ARIES Project Meeting Minutes
13-14 October 2011
Documented by M. Tillack
Ref: Agenda and Presentation Links: Meeting Agenda
Welcome/Agenda -- Mark Tillack welcomed the ARIES team to the Gaithersburg Hilton hotel. He reviewed the day and a half agenda and discussed the topics to be presented.
Plans and General Scope
Updates from OFES -- Al Opdenaker discussed FY12 funding and the likely impact of the continuing resolution. If anyone is in dire need of 1st quarter funds, they should contact Al.
FESAC activities were summarized (see the FESAC web site). Several ARIES members are leading or contributing to the current charges, which include:
Comments on the current ARIES project -- Farrokh Najmabadi said that the 1st iteration on our conceptual designs is needed ASAP. Some remaining issues remain (preventing the detailed designs to be completed), to be discussed during and at the end of this meeting and resolved in the next few months. This includes in particular some issues regarding configuration choices and safety implications of design choices.
We should reconsider the naming of our current designs, which are confusing to people outside the project (and some inside).
ARIES Task Results
ARIES systems code results and further development -- Lane Carlson described code improvements and the implementation of DCLL modules.
First he explained the current logic behind the ACT designations, which are likely to change.
By examining the power balance algorithms in detail, and creating a comprehensive power flow diagram representing the contents of the system code, some errors were found, reported and corrected. The errors were large enough to require a revision in the strawman.
The "physical build" part of the code also was corrected.
Preliminary DCLL strawmen (ACT-3 and -4) were proposed. Scan parameters and filtering used to obtain these were described. The initial DCLL point designs displayed for discussion.
Don Steiner noted that the COE's all look similar - we need a different parameter to choose a point. Najmabadi felt that the best point appears to be the lowest BT point. Kessel said the wall loading is another parameter to look out for. He is looking for βn around 4.5. El-Guebaly said the radial and vertical builds still need refinement. Following that comment, Najmabadi suggested we complete the SCLL designs first, and work on the build for the DCLL designs before releasing the first set of DCLL strawmen.
Carlson raised some areas where he needs guidance. He said the pumping power split would be most helpful. Kessel provided guidance on the IB/OB split for radiation: 80/20 OB/IB power flow with 90% radiated.
El-Guebaly asked if we can reduce triangularity to make room for shielding behind the IB divertor. Kessel answered we can't reduce it enough to make a difference. Maybe we could reduce the distance from the separatrix to the inboard divertor plate. (Note, the geometry of the divertor slots, strike points and surrounding structures is an ongoing task, discussed again under "edge physics modeling".)
Studies of finger-type divertors with helium -- Minami Yoda described initial experimental results for the finger-type divertor using a single impinging jet (rather than an array of jets). The nondimensional heat transfer coefficient, or Nusselt number Nu, was plotted against the nondimensional coolant mass flow rate, or Reynolds number Re, and semi-empirical correlations relating Nu to Re were presented.
The initial single-pass experimental results with helium and argon showed that the Nu for helium was significantly less than that for air. Yet the Nu results for argon were similar to those for air. To better understand the discrepancy between the helium and air results, the experiments were simulated using ANSYS FLUENT.
Based on these simulation results, it appears that the discrepancies are due to conduction in the side walls. Some discussion ensued whether the correlations should be modified to include conduction (as presented by Yoda), since the heat incident upon the divertor is removed by both convection and conduction, or a conduction correction should be applied to the data, so that the correlation is a "pure" convective heat transfer correlation (that could then be used in future design and analysis).
Najmabadi asked how the turbulence model was chosen. Yoda said the Spalart-Allmaras model gave a better match to the measurements of cooled surface temperature, with the k-ε model was predicting temperatures that were up to 40K lower than the measurements. Spalart-Allmaras should give better results in flows with separation, such as the impinging jet stagnating on the cooled surface.
Steiner asked if the helium used in the experiment was pure. Yoda said that since the experiments are single-pass and the helium comes from a bottle, it should be at least 99% pure, so impurities are unlikely to significantly affect the properties.
Najmabadi asked why don't you account for conduction losses in your calculations to correct the correlations? Yoda said she plans to look into this.
