ANSYS 创始人John Swanson 的访谈
n5321 | 2025年12月5日 19:41
Rajesh Bhaskaran: 康奈尔大学 Swanson 工程仿真项目负责人
John Swanson: ANSYS 创始人，有限元分析先驱
Rajesh Bhaskaran:
John Swanson: And I'm John Swanson. I'm one of the pioneers of Finite Element simulation. I started the ANSYS software program many years ago, and Rajesh wants to ask me questions about how that came about.
Rajesh Bhaskaran: So here's the first question: What were the early years of Finite Element Analysis and simulation like?
John Swanson: Well, simulation has gone hand-in-hand with computer technology. So, part of the answer is, of course, what was computer technology like? Back then, computing was done with large mainframes. The source code was on punch card decks. You got one or two turnarounds a day if you were lucky. A deck of cards for software might be 100,000 to 200,000 cards.
The prototype, or the most desirable engineering computing, became the minicomputer—the VAX-11/780—whose speed was a nice steady one megaflop. Today's computing, of course, is up into gigaflops, teraflops, and so on, but one megaflop was a nice machine at that point.
Rajesh Bhaskaran: And what are some of the early stories that stick in your mind?
John Swanson: That's a pretty general question. Um... the shuttle launch story. It was early in the morning of the first shuttle launch, and I got a very early call. The question was heat transfer. The object was the shuttle tiles, and the desire was to do a three-dimensional simulation because they'd only done one-dimensional simulations. They really wanted to have a little more assurance because they were launching within hours.
Rajesh Bhaskaran: How did you get the idea to start the ANSYS program?
John Swanson: Well, the ANSYS program started because I had a problem that I needed a solution for. The problem was a simple stress concentration problem in an axisymmetric structure, and there were no tools for doing that. So I developed a network of springs to simulate the stress concentration, and I got what looked like plausible results.
So I showed them to our government sponsors on the particular project and I said, "Hey, that looks like what Ed Wilson is doing out at the University of California. Why don't you go talk to him?" So I went to talk to Ed Wilson, and he had written an axisymmetric finite element program.
I worked all evening into the night writing up the coding for the punch cards for that particular problem. At 3:00 in the morning, I found someone who could punch the cards, feed them into the computer, and by 5:00 I had a really good-looking solution—it was all numbers on paper. By 7:00, I had a big box of cards under my arm and I was heading back toward the airport. Of course, I rewrote the whole thing as what I wanted, but that was my first interest in finite element analysis: to solve that simple stress concentration problem.
Rajesh Bhaskaran: And which year was that?
John Swanson: That would have been... 1964 probably. Let's pin it down to '64, maybe '65.
Rajesh Bhaskaran: And then in the 1970s?
John Swanson: So yeah, as I went from that period on, I began adding more and more capability. That was two-dimensional axisymmetric. I added shells, I added solids, I added dynamics. I had separate programs for shells, for solids, and so on. And I began to realize I was doing the same thing over and over again, and all I was doing was changing the element type.
So [I thought], "Well, I can put together one software package where you can just specify which element type and save myself a lot of work." And that was the basis for the program that eventually evolved into ANSYS. That program was done on a government contract and became public domain. ANSYS, of course, picked up from that and went on.
Rajesh Bhaskaran: How did you distribute the first version of ANSYS?
John Swanson: Well, the first version was done in what was called "time sharing." We mounted the program on a computer, and then you would sign up to buy time on the computer. You would pay for the cost of the computing time with a surcharge for running ANSYS.
Now, we looked at several different pricing mechanisms. We did a pricing per degree of freedom or a pricing per run, but eventually, computing time became the common theme. As we went into the minicomputers and smaller computers, then we just started charging a fixed fee. Eventually, that evolved into a "seat fee"—in other words, so much for each user.
You know, pricing has always been an issue. Our pricing originally was based on the speed of the machine: the faster the machine, the more expensive the pricing. Now, of course, the machines got faster and faster, so the price kept going up and up. So every year or so, I cut the price in half, and everybody loves the price cut, so that worked out well for all concerned.
Rajesh Bhaskaran: So what was meshing like in the first version of ANSYS?
John Swanson: First version... well, let's go back to Ed Wilson. When I wrote up my simple problem, the mesh generation was: you could specify Node 1 and Node 5 and fill in between. And then you could generate Node 6 and Node 10 and fill in between. And then you can generate Element Number 1 from 6 to 5 to 10 to 2 and make five of those, incrementing by one. That was the mesh generation.
After that, I generated a series of mesh generators for shells and solids. Mesh generation technology now is much, much advanced. You can do huge three-dimensional solid models in minutes using bricks or tetrahedrons. Yeah, meshing is not an issue anymore, whereas it used to be the issue. You would spend days, weeks, or even months doing a mesh for, say, an automobile engine. Now you take the CAD part, you just say "mesh all" with this size, and you get these huge models which, with today's computing technology, run in minutes or hours. Vast changes both in individual productivity and in computing productivity.
Rajesh Bhaskaran: And how did the graphical user interface come about and how did it evolve?
John Swanson: Hey, my first paper... I had done a plane section of a hexagon with some holes in it. I printed out the stresses at every point in a rectangular lattice. And then I drew lines for where the holes were. Then I took my colored pencils and I said, "Well, the 1000 contour will go here and over here, and here's a 450 contour here," and so on. That figure in my published paper is this grid of numbers with these lines drawn.
