An artificial intelligence system capable of estimating the temperature history of equipment aided in determining the useful life of strategic components essential for the renewal of the operating license for the Angra 1 nuclear power plant, located in the municipality of Angra dos Reis, on the coast of the state of Rio de Janeiro. The operating license of the installation was set to expire on December 31, 2024, and the state-owned company Eletronuclear had requested an extension of the license for an additional 20 years from the National Nuclear Energy Commission (CNEN). At the end of November, after a technical and safety assessment, the commission authorized the extension of Angra 1’s operational lifespan until 2044.
Named the Sistema de Vida Qualificada (SVQ), the software was developed by researchers from the Process Monitoring Laboratory (LMP) of the Nuclear Engineering Program of the Alberto Luiz Coimbra Institute for Graduate Studies and Research in Engineering at the Federal University of Rio de Janeiro (COPPE-UFRJ). “The SVQ uses data collected by current sensors to determine the past temperatures at which the equipment operated during periods when these types of sensors did not exist in nuclear power plants,” describes physicist Roberto Schirru, coordinator of the LMP. “The SVQ will also be capable of predicting future temperatures.”
Operational temperature is the main determining factor for the obsolescence of equipment and electrical devices in a nuclear power plant, including motors, valves, pumps, and cables. Exposure to radiation and humidity is also a relevant parameter, but to a lesser degree.
Angra 1 began commercial operation in 1985 and has an installed capacity of 640 megawatts (MW), enough to power a city with 2 million inhabitants. When it was built, nuclear power plants were not yet equipped with temperature sensors fitted on the equipment inside the building housing the nuclear reactor, as is the case today. The Brazilian power plant only implemented these devices in 2015.
The LMP team used data obtained from the sensors between 2015 and 2020, the year the first version of the SVQ was developed, to create mathematical models of the operational temperature variations of the equipment. They supplemented this data with information on the historical ambient temperature inside the reactor building since it began operating, obtained from the Angra Integrated Computer System (SICA), which was also created by the LMP. In addition to ambient temperature, SICA monitors over 6,000 safety parameters of the plant.
The SVQ, Schirru explains, was trained using deep neural networks to correlate data from the sensors installed in the devices since 2015 with the information provided by SICA to establish patterns and, thereby, infer the temperature in the equipment during the period from 1985 to 2015—the plant’s first 30 years of operation. To validate the data, the LMP team also calculated the equipment’s temperature between 2015 and 2020 using the SVQ. They then compared the inferred values with the actual data obtained from the installed sensors.
In an article published in 2023, in the journal Progress in Nuclear Energy, researchers from UFRJ reported that the results obtained showed an average error of less than 2 degrees Celsius (ºC) in 89% of the cases studied. “With information about the operational temperature history, the engineers at Angra 1 now have a more precise reference to assess the equipment’s useful life,” emphasizes Schirru.
The equipment from a nuclear power plant is designed to have a useful life of 40 years. As detailed by electronic engineer Marcos das Candeias da Silva, of the Department of Electrical Systems and Instrumentation and Control at Eletronuclear, responsible for the operation of Angra 1, the manufacturers determine the nominal lifespan of the equipment based on the established temperature and radiation limits within which they will operate.
The inspection process of the plant boasts a radiation-assessment system developed in Brazil
“A piece of equipment certified for 40 years of use at a hypothetical ambient temperature of 50 ºC will have a longer lifespan if the actual temperature at which it operates is lower, for example, at 35 ºC or 40 ºC—and a shorter lifespan if the temperature exceeds the projected 50 ºC,” explains the engineer.
To estimate the lifespan, the technicians use the Arrhenius equation, which allows them to calculate the variation of the rate constant of a chemical reaction—those which cause the parts to wear, in this case—with temperature. The method was developed by Swedish physicist and chemist Svante August Arrhenius (1859–1927), who won the Nobel Prize in Chemistry in 1903.
Marcos das Candeias da Silva is responsible for the Environmental Qualification Program of Electrical Equipment at Angra 1. He states that the SVQ has shown that all the electrical equipment from the power plant has operated over the past 40 years at lower temperatures than those for which they were originally designed. “The system has allowed the environmental qualification of the equipment to be maintained beyond its expected nominal lifespan,” he says.
Around 1,500 pieces of equipment were evaluated, and over 100 had their lifespans extended. There are multiple benefits to extending the use of the equipment. “In addition to the financial savings, no operational downtime is required for replacements at the plant, unless they are necessary. Finally, we prevent the unnecessary disposal of materials that are still useful,” reports the engineer.
View of the Almirante Álvaro Alberto Nuclear Center, made up of the Angra 1, 2, and 3 (under construction) plantsEletronuclear
Without the SVQ, the managers at Eletronuclear would have to replace the equipment based on the certified lifespan of each part, or rely solely on the inspection and testing of the items. “Angra 1 is the only power plant in the world with an AI system supporting technical decisions on whether to replace its parts, or not,” says Candeias da Silva.
In July 2024, after 10 days of inspections, a team of specialists from the International Atomic Energy Agency (AIEA) published an operational safety assessment of Angra 1, which positively highlighted the plant’s program for managing aging and emphasized the importance of using the SVQ artificial intelligence software, developed at COPPE-UFRJ, to support this process.
The team led by Hungarian physicist Gabor Petofi, senior director of nuclear safety of AIEA, further recommended that the practices of Angra 1 be shared with the global nuclear industry. According to the World Nuclear Association, there are 437 nuclear reactors in operation, the majority of which were built in the 1970s, 1980s, and 1990s. The international standard is to license these power plants for 40 years, with subsequent renewals every 20 years.
“The SVQ is of great importance. To guarantee safety, a nuclear power plant needs to know the operational history of its equipment from the moment it became operational. Older facilities, such as Angra 1, lack this history because they did not have sensors that could provide such information. The SVQ is capable of estimating the past data with high accuracy,” says electrical engineer Cláudia Márcio do Nascimento Abreu