The manufacturing world is undergoing a profound transformation based on new design technologies that couple 3D representations of highly complex structures with artificial intelligence, model-based reasoning, and data-driven learning. Design representations of the future will be hybrid, fusing complex geometric information with physics and machine-learning models.
At the same time, we are seeing the introduction of intricate new materials into a broad range of manufacturing processes. In many ways, design technologies have not been able to keep up with this rapid pace of change. As a result, manufacturing capabilities are driving the evolution of design technologies, not the other way. For example, hybrid manufacturing technologies that allow seamless interleaving of additive and subtractive processes are already available, but there are very few designs that are truly enabled by these technologies.
Product designers today are forced to make key tradeoffs in the preliminary stages of the design process. Standard computer aided design and computer aided manufacturing (CAD/CAM) software and product lifecycle management (PLM) systems are still useful for describing design geometries and materials. But the use of aging legacy tools and materials causes manufacturers to think very narrowly about their designs, limiting their ability to innovate.
In other words, legacy designers and manufacturers are trapped by their current software tools that cannot scale up to meet the escalating levels of hardware and process complexity.
Palo Alto Research Center (PARC) researchers are working with a group headed by the Defense Advanced Research Projects Agency (DARPA) to expand the limits of existing CAD technologies. The goal is to enable product designers to create novel designs that exploit the geometric and material complexity enabled by additive and hybrid manufacturing. DARPA’s TRAnsformative DESign (TRADES) research program is developing technologies to facilitate the creation of more complex multi-material objects that optimize multiple design objectives. The goal is to harness the tidal wave of new materials and fabrication methods out there to enable designs that are unimaginable today.
The platform PARC, a Xerox company, is developing under the TRADES program aims to streamline the production process from initial mock-ups to final parts production. It is pre-programmed to work with a range of materials and composites, with specific tools integrated for additive and hybrid manufacturing (combined additive and subtractive manufacturing). This breakthrough approach has an ability to cater to objects with billions of geometric attributes such as jet engines or gas turbines. The programs can automatically optimize shape and material layout along with some design parameters for an object and determine the best settings for fabrication.
“We have reached the fundamental limits of what our computer-aided design tools and processes can handle, and we need revolutionary tools that can take requirements from a human designer and propose radically new concepts, shapes and structures that would likely never be conceived by even our best design programs today, much less by a human alone,” explained Jan Vandenbrande, DARPA program manager, at the kickoff of the TRADES project in June 2017.
Toward Hybrid Manufacturing Processes and Hybrid Materials
We envision that the future of manufacturing will be AI-enabled with hybrid design representations, hybrid processes and hybrid materials. Hybrid manufacturing approaches will combine subtractive manufacturing and additive manufacturing techniques by incorporating the widespread use of 3D printing and design tools.
Additive and subtractive manufacturing each offer certain advantages and disadvantages, so we should leverage the best aspects of both approaches to dramatically increase efficiency. For example, a hybrid planner can additively manufacture new product features. Then while printing the design, it may require some support structures to facilitate the process. Later, the AI engine can automatically direct the system how to subtract those support structures without introducing too much complexity. In other words, an AI planner can ask an additive process to deliberately add excess material knowing that a future subtractive process will remove this material. This addition and deletion of excess material may be the key step to making the design manufacturable.
We expect a similar adoption of hybrid materials. For instance, to create a composite layer of materials with current systems, each layer must be designed separately and then stitched together somehow. Such complexity creates a restrictive pain point requiring extensive design planning that drives up costs. By contrast, 3D printing systems today are moving towards being able to fabricate smoothly gradient material properties from hard to soft materials, which standard PLM software simply cannot represent.
Manufacturers can also take helpful cues from the animation industry, which deploys extreme computational power to render highly complex animated scenes for blockbuster movies. In a similar manner, material scientists can now apply animated computer graphics tools to render a high-resolution CAT scan of a patient’s femur bone, for instance. 3D printers can then replicate the resolution of that specific bone structure to manufacture it accurately. Again, the ability to represent such a precise level of detail is just not possible in current PLM software and PARC is diligently working on enabling this.
Palo Alto Research Center
|Femur bone representation, cutaway view, and closeup. No other design software can represent designs with multiple materials at this resolution without requiring many GBs if not TBs per model. Applications for the medical, aerospace and automotive industries are enabled by this representation.|
We can also inspire fresh thinking by adopting AI planning tools and model-based reasoning systems. No legacy computer-aided design system today can automatically determine how to set up a tools platform and connect that geometry to model-based AI and planning. But future manufacturing systems will take in diverse 3D geometries to suggest extensive options for the creation of cost-effective designs with existing tools. In this way, we can fully grasp the material properties of original product designs and thus understand all the physics that will be required for manufacturing them.
To succeed, we will need to enhance a human engineer’s role by having our tools represent, plan, and manage complex, graded geometries and multiple length scales for materials, while asking the engineer to include domain-specific expertise and experience to curate designs efficiently. Doing this effectively will require incorporating material and manufacturing uncertainty into the physics analysis of all functional parts.
Manufacturers today face a clear need to move beyond current siloed design tools. The industry’s increasing levels of complexity will require smarter systems that can guide manufacturing decisions much earlier in the design process. What’s needed is an integrated view of all possible manufacturing options, materials and parts at the start of the design process. Artificial intelligence and material physics are quickly converging to give us that clearer picture by incorporating necessary processes and parts to drive real manufacturing innovations – at the earliest possible stages of a product’s design.