Record Details

Risk-Aware Framework Development for Disruption Prediction: Alcator C-Mod and DIII-D Survival Analysis

Harvard Dataverse (Africa Rice Center, Bioversity International, CCAFS, CIAT, IFPRI, IRRI and WorldFish)

View Archive Info
 
 
Field Value
 
Title Risk-Aware Framework Development for Disruption Prediction: Alcator C-Mod and DIII-D Survival Analysis
 
Identifier https://doi.org/10.7910/DVN/EFAWUW
 
Creator Zander Keith, Chirag Nagpal, Cristina Rea, R. Alex Tinguely
 
Publisher Harvard Dataverse
 
Description In this paper, we investigate the usage of survival analysis for disruption prediction and avoidance in tokamaks. Determining the optimal action to minimize damage from an oncoming disruption requires the plasma control system to take into account both the length of warning time and the associated risks of available actuator responses. Making time-to-event predictions from time-series data can be achieved with a survival analysis statistical framework, and there have been many tools developed for this task which we aim to apply to disruption prediction. Using the open-source Auton-Survival package we have implemented disruption predictors with the survival regression models Cox Proportional Hazards, Deep Cox Proportional Hazards, and Deep Survival Machines. To compare with previous work, we also include predictors using a Random Forest binary classifier, and a conditional Kaplan-Meier formalism. We benchmark the performance of these five predictors on experimental data from Alcator C-Mod and DIII-D. We observe only minor differences when comparing receiver operating characteristic scores. However, survival regression models tend to achieve longer warning times in the low false positive rate regime.
 
Subject Physics
Alcator C-Mod
DIII-D
disruption avoidance
disruption prediction
machine learning
survival analysis
 
Date 2024-02-26