AUTODESK GENERATIVE DESIGN: THE FUTURE OF MANUFACTURING

Think about being tasked with designing a chair. How would you go about it? Would you start off with a few rough sketches of general designs before choosing one to focus on? How many designs can you think of and consider? Maybe two or five or even 10? Once there’s a batch to work with, it would be time to do the calculations and considerations, to eliminate design candidates that might not support the required weight, or perhaps not provide the necessary surface area.

But what if instead of taking these steps to come up with a few designs, you could have thousands of designs to choose from—all of which fit your stated parameters? This would expose possibilities that you might never have imagined.

This is where generative design comes in. Think of it as "another brain in the design studio"— a computer program that uses algorithms to explore and populate the design space. As shown in the image below, you might end up with something fairly standard, or your chair may end up looking like something straight out of the Museum of Modern Art, while meeting all of your design requirements.

What if a CAD system could generate thousands of design options that all meet your specified goals? It’s no longer what if: it’s Autodesk Project Dreamcatcher, the next generation of CAD. Dreamcatcher is a generative design system that enables designers to craft a definition of their design problem through goals and constraints. This information is used to synthesize alternative design solutions that meet the objectives. Designers are able to explore trade-offs between many alternative approaches and select design solutions for manufacture.

Recent advancements in Artificial Intelligence (AI) and the simulation of complex phenomena have enabled software to play an active, participatory role in the invention of form. Autodesk Project Dreamcatcher is an experimental design platform with focused research probes into generative design systems.

Figure 1: The Autodesk Project Dreamcatcher Workflow.

Figure 2: Autodesk Dreamcatcher example designs for a chair.

The Dreamcatcher system allows designers to input specific design objectives, including functional requirements, material type, manufacturing method, performance criteria, and cost restrictions. Loaded with design requirements, the system then searches a procedural synthesized design space to evaluate a vast number of generated designs for satisfying the design requirements. The resulting design alternatives are then presented back to the user, along with the performance data of each solution, in the context of the entire design solution space. Designers are able to evaluate the generated solutions in real time, returning at any point to the problem definition to adjust goals and constraints to generate new results that fit the refined definition of success. Once the design space has been explored to satisfaction, the designer is able to output the design to fabrication tools or export the resulting geometry for use in other software tools.

Define: Tools for Defining Problems

Dreamcatcher's problem definition is a format for designers to describe design problems. Through pattern-based description, solutions become modular and accretive, thereby expanding the quality and number of alternatives that are searched in a Dreamcatcher design session. The Dreamcatcher design knowledge base, created through machine learning techniques, is a classified index of pre-existing objects that perform functions, or satisfy constraints, similar to those the user has defined in their problem definition.

Diversifying Input Modalities

Mimicking the variety of reference material in a typical design brief, in Dreamcatcher the designer explicitly and implicitly documents goals and constraints through a number of input modalities including natural language, image inference and CAD geometry. An individual or team may manipulate the problem definition through these multiple modes of input and verify or modify the inferred changes to the problem definition document. Focused efforts on modeling problem definitions and performing design synthesis on full system models rather than individual parts is an active area of investigation for the team.

Figure 3: Topology Optimization (TO) go through the Bike Frame and provide a surprisingly new design freedoms.

Generate: Shape Synthesis

The Dreamcatcher team is developing several, purpose-built design synthesis methods that algorithmically generate designs of different types from a broad set of input criteria. Synthesis objectives include structural, thermal and fluid physical requirements. Dreamcatcher's design synthesis methods compete against each-other to solve problems most effectively through Dreamcatcher's high-performance computing servers. A focused research effort into incorporating manufacturing constraints for various methods of fabrication are incorporated into the design synthesis process itself, so that only manufacturable designs are returned to the design team. The Dreamcatcher system enables designers to truly leverage an emerging class of manufacturing tools that release designers from hundreds of years of predicating design decisions on tool based constraints.

