In the LHD, typically one time every three minutes a plasma that lasts for a few seconds is generated. During those few seconds, plasma temperatures, density, and other plasma parameters are diagnosed by several diagnostics devices. Because each diagnostics device records diagnostics data with an exceptionally high temporal-resolution, when combining all of that massive amount of data, this becomes big data. Analyzing this big data by using computers, we are investigating plasma performance and phenomena occurring inside a plasma. In order to effectively perform experiments, at present, similar to the generation of plasmas, we conduct analyses one time every three minutes. However, it was not simple to conduct analyses of big data in short periods of time. Here, we will discuss the development of a high-speed plasma analysis system.
In this analysis system, using experimental data obtained from numerous diagnostic devices as input data, we run a computational program on computers. However, even if we forward these experimental data as they are to the computers, we cannot run the program. An LHD plasma is not in a simple circular configuration; it takes a complicated helical shape, and numerous diagnostics devices observe the plasma from various differing directions. For that reason, the data obtained is represented based upon differing coordinates for the various diagnostics devices. In order to undertake analyses with the computers, we must convert all of the coordinates of the experimental data so that they are expressed by coordinates that are suitable for the computer calculations. Further, in order to incorporate the various physical processes of a high-temperature plasma, the computer calculations are performed in close cooperation with numerous programs. Because the various physical processes influence each other, there are complicated dependency relationships also among the programs that treat those physical processes. For example, in the case of conducting calculations regarding the transport of heat within a plasma, we input experimental data regarding the temperature profile in the plasma, and the information regarding the magnetic field that is confining the plasma is necessary.
Because the magnetic field, too, changes according to the plasma condition, in order to gather that information we must input density and temperature profiles and then conduct calculations using a different program. Because there have been such difficulties, in the past, the collection of experimental data and the computer analyses were conducted separately. It took much time to perform analyses, and the number of analyses that could be undertaken in one day was limited.
In the high-speed plasma analysis system that has now been developed, we are now able to automatically execute converting of coordinates for diagnostics data and computer calculations using the numerous necessary programs. Further, by using several computers simultaneously and making a database of analysis results and then consulting this database, we have succeeded in greatly shortening the time necessary for analysis. From this, it is now possible to undertake analysis through the computer while collecting experimental data. We are now able to provide analysis results to researchers within three minutes after each plasma shot. With the success of the high-speed plasma analysis system, we can anticipate still further acceleration of high-temperature plasma research.