Improved training framework in a neural network model for disruption prediction and its application on EXL-50

A neural network model with a classical annotation method has been used on the EXL-50tokamak to predict impending disruption.However,the results revealed issues of overfitting and overconfidence in predictions caused by inaccurate labeling.To mitigate these issues,an improved training framework has been proposed.In this approach,soft labels from previous training serve as teachers to supervise the further learning process;this has lead to a significant improvement in predictive model performance.Notably,this enhancement is primarily attributed to the coupling effect of the soft labels and correction mechanism.This improved training framework introduces an instance-specific label smoothing method,which reflects a more nuanced model assessment on the likelihood of a disruption.It presents a possible solution to effectively address the challenges associated with accurate labeling across different machines.

Characteristics of the SOL ion-to-electron temperature ratio on the J-TEXT tokamak with different plasma configurations

Accurate measurement of the average plasma parameters in the edge region,including the temperature and density of electrons and ions,is critical for understanding the characteristics of the scrape-off layer(SOL) and divertor plasma transport in magnetically confined fusion research.On the J-TEXT tokamak,a multi-channel retarding field analyzer(RFA) probe has been developed to study average plasma parameters in the edge region under various poloidal divertor and island divertor configurations.The edge radial profile of the ion-to-electron temperature ratio,τ_(i/e),has been determined,which gradually decreases as the SOL ion self-collisionality,v_(SOL)*,increases.This is broadly consistent with what has been observed previously from various tokamak experiments.However,the comparison of experimental results under different configurations shows that in the poloidal divertor configuration,even under the same v_(SOL)*,τ_(i/e) in the SOL region becomes smaller as the distance from the X-point to the target plate increases.In the island divertor configuration,τ_(i/e) near the O-point is higher than that near the X-point at the same v_(SOL)*,and both are higher than those in the limiter configuration.These results suggest that the magnetic configuration plays a critical role in the energy distributions between electrons and ions at the plasma boundary.

Towards advanced divertor configurations on the J-TEXT tokamak

Developing advanced magnetic divertor configurations to address the coupling of heat and particle exhaust with impurity control is one of the major challenges currently constraining the further development of fusion research.It has therefore become the focus of extensive attention in recent years.In J-TEXT,several new divertor configurations,including the high-field-side single-null poloidal divertor and the island divertor,as well as their associated fundamental edge divertor plasma physics,have recently been investigated.The purpose of this paper is to briefly summarize the latest progress and achievements in this relevant research field on J-TEXT from the past few years.