Implementation of the Disruption Predictor APODIS in JET’s Real-Time Network Using the MARTe Framework
Juan Manuel López, Jesús Vega, D. Alves, S. Dormido-Canto, Andrea Murari, J. M. Ramírez, Robert Felton, Mariano Ruiz, Guillermo de Arcas · IEEE Transactions on Nuclear Science · 2014
This paper describes the implementation of a real-time disruption predictor that is based on support vector machine (SVM) classifiers. The implementation was performed under the MARTe framework on a six-core x86 architecture. The system is connected via JET's Real-time Data Network (RTDN). The online results show a high degree of successful predictions and a low rate of false alarms, thus confirming the usefulness of this approach in a disruption mitigation scheme. The implementation shows a low computational load, which will be exploited in the immediate future to increase the prediction's temporal resolution.