Advances in Cloud-Based Computing and Optimization

Through a purpose-built, scalable and parallelized cloud computing framework code-named Saturn, Dreamcatcher is able to generate and evaluate solution sets with complexity well beyond that of Generative Design Systems of the past. Saturn provides the high-performance computing infrastructure necessary to run the computationally intense optimization and analysis engines, including multi-physics simulations.

Figure 4: SolidThinking Inspire design generation flow.

Explore: Design Space Visualization

After a number of solutions have been computationally generated from a problem definition, the Dreamcatcher design explorer presents to the user a set of possible solutions and their associated solution strategies. This user interface provides a sense of the shape of the valid design space and variable interactions. It also assists users in building a mental model of which alternatives are high performing relative to all others in the set. Once the solution has been adequately explored, the designer can modify the problem definition to iteratively generate more relevant solutions.

Traditional optimization workflows like that of the NASA ST-5 antenna are 'bottom-up' where a design space must be defined by the user and then searched by a genetic algorithm or similar optimization function. By contrast, Dreamcatcher uses a 'top-down' approach where higher level goals are specified. This is the major differentiation between design optimization tools and Dreamcatcher's exploratory design synthesis process.

Figure 5: Antenna bracket, redesigned using SolidThinking Inspire.

Arguments for the incorporation of AI into design often default to concerns around replacing the human designer. While many elements that are commonly modeled from scratch such as brackets, adapters and stiffeners may be created more effectively by a system such as Dreamcatcher, complex elements and aspects that are difficult to quantify will require new types of interaction to leverage human intuition and computational rigor in partnership. Dreamcatcher is pioneering new methods for interactive synthesis and optimization with industry leaders from the automotive, aerospace and manufacturing fields.

So what exactly Generative Design is?

Generative design leverages machine learning to mimic nature’s evolutionary approach to design. Designers or engineers input design parameters (such as materials, size, weight, strength, manufacturing methods, and cost constraints) into generative design software and the software explores all the possible combinations of a solution, quickly generating hundreds or even thousands of design options. From there, the designers or engineers can filter and select the outcomes to best meet their needs.

Imagine if instead of starting a "Drawing" or "CAD design" based on what you already know or ideas that are in your head, you could tell a computer what you want to accomplish or what problem you are trying to solve. For example, say you want to design a chair. Instead of drawing two or three options (maybe 10 if you’re really creative), you can tell the computer you want a chair that supports X amount of weight, costs X much, and uses X material. The computer can then deliver hundreds, if not thousands, of practically and easily manufacturable design options that all meet that criteria and are likely options that you could not conceive on your own. That’s the power of generative design.

Figure 6: Generative design replicates natural world's evolutionary approach with cloud computing to provide thousands of solutions to one engineering problem.

What is Generative Design not?

Generative design is software that augments an engineer and uses the power of cloud computation and machine learning to explore a whole set of new solutions. It expands the engineer or designers known universe of valid solutions to their design challenge. By contrast, many of the technologies that masquerade as generative design - topology optimization, lattice optimization, parametrics or similar technologies -- are focused on improving a preexisting design, not creating new design possibilities like with generative design. The confusion arises because the inputs to generative design are similar to the inputs to many optimization tools. However, generative design produces many valid (high performance yet cost-effective) designs or solutions instead of one optimized version of a known solution.

In addition to creating entirely new solutions, another area where generative design differs and stands out is that it takes manufacturability into account. That means the process of testing products and going back to the drawing board is drastically reduced. Traditional optimization focuses on refining a known solution, which usually involves removing excess material without any notion of how something is made or used.  Additional modeling, traditional simulation and testing are then required steps at the end. With generative design, the simulation is built into the design process. You can specify manufacturing methods like additive, CNC, casting, etc. at the outset and the software only produces designs that can be fabricated with your specified manufacturing method. Or you can explore designs for multiple manufacturing methods.

