Principia Cryogenica

Principles of mechanical design of large scale refrigeration systems

Mechanical design considerations from a process basis

Helium gas storage

The function of the helium storage tanks is to:

1. Short-term: provide headroom for upset conditions. "Headroom" means excess capacity in both directions: some types of upsets (loss of vacuum, quench, etc) require the system to absorb and store lots of inventory, while other upsets (2K cold box trip) require the system to remove lots of gas from storage for backfilling the linac.
2. Medium-term: provide a buffer of inventory for proper operation of the floating pressure process
3. Long-term: maintain a buffer of inventory to offset losses due to leaks

From (1) above, it's beneficial to have tanks that are statically bottled up at some pressure so that they can be put into action during the response to an upset condition. At JLab, we typically have 2 tanks isolated at high pressure (~10 atm) and 2 tanks isolated at low pressure (~2 atm). The remaining tanks are connected into the operating system and are allowed to float at some intermediate pressure.

Some reasons why we have maintained the clean line and purifier discharge as separate lines, with individual valved connections to each tank:

1. You can use the recovery system to scrub gas from one (otherwise isolated) tank into another (otherwise isolated) tank. This is critical for recovering from an instance of tank contamination. It's also necessary to allow us to replace the tank relief valve as we've planned.
2. You can use the recovery system to pump an (otherwise isolated) tank to a relatively high pressure so that it may be kept in reserve for use during the response to an upset condition.
3. You can isolate the recovery system from the main compressor system to better identify the source of helium leaks.

Design for manufacturing

Pipe bending

A bending process is sometimes used to fabricate warm and cryogenic pipe spools in lieu of welding an assembly of various fittings and short lengths of straight pipe.

A significant fraction of the JLab cryogenic fabrication shop technicians (welders and pipefitters) know how to use the tool and most of them have completed pressure system projects using it. There is an experiential learning process that is required before people become proficient at creating correctly formed parts. To a project manager, that means you must budget time and material to allow trial and error to occur. Whenever we train a new user we probably generate 40℅ usable parts for the first few dozen bends, and 80-90℅ thereafter.

One reason you would choose to use a bend as opposed to a welded pipe elbow is because you want to increase flexibility in your assembly. That is, you want your piping spool to conform to the dimensions needed to satisfy the loads on the system.

It takes 3-4 parameters to fully define a typical bend. As you can see, none of these are very tightly controlled.

• radius
• bend angle
• position along pipe
• (if necessary) rotation angle w/r/t other features

Bend radius

The bend radius is fixed by the tooling chosen for the bend. When you purchase bender tooling you must specify pipe diameter and bend radius. For our bender, there is a very small set of available radius options for each given pipe size. Generally thinner-walled pipes must be bent using larger-radius tooling or else the intrados will collapse.

This is not to say that the bend radius matches the tooling radius. Because of springback (discussed next), the actual bend radius always exceeds the tooling radius by some small amount.

Bend angle

Because of material elasticity, pipes in their loaded condition in the bender will spring back some amount when they are returned to a free state. Because of this, some "overbend" is required to achieve the desired bend angle. An experienced operator will have a feel for this, and an inexperienced operator will iterate until he finds a solution. I believe our machine comes preloaded with a program that helps estimate the overbend but I don't know how effective it is. Often I have seen the technicians intentionally underestimate the overbend, then creep up on the intended angle by adding increments of bend.

The exact required overbend is a complex function which would be really fun to characterize. It appears to be dependent on

• nominal bend angle
• wall thickness
• choice of material
• choice of lubrication

Bend position along pipe

This becomes important when putting 2 or more bends on the same spool. This is critical for building a U-tube, for example.

On our machine, this is left entirely up to the operator. Our technicians usually make a mark on the pipe and line it up with one of several notches/pointers on the frame of the bender. It takes iteration and experience to get this right, but once a solution is found it is generally repeatable.

With the addition of a couple roller stands and a string potentiometer, it's possible to gain a digital readout of axial pipe position. In fact on our machine this is an option that we haven't yet purchased. Clearly the footprint of the machine would increase if we did this, and for the moment we value our bender for its easy portability.

If you add motors to drive the pipe along its axis, you can gain full CNC control over this parameter.

