Startup Accelerator: Will PanX Disrupt Simulation for 3D Printing?

I usually try to stay grounded and err on the side of little to no enthusiasm for new things. This has historically been a good approach in 3D printing. However, in my exploration of startup PanOptimization and its PanX software line for additive manufacturing (AM) simulation and optimization, I’m growing increasingly enthusiastic at every step.

PanX is different in all the right ways and, without too much fanfare, is fundamentally disrupting finite element analysis (FEA). PanOptimization is refreshingly different because it is a company of just three persons. Founder Pan Michaleris, who founded Pan Computing and sold it to Autodesk in 2016, is an authority on thermo-mechanical modeling with 30 years of experience. Pan Computing’s FEA solver, CUBES, became Netfabb Simulation and was also integrated into Fusion. The other two employees are Erik Denlinger and Tyler Nelson. Erik researched the thermal behavior of directed energy deposition (DED) for a decade and is the editor of the unputdownable book Thermo-Mechanical Modeling of Additive Manufacturing. Tyler worked at GE Research Center and GE Aviation in modeling and simulation. He developed GE Aviation’s SimComp software and helped develop the GE Amp software.

The company may hire a few more people over time but wants to stay small. Thanks to the previous exit, they are self-funded and do not seek investors. They are also profitable already and want to slowly build out and improve their product rather than “blow up” or chase revenue. It’s a refreshing plunge into a different entrepreneurship model than the conventional startup model. With deal flow treacle slow, perhaps in the current market, PanOptimization is a small but beautiful model. What if you just got three incredibly good people and then paid them for five years until a product emerged? Growth hackers be damned.

I think PanX’s more conservative model is refreshing and should be explored by people contemplating a new business venture. But what PanX is doing is perhaps even more interesting. FEA has been a bit of a clash of the titans, with ANSYS, Altair, Siemens, Dassault, COMSOL, and others ruling the roost in certain industries and applications, battling it out across the board. This teeny tiny firm aims to tip over that apple cart. PanX has been built to cater to the additive market, enabling users to do things on desktop PCs that typically require cloud or HPC systems, all while meshing and solving parts efficiently.

Daniel Driscoll, a Manufacturing Engineer at Velo3D, works extensively with PanX. For him the key thing that PanX does is,

“The ability to specify different processes volumetrically is crucial. We use significantly different processes in various areas of our builds, involving different powers and resultant built-in stresses. Without this capability, no simulation tool can accurately simulate parts printed with Velo3D’s parameters. Also, the speed of simulation at relevant resolutions is essential—our customers are printing large, complex parts with our Sapphire XCs. Thus, we need to accurately resolve thin features (less than 1mm) while also simulating a volume 600mm in diameter and up to 1000mm tall. Traditional additive simulation software tools either homogenize the thin features to the point where they significantly negatively affect accuracy, or they are unable to mesh and run with commercially practical compute resources (i.e., not specialty HPC/supercomputers). A better balance of assumptions is necessary. PanOptimization has collaborated with us to add complexity to the simulation where required; with larger machines having higher deposition rates, accuracy on the thermal side of the simulation is critical to building successes, as well as vital for achieving anything accurate in deformation simulations. Being able to reasonably include the powder in a simulation, rather than having an ‘infinite slow heat sink,’ has allowed us to replicate thermal and distortion issues that arise from trapped powder volumes or adjacent parts creating locally higher heat conditions.”

From a usability perspective, large printers and large parts flummox many traditional tools. These are precisely the parts—like rocket combustion chambers—that are driving the adoption of large machines and multi-laser systems. The Air Force and New Space companies are looking for huge metal parts that they want to make quickly. A failure on a one-week build on a $3 million machine is an expensive one. However, missing a launch schedule or disappointing your local military satellite agency could be even more costly. As Daniel points out, previously, you’d need a supercomputer to perform many of these calculations. A defense contractor isn’t likely to pop this stuff up on AWS either. Creating meshes with the requisite amount of detail while being practical is also crucial. With complex parts, such as heat exchangers, trapped powder is a significant complexity problem as well.

Mark Kirby has worked for Rolls Royce, Renishaw, and the University of Waterloo in an aerospace and additive career spanning over 36 years. He’s now at Tronosjet, a Canadian firm with AS9100D certification, capable of approving and certifying commercial aviation repairs and replacements. He also uses PanX. Mark grew up with Delcam and Powermill and hasn’t seen a tool path he didn’t want to simulate. In additive, however, many people are “landing with your wheels up” because they don’t have the time to simulate 3D printed parts. Previously, he found simulation for additive to be “too slow, inaccurate, or both and not practical to do.” Now it runs on his laptop, and besides mitigating mistakes, he also sees how the software lets him improve.

Mark runs three Renishaw machines, building mostly in titanium, copper, and aluminum. For him, PanX has been valuable, especially when it lets him compensate for distortion and check for recoater interference. It also helps him search for smart support strategies. He likens PanX to running a preflight checklist and uses it for around 80% of all of his builds.

¨Because it is so expensive to build again, we tend to over support parts, now we can reduce that and predict other problems in builds such as cracking and look for hot spots in builds. We can iterate in software with different strategies quickly to see which one is best.¨

PanX uses a novel multi-grid modeling approach that allows for the generation and solving of FEA meshes consisting of 500 million nodes. The company aims to produce high accuracy and resolution results while also computing sensitivities. Their goal is to offer an FEA product line that not only serves as a “preflight checklist” but also improves build outcomes at any scale. Their multigrid approach and status as a new player in a known market are sure to attract attention. Will PanX’s small team sustain the promise into a disruptive business? This is something we have yet to see. I’m definitely perked up and paying attention—are you?