Research progress of nuclear Plasma becomes faster thanks to the artificial intelligence (AI)

The discipline of artificial intelligence knowledge (Artificial Intelligence, AI) still be warm discussions and debates among dibanyak. One of the most feared is if AI is used for the purpose of weapons in the military, see Slaughterbots, the short Film is terrible About Autonomous Robots capable of Slaughtering human beings. But AI can be applied in many fields, especially complex problem solve any that can't be resolved in a short time. The AI itself defined as intelligent machines and the branch of computer science that aims to design and realize the AI. Textbooks define AI as "the study and design of intelligent agents," where an intelligent agent's intent with is a system that looked at their surroundings and take actions that maximize the chances of success (probability). According to its creator John McCarthy in 1956, AI is defined as "the science and engineering to design intelligent machines [1]."

Figure 1. Nuclear fusion reactors misbehaving (left) (source: https://phys.org/news/2017-12-artificial-intelligence-efficient-fusion-reactions.html)

One of the many applications of AI, in this article will be discussed about the utilization of AI in the field of plasma physics. Fusion energy research, scientists must study predicts major disruptions that can stop the fusion reactions and also the potential of producing fusion devices can damage the doughnut-shaped tokamak called. This research is very important before nuclear fusion reactors used for commercial purposes. Treatment that can predict exactly when the onset of the disorder that causes a sudden loss of control of the heat (plasma charge that can trigger the reaction of fuel) will be very important. This prediction is required in taking measures to avoid or mitigate events that can impact of large-scale disasters.

On December 14, 2017, a collaboration of scientists from Princeton Plasma Physics Lab (Princeton Plasma Physics Laboratory, PPPL), Department of energy (U.S. Department of Energy, DOE), and Princeton University are using artificial intelligence in improve the prediction of interference in the process of nuclear fusion. Researchers led by William Tang who was a physicist and Professor PPPL and Professor at Princeton University, is developing the program code for the prediction research reactor ITER (International Thermonuclear Experimental Reactor ). ITER is an international nuclear fusion research project which is located in the South of France.

Read also the article titled Project ITER nuclear fusion Reactor for research has reached the half way

The scientists in the research of using the software (software) code Fusion Recurrent Neural Network (FRNN). FRNN form of "learning in depth (deep learning)" which is a newer and more powerful than the software for the machine learning of modern applied artificial intelligence applications. According to Tang, "the learning algorithms in depth (deep learning) is a new path that is used to predict the disturbance on plasma systems and also now with this ability can handle data that are multi-dimensional in nature".

FRNN is a deep learning architecture that has been proven as the best way to analyze the data with sequential pattern recognition algorithm remotely (not directly dealing with the plasma). Member of the research team and the PPPL machine learning (machine learning) Princeton University is the first to systematically apply a deep learning approaches to the problem of forecasting disruptions in tokamak fusion Plasmas.

With the use of deep learning as a prediction model applied to plasma physics, from the obtained research results that the ability to predict the disorders become more accurate than previous methods. By using images from a large database on the facilities of the Joint European Torus (JET) located in the United Kingdom (United Kingdom) which is the largest and most powerful tokamak in research has been able to significantly improve the prediction of disorders and reduce the alarm warning. FRNN will be applied to the project if ITER were completed that later will be used to memprediski exactly up to 95% when there is interference, along with it can give an alarm warning only 3% when not happening ganggaun.

The most important thing in the application of the FRNN on nuclear fusion reactors according to Alexey Svyatkovskiy (big data Researchers from Princeton University) is "an artificial neural network Training in depth (data input and many many iterations) is a task intensive computing that requires the involvement of a high-performance hardware (hardware) used computers ". He further explained that "since that's a big part of what we do is develop and distribute new algorithm dibanyak processors to achieve parallel computing systems become very efficient. The computing processes will deal with the increase in problems taken from the data base of the JET tokamak disruptions more [2] [3] [4].

References:

  1. Phys Org, Artificial Intelligence (AI) (https://phys.org/tags/artificial+intelligence/) accessed on 2 January 2018
  2. Princeton Plasma Physics Laboratory. 2017. "Artificial intelligence helps accelerate progress toward efficient fusion reactions". Phys Org, December 14, 2017 (https://phys.org/news/2017-12-artificial-intelligence-efficient-fusion-reactions.html) accessed on 2 January 2018
  3. _______________________________________. 2017. "Machine learning technique offers insight into the plasma behavior". Phys Org, 13 July 2017 (https://phys.org/news/2017-07-machine-technique-insight-plasma-behavior.html) accessed on 2 January 2018
  4. Parsons, Mathew s. 2017. "Interpretation of machine-learning-based models for control plasma disruption". IOPscience, Plasma Physics and Controlled Fusion No.8 Vol 59, June 5, 2017
Artikel Berhubungan:

Sponsor Warstek.com:

Leave a Reply

Your email address will not be published. Required fields are marked *