design can only be done by the designer, not the computer.

Generative design is still in its infancy.  Current technology is not able to handle assemblies of parts, only single components. The software is largely in its development stage, exclusively used by companies in their own internal testing. It is yet to become accessible to the wider design world.

The next stage of generative design technology should be able to handle different types of stock material, using existing standard raw material sizes to generate designs. Only then can it harness the cost savings of mass production by using mass produced stock materials in its designs.

Leave a Reply

Your email address will not be published. Required fields are marked *