1 on 1 with: John Chawner

Most people know John Chawner as the president and co-founder of Pointwise where for a little over 26 years they developed, deployed, and supported mesh generation software for computational fluid dynamics. You may be familiar with some of his writing on the Another Fine Mesh blog.

In April of 2021, Pointwise was acquired by Cadence Design Systems. These days John continues his work with mesh generation and CFD in the Multiphysics Systems Analysis Group. It’s been several months of excitement as he integrates CFD into Cadence’s legacy of computational software development for electronics design.

Let’s learn a little more.

Thank you for taking a few minutes to share your thoughts on a few simulation-related topics with us.  Can you tell us a little about your background and how you got to where you are today?

I am an aerospace engineer by degrees (BS from Syracuse University and MS from the University of Texas at Arlington). I’ve spent my entire career in the CFD software business beginning with applied CFD for propulsion aerodynamics (inlet and nozzle flowfields). Later, I was one of the original developers of the Gridgen software for mesh generation. That work began at my first job out of college at General Dynamics (now Lockheed Martin Aeronautics) in Fort Worth. After that, I worked for three years at Prof. Dale Anderson’s consulting company doing research contracts and funded development of Gridgen.

Is there anyone who was a significant influence on your career?

As an undergrad at Syracuse, Prof. John LaGraff took me under his wing for an independent study project. Another student and I rebuilt a very large shock tube from parts scattered throughout the sub-basement of the engineering building. Believe it or not, it actually worked when we tested it. Prof. LaGraff and I have maintained contact throughout my career. The student I worked with on that project now serves with me on the Mechanical and Aerospace Engineering Advisory Board.

During a summer job at NASA Lewis (now Glenn) Research Center in the 10×10 supersonic wind tunnel, I was mentored by Bobby Sanders. He gave me the opportunity to learn Fortran programming and run my first CFD code (a method of characteristics code). Bobby and I had the opportunity to interact on the X-30 project in the late 1980s when I was at General Dynamics.

I see that you’re a member of the CFD Vision 2030 Integration Committee. Can you tell us about this?

The Integration Committee (IC) fosters and promotes the vision of CFD in 2030 first defined in the NASA CFD Vision 2030 Study published in 2014 by a blue-ribbon panel of experts. The IC is an activity within the American Institute of Aeronautics and Astronautics (AIAA) that I’ve been a member of since 1980.

The Vision captured my imagination immediately upon my first reading because it aligned so well with our thoughts at Pointwise about the role of geometry modeling and mesh generation. When the IC was formed in 2018, I was thrilled to become a charter member and I remain fully committed today as a member of the steering committee.

The Vision calls for a future in which a single engineer can conceive and execute a large number of CFD simulations in a single day and complete them all to within a prescribed level of accuracy. The Vision looks at CFD through the lenses of high-performance computing, numerical algorithms, physical modeling, geometry modeling and mesh generation, and multidisciplinary analysis and optimization.

The IC is in the early stages of creating grand challenge problems in the areas of aerodynamics, propulsion, and space launch vehicles that can be used to motivate, monitor, and assess the industry’s progress toward 2030.

Where can people learn more about this?

At the IC’s website (www.cfd2030.com) you can find links to the original CFD Vision 2030 Study and a recently completed and quite extensive update of the Vision’s roadmap.

What are some of the CFD industry’s biggest challenges from an end-user perspective?

I’m willing to bet that the challenges CFD users face are similar if not identical to those from other computational disciplines. That’s one reason why I’m also involved in the ASSESS Initiative that looks at simulation from a higher-level and broader perspective.

Here’s what I’ve heard from CFD users. First, simulations take too long – person time and run time. This limits their ability to impact the design process. Second, the expertise required to effectively and efficiently use CFD software is kinda high. This narrows the talent pool for performing simulations. Third, extracting engineering information from the terabytes of simulation output is becoming increasingly burdensome. Again, another time and expertise problem.

These challenges all make perfect sense. We want things to be as fast as possible. We want to remove barriers to entry and make technology as broadly useful as possible. And we want the tools to deliver precisely the information we require to do our jobs.

What about vendors?  Where do their challenges lie?

