The field penetration threshold of magnetic perturbations has been observed to vary non-monotonically with an increase of density in ohmic plasmas on the J-TEXT tokamak. This observation appears contradicting the previous empirical density scaling law. Disentanglement of plasma density and rotation dependences of the field penetration threshold has been carried out. It shows that the field penetration threshold depends only weakly on the density but linearly on the plasma rotation. This result is not only important for the prediction of error field tolerance in fusion devices, but also opens a question on the role of density in the forced magnetic reconnection process in magnetized plasmas.

The magnetic diagnostic of Mirnov probe arrays has been upgraded on the J-TEXT tokamak to measure the magnetohydrodynamic instabilities with higher spatial resolution and better amplitude-frequency characteristics. The upgraded Mirnov probe array contains one poloidal array with 48 probe modules and two toroidal arrays with 25 probe modules. Each probe module contains two probes which measure both the poloidal and the radial magnetic fields (B and B). To ensure that the Mirnov probe possess better amplitude-frequency characteristics, a novel kind of Mirnov probe made of low temperature co-fired ceramics is utilized. The parameters and frequency response of the probe are measured and can meet the experiment requirement. The new Mirnov arrays have been normally applied for a round of experiments, including the observation of tearing modes and their coupling as well as high frequency magnetic perturbation due to the Alfvén eigenmode. In order to extract useful information from raw signals, visualization processing methods based on singular value decomposition and cross-power spectrum are applied to decompose the coupled modes and to determine the mode number.

No abstract

Using machine learning (ML) techniques to develop disruption predictors is an effective way to avoid or mitigate the disruption in a large-scale tokamak. The recent ML-based disruption predictors have made great progress regarding accuracy, but most of them have not achieved acceptable cross-machine performance. Before we develop a cross-machine predictor, it is very important to investigate the method of developing a cross-tokamak ML-based disruption prediction model. To ascertain the elements which impact the model’s performance and achieve a deep understanding of the predictor, multiple models are trained using data from two different tokamaks, J-TEXT and HL-2A, based on an implementation of the gradient-boosted decision trees algorithm called LightGBM, which can provide detailed information about the model and input features. The predictor models are not only built and tested for performance, but also analyzed from a feature importance perspective as well as for model performance variation. The relative feature importance ranking of two tokamaks is caused by differences in disruption types between different tokamaks. The result of two models with seven inputs showed that common diagnostics is very important in building a cross-machine predictor. This provided a strategy for selecting diagnostics and shots data for developing cross-machine predictors.