Figure 7: On the top, the preconceived bracket design from the 2013 GE jet engine bracket challenge is refined using traditional topology optimization. Because there is no awareness of manufacturing processes, it will need to be remodeled again ‘by hand’ in CAD software. 

On the bottom, Autodesk Generative Design software uses attachment points, strength requirements, weight, materials, and manufacturing method as constraints to produce multiple geometric solutions for the bracket. There is no preconceived geometry as a starting point. In this case, generative design produces 30 design options while topology optimization offers one.

Another often overlooked benefit of generative design is the ability to consolidate parts. Because generative design can handle a level of complexity that is impossible for human engineers to conceive – and because additive manufacturing can enable the fabrication of the complex geometries that generative algorithms often produce – single parts can be created that replace assemblies of 2, 3, 5, 10, 20 or even more separate parts. Consolidating parts simplifies supply chains, maintenance and can reduce overall manufacturing costs.

With its ability to explore thousands of valid design solutions, built-in simulation, awareness of manufacturability and part consolidation, the reality is that generative design impacts far more than just the traditional notion of design. It’s really about the entirety of the manufacturing process. In some ways, you could argue ‘generative manufacturing’ would be a more apt term. Because of its expansive impact on the manufacturing process, generative design delivers a leap forward in real-world benefits. It can lead to major reductions in cost, development time, material consumption and product weight.

Topology Optimization ≠ Generative Design

Topology optimization is often mentioned when discussing generative design, but the terms are not interchangeable. Topology optimization is an aspect of generative design, and involves designing based on various stress points to reduce the amount of material used and thus increase the speed of production while reducing material costs and weight. Topology optimization uses an existing design, whereas generative design can involve inputting more general values and building the design from the ground up. ProtoCAM Engineering Manager, Ed Graham, has his own way of remembering the difference between generative design and topology optimization:

“I always think of the human body when it comes to the generative design approach; through generations, the human body has evolved into what we are today, although every body type has its own constraints to fit the exact body type, i.e. short people, tall people etc. Topology optimization is based more around science, and involves engineering based on fixed constraints of a specific design,” Graham says.

Difference between Topology Optimization and Generative Design Video Link: https://www.youtube.com/watch?v=uqWEwtJuRa4

Topology Optimization asks, what aspects can I remove to optimize this design? Whereas as Generative Design asks, what can I create based on these specific constraints and requirements?

How generative design exactly works.

The generative design process starts with a designer defining a bounding/design area, connection points and obstacles, as well as a variety of constraining parameters. An example of the first three can be seen in Figures 8, 9 and 10, where these parameters are defined for a motorcycle swingarm design. Optimization is based on desired materials, manufacturing technologies (method of production), temperature tolerance, cost, and strength of the part and its ability to withstand specified forces.

Figure 8: Bounding space for the swingarm.

Figure 9: Connection points for the swingarm (where the part connects to the wheel and the m​otorcycle).

Figure 10: Obstacles for the swingarm (wheels and chain) where a ch​ain is placed on both sides to create a symmetrical result.

Figure 11: Resulting design for motorcycle swingarm.

Normally, a designer would have to spend several hours trying to calculate whether their proposed design would fit these constraints. In contrast, this is done by default using generative design and all of the resulting designs will fit the defined constraints, and often exceed them. Because of this effective automation of tasks through software, there is room for the designer to play around with aesthetics of the part and invest more time into exploring creative and effective designs.

Optimization through constraints.

So, what about all of these constraints? You might not care about parameters like temperature, flex or weight to a great extent for a chair. But you would for brackets, winglets or gas turbines in planes, or bridges that you walk and drive over. When airplanes fly, for example, every ounce of weight translates into huge incurred fuel costs. Parts must be able to withstand tremendous forces and large variations in temperature, with proper heat dissipation.

