Maria Katiuscia Zedda

A prediction tool for real-time application in the disruption protection system at JET

A disruption prediction system, based on neural networks, is presented in this paper. The system is ideally suitable for on-line application in the disruption avoidance and/or mitigation scheme at the JET tokamak.A multi-layer perceptron (MLP) predictor module has been trained on nine plasma diagnostic signals extracted from 86 disruptive pulses, selected from four years of JET experiments in the pulse range 47830–57346 (from 1999 to 2002).The disruption class of the disruptive pulses is available. In particular, the selected pulses belong to four classes (density limit/high radiated power, internal transport barrier, mode lock and h-mode/l-mode).A self-organizing map has been used to select the samples of the pulses to train the MLP predictor module and to determine its target, increasing the prediction capability of the system.The prediction performance has been tested over 86 disruptive and 102 non-disruptive pulses. The test has been performed presenting to the network all the samples of each pulse sampled every 20 ms. The missed alarm rate and the false alarm rate of the predictor, up to 100 ms prior to the disruption time, are 23% and 1%, respectively.Recent plasma configurations might present features different from those observed in the experiments used in the training set. This novelty can lead to incorrect behaviour of the predictor. To improve the robustness and reliability of the system, a novelty detection module has been integrated in the prediction system, increasing the system performance and resulting in a missed alarm rate reduced to 7% and a false alarm rate reduced to 0%.

Automatic disruption classification at JET: comparison of different pattern recognition techniques

In this paper, different pattern recognition techniques have been tested in order to implement an automatic tool for disruption classification in a tokamak experiment. The methods considered refer to clustering and classification techniques. In particular, the investigated clustering techniques are self-organizing maps and K-means, while the classification techniques are multi-layer perceptrons, support vector machines, and k- nearest neighbours. Training and testing data have been collected selecting suitable diagnostic signals recorded over 4 years of EFDA-JET experiments. Multi-layer perceptron classifiers exhibited the best performance in classifying mode lock, density limit/high radiated power, H-mode/L-mode transition and internal transport barrier plasma disruptions. This classification performance can be increased using multiple classifiers. In particular the outputs of five multi-layer perceptron classifiers have been combined using multiple classifier techniques in order to obtain a more robust and reliable classification tool, that is presently implemented at JET.

Geometrical Kernel Machine for Prediction and Novelty Detection of Disruptive Events in TOKAMAK Machines

This paper presents a so called Geometrical Kernel Machine used to predict disruptive events in nuclear fusion reactors. Here, the prediction problem is modeled as a two classes classification problem, and the predictor is built by using a new constructive algorithm that allows us to automatically determine both the number of neurons and the synaptic weights of a Multilayer Perceptron network with a single hidden layer. It has been demonstrated that the resulting network is able to classify any set of patterns defined in a real domain. The geometrical interpretation of the network equations allows us both to develop the predictor and to manage the so called ageing of the kernel machine. In fact, using the same kernel machine, a novelty detection system has been integrated in the predictor, increasing the overall system performance.