[adfergusson]

I'm putting together an experimental testing rig that is going to be driven by a fairly big servo/drive combo (three phase, 12A, 415V drive). The drive is also going to be dumping a lot of energy during deceleration through its regen/brake resistor. However, due to space/layout limitations the setup is going to be located near a robotic machining cell that is mostly used to machine, drill, trim, etc... carbon fibre composites and some of this dust does get out the cell. As a result, I'm worried about the servo drive (or something else) blowing up if carbon fibre dust gets sucked into the cabinet.

An idea to prevent this was to get a fully sealed enclosure and then put together a cooling loop like the ones used in custom/high end PCs. I'd have a radiator, with fans, the pump and coolant reservoir inside the enclosure and another radiator outside the enclosure, connected with pipe through some glands that I could seal. As a result, no dust should be able to gt into the enclsoure

Building a cooling loop that could handle the heat won't be a problem (made plenty before, but for workstations), but I haven' seen any examples of people doing this sort of things for DIY CNC machines so I thought I'd check here to see if anyone might be able to think of reasons why this wouldbe a bad idea? Potential problems, and mitigation, that crossed my mind were:

Amount of power to dissipate; some cooling loops for high end gaming PCs can dump out over a 1000W of heat without much difficulty
Cooling component failure; redundancy, just go for two independent loops and have an alarm buzzer trip if a pump/fan stops drawing current
Space; get a bigger cabinet!
Regulations; IP5x and 6x cabinets are liquid resistant/proof so would having a cooling system which pumps liquid in and out of the enclosure make this a non IP5x or 6x cabinet (even if the coolant is trapped in a closed system)?
Leaking; coolant for water cooled PCs is non-conductive so shouldn't bother the other components; redundancy and alarms for low coolant levels (also available for custom PC setups)

Can anyone think if I'm missing something?

[underthetire]

If you are truly creating a lot of cf dust, you may end up in the class 2 div 1 or class 2 div2 category of NFPA . check in to this before you begin. I've seen a lot of machines with water cooling, they all leak eventually. If you don't meet the class 2 conditions (hope not!) Why not just mount the resistors outside the cabinet with a shield and maybe a fan. That's what most companies did when they used braking resistors back in the 80s and 90s.

[adfergusson]

Thanks underthetire. We're definitely not operating in anything like the hazard categories you're talking about but it only takes a tiny amount of carbon fibre dust to blow out a computer (we know this from experience). The main issue is that the cost of something going wrong, and writing off the test, would be huge as the tests take a week to run and the samples about a week to prepare and instrument. If the dirve did blow then the lead time for replacement/repair would also be large. Consequently, I'm probably being a bit paranoid about my fear of a drive blowing during use.

The servo drive's built in brake resistor should be capable of dealing with the heat from deceleration so we haven't got an external one yet. However, if getting an external resistor results allows us to use a sealed electrical housing then that would probably be worth a look.

As a matter of interest, where did leaks in the cooling systems you saw came from? Did the leaks look vibration, corrosion or galvanic action induced, or something else (first three are a non-issues for pc based water cooling, if its something else then I'll need to consider it), We've a number of workstations (some of these can draw up to a kilowatt during some simulations) that have been liquid cooled and leak free for about four years now, which is why I was looking at using something similar but would be grateful if anyone can alert me to something I might be missing?

[underthetire]

I know what you mean with carbon fiber, we do a lot of experimental stuff at work, its always a mess. Heat cycling seems to cause failures from stuff getting brittle or just corroding, even stuff with Dow frost gets brittle over time. You can get sealed ac units, but that gets expensive. You might get away with a sealed heat exchanger sized correctly. Most machine tools with multiple large drives use them.

[Fenza CNC]

Something along the lines of this might be of help to you, Aqua Computer Aquaduct 360 Eco Mark II - 12V Pump [11232] it's a PC radiator designed for mounting outside of the computer case.

When PC's handle 1000W of heat they're taking that heat directly from the source. An air to air setup like you describe wouldn't be as efficient but I can't see why it wouldn't work! Mounting the braking resistors outside the cabinet would help as well.

Hope this help,
Fenza

[UA_Iron]

I design large medical device instrumentation with a lot of fluid handling components surrounded by electronics, VFD's etc. We have to meet some fairly stringent regulations in terms of safety - IEC 61010 and other components. Believe me when I tell you that we have dealt with the pains of leaking tubing, fittings, orifices etc... We recently did a dFMEA on an instrument and found over 200 leak points contributing to potential hazard modes.

Here are my suggestions:
1. Do as other have recommended and locate heat generating non-carbon fiber dust sensitive components outside your control box - we aim to make your cooling scheme as efficient as possible.
2. Design your control box such that liquid connections are made at the bottom of components (fitting coming into the bottom of a heat exchanger) with no components underneath - you want to orient things in such a way that there will be no dripping onto other components.
3. You want thermal mass - or "cooled mass" within the cabinet. You can still have fans actively blowing over components, but you will want some mechanism to get that heat to the liquid cooling elements. The more cooled mass and surface areas you have, the better you can evacuate heat. Something like this comes to mind:
https://www.alliedelec.com/aavid-the...000g/70115274/
4. One-touch (aka push to connect) fittings are the best. Do not use barbs - tubing will eventually fatigue at the interface. Under higher pressures one-touch fittings create a better seal where barbs would be compromised.
5. Shield areas where connections are questionable - like the thermal plate item listed in #3 - this would require some kind of compression fitting around the copper tubing to mate with a one-touch fitting. A compression fitting will likely not seal as well as a one-touch fitting, so the chances of an interface failure at the compression fitting is higher than that of the one-touch. By putting a sleeve around the fitting, and then ensuring it can drip to safe locations is a good strategy for reducing risk.
6. Observe material compatibilities. If you're using a coolant, make sure that it is compatible with all your components - copper tubing, o-ring seals within one-touch fittings and bulk-head fittings, tubing material etc.
7. Design a single tub, or fluid pooling point at the bottom of the control box and put a capacitive sensor at the lowest point for leak detection: Products

I bet something like an aftermarket transmission cooler would work well for the application.

[adfergusson]

Thanks, that some really helpful advice. Will give it a gentle run with light load to see if more cooling and/or external brake resistor is needed. Transmission cooler kit sounds like a good idea if I end up needing to go down this route. Thanks again