The CFD-specific technology challenges are shared by both users and vendors. The bridge between geometry models and CFD continues to vex all parties involved. (I like to say that geometry modeling is to meshing what turbulence is to CFD.) This is a process-related issue.

Speaking of turbulence, the accuracy of our simulations suffers near the boundaries of the flight envelope where the flow may be turbulent or separated or where the other physical models we employ tend to break down. (I also like to say that mesh generation exists in order to make turbulence modeling seem respectable.) This is an accuracy issue as well as an issue of range of applicability.

Other issues are more germane to the vendor although the dividing line can be gray. Advances in computing are outpacing our ability to refactor and rewrite our codes, whether that’s for the next generation of heterogeneous HPC platforms or the more exotic exascale platforms on the horizon and beyond. Programming models need to be established that allow codes to be moved from platform to platform with the minimal amount of recoding or platform-specific code. I’d call this an environment issue.

Then there are new technologies like machine learning that weren’t on the radar back in 2014 when the CFD Vision 2030 Study was written. Determining how best to implement and deploy technology like that and properly setting expectations about what it can and cannot do are going to be critically important. This is a disruption issue.

As CFD has moved past the “one and done” phase of computation, we’re now faced with running both hundreds of cases involving toolchains involving multiple codes. Whether we’re doing design optimization or a multidisciplinary simulation, we need to stop reinventing the wheel on how all these codes can work together efficiently. This is an ecosystem issue.

Before you think that sounds very negative, the good news is that advancements are being made in all the areas cited above. As author William Gibson wrote, “the future is already here; it’s just not evenly distributed.”

As far as engineering simulation tools go, what are some of the more noteworthy advancements you have seen?

Keep in mind that I started working in CFD in 1984. One of my first CFD applications was a two-dimensional exhaust nozzle for a jet engine. I spent a year working on it to get results that were, let’s say, not suitable for engineering decision making. That simulation would now be an undergraduate homework problem.

So would it be fair for my answer to be “everything?”

I would say one of the most noteworthy advancements has been the evolution of computing platforms. Back in the day we were using only specialized graphics workstations and supercomputers. Today we can use commodity hardware that’s much more affordable. And then there’s the cloud – you don’t even need your own computer anymore. (I was doing cloud before it was cool, running jobs on a Cyber 205 at NASA Ames Research Center from Fort Worth in 1986.)

Speaking of affordability, the rise of open-source CFD software has certainly been a game changer in terms of lowering a barrier to entry while also delivering top notch simulation capabilities.

And while commercial CFD vendors have been aspiring to multidisciplinary simulations for just about as long as I can remember, today we see not only multiphysics simulation but optimization and CAD integration and in situ visualization and so much more integrated computational capability.

Where do you see things five, ten years down the road?

Making predictions is difficult, especially about the future. Yoggi Berra supposedly said that.

I was recently talking to another CFD veteran and two things are clear. Doing a CFD simulation to understand a flowfield is behind us. Simulations are done to produce engineering data – lift, drag, temperatures, loads – to drive a design forward.  The same goes for doing a single simulation; it’s about ensembles of simulation now, about mapping the flight envelope or exploring a design space.

These scenarios imply something about how we’re going to manage all that data – input and output – to best support our engineering processes. And they imply something about the efficiency of our processes.

You may infer that the above statements are more about process and less about processors (i.e., computers). Not that computing platforms aren’t important, but our vision of future CFD can’t be based on hoping that computers get “better.” We can control only what we can control.

In terms of the core of CFD technology, I believe we will see new solver algorithms come to the foreground that give us a leap forward in terms of accuracy and simulation efficiency. I believe we will see scale-resolving simulation techniques become more widely used. All of this is about moving off the plateau that Tony Jameson famously said CFD solvers have been stranded on for a decade or so.

And I would be remiss without making a prediction about mesh generation. I believe that mesh adaptation will become a standard component of most CFD simulations.

I am hoping that by making so many predictions at least a few of them will turn out to be true.

Democratizing Simulation is all about making simulation more accessible for non-experts while allowing companies to fully leverage CAE resources and investments.  What do you see as the primary barrier(s) to democratization of CFD simulation and for simulation adoption/usage in general?

