New research from National Laboratory of the North West Pacific (PNNL) Use automatic learning, data analysis and artificial intelligence to identify potential nuclear threats.
Non-prolifction analyst PNNL Benjamin Wilson On a unique opportunity to combine these data exploration and automatic learning techniques with nuclear analysis.
According to Wilson: “Prevention of nuclear proliferation requires vigilance. It involves labor, from nuclear materials to investigations that manages nuclear materials. Data analysis techniques can be used to facilitate this. ”
With the support of the National Nuclear Security Administration (NNSA), the Mathematics for Artificial Reasoning in Science (March) and the Ministry of Defense, PNNL researchers work on several projects to improve the efficiency of nuclear non-proliferation and security measures. You will find below the main theses of some of them.
Nuclear material leak detection
Nuclear reprocessing installations collect worn nuclear fuel and separate them into waste. The products are then used to produce compounds that can be transformed into new fuel for nuclear reactors. These compounds contain uranium and plutonium and can be used to make nuclear weapons. IAEA monitors nuclear installations to ensure that none of the nuclear materials is used for nuclear weapons. These are long -term regular inspections, as well as the collection of samples for more in -depth analysis.
“We could save a lot of time and labor costs if we could create a system that automatically detects anomalies from the data process data,” said Wilson.
In a study published in The International Journal of Nuclear Guarantees and Non-ProliferationWilson worked with researchers from Sandia National Laboratories to create a virtual replica of a reprocessing installation. They then formed an artificial intelligence model to detect models in the process data which represented the leak of nuclear materials. In this simulated environment, the model has shown encouraging results. “Although it is unlikely that this approach will be used in the near future, our system provides a promising start to supplement existing guarantees,” said Wilson.
Analyze texts to seek signs of nuclear weapons proliferation
PNNL data scientists have developed an automatic learning tool based on Google Bert: a language model formed on Wikipedia data for general requests. Language models allow computers to “understand” human languages - they can read texts and extract important information, including context and nuances. People can ask Bert questions, such as: “The population of Switzerland?” And get the right answer.
Although the model formed by Wikipedia is excellent to answer general questions, it lacks knowledge in the nuclear field. The team has therefore created Ajax, a help to fill this lack of knowledge.
“While Ajax is still at its beginnings, it has the potential to save analysts of many hours of work by providing both a direct response to requests and proofs of this response,” said Subramanian. The evidence is particularly intriguing for researchers, because most automatic learning models are often called “black boxes” which leave no trace of evidence of their responses, even if they are correct. Ajax aims to provide auditability by receiving documents containing evidence.
According to Subramanian: “When the field is as important as the detection of nuclear proliferation, it is essential for us to know where our information comes from”.
This development was published in the International newspaper of nuclear guarantees and non-proliferation.
Currently, IAEA analysts spend many hours reading research articles and manually analyzing data trains containing information on nuclear proliferation. The researchers hope that in the future, it will be possible to ask Ajax questions and obtain not only an answer, but also a link to the source of information. This will considerably simplify the task of analysts.
Image analysis to determine the origin of nuclear materials
Sometimes the police officers meet nuclear materials which are outside of regulatory control and of unknown origin. It is extremely important to know where the material comes from and where it was created. After all, it is always possible that the extract sample is only part of the material in illegal circulation. The forensic analysis of nuclear materials is one of the analysis tools used in this vital work.
PNNL researchers, in collaboration with the University of UTAH, Lawrence Livermore National Laboratory and Los Alamos National Laboratory, have developed an automatic learning algorithm for the medico-legal analysis of these samples. Their method uses images under an electron microscope to compare the microstructures of nuclear samples. Different materials contain subtle differences that can be detected using automatic learning.
“Imagine that the synthesis of nuclear materials was like cooking cookies,” said Elizabeth Jurrus, chief of the March initiative. “Two people can use the same recipe and end up with different appearance cookies. It's the same with nuclear materials. ”
The synthesis of these materials can be influenced by many factors, such as local humidity and the purity of the starting materials. Consequently, the nuclear materials produced in a particular company acquire a special structure – a “signature aspect” which can be seen under an electron microscope.
Research is published in the Journal of Nuclear Materials.
The researchers have created a library of images of various nuclear samples. They used automatic learning to compare the images of their library with unknown samples, and thus determine the origin of the unknowns.
This will help nuclear analysts determine the source of the equipment and direct additional research.
It will probably take some time before agencies like the AIAA integrate automatic learning techniques in their methods of detecting nuclear threats. However, such research can certainly influence this process and optimize it.
“Although we did not expect that automatic learning replaces anyone's work, we see it as a way to facilitate their work,” according to the researchers. “We can use automatic learning to identify important information so that analysts can focus on what is most important.”
