题目：Developing Disruption Warning Algorithms Using Large Databases on C-Mod, EAST, and DIII-D
共同作者：R.S. Granetz, C. Rea, R.A. Tinguely
MIT Plasma Science and Fusion Center, Cambridge, MA, US
D. Chen, Y.M. Duan, S. Gu, X. Gu, Z.P. Luo, J. Qian, B. Shen,
B. Wang, B.J. Xiao, F. Yang
Institute of Plasma Physics, Chinese Academy of Sciences, Hefei, China N. Eidietis, A. Leonard General Atomics, San Diego, CA, US
In order to mitigate disruptions, we must first be able to predict, in real time, if a disruption will be happening soon. We also want to know if one universal disruption prediction algorithm can work on multiple tokamaks. Toward this goal, we have developed large databases from 1000’s of discharges on C-Mod, EAST, and DIII-D, containing disruption-relevant signals vs time for both disruptive and non-disruptive plasmas. These databases are being used to learn which plasma parameters are useful for predicting disruptions, and to develop machine learning algorithms for disruption prediction.
In this presentation I will focus on two sub-topics:
1) Comparison of several plasma parameters on EAST and C-Mod, showing that disruption warning signals behave differently on the two machines
2) Machine learning: decision trees and random forests for disruption prediction using the DIII-D database
授课人介绍：Robert Granetz is a research scientist in the Alcator group at the MIT Plasma Science and Fusion Center. Much of his research over the last few years has concentrated on disruptions and mitigation."