Article originally published in Benchmark Magazine by NAFEMS, the International Association for the Engineering Modeling, Analysis & Simulation Community. Find out more at nafems.org.
In the late 1980s and early 1990s, the engineering research community began to buzz about a new concept called topology optimization. The homogenization method for topology and shape optimization was first introduced by Martin Bendsoe and Noburo Kikuchi in 1988, and it had the entire research industry talking.
The concept of mathematical optimization allows a user to provide a guided objective and constraints, then let the computer run the study in loop to find the ideal answer. Topology optimization algorithms follow this same process while optimizing the shape and topology of a structure. The resultant outputs produce parts that meet size and functional requirements within the allotted design space while using the minimum amount of material.
Jeff Brennan, Chief Product Officer, Altair 365, was in the thick of this budding movement, first learning about how to apply optimization to engineering problems early in his college career.
“Everyone in my engineering class was pulling an all-nighter to solve a mechanical dynamics problem with Fortran,” said Brennan.
“As I’m leaving to go to the computer lab, my roommate was coming into the dorm with a six-pack. I said ‘Come on Tom. You’re not going to spend the rest of the night in the computer lab like the rest of us?’ He said, ‘No man, I’ll leave that to you guys.’
I found out he had written the algorithm and made the position of each of the statements in the Fortran program a variable in an optimization loop. He hit ‘go’ on that optimization program and it would reorient the various operations until it came to the fastest answer. And, he sat that night and drank all six of those beers. Somebody figured it out. Somebody knows how to use numerical algorithms to do less work and get the best answer. And I thought ‘Bing!’ There’s something here.”
Brennan went on to the University of Michigan, studying under Dr. Noboru Kikuchi, one of the fathers of topology optimization, at his mechanical engineering laboratory and applied mechanics group. One of the first applications of topology optimization came in the study of biomimicry for osteoporosis and prosthesis research.
They studied the factors that encouraged bone growth in healthy individuals with a goal of replicating the same excitations in elderly patients, as well as to encourage bone growth around an implant in the body to create a stronger and more natural bond. The underlying theory; the body grows bone in an optimal manner. It’s an extension of Wolf’s law, which posits that the body responds to mechanical stresses by increasing bone material and density where it is needed to support the stress.
After graduating, Brennan interviewed with Altair, then a small engineering consulting company that had started seeing success with its HyperMesh pre-processing tool. In his interview, Brennan showed Altair the topology optimization work he was applying as a student. Seeing an opportunity to commercialize topology optimization, Altair hired Jeff Brennan.
“We were very excited by the technology,” said Jim Scapa, Founder, Chairman and CEO Altair. “We ended up coming to an agreement with Professor Kikuchi and his partner Alejandro Dias to resell their software into the commercial market. We really believed in it and wanted to take it further.”
Landing OptiStruct’s First Customer
Brennan became OptiStruct’s first evangelist in 1992, traveling around the country, and later the world to pitch this new technology.
“The amount of rejection that I got was tough,” said Brennan. “I was glad I was young when I was starting to sell OptiStruct because it didn’t fit into people’s processes, even if they understood the concept behind it. There was no place to put it.”
Despite the early hurdles, OptiStruct started to gain notoriety in the engineering community and stack up its first major wins. By 1994, OptiStruct was recognized by Industry Week magazine as its ‘Technology of the Year.’
Also, in 1994 Altair approached General Motors and pitched OptiStruct. Dr. Keith Meintjes, currently the practice manager, simulation and analysis at CIMdata, Inc., was the simulation manager of GM Powertrain at the time.
“Jeff showed up at GM Powertrain and somehow sold me the first-ever copy of OptiStruct,” said Meintjes.“He didn’t tell me until years and years later that they’d never made a commercial sale of the software before. I can pride myself on being customer 001. The software at that point was very difficult to use, but in the hands of talented engineers, you could essentially make magic.”
Back to the Drawing Board
As Altair built on its success with OptiStruct, Scapa sought to negotiate a deal with Kikuchi and Dias, the original authors of the software, to purchase the technology. However, an eleventh-hour disagreement over ownership of the intellectual property put the deal in jeopardy. Scapa decided that without the IP, he was going to take the lab code and have Altair develop a new commercial software from the ground up.
“Dias took a hard stand, and so did I,” said Scapa. “Dias didn’t think I could do it…I didn’t really know if we could do it either, but I figured, with enough perseverance, we could.”