Manufacturers are always trying to optimize their parts to account for weight and harsh conditions. General Electric, in partnership with GrabCAD, went so far as to launch a 3D-Printing Design Quest. They challenged the public to redesign a metal jet engine bracket, making it 30 percent lighter while preserving its integrity and mechanical properties like stiffness. The winning design sheared off 84 percent of the weight of the original bracket, taking the total weight from 4.48 pounds to just 0.72 pounds (Figures 12, 13).

Figure 12: GE’s original bracket.

Figure 13: GE’s redesigned jet engine bracket.

Using generative design, companies could duplicate this crowd-sourcing effort internally, using the software program algorithms to arrive at numerous solutions that fit desired parameters.

Generative design and Additive Manufacturing.

Using Additive Manufacturing or 3D-Printing allows for even more design possibilities. This can be done by varying the parameter of “manufacturing technology” in Dreamcatcher. Additive manufacturing technology has been evolving at a rapid pace. It is now to the point of producing parts of industrial quality due to the likes of FIT AG9 for metals and Carbon3D10 for polymeric plastics.

Autodesk also offers a generative software product called Within-11, which makes use of additive manufacturing capabilities to create lightweight, latticed designs for automotive, medical implant, aerospace and industrial equipment applications. These designs would not be producible by traditional means. Using Within, engineers were able to create a lightweight, load-bearing engine block that had better heat dissipation and superior performance (Figure 14); a lightweight roll hoop for Formula One racing cars (Figure 15); a customized implant for cranioplasty (Figure 16), and micron-accurate rough lattice surfaces for medical implants to aid fixation with bone.

Figure 14: Automotive: Load-bearing engine block.

Figure 15: Automotive: Lightweight roll hoop.

Figure 16: Medical Customized implant for cranioplasty.

Figure 17: Autodesk Medical Implant containing lattice structures.

Additive manufacturing combined with the Within software has been even more helpful when it comes to traditional injection molding. Figure 18 shows the traditional solid mold on the left side, and a re-designed latticed mold on the right, with better simulated heat dissipation properties. In addition, this mold is more lightweight and requires much less material to produce.

Figure 18: Injection mold optimization using Autodesk Within.

With the technology constantly evolving and reshaping itself (pun fully intended), the number of design possibilities will explode and manufacturing efficiency will further improve. Think lighter, faster planes with optimal engine heat dissipation. Self-building and self-repairing bridges that look organic and last longer. Faster bone growth thanks to biodegradable, bone growth-stimulating optimized and 3D printed implants. And this is just the beginning.

I have taken two important generative design case studies which are to be discussed in the upcoming headings. These studies will help us in understanding the implications of generative deign into normal production schedule so that design optimization for the near future will be at different level.

Case Study 1.

Airbus Cabin Partition.

Airplane manufacturer 'Airbus' used generative design to reimagine an interior partition for its A320 aircrafts and came up with an intricate design that ultimately shaved off 45 percent (30kg) of the weight off the part using 3D printing in a high-tech alloy called “Scalmalloy”. That weight decrease will result in a massive reduction of jet fuel consumed and a reduction of hundreds of thousands of tons of carbon dioxide emitted when applied across its fleet of planes -- equal to taking 96,000 passenger cars off the road for a year:

(1) The design that was arrived at by using an algorithm based around mammal bone growth.

(2) Stress testing compared to existing partitions.

(3) The 3D-printed parts being cleaned up.

(4) The resulting partition, which is 45 per cent lighter than the existing panel.

Figure 19: Airbus Cabin Partition.

Figure 20: Cabin Partition optimized using generative design.

Figure 21: Airbus Cabin Partition Generative Designed and Successfully.

Figure 22: Comparison of Building Limits in EOS M400 and Concept Laser M2 machines.

Project Video Link: https://vimeo.com/151334118

Case Study 2.

X VEIN-The Life Saving Drone.