Rotation angle

This is important only when making 3-dimensional pipe spools. Like the previous parameter this is also left entirely up to the operator on our machine. Unfortunately there are no indicators to allow the technicians to repeat the setup, so it must be reestablished each time. Usually they level the top of the machine, then use an angle finder on the previous bend to set the orientation of the spool.

You also have the option of buying machines with NC or CNC control over this parameter.

There's also mandrel bending machines, which use an expanding tool to press against the ID of the pipe at the business end of the bend. This prevents collapse of the intrados and allows you to achieve tighter radii with thinner wall material vs non-mandrel bending. Mandrel benders have a necessarily larger footprint because the mandrel must enter through the end of the pipe.

It's worth noting that when you cold work 304 stainless steel, you'll induce some small amount of magnetism. It would be fun to measure this on our machine, but we haven't had the opportunity.

There's a soft brass wear component (called the wiper if I remember correctly) as part of the tooling. The manufacturer recommends heavily lubricating this part, as well as the pipe, with a white lithium grease before bending. Since it takes effort to degrease afterwards, we typically skip this step for pipe sizes less than 1.5" although this is probably not good practice. The ungreased wipers are wearing faster (although we have not had to replace them yet). Potentially worse, they're probably leaving behind brass particles which could lead to weld contamination if they're present in enough volume.

Material selection

Surface coatings

Galvanizing should be avoided, especially for structural components where it's conceivable that we may need to weld to it in the future.

Component selection

Relief valves

Process relief valves should be ASME type. Reliefs should be sized such that the effective orifice area K*A is sufficient for the flow to be relieved. The flow regime may be either critical or sub-critical.

The mounting design should take into account the reaction forces due to the momentum change of the relieved flow, evaluated at the steady-state relief condition. In most cases, it is wise to install a device on the exhaust that redirects or (attempts to balance) the exhaust flow. Reliefs should not be exhausted towards where personnel might stand or walk. It is imperative to design the exhaust spool such that any moisture can drain out by gravity, especially if it is attached to a cryogenic relief that may attract moisture from the air when cold. The exhaust spool must also prevent entry of creatures who might like to live in it (birds, wasps, etc.).

To minimize time to replace, reliefs should have flanged interfaces to the inlet piping. Where they are exhausted remotely by some lengthy outlet piping, the outlet piping interface should be flanged as well.

Pilot relief valves should have a modulating pilot to minimize blowdown, thus mitigating helium inventory loss. The pilot should relieve into the main body of the relief (instead of having a separate exhaust to atmosphere).

Project documentation

P&I development

The JLab P&I design "style" originates from the following principles:

1. Equipment in a thermodynamic process should be arranged on the P&I in generally the same location and orientation as they are found on a TS diagram. For example, most warm compressors are shown on the P&I with the flow direction from right to left, just as they are on a TS diagram. Likewise, expansion processes are shown happening from left to right. Also, the 'main process' of the P&I will generally be oriented so the the piping with the warmest temperature located at the top of the drawing, and the coldest at the bottom.

2. Piping layout and vessel shapes should loosely resemble the actual geometry, within reasonable artistic interpretation (e.g. that jog in the line may actually represent a "thermal stability loop").

JLab P&IDs are traditionally created in the ME10 drafting system. Thus, much of our "style" results from normal behavior of ME10. There are considerable limitations; for example, it's especially tough to maintain consistency between drawings.

Very rarely we feel the need to create color drawings. More often we use only two or three different line thicknesses to distinguish the main process piping from the auxiliary systems. When these lines cross, we use a "jump" to make it clear that they don't connect. It's sometimes hard to decide which line should jump the other line, but typically the auxiliary piping will give way to the main process.

Other "best practices" of JLab P&I development include:

• maintain no more than one vacuum enclosure per drawing
• instrument valves (e.g. isolation valves, equalization valves, bayonet purge valves, etc.) fall below the threshold of the drawing scope and are not shown
• flags (which work like GOTO statements) are used to join far-away groups of piping
• reducers are never assumed; a reducer is shown explicitly wherever there's a reducing tee
• identical, repeating components are shown once; the example cases are labeled with variables in their names