1 August 2001

Presentations

Minutes

Our weekly telecon was held again this week.  ORNL presented a progress report since last week's project meeting.  Reiersen reviewed the findings from last week's project meeting.  There was considerable discussion about how these recent developments should affect what we do next.  Significant finding included:

More details on these findings is provided in the discussion below.

HM Fan's transient analysis of cooldown was reviewed.  In the analysis, it was assumed that the septum was solid copper and that the temperature at the base of the septum was clamped at 80K.  The analysis showed a ratcheting of only 5K after 16 pulses and no problems in meeting the 15 minute cooldown requirement.  The problem is that the base of the septum is not clamped at 80K and it is difficult with gas cooling to adequately edge-cool the base of the septum.  The reference design has transverse cooling through a labyrinth in the septum and is cooled with pressurized helium gas.  There are 20 of these cooling labyrinths on each side of the coil and therefore, 80 cooling connections per coil.

Helium gas has the nice feature that we can cool it to 80K and it will still remain in the gaseous state.  The concern about helium cooling is that if we develop a significant leak, we will have to fix it.  This would probably require dismantling the whole machine.  Gas cooling will require complex manifolding with many connections - on the order of 500 for the modular coils alone.  Given our experience at PPPL with cooling leaks on TFTR and NSTX, we cannot afford to be cavalier about this possibility.  Another downside of using helium is the cost of the compressor, which is in excess of $400K.

We could tolerate leaks better if we cooled with nitrogen gas because it would probably be an open loop system - the boil-off from the liquid nitrogen supply (our cryogenic heat sink) is already paid for.  The concern about nitrogen cooling is that we need to run at a higher coolant temperature in order to ensure single phase flow.  The nitrogen saturation curve is shown in the figure below.  To ensure single phase flow at 5 atmospheres would require keeping the the coolant temperature above 95K (versus 77K at 1 atmosphere).  A higher inlet temperature would mean a higher starting temperature for the coils.  This translates into greater Joule losses (making cooling harder) and a larger temperature excursion over the pulse (increased thermal stresses or lower current requirements).  We also would have to deal with the temperature differential between the structure, which is cooled by the boil-off at 80K (slightly above atmospheric pressure environment in the cryostat), and the coils.  With an 80K inlet temperature, the starting temperature is assumed to be 85K, the maximum temperature rise is 43K, and the resistive dissipation in the coils is 78MJ.  For each increase in starting temperature of 1K, the temperature rise (DT) increases by 0.6K and the resistive dissipation by 2MJ.

Goranson compared gas cooling with nitrogen for edge-cooling versus transverse cooling through the septum.  He allowed the starting temperature to be substantially higher (10K v. 5K) than the inlet temperature of the coolant, thereby cutting off the long "tail" at the end of the cooldown.  For this calculation, he assumed an inlet pressure of 3atm and an inlet temperature of 88K.  The cooldown time with transverse cooling was estimated to be 11 minutes if the coil was only cooled down to 98K.  Edge cooling required a longer time of 18 minutes.  Goranson indicated that the number of cooling loops in either case would be comparable to the reference design with helium.  The results are encouraging in that we might be able to get by with gaseous nitrogen instead of helium.  We should sharpen our pencils to determine if we want to change from gaseous helium cooling to gaseous nitrogen and/or from transverse cooling to edge cooling.