Yuki Ogasawara and Ryo Kumeda, aka team ROK, were just 15 years old when the Great East Japan earthquake and tsunami caused immeasurable destruction and damage on March 11, 2011. Like many of the country’s citizens, the two friends considered how they could help others in the disaster’s wake, starting them down a creative path toward drone development.

Drones were just gaining attention, and Ogasawara and Kumeda sought to apply this emerging technology toward disaster efforts. They had already won the Fighting Spirit Prize at the National High School Programming Contest (with a robot arm controlled by a smartphone) and taken seventh place in the RoboCupJunior Japan Open rescue division, and they were ready for the challenge. So in 2012, team ROK began developing a multicopter drone. And in 2015, they won the multicopter division of the National Student Indoor Flying Robot Contest.

Building on that success, team ROK went to Maker Faire Tokyo 2016, where they exhibited X VEIN, a drone built for disaster conditions and search-and-rescue missions. Featuring extended flight times, a reinforced frame, and propeller guards to prevent damage from crashes, X VEIN impressed visitors with its X-shaped body and lattice-structure design reminiscent of veined dragonfly wings.

But X VEIN’s appearance at Maker Faire Tokyo might not have happened if not for an event held at DMM.make AKIBA, a Tokyo hub of the maker community. There, Ogasawara and Kumeda first encountered generative design software and knew it would be essential for developing a drone with sufficient body strength, lightness, image-capturing capability, and safety features.

Figure 23: X VEIN had full 3D-Printed body with Generative Design wings.

Generative Design Breakthrough:

“There are many reasons existing drones are not used in disaster-hit areas, including their lack of safety features, their size and weight, and the low potential for customization,” says Ogasawara.

Clearly, a tool that can solve for those requirements is key to any drone-development process. Generative design provides computer-generated schematics and structural analysis optimized to fulfill predetermined conditions, which can then be output to a 3D printer. This allows for lighter designs than previously possible with conventional manufacturing methods—essential for producing a drone that would meet the necessary weight requirements.

“For a drone to hover in midair, the lift it generates must exactly match its own weight,” says Ogasawara, who is responsible for X VEIN’s mechanical design elements. “Variations of even 5 percent of overall weight change how operators must control the drone. It is crucial we make our drone as light as possible.”

Figure 24: Fusion 360 illustrating Generative Design wings for X VEIN.

X VEIN Gets Its Wings:

Currently a student at Saitama University’s Department of Electrical and Electronic Systems, Ogasawara joined hardware startup Exiii as an intern in early 2016. At Exiii, he worked with chief creative officer and designer Tetsuya Konishi. By April, they had completed design sketches for the X VEIN concept. The generative design-based lattice was integrated into the drone’s body, leading to the unit’s current design.

Figure 25: X VEIN in operation mode.

On the modeling of the drone’s body, Ogasawara says: “The design had a lot of free-form curves, which are hard to translate from sketches to a 3D model. Through Wacom, we were given access to a Cintiq pen display tablet. Using it to build up a model in Autodesk Fusion 360 with input based on our sketches made it possible to follow our design while smoothly re-creating it in 3D space.”

“With a 2.4 GHz frequency used for control, we have an operational range of about 500 meters without any obstructions,” Kumeda continues. “Current laws in Japan require operators to stay within visual contact of their drones, so our effective range comes to about 100 meters.” The drone’s camera is mounted on a gimbal, isolating it from vibration and tilting, stabilizing the image. The images can be viewed in real time using a smartphone, accessing areas made impassible by disaster damage, confirming conditions, or performing other tasks. The drone could also carry thermographic and infrared-imaging equipment to locate survivors. Because most of X VEIN’s components are 3D printed, replacement parts also can be procured on-site—a huge advantage in disaster situations.

Figure 26: Team ROK's Yuki Ogasawara (left) and Ryo Kumeda (right) won the the National Student Indoor Flying Robot Contest with the blue drone on the right.

How 3D-Printing works?

Capabilities of A 3D-Printer.

Housing Crisis and 3D-Printing.