Liquid metal flow loop design -- Mark Tillack discussed his attempts to design liquid metal cooling circuits with minimum MHD effects by controlling 3D currents that are generated by perturbations to the flow. Certain perturbations are acceptable because they do not lead to large 3D currents, whereas certain other perturbations should be avoided. Both pressure drop (and primary stress levels) and flow distribution are concerns.
A complete flow circuit for the inboard and outboard blanket was described. It was assumed that the divertor and high-temperature shield would be He-cooled. The primary manifolding was moved to the ring header outside the field, where MHD effects should be eliminated.
Kessel suggested that the field strength at the location of the ring header may not be negligible, and Tillack agreed to check into it (using a simple dipole approximation to the outermost PF coil).
Malang mentioned that the small FW channel size together with a velocity of 4 m/s would result in a friction pressure drop of some MPa. Tillack agreed to calculate friction forces and compare with MHD and inertial forces.
Steiner suggested that experiments are needed to quantify modeling results, and generally questioned why we are still considering liquid metal blankets if MHD is still not understood. Tillack explained that a lot of data and modeling exists for simple geometric elements, but the complex designs such as Tauro and ARIES-AT are very difficult to model and very expensive to test. Steiner asked how bad are 3D effects, and Tillack agreed to do a better job communicating this to the Team (note, some of this was done in 2010 for the system code models of MHD pumping power).
Najmabadi made some suggestions for design variations, such as the use of inserts to reduce 3D currents and full-width access pipes. He agreed that we should have better estimates of 3D effects, and stressed that we should not characterize ARIES-AT as "flawed", but rather incomplete in certain areas.
Power core configuration -- Xueren Wang reviewed ARIES-AT design features as a starting point. He mentioned the option of using He cooling the HT shield as well as the divertor for the PbLi/SiC blanket option. In this concept, steel is used for structure as well as shielding material.
Najmabadi suggested we also consider the option of using SiC channels (with He coolant) in the shield. This might improve the compatibility and temperature matching with the other coolant streams.
Pros and cons of different options were discussed. We need a more detailed design and some analysis of the He-cooled steel shield option and its variants before we can downselect.
El-Guebaly cautioned that we need to consider afterheat in steel.
Wang then continued to discuss the integration of the He-cooled divertor into the power core. A modified transition joint design with thermomechanical analysis was presented. This design attempts to reduce strains by eliminating sharp corners. The fabrication steps all use brazing, as opposed to the original joint design that used some welding.
The power core configuration and flow path analysis were presented for the PbLi/SiC design with He-cooled divertor and PbLi cooling of the shield (the He-cooled shield option is further discussed below).
Various options for access pipe routing were reviewed. Access pipes need to either penetrate the TF coil structure at the bottom of the power core, or the inner vacuum vessel door. A 3rd option was suggested in which the pipes are routed above the lower PF coils and below the port.
Najmabadi mentioned that this is related to the vessel design, which needs further examination. There is likely a need to adjust the port doors to properly account for sector removal, which may also open up space for routing access pipes differently.
Najmabadi prefers removing water from the vacuum vessel, as a result of recent discussions on the safety implications of tritium containment. Some discussions ensured on the vessel thickness, and the idea of moving the shielding function from the vessel to additional shielding material inside the vessel. Laila will provide a radial build for the two options. Issues of decay heat removal, shielding and radial build were discussed further.
Two options were presented for control coil maintenance: demountable or moveable. Najmabadi clarified the situation. Two sets of control coils are needed: two near the midplane and two at the upper and lower ends of the port door. The upper and lower coils are probably multi-turn (no joints) and have to be moved. Those near the midplane are saddle coils and can be attached to the door. Both are cooled copper; either water or helium could be used (depending on tritium, hydrogen generation, etc.)
It was agreed that we will use a flat inner port door to allow upper and lower coil vertical motions. This design is simpler; no serious penalty was identified if we replace the curved door with a flat one.
Najmabadi noted that we need to ensure there is clearance for sector removal after cutting pipes.
The alternative design option using He-cooling for the shield was discussed. PbLi coolant routing was again displayed. Najmabadi suggested we consider using the full width of the sector for the access ducts.
El-Guebaly noted that we need to specify the thickness of He and steel in the divertor region in order to perform neutronics analysis.
Najmabadi reiterated that we need more analysis of the design options before we can downselect.