It was a bold step, which came with additional pressure. Altair now had a portfolio of OptiStruct customers, all waiting for the next version of their software, which it now had to essentially re-build from scratch.
“I hired Harold Thomas and Yaw-Kang (YK) Shyy to lead the development of the next generation of OptiStruct. Somehow, they were able to put together the new software in about six months. They were brilliant, brilliant coders. They became our master coders.”
Soon after the hiring of Thomas and Shyy, Altair sought out another innovator in the research community, Ming Zhou, to help take OptiStruct to the next level.
Scapa said, “Harold came to me and said, ‘Pretty much the best optimization guy on the planet is coming out with his PhD. All the most innovative papers are coming from this guy.’ Ming joining was key because he brought this creativity to everything we were doing.”
“Ming is really a pioneer of the technology,” said Uwe Schramm, chief technology officer at Altair.“He’s been there since the whole thing started in academia. Ming, along with Harold and YK, they were the architects of this approach. They are pioneers of commercial topology optimization.”
A Modern Finite Element Analysis Solver Takes Shape
In 1997 and 1998, the development team was implementing key functionality that would shape and enable OptiStruct’s future growth. They shifted the code from the homogenization method to the density methodology and began focusing on adding finite element analysis (FEA) solver functionality and manufacturing constraints.
“If you want to do optimization, you have to do good analysis,” said Schramm. “Customers wanted to take their more complex models and run them, so we had to add features to keep them happy.”
“OptiStruct really did become both an optimization code but also a super-capable solver code,” said Brennan. “Now the third generation of that solver code is handling nonlinear problems like a champ- material nonlinearity, geometric nonlinearity, all kinds of gap constraints, contacts, you name it. It’s become world class in terms of the greatest implicit linear, nonlinear and optimization code on the market.”
Next-Generation Optimization Takes Flight
Having won some key automotive accounts, Altair began to set its sights on other markets, especially aerospace. One of OptiStruct’s most crucial aerospace wins was the Airbus A380 light-weighting project (Figure 2).
Figure 2: Simulation of the Airbus A380 wing ribs.
“The Airbus structures group was working with our consulting group in the UK and they had a real need,”said Brennan. “The A380 wing structure was way overweight and this baby was not getting off the ground. They had certain manufacturing constraints that they needed. They didn’t want to have 13 different wing ribs that went from inboard to outboard with totally different topologies, totally different truss structures. That would be a nightmare for the wiring harness guys.
The OptiStruct development team developed a methodology for pattern recognition and repetition so we could basically come up with a modified solution that each of the wing rib sets looked similar and had a similar number of holes. That really made the difference to create a manufacturable, workable solution.
The interaction between software development and the applications was one of the core reasons why OptiStruct was successful early on. That flexibility to take customer requests right as they needed them, and sometimes even overnight, code those things, give them back, and solve the problem. That cemented Altair’s reputation as not just an innovative company but a company that delivers.”
Early adopters were winning with OptiStruct. Altair was stacking up OptiStruct wins at large OEMs, but small customers were also starting to see its potential as a competitive differentiator.
“I’ll never forget going out near my hometown, Kalamazoo Michigan, to a company called Nelson Metals,”said Brennan. “They were a small metal casting house. They had a real competitive advantage over the other casting companies because they’d show up at concept meetings and they would say, ‘We can make a part 20 percent cheaper, with better performance, 30 percent lighter.’ and people like be like ‘How did you do this?’
Because of that open mind, these early adopters had a huge advantage. I’d say some of them still have an advantage because they started adopting optimization earlier than everyone else and they’ve probably outpaced their competition since.”
Crossing the Chasm
Recognizing OptiStruct’s commercial potential, Jim Scapa and Altair’s management team realized the next challenge, how to scale the success.
“There’s a book called Crossing the Chasm by Geoffrey Moore , which fundamentally talks about technologies and product life cycle,”said Scapa. “There’s an initial upward curve for early market technologies, where innovators and early adopters start to discover the product, but then a chasm divides these early technologies from reaching a mainstream market. It’s very difficult for true technology companies to make this leap and cross into a more mature and stable market, despite many product’s early successes. Most products end up falling into this chasm.
I saw OptiStruct follow this curve and make the leap across the chasm. It actually took a long time to establish OptiStruct among the product development community at large. We had to develop all these manufacturing constraints so that you could create parts that were truly useful. And, we had to convince engineering organizations that this made sense to use in the design phase.”
