Deep learning to design Z-FFR device models

Abstract

Z-Pinch fusion centre, encased by a fission envelope, serves as an individual neutron source. It can expeditiously catalyze fission reactions in 238U and 232Th nuclear materials, which are hard to use in current commercial nuclear reactors. This is the essence of the Z-Pinch Driven Fusion-Fission Hybrid Reactor (Z-FFR). The fusion core acts as a stand-alone neutron source, efficiently driving fission reactions in nuclear energy materials that are difficult to use in existing commercial nuclear reactors, such as 238U and 232Th. Then it can deliver enormous amounts of energy in a stable and controlled manner. This new type of reactor uses the fact that the fission discharges energy (∼200 Megaelectronvolts) and the neutrons’ number released is much greater than that released when the fusion discharge energy (∼17 Megaelectronvolts). Moreover, the neutrons’ number is released to achieve energy amplification and neutron amplification, significantly makes it less difficult in implementing fusion technology applications, and increases the utilisation of nuclear energy resources by more than one order of magnitude. The Z-FFR has a complex design and covers a wide range of physical processes. The use of deep learning to design the device model allows for a more closely engineered model. Deep learning allows the model design to be decomposed, the Z-FFR design data flow to be analysed and optimised, and the tedious physical process to be turned into a deep learning network layering so that we can obtain an accurate physical model. The deuterium-tritium combustion depth parameters obtained by deep learning reach around 30%, demonstrating the ability to achieve fusion self-sustained combustion.