3D divertor model -- Lane Carlson discussed the production of 3D models using ABS plastic, and passed the results of a small test run to the Team. This sample represents a small part of a finger-type divertor at 1:1 scale. It is very helpful in visualizing the configuration and potential issues.
The cost and effort were modest, so the Team agreed that continued use of the 3D printer is worthwhile. Tillack noted that a large model (e.g., a full tokamak power core) would be difficult to produce without our own printer, which would cost in the range of $15k.
One of the issues uncovered is the need to include clearance in the parts drawings to allow for assembly.
Update on fracture analyses -- Jake Blanchard described the extension of his crack propagation modeling from 2D to 3D. The results show a factor of ~2X improvement (reduction in stress intensity) in the 3D analysis as compared with 2d.
Najmabadi asked how the locations of initial cracks are decided? Blanchard explained that the highest stress values obtained in analysis without a crack are used.
Tillack asked how design limits like fracture toughness are to be interpreted in the context of the observed stress intensities. Do we only need to stay below the toughness limit? Blanchard said that is true for a single load cycle, but you need to make adjustments (e.g. crack growth rate measurements) to evaluate high-cycle fatigue.
He agreed to look at plastic strain and fracture mechanics under cooldown conditions by the next meeting.
Literature review for ferromagnetic loads -- This became an issue when we started considering ferritic steel for the vacuum vessel (which experiences large gradients in field strength). Jake reviewed the literature, going back to Starfire. ITER has done a lot of work more recently.
Some analysis with ANSYS was presented and compared with ITER calculations. A general conclusion is that loads can be tolerated, but need to be considered.
The next step is to include transients by placing a FM block in the blanket region and dump the TF coil current (and later it was decided to also examine disruption loading conditions).
ITER vacuum vessel loads -- Jake Blanchard's second talk started with a review of vacuum vessel loads considered for ITER. A thorough review of normal and off-normal scenarios was shown.
The time scale for an ITER "fast discharge" of the coils is 25 s. The worst case loading condition is a disruption, leading to a 300 MPa stress.
Tillack asked how relevant is the ITER case to ours? Najmabadi said our design probably would have much lower stresses on the vessel and higher stress on the sector strongback, because we don't use the vessel for power core support. Kessel added that our vessel also has much weaker toroidal continuity, which would tend to reduce currents and stresses. If the strongback is conducting, then it will help define the location of forces under transients and confine halo currents.
In response to Arthur Rowcliffe's question from the previous meeting, we considered that the worst case for ITER is ~300 MPa under disruption loading, but this is highly conservative for us. Probably in our case it will be 100 PMa or less. Analysis will continue.
Vessel vacuum tritium permeation & PbLi/water safety issues -- Brad Merrill presented the talk for Paul Humrickhouse, who was unable to attend the meeting due to illness.
He reminded us that water was added to the vacuum vessel in ARIES-CS (duplicating ITER) in order to help remove decay heat. Since then, the use of the ITER vessel for passive (natural convection) decay heat removal was dismissed by the French authorities. They didn't like passing pipes through a containment boundary.
Steiner asked how can we can have such high tritium pressure at the VV. Merrill said that breaks in the sector allow several percent "leakage" of T to the vessel. Steiner wondered, why can't we load H into the water and reduce the rate of exchange of T into HTO? He asked Merrill "on balance, do you still favor the use of water? Merrill answered "no".
El-Guebaly remarked that water still has advantages in shielding and a compact radial build.
Kessel asked if we lose site power, does the SiC design survive? Merrill said that with a tungsten divertor, we may have a problem. El-Guebaly replied that analysis in ARIES-AT showed it was OK. LOFA was worse than LOCA, due to PbLi activation.
Edge model results -- Marv Rensink presented by call-in the latest results from LLNL. He described two scenarios they have been exploring: mostly attached and mostly detached. Analysis was presented and the features and differences compared.
The LLNL recommendation for the "reference" case is weak radiation (mostly attached plasma) even though the peak heat flux is clearly too high. Incremental improvements may produce an acceptable (but still high) heat flux. Strongly radiative (mostly detached) plasma gives very low heat flux (~3 MW/m2 peak), but requires edge high density and is difficult to access.