Then we got graphic terminals. The first graphic terminals were the Tektronix—green line on green screen. It was storage tube technology, so a moving light left an image on the phosphor screen. I was at a technical conference one time... it was the end of the day, we'd had a show. I had one of these display tubes, and this particular device, when it wasn't being used, it would run a colored rectangle to just keep the screen uniformly refreshed. We were so tired at the end of the day. I had a crowd of people sitting behind me watching the line go back and forth, back and forth.
Then we got raster technology where you've got every pixel... black and white first. That was easy because you just took your figure you had on your vector screen and you made a black and white image. Straightforward.
Then they came out with color raster screens. I said, "Oh, I'll color the lines." And I did, and I looked at it and said, "Well, that's not very interesting. It looks pretty much same as it was before." I said, "I wonder what would happen if I colored the spaces in between the lines?" And I did that. It took me 10 minutes or so to code. Up came the picture and I said, "Bingo." That new technology... it popped. You could just see everything. It's the standard contour display now, but it was the change from the color of the line to fill the spaces that made all the difference.
Rajesh Bhaskaran: And how did the educational program start?
John Swanson: Well, the educational program was my idea, and it was the only idea I ever had that got vetoed by my Advisory Group, which was my managers. To a person, they said, "Bad idea. You're going to compete with the industry. You're going to compete with our consultants." I said, "No, we're doing it anyway." It is the only time I ever overruled unanimous opposition.
The first price on the educational version was $1 because basically I wanted to support education. That was early 1980s. A year or so later, I raised the price to $100 because I wanted the university to sign on for security purposes, not just the individual professor. One person screamed bloody murder because I raised the price by a factor of 100.
But ever since then, the ANSYS software has been widely used in education. I think the last number was 2,000 universities worldwide use ANSYS.
One of the stories that go along with that: my marketing manager was a woman, but her husband worked at United States Steel and he was head of the computer center. So he was at IBM meetings about computing. He was walking down the corridor and he looks into an office and there's a whole row of ANSYS manuals. He asked, "What's that?" They said, "Well, we hired this guy from the University and he said: I know how to do simulation, I use ANSYS, and away we go." So IBM became one of our bigger customers, but it was from the educational program. So it's a good business strategy as well as a good educational strategy.
Rajesh Bhaskaran: What is your educational background and how did it prepare you?
John Swanson: Cornell at the time was a 5-year program. My scholarship was a National Merit Scholarship—it covered four years. I'm always grateful that Cornell came up with the money for the additional year, plus money for a half year to get my Master's Degree.
Then I went off to work for Westinghouse in Pittsburgh on the nuclear rocket program. I was there only six months when my manager said to me, "I need to put somebody in for the PhD program." I said, "Okay, sure." A couple of months later they said, "You're in." So at that point, I went to night school at the University of Pittsburgh for three years... three courses per trimester, three trimesters a year. I got my PhD degree in 1966.
Rajesh Bhaskaran: Were you also developing ANSYS code at this point?
John Swanson: No, at that point I was working on Boundary Point Collocation methods. I had gotten into early finite element work, but I'd been doing boundary point collocation work as well, so there was an overlap there.
I worked at Westinghouse for another three years after that. The program was showing signs of defunding—it was a government program and the handwriting was on the wall. Besides which, I was much more interested in doing finite element work than I was in managing my stress group. My management philosophy was: if my in-basket got too high, somebody would call and tell me what was important, and I didn't have to deal with that. So I often claimed that my departure was a safety measure because [the basket] was high enough to collapse on me and kill me.
I went out looking for a job that would pay me what I wanted to be paid—namely Aerospace wages—to do what I wanted to do—namely develop software. I could not find both. So I started my own company, which was Swanson Analysis Systems, eventually to become ANSYS Incorporated.
Rajesh Bhaskaran: How long was that before you were selling the software?
John Swanson: Well, getting Aerospace wages took a long time. The funding originally was from my own investments. That was $1,282 as I recall, for office space. But I did consulting for Westinghouse to earn the money to pay for the computer time at United States Steel to develop the software, which by the end of the next year I was licensing back to Westinghouse. So Westinghouse was my major support and one of my first customers. Westinghouse has been good to me, and I think I've done good things for them as well.
Rajesh Bhaskaran: Why do you keep supporting Cornell Engineering still?
John Swanson: Well, Cornell is a good school. And you know, I'm in the phase of life now where I'm doing "outgo" as opposed to "income." So most of my work these days is charitable work, and Cornell is obviously a good place to do that work. Both with the engineering school myself and the veterinary school, which is what Janet, my wife, supports. So my wife is actively involved in philanthropy as well. Our objective is to give it all away and to die at the same time so it all comes out even. But as I pointed out last night, we're not going to do that [die yet]; we're staying. That's the way we want it to be.
Rajesh Bhaskaran: And what do you see as the future of simulation?
John Swanson: Well, simulation is great because it keeps evolving and changing. You know, at our meeting today we were talking about worldwide networks and SimCafe and so on... Simulation is enormous as far as the market and as far as the impact on engineering. As you're aware, I got the John Fritz Medal, and that's the top engineering award in the country. That was based not just on what I did, but what simulation has done: our products are better, our products are higher quality, our products are faster to market. All those things come from simulation.