Full disclosure: In the past, I’ve been publicly critical of the word “democratization” from the standpoint of diction. It just doesn’t feel like the right word to describe making software more broadly accessible.

Regardless of my issues with word choice, I’m all for making CFD available outside of the “priesthood” of simulation. I know nothing about what makes an internal combustion engine work but I’m certainly a capable driver. That’s an analogy for how simulation software should be.

Before anyone trots out the “it’d be like giving a gun to children” complaint (which I have actually heard), it’s our responsibility as software developers to put the appropriate guard rails on the tools.

One of the main barriers to achieving this goal is the simple fact that simulation software is mostly general purpose. You can compute any flowfield around any object with general purpose CFD. However, users of software rarely compute flow around anything; they compute the flow around the specific types of widgets their organization manufactures. Somehow, CFD needs to learn the language of the various application domains. There are plenty of ways to accomplish this from scripting all the way up to dedicated “sim apps” that target specific tasks.

I would be remiss if I didn’t point out that a key to democratizing simulation is making meshing invisible to the user (a goal of the CFD Vision 2030). If we’re concerned about non-experts being flummoxed by the CFD flow solver, we certainly don’t want to expose them to meshing.

How essential are democratization and automation to the wider adoption of CFD simulation?

Automation is absolutely vital.  We’ll never get to the time when all CFD is automated. There will always be “hero” cases and new developments that require manual intervention. But there’s no reason why we shouldn’t be able to automate CFD for most daily applications. It’s being done now in a variety of ways.

And it may not be as difficult as it seems. I’ll take a little poetic license with a line from Fred Brooks’ classic book on software development The Mythical Man Month. “Automatic has always been a euphemism for better than we’re currently doing.”

As for democratization, if we take it to mean improved usability then that’s probably something software vendors should be doing with each and every release.

What technologies or products do you see as missing for the wider adoption of simulation? How important is robust automation for this to happen?

If I take your question literally and really focus only on wider adoption of simulation, it’s not a matter of tech or products. It’s a cultural issue. It’s about raising awareness of simulation’s benefits, setting expectations for what it can and can’t do, and quantifying its value (not justifying its cost).

Many readers may be wondering “Aren’t all those aspects of simulation obvious?” For people in engineering simulation today, those aspects are obvious. But our larger audience, the part of the iceberg below the surface, doesn’t have that understanding yet. It’s important for us to remember the “curse of knowledge:” it’s hard for us to remember what it’s like to not know the things we’ve worked hard to learn.

Why do you think it is important to participate in Rev-Sim and how does an organization like ours provide value to the broad simulation community?

I believe it’s always a good idea for like-minded organizations to assemble, share, strategize, and amplify their ideas and successes. For Rev-Sim, it’s democratization. For ASSESS, it’s the entire CAE ecosystem. For CFD 2030, it’s the future of computational aerosciences.

Thank you for taking some time out of your schedule to talk with us. One final question. Enough about work… tell us a little about yourself.  What you enjoy doing in your spare time, interesting hobbies, and so on.

I started running at age 54 and after five years of trying, this year I finally hit my goal of running 600 miles in a year. My wife and I, pandemics notwithstanding, try to run two half marathons per year.

Other than running, my hobbies are a lot more sedate. I’m a music lover, from industrial ambient to classic jazz and everything in between. I got to see King Crimson perform recently here in Fort Worth which was great. I’m an art lover, especially modern painting. That makes the Fort Worth Museum of Modern Art a favorite place for me to visit. And even though the pandemic has decimated my leisure reading (no daily commute for audio books, no airline flights for hardcovers), I enjoy reading everything from Tom Clancy to William Faulkner.

(Note: The opinions expressed in this interview are John Chawner’s alone and are not forward-looking statements regarding Cadence’s business or products.)

Member Login

Welcome to Revolution in Simulation, an online community. A one-time registration is needed to provide you with free unrestricted access to all Rev-Sim content and to the quarterly newsletter featuring the latest updates, news, events and more.

Not a Registered Member yet? Register once for a RevSim membership