Kessel asked what causes such a high density? Is it a boundary condition, particle source? Rensink answered that high density is caused by weak particle pumping when the plasma is detached. The flux from the core of the plasma adjusts to match the particle pumping rate in the divertor.
Kessel asked what explains the density difference in the two cases? Rensink said particles are removed when they strike the plates. 1% of the incident ions are pumped on the plate. Neutral particles are not pumped in the model results presented here. Perhaps they should be. Inclined plates tend to push recycled neutrals into the private flux region.
Steiner asked if is pressure balance is somehow involved. Higher temperature implies lower density. Rensink agreed with this assessment.
Kessel asked whether, even with a high tilt angle, is it still desirable to allow space for density buildup on the non-private side? Najmabadi and Kessel suggested it may be better to place the strike point on the dome so that gas is not lost to the private region. Rensink said this is possible, but the disadvantage is that the dome then will see higher heat fluxes. We could consider a dome that doesn't follow the flux surfaces. The mesh would be distorted, causing possible computational problems.
Kessel wondered how to create the right geometry for the slots. Where precisely should the plates be located? Rensink said he doesn't know. Past models assumed flat plates. Curved plates also could be used to tailor the geometry.
Kessel asked if ITER is sufficiently optimized. Should we use that configuration? Rensink noted they have a vertical plate and a private flux dome to enhance the neutral pressure in the OB divertor.
In his summary, Rensink mentioned "particle throughput is significantly lower for detached plasmas". Kessel said that detached plasmas have higher density and should pump better. There seems to be an inconsistency. We need better tracking of neutral and charged particle flows.
ARIES-ACT1 preliminary plasma description -- Chuck Kessel showed results of JSOLVER and his preliminary fixed and free boundary equilibria. Ideal MHD stability of high-n ballooning and external kink modes was also discussed. Steiner asked if the cases with high edge density (with detachment) presented by Rensink are consistent with his solutions. Kessel said he has no major concerns. His density is very similar. They need to track one another (as well as GA analysis) to ensure consistency.
Turnbull stated that the reason we don't achieve βN of 5 experimentally is because profiles can't be achieved due to current drive limits, and not wall stabilization limits. Kessel remarked that the benefit of pushing beyond 4.5 seems small, and it pushes beyond the database (so he doesn't support it).
The conservative physics case (ACT2, now called ARIES-SC) will be examined next.
Optimization Results for Consistent Steady-State Plasma Solution -- Alan Turnbull described progress in the GA self-consistent simulations of the ARIES-ACT1 (now ARIES-SA) baseline. One of the problems they have been struggling to resolve is profile collapse when running transport codes GLF23 and TGLF. He also briefly described recent code improvements.
Kessel asked why the boundary condition (the pedestal) is so low (below 4 keV)? It may be the reason the profiles are collapsing. Turnbull responded that βp is set in the pedestal. With higher density, the temperature is halved to keep βp fixed. This is somewhat arbitrary and needs to be fixed. Kessel said we need to radiate from the core, probably leading to a higher Ar density.
Steiner asked how the solutions are optimized - around what parameter? Turnbull said he iterates on density to obtain a steady state, then varies q, pressure, beta, shape and current drive mix. Steiner continued: why will this give different results than ARIES-AT? What has changed in the models? Turnbull said the tools are the same. We're just looking at different parameter cases. βN is a key parameter, which is now lower. The shape and density have changed, and a lot more Ar was added. Getting the right pedestal height may solve some of these problems. Pedestal density now is quite high (and temperature low).
Kessel noted the EPED analysis is the most critical thing to do next. Steiner reiterated, why are the profiles so different than earlier results? Kessel said we are trying to treat the pedestal more accurately and to match experiments. Turnbull added that ARIES-AT didn't really have a pedestal at all, so this is all new.
Review of Past and Current Action Items -- Mark Tillack led a discussion of action items and remaining issues, starting with notes tane during the meeting by himself and Lane Carlson. Some revisions were made and the final version posted with the presentations (at the end of the agenda).
Design designations -- Following the meeting, options
for the naming convention for our latest designs were offered via email.
At present, the leading candidates are:
Next meeting -- The next meeting will be held at UC San Diego on January 23-24, 2012 (half a day Monday and all day Tuesday).