All of the cooling analyses performed thus far have avoided 2-phase flow.  There are two worries commonly associated with 2-phase flow: uneven flow distribution among parallel channels and coolant channel "burnout" due to flow stagnation arising from a vapor lock.  It is not clear that these concerns are relevant to our situation.  Concerns about uneven flow distribution arise only when the flow rate in one channel affects the flow rate in a neighboring channel.  Proper sizing of the manifold and flow passages and active control of the flow rates in each channel should easily mitigate these concerns.  "Burnout" is typically a concern when we have a heat load to the coolant channel that is independent of the cooling.  With a loss of flow situation that might arise from a "vapor lock", the temperature of the channel could be unstable and climb until a failure occurred.  We on the other had, do not have a steady state heat source.  In fact, there is no heat being generated during our cooldown.  We just have to drain the reservoir of heat stored in the winding during the previous pulse.  The temperature of the winding pack is at most 40K higher than the cooling channel.  Without flow, the temperature will drift up towards the bulk temperature of the winding pack.  It will not be unstable.  Furthermore, we have excellent thermal conduction along the length of the winding pack.  If we are not taking the heat out locally, we can take it out a meter upstream or downstream.  It appears that the situation might be self-healing rather than unstable.   Therefore, we should consider using 2-phase LN2 cooling as an option, in addition to the gas cooling option.  This might give us the lowest starting temperature (best I²t capability), fastest cooldown time (best rep rate), simplest and most reliable design (perhaps one cooling circuit (?) per winding pack instead of 20 and improved fault tolerance), and lowest cooling system cost (gas compressors not required).  It is not clear it would work but it's worth a look.

ORNL received a conductor sample from New England Wire (NEEWC) that is very close to our projected size.  Our current design criteria of 3X the conductor width was based on a small sample (7mm square) that was fabricated at ORNL.  The testing was done on uninsulated conductor.  The new sample is much closer in size to what NCSX requires and was insulated after squaring it up with Nomex.  The finished conductor size is 11mm x 15mm (versus our 13mm x 16mm).  The good news is that the packing fraction is 78%, significantly higher than the 70-75% previously assumed.  NEEWC suggested that it could be even higher if the strands has not been varnished. (The sample was for a high energy physics application.)  The bad news is that our design criteria of 3X the conductor width may need to be changed.  The conductor is very flexible and bends up to a point beyond which it locks up, becoming very difficult to bend with the insulation crinkling.  A more appropriate criteria appears to be in the range of 4X-5X the conductor width.

Nelson reported on different vacuum vessel segmentations.  The motivation for changing the vacuum vessel segmentation was to make it possible to assemble the machine if the vacuum vessel was expanded to provide space for an inboard RF launcher.  The options considered basically involved going from the reference design of three 120 degree segments to six segments, probably of unequal angles.  There was no obvious winner identified.  It is very tedious to determine if a coil module can indeed be slid into place over the vacuum vessel.  PPPL will investigate automating a way to determine the optimal trajectory for assembly.

Williamson confirmed the interference between the modular coil and the vacuum vessel at the 30 degree cross-section that Brown had previously identified.  Williamson pointed out that this was actually indicated in Strickler's table of parameters for the M12 coil set in which the minimum coil-to-plasma distance was 17.46cm, far less than the 19cm value required for the nominal winding pack.  Following discussions centered around the engineering metrics in CoilOpt.  There is nothing sacrosanct about 19cm.  If the distance is less than 19cm, the radial build of the winding pack is reduced, increasing the current density.  Likewise, if the minimum coil-to-coil spacing is reduced, the lateral build of the winding pack is reduced, also increasing the current density.  The current density on the other had is a fundamental constraint (although we may be a bit fuzzy on what exactly that limit should be).  The consensus was that CoilOpt should be modified to determine the radial and lateral build of the modular coils by the minimum coil-to-plasma distance and the minimum coil-to-coil separation.  From the winding pack dimension and the coil current, the current density could be calculated.  Similarly, the radius of curvature constraint would be calculated based on the winding pack dimensions.

Nelson discussed progress in determining the stay-out zone and divertor geometry.  ORNL will lay out a stay-out zone and divertor geometry for use in the next iteration of boundary physics and machine configuration studies.  Presently, CoilOpt calculates the minimum coil-to-plasma separation.  The physical constraint is actually the coil-to-vacuum vessel separation.  Once the stay-out zone and divertor geometry is specified, ORNL will lay out the vacuum vessel outside those envelopes.  PPPL (Brooks) will describe a pseudo-vessel surface in Fourier harmonics that closely matches the vacuum vessel geometry on the inboard side and roughly conforms to the plasma (but offset) on the outboard side.  Strickler will modify CoilOpt to calculate the minimum separation from this surface, rather than the plasma.

Please forward any comments or corrections to reiersen@pppl.gov

(last edited on 09/25/2001 03:47 PM )