Establish a sound disruption precursor signal to be used for disruption avoidance

An algorithm based on Singular Value Decomposition analysis of the signals proportional to the magnetic fluctuations dB/dt collected by an array of MIrnov coils has been developed, implemented and tested both on the FTU database and in plasma operation in its real-time version. Normalized signals are used, making the analysis fairly independent on the signals amplitude and on the sensors calibration. The algorithm provides early and reliable detection of the MHD activity, real time mode numbers m,n classification and a candidate estimator for the probability of disruption based on the entropy H. Demonstration of the last as an useful ingredient to develop a working disruption precursor requires further analysis. Classification of the trajectories of the representative point of the discharge in suitable sub-spaces of the parameters can provide useful hints towards this goal. Moreover, in the perspective of experiments of disruption avoidance/mitigation the interaction between sensor-actuator need to be properly considered.