In 1999, OptiStruct became part of Altair’s newly minted units-based licensing model, in which customers purchase a pool of recyclable tokens that can be applied to any of Altair’s CAE applications. There was early reluctance, especially from the sales and finance teams, who feared a loss in OptiStruct revenue.
“Our primary product was HyperMesh, but we clearly saw OptiStruct as a large opportunity,”said Scapa. “My problem at that time was ‘How do I take this new product and get a lot more traction around it?’ And that’s where I first came up with this idea of Altair’s units-based licensing model. With the unit model, I could basically stop selling HyperMesh and instead start selling units, allowing customers to immediately have access to OptiStruct. That way I eliminated the friction of having to sell a second product into the account and it started to grow from there. It was huge because OptiStruct may have fallen into the chasm, quite frankly, if I hadn’t done this. In the long run, I think the unit model is a big reason why we’ve been as successful as we are as a company.”
The Additive Manufacturing Revolution
“The technology was always super-cool,”said Scapa. “But our competitors stayed away from it for a long time. The competition woke up when additive manufacturing became interesting. There was all this hype around additive and topology optimization, and the solutions that we were offering with OptiStruct were perfect for additive because you could make parts with internal voids and would make all kinds of shapes that you couldn’t make with castings or stamping or other traditional manufacturing processes.”
With the rise of additive manufacturing, competition began to join the marketplace. With that came challenges to OptiStruct’s topology optimization throne, but also exciting opportunities.
“There’s still so much out there,”said Brennan. “The ability to tie a digital twin directly to 3D printing and be able to adjust shapes quickly, evaluate them quickly. Part replacement could be a major opportunity with topology optimized structures for things that are aging. You might not ever have to have a CAD file for a part that needs to be replaced from 1960, you’d just need to know its position, its volume, its load cases, and you could quickly generate the ideal shape, have it printed and you’ll have a replacement overnight from someplace like Amazon. That’s just Jetsons stuff.”
Legacy and Future of OptiStruct
Worldwide, Altair now has more than 3,000 companies using OptiStruct.
“We’ve been on the leading edge of where simulation is going, promoting this idea of ‘simulation-driven design’ and the possibilities offered by pervasive optimization,”said Scapa. “OptiStruct has had such a widespread impact over the last 25 years. I think it helped launch this whole light-weighting movement that’s going on.”
“OptiStruct has changed companies and certainly has put them into a much more competitive position, which is tough to do in the global marketplace now,” said Brennan.
Although OptiStruct has reached many milestones that would have seemed impossible back in 1993, the Altair team continues to look forward at the potential evolution of the tool.
“Today, a lot of people are excited about lattices, new materials, mixed materials, and mixed topology with shape which we were always trying to do but are now doing in a more seamless integrated process,” said Scapa.“We’re also adding a tremendous amount of nonlinear simulation to OptiStruct. We have displaced a lot of the traditional linear analysis solutions with OptiStruct over the years and we’re now starting to do the same with the nonlinear analysis solutions. A lot of the reason for that is because of how integrated optimization is within all our solutions.”
Machine learning and the Internet of Things (IoT) also present exciting opportunities in combination with advanced structural analysis and optimization.
“We are beginning many, many projects where we’re trying to apply AI and machine learning mixed with our shape and topology optimization algorithms to be able to solve more complex, highly nonlinear problems,” said Scapa.
“The ability to take data from postproduction in the field about how it’s really operating and use that to inform the digital twin has great potential,” said Brennan. “The idea that your topology or your design shape is a living thing and it gets feedback from its environment through this digital twinning is mind blowing to me, but I hope it happens.”
From humble beginnings, OptiStruct pioneered the concept of topology optimization in the commercial market. Across virtually every industry around the globe, it has not only enabled weight savings and performance improvements that were previously thought to be impossible, but fundamentally changed how designers and engineers approach product design. No one knows exactly what the future will hold, but OptiStruct is striving to adapt to advances in manufacturing, computing, and data intelligence in order to continue its legacy of innovation.
 G. A. Moore, Crossing the chasm: marketing and selling disruptive products to mainstream customers. New York, NY: HarperBusiness, an imprint of HarperCollinsPublishers, 2014.
- The Cost of a Millisecond: HPC Job Scheduling for Semiconductor Design - February 14, 2020
- Altair Enlighten Awards Now Open – Accepting Submissions for Global Automotive Lightweighting Advancements - February 3, 2020
- Crash Simulation with Arm and the Catalyst UK Project - January 24, 2020