The baking stage of the vacuum chamber has ended, and we have achieved a very good vacuum state. Recently, the new generation artificial sun "China Circulation III" launched a new round of physics experiments, and the digital twin system that was first put into use in the experiment has attracted attention. On November 8th, a reporter from Science and Technology Daily exclusively interviewed Liu Xiaolong, a key member of the R&D team and a researcher at the Southwest Institute of Physics of China National Nuclear Corporation. He stated that with the precise monitoring of the digital twin system, significant results have been achieved in the vacuum chamber baking process, and the vacuum degree of the plasma operating environment has been increased by about an order of magnitude, from 10-4 Pa to 2 × 10-5 Pa. China Circulation III achieved high confinement mode operation for the first time under a plasma current of 1 megaampere on August 25, 2023. In the new round of experiments, it will challenge high constraint mode operation at 1.6 megaamperes of current and move towards the goal of plasma temperature exceeding 100 million degrees Celsius. In order to achieve temperatures unimaginable to humans, plasma particles in the microscopic realm move at high speeds. If atmospheric molecules collide with the vacuum chamber at this time, it will cause the particles to lose kinetic energy Liu Xiaolong explained that only an environment close to vacuum can provide guarantees for the high-speed movement of plasma. The purpose of vacuum chamber baking is to obtain an extreme vacuum environment Liu Xiaolong said that the entire baking process heats up by 5 degrees per hour, and the suspended impurities in the baking are removed by a vacuum pump until the vacuum state meets the operating requirements of the magnetic confinement fusion device. The more extreme the conditions, the higher the requirements for equipment safety, and the safe operation during the experimental preparation phase is crucial. Whether the equipment can withstand high temperature baking, such as metal thermal expansion outside the safe range, and whether local stress will be excessively concentrated, all need to be closely monitored, "said Liu Xiaolong. In the past, sensors were used to monitor anomalies and make adjustments based on data, but the distribution of temperature measurement points was very limited, making it difficult to conduct a" carpet like "investigation. If there were omissions, even minor damage to the equipment structure would cause irreversible losses to the entire experiment. In the pursuit of extreme vacuum, comprehensive monitoring is the foundation for ensuring safety. In terms of accuracy and computational speed, artificial intelligence technologies such as neural network algorithms have been difficult to meet the requirements of scientific heavy tools in the past, "said Liu Xiaolong. However, in recent years, machine learning, deep learning, and other fields have continued to mature, and it is expected to" twin "reproduce different experimental processes. The first digital twin system put into use is to construct a digital model in virtual space that is exactly the same as a physical entity, achieving real-time and accurate monitoring of the vacuum chamber baking process in all directions. At the beginning of model building, the most difficult problem the team encountered was insufficient data. The baking process only runs for about 15 days a year, which is a drop in the bucket for the data sample size required for machine learning, "Liu Xiaolong admitted, having only a small amount of real data on hand. With a solid grasp of simulation technology, the team designed a simulation model that generates a large number of virtual datasets based on a small amount of real data, and through complex processing such as data cleaning, calibration, and model correction, obtains a sufficient amount of high-precision virtual data, solving the "grass" problem in model training. After we have the data, let's go to 'Twin' Liu Xiaolong introduced that relying on advanced intelligent algorithms, the model created by the team can deduce the rules of temperature at each point, and then simulate the temperature distribution of the entire vacuum chamber, making virtual and real highly synchronized. The visualization achieved through 3D rendering also makes the experimental process easier to control. This way, we can achieve comprehensive and precise control over the vacuum chamber Liu Xiaolong said. The temperature twin is just a small step taken by the team towards an artificial intelligence system. Liu Xiaolong told reporters that simulations of different parameters such as stress and displacement will be carried out in the future, changing the previous situation of only obtaining "point position" data through real-time detection by sensors, and obtaining "infinite" signals through digital twins, providing rich data and information for the next theoretical research. When it comes to the future, Liu Xiaolong said that the "twin" of vacuum chamber temperature is just a "small trial knife". The goal of China Circulation III is to thoroughly study controlled nuclear fusion reactions. To help achieve this goal, digital twins can recreate a virtual fusion reactor that corresponds to the operating fusion reactor. (New Society)