출처 : /www.sae.org/news 2018.3.7 / BRUCE MOREY​

<내용>

생성 디자인 소프트웨어가 AI를 활용하여 어떻게 차,장비설계,디자인들을 바꾸는지

- 컴퓨터 알고리즘을 활용하여 자주 충돌하는 요구사항을 평가하고 반복하도록 하여 최적의 설계를 선택

- 인간이 가능한것보다 더 많은 설계대안을 평가할수 있음 

- 클라우딩 컴퓨팅과 생성 디자인의 도래는 막을수 없고, 엔지니어링 공동체의 많은 관심이 필요한 시점.

- 인공지능은 엔지니어가 컴퓨터와 상호 작용하고 소프트웨어 디자인을 방법을 변화시켜 디자인의 패러다임 변화를 촉진할 수 있음 

What if a computer program could create a design for you?

As computers get more powerful and algorithms more clever, design software is now generating potential solutions rather than simply evaluating designs proposed by humans. In fact, today’s generative design tools can offer hundreds, even thousands, of alternatives for a human engineer to evaluate.

Aids to design efficiency are becoming increasingly critical. The set of constraints and requirements for off-highway equipment, heavy-duty trucks, and light vehicles is growing enormously. Requirements are often conflicting, forcing engineers to think of trade-offs between say, fuel economy and safety. Cabin comfort and mass. Design complexity versus manufacturing cost. It is becoming more than a human can handle.

“Generative design for us is a way to use the power of computing algorithms to evaluate and iterate on these often-conflicting requirements. Requirements that often span not only engineering, but materials and manufacturing requirements as well,” explained Tod Parrella, NX Design Product Manager for Siemens PLM.

Level of focus – systems vs. parts

Tools that Siemens offers iterate on designs to explore all possible solutions, according to Parrella, based on requirements and constraints. The human engineer then chooses the most optimal design. “The engineer becomes more the requirements driver, and the algorithms and software help derive the optimum product form based on those inputs and requirements,” he said.

In the briefest of summaries, generative design tools rely on CAE simulation to evaluate successive designs against performance requirements. There are many variations on that theme, two of which Siemens offers—Topology Optimization and Multivariate Design Optimization.

Topology Optimization tries to find the best shape for a part, typically to minimize stress and strain hot spots in response to loads, while also minimizing weight and mass. Brackets, door hinges, or crane booms are examples of parts designed through topology optimization, often resulting in organic looking designs. While noting that this technique is especially useful in designing parts for additive manufacturing, it is just as useful for traditional manufacturing techniques. According to Parrella, Siemens can accept manufacturing constraints as well to ensure a design is say, castable or machineable.

“We offer two versions of this, one for the designer or engineer who needs tools that are well integrated into the CAD environment and another for more advanced uses. The advanced is integrated into the simulation environment offering much more flexibility and analysis types,” he said.

Multivariate Design Optimization explores a design space that includes many different variables. It is much broader than topology optimization and can holistically derive system solutions even up to and including vehicles. Siemens offers its HEEDS platform multivariate design simulation, which provides solutions using any number of simulation tools, including third-party simulations or a customer’s own.

A typical use in the commercial-vehicle domain is using these tools to optimize the design of large chassis components. “They are not necessarily optimizing for the lightest weight design, but want to minimize mass and material, while maximizing strength,” he said. “Generative design is making possible the ability to realize new designs that would never have been imagined without it.”

According to Parrella, it is making possible the ability to evaluate many more design alternatives than humanly possible, offering more complete designs that account for multi-disciplinary requirements and constraints.

Artificial intelligence and cloud data storage

Autodesk is another company advancing new tools for generative design. It offers two tools of especial interest to off-highway and vehicle designers: Fusion 360 for general topology optimization, and Netfabb for lattice and surface optimization, especially tuned for the needs of additive manufacturing designers.

“While we offer those tools, it is important to note that generative design is not just topology optimization,” stressed Doug Kenik, product manager for Autodesk. “It is software that helps designers solve problems in balancing requirements for both business and performance, as well as materials and process.”

That is where Autodesk Generative Design enters the picture. It runs artificial intelligence-based algorithms to produce a wide range of design alternatives, according to Kenik.

“It is currently in a beta release and we are getting feedback on it right now,” he said. Without revealing too many specifics, he did share that AI techniques like Deep Learning, and possibly Convolutional Neural Nets, can be leveraged in such a system.

But, according to Kenik, the algorithms are not as important as the data. Importantly, he ties AI design with cloud data storage. “Data is the key, that is when AI becomes feasible,” he stated. “Generative design techniques can now produce tremendous amounts of data, which you can store in the Cloud.”

Autodesk Generative Design alone could generate thousands of design options, based on a single set of requirements and constraints. That is why Autodesk is keen on cloud computing and data storage, since it is fast becoming the lowest-cost option for many companies.

The advent of cloud computing and generative design is an unstoppable force, according to Kenik. It needs attention from the engineering community. AI is currently making itself known as applied to social issues, like identifying human faces and interpreting natural language.

“There is so much opportunity there to make the engineering process and design process go so much smoother,” he said.

AI could jumpstart a paradigm shift in design, changing how engineers interact with computers and design software. “Instead of simply acting as a repository for easy access to designs, let the system work with us in generating these designs,” he said.

He offers cautions for the future: “If you’re not looking at it right now from an engineering and design standpoint, there could come a time—perhaps soon—when your competition, using these tools, will have a leg up on you.”