this is for holding javascript data
Jakub Urban (minor) corrections from Ivan
over 9 years ago
Commit id: 8634c4a43d213ad9d4b8b09e5cb67504b295f430
deletions | additions
diff --git a/JUrban_SOFT2014.pdf b/JUrban_SOFT2014.pdf
index 0d31acb..aa7f4c9 100644
Binary files a/JUrban_SOFT2014.pdf and b/JUrban_SOFT2014.pdf differ
diff --git a/Results.tex b/Results.tex
index ec3c98e..98d9fad 100644
--- a/Results.tex
+++ b/Results.tex
...
VacTH is performing reasonably well for its most favourable diagnostic set of 8 magnetic probes and 16 flux loops and $n_P = n_Q = 4$. With a higher number of harmonics or with less flux loops, VacTH becomes unreliable and yields significant errors. Unfortunately, only 4 flux loops are currently available on COMPASS. In fact, it is easier for VacTH to fit magnetic probes than flux loops while magnetic probes do not fully determine the magnetic flux. This is also apparent in $E_\mathrm{mp}$ and $E_\mathrm{fl}$ values in Tables \ref{table:ex4275} and \ref{table:ex6962}. An additional optimization of the fitting weights or algorithm is probably needed. The current behaviour might be quite anti-intuitive as VacTH performs significantly worse with 16 flux loops and 16 or 64 magnetic probes in comparison to 16 flux loops and only 8 magnetic probes.
EFIT++ internal plasma parameters reconstruction results are shown in Fig. \ref{fig:kinetic_stats}. It shows that purely magnetic reconstruction with $n_{p'}=n_{FF'}=1$ introduces (except for $I_{\rm{p}}$) a systematic error for realistic pressure profiles, i.e. for plasmas that do not have the same profile parametrization.
It is known that magnetic reconstruction with EFIT is difficult for
limited limiter plasmas (without additional constraints, particularly the
store stored energy) \cite{efit1985}.
This suggests that using $n_{p'}=1$ for limited plasmas and $n_{p'}=2$ for diverted plasmas might lead to better results. This is demonstrated in the bottom row of Fig. \ref{fig:kinetic_stats}.
Reconstructions with such optimized parameters do not suffer from the systematic error; however, they generally increase the error bars for target equilibria with linear $p'$ and $FF'$, especially for $q_0$. It is also notable that $\delta l_{\rm{i}} \cong 0.1$ for all $n_{p'}=n_{FF'}=1$ reconstructions.