Fast plasma shutdown without runaway electron generation by gas puffing is investigated in the JT-60U tokamak. Argon-only injection enables a fast shutdown; however, it induces runaway electron generation. Hydrogen-only injection generates much less runaway electrons; however, the shutdown time is considerably longer. Mixed injection of hydrogen and argon achieves a fast plasma shutdown without runaway generation. Argon atoms contribute to radiate the energy of plasma leading to a fast plasma shutdown, whilst hydrogen atoms contribute to increase the electron density for avoiding runaway electron generation.
The effect of neutrals on plasma resistivity due to electron-neutral collisions is studied with respect to its effect on tokamak disruptions. The resistivity of the tokamak plasma after the thermal quench is critical in determining the current quench rate, the plasma temperature, and runaway electron generation in tokamaks through the electric field, all features which are important for mitigating the damaging effect of disruptions. It is shown that the plasma resistivity during tokamak disruptions is a time-dependent parameter which may vary with disruption time scales due to the increasing fraction of neutrals. However the effect of neutrals on resistivity is found to be small for the expected neutral fraction, mostly due to power balance considerations between radiation and Ohmic heating in the plasma.
Federated learning has faced performance and network communication challenges, especially in the environments where the data is not independent and identically distributed (IID) across the clients. To address the former challenge, we introduce the federated-centralized concordance property and show that the federated single-mini-batch training approach can achieve comparable performance as the corresponding centralized training in the Non-IID environments. To deal with the latter, we present the federated multi-mini-batch approach and illustrate that it can establish a trade-off between the performance and communication efficiency and outperforms federated averaging in the Non-IID settings.