Subtitle: Simulation vs. Design Analysis of OHPs
Brent Cullimore
In 1990, Hisateru Akachi showed in his patent that if you created flow circuits from a coiled, tiny-diameter loops, and filled them part way (say 30-70%) with a volatile working liquid, they could transport heat from one set of turns to the other. Without moving parts, against gravity, and with tremendous design flexibility. They are also easy to 3D print, though additive manufacturing wasn’t a consideration then of course. And they work with many working fluids, including in temperature ranges that conventional heat pipes can’t reach.
He called them “Loop Type Heat Pipes,” but that was a name that was not to last because the new Loop Heat Pipe (LHP) technology that was coming out of the former Soviet Union at the same time was to dominate for the next 30 years. Akachi also called his novel devices heatlane and kenzan fins in later publications. These early devices looked a lot more like a display case for a paper clip collector (and don’t you just know that those people exist). Which is perhaps why Manfred Groll informally and affectionately referred to them as spaghettipipen.
One very useful dissertation (Sameer Khandekar, Thermo-hydrodynamics of Closed Loop Pulsating Heat Pipes, 2004) earned my affection by pointing out the similarities to the even earlier inventions of the drinking bird toy and the bubble pump.
I first learned about these strange new paper clip loops in the early 90s thanks to the presentations I attended that were given by Groll and his students. He referred to them as pulsating heat pipes (PHPs), at least in more formal conversations. Now, as they are catching on more and more, I hear the term oscillating heat pipes (OHPs), so I’ll more or less stick to that term here. But if you get to know them, you’d be forgiven for thinking of them as Indecisive Loops or Frantic Loops. And even though that last one sounds like a breakfast cereal, I’d try a bowl.
If you think that the controversy ends with what to call them, you haven’t seen how they are also controversial as to how and why they operate, and when they don't. Think of a schoolroom full of little attention-deficit percolators sitting at their desks, charged up by their morning bowls of Frantic Loops. Now add the fact that the chugging in one of those little percolators can both stimulate and suppress the chugging of neighboring percolators.
Do OHPs transfer more by sensible heating and cooling of liquid slugs, or by latent heat by converting liquid slugs into vapor plugs and vice versa? The answer depends on who you talk to, and which loop you are considering and what operating state it is in. While the answer is inevitably going to be “both,” it is very hard to estimate how much of each mode is happening in any one instance.
As a liquid slug oscillates within a heated zone or flies past it, is nucleation within that slug important for sustained OHP operation, or is it just needed to break up the large initial slugs during start-up? That one seems to generate even more controversy, and the answer even affects the criteria you design against in terms of operational limits.
These devices are so complex and chaotic that they defy currently-available simulation tools. CFD runs take days or weeks to produce a few seconds of simulation, but of course thermal time scales are much longer. Thermohydraulic tools (such as SINDA/FLUINT) are orders of magnitude faster even with full two-fluid two-phase solution methods, but they too are currently unable to produce design information in a useful timeframe without more specialized modifications. As a result, both of those approaches might be relegated to a research support role: helping settle some of the above controversies. Perhaps helping to predict when OHPs will start and when they won't. But they are unlikely to help a potential OHP user accept and integrate this technology.
Currently, the most successful simulation approaches are “spring and mass” based, modeling each slug. Such methods have limited ability to resolve the research controversies and yet they are still too slow to be useful for design purposes. A design tool must be fast enough to work at the system level, over long thermal time scales, and most importantly: with repeated runs to bracket worst cases, to assist in design sizing, and to characterize uncertainties. Steady-state solutions are key tools for system-level integration studies, but they simply don't exist for OHPs. At most, we can hope for a time-averaged solution.
What do you do with a technology that defies simulation, other than toss a Hail Mary pass to future quantum computers? What do you do with a drunken loop that won’t walk a straight line, when all your simulation technologies work best when linearized line segments are assumed?
Of course, you could rely on the methods of early helicopter prototyping: find out when they crash and try again. The joke I’ve heard repeatedly, and half believe, is that no one would have accepted the helicopter as a new concept if it had been invented between about 1990 to 2010. In those years, the need for predictability via design simulation was coming on strong, yet the CFD tools required were still far from handing “radial flow over a wing that flies into its own wake.”
OHPs are coming on strong in electronics packaging because they are cheap and highly customizable. But in aerospace, a lot more is on the line than an angry review on Amazon.
So what do we do? Do we have to disregard OHPs as design options while they remain so unpredictable? Do we create huge empirical databases for each possible design? Do we wait until a few of them have “crashed” and then develop a collection of both angry and defensive opinions (and is there any other flavor these days)?
I think a big hint comes from more conventional heat pipes. Few designers need to solve for the vapor and liquid flows within such a heat pipe. As long as they have assurances that key limits (boiling, entrainment, nucleation, sonic, and wicking) are not exceeded, you can describe the operation of a heat pipe in much more simple approaches as outlined here. And no, I’m not talking about the “very conductive bar” approach.
Still, the approach I am talking about, the “vapor node approach,” is as disrespected in R&D circles as it is respected and widely used for decades by engineers needing to apply heat pipes to their design. Each group has very different needs, so that this gulf exists is hardly surprising. One group is trying to push the limits of possible designs, and the other is just trying to make sure that Heat Pipe Design X will work just fine under all of the circumstances that they can analytically throw at it. One needs simulation tools, the other needs design tools.
For OHPs, we just happen to be currently unable to fulfill either end of the spectrum. As a result, I think we will have to work on both needs separately … while keeping watch on each other’s successes or failures.
Approximating OHP performance won’t be as simple as approximating conventional heat pipe performance. OHPs will take more complex calculations under the hood and will have to report the OHP’s performance in ways that can be more easily checked against limits (once we all agree what those are!). Stay tuned for news of any progress. Lack of progress is more likely, but you are less likely to hear about that.
I dearly hope that this blog post will soon be updated if not outdated as answers to these quandaries appear and as basic research controversies are resolved.
But I’m also hoping that the community has converged on a single name for PHPs/OHPs/spaghettipipen by then. Until it has, I see no reason not to promote “DLs” for Drunken Loops. I won’t push for Frantic Loops, however, because Kellogg’s has too many lawyers on staff.