Contents An Analysis of the Multi-machine Pedestal and Core Databases by J.G. Cordey, O. Kardaun, D.C. McDonald and members of the ITPA Pedestal and Global database groups Fitting Pedestal data to a) Thermal Conduction Model. b) MHD Limit Model. Determining scaling of Core. Predictions of the estal temperature and τ ε in ITER.
Joint Pedestal-Core database DATABASES Consists of 239 pulses from Asdex Upgrade(63), CMOD(19), DIII- D(11), JET(74), JT-60U(62). Similar selection to paper by K. Thomsen at H-mode workshop. Type I ELMs + CMOD. Expect database to be roughly doubled in size in the very near future, 180 pulses from DIII-D, 60 from JET, CMOD? So the analysis is on going. Global Confinement ELMy H-mode Database DB3v10 Consists of 2677 pulses from 14 devices, the selection is the same as used in the IAEA Sorrento paper by O.Kardaun 2
Background T, n etc. averaged over several ELM cycles. Pedestal Models a) Thermal Conduction model W & 1 6 W= P - - gbn, * P t econd Thermal conduction through estal ELM loss term W core W T, n, P b) MHD limit model bc ac Dac - = Ã bc where r a 1 ~ D ~ i R a r 2 2 Rq Rq Main difference between models is the dependence on P in model (a). JG02.148-9c 3
Pedestal Scaling - Thermal Conduction Model W e I R P M q = - 376. 01. 176. 0. 05 117. 0. 06 0. 31 0. 06 0. 28 012. 133 sh. 019. q Ÿ q 95 / q sh cyl No dependence on B or n Dependence on e, k a is indeterminate. Satisfies Kadomtsev, Connor-Taylor constraint. 4
Expressed in Dimensionless Variables Bt ~ r *- 34. b - 17. Close to gyro-bohm but with a large degradation with respect to β. For small power loss by ELMs can be re-expressed approximately as P BW b ~ + 17. -1-3 -17. r b b * o o P Pure gyro-bohm heat loss Elm loss term β o Type III/Type I transition β. 5
MHD Limit model If a ballooning mode formalism was to apply then p š I R 2 2 r * a š I R 2 2! 3mT 8 I 12 / " $ # a Fitting this type of expression to the data is statistically quite difficult since there are strong correlations in the database between n and R and between T and I. Two techniques have been used an Errors in Variables technique and the simpler technique of changing the variables to a set that are not correlated. Both techniques give the same result. 6
The resulting regression is W = exp - 265. 010. I R * q m 1 6 r n 2 0 - *. 17 0. 05 0. 14 0. 01 1 sh. 40 0. 18 0. 24 0. 12 Note both r* and n* dependence is very weak. Equivalent to n ~ in variables technique. T -056. almost identical to result from Errors 7
Comparison of fits of estal data to models Thermal Conduction Model RMSE = 21% MHD Limit Model RMSE = 32% 8
Scaling of the Plasma Core Use expressions derived for W to obtain W core = W th - W and then obtain fit to full ELMy H-mode database. For the Thermal Conduction Pedestal Model W = exp -335. I B n R m P core 0 5 e k 0. 6 0. 17 0. 57 2. 24 0. 88 0. 8 0. 18 0. 34 a The combined two term model W th = W + W core fits the ELMy H- mode database with an RMSE = 15.5% compared with 15.9% for the IPB98(y,2) one term scaling. The two term model with W from the MHD limit ρ* model has a somewhat worse fit 16.5%. 9
Fits to full ELMy H-mode DB Two Term Thermal Conduction Model 10 ASDEX D3D PBXM TFTR AUG JET PDX CMOD JFT2M TCV COMPASS JT60U TDEV 1 τ ε (s) 0.1 0.01 0.01 0.1 1 τ ε cond (s) JG02.148-5c 10 RMSE = 15.5% 10
ITER reference one term model IPB98(y,2) 10 ASDEX D3D PBXM TFTR AUG JET PDX CMOD JFT2M TCV COMPASS JT60U TDEV 1 τ ε (s) 0.1 0.01 0.01 0.1 1 τ ε IPB98 (y,2) (s) JG02.148-7c 10 RMSE = 15.9% 11
Profile Stiffness? Although scaling of W and W core with P and B is similar, their scaling with I, R and n is very different. W š I R W š I n R 18. 12. 08. 056. 21. ; core Recent calculations by X.Garbet however suggest that this core model is partially stiff. ITER Predictions T (KeV) τ ε (s) ITER reference scaling IPB 98(y,2) - 3.6 Thermal Conduction model 5.7 4.1 MHD Limit model 2.7 3.5 Both two-term model predictions are within the 95% confidence interval of the IPB98 scaling The FIRE T predictions are 3.5 and 1.6 KeV 12
SUMMARY For the present estal DB, the thermal conduction model gives a better fit to the data than the MHD limit model. The origin of the degradation in τ ε with β seen in the one term models, eg. IPB98(y,2) Bτ ε ~ ρ*-2.7 /β, is mainly from the estal and is probably a consequence of the ELMs. A two term model of the estal and core has been develo which is as good a fit to the full ELMy H-mode database as the one term models. The prediction of the estal temperature in ITER have been given for the two models, the global confinement times from both two term 13 models are within the confidence interval of the IPB98(y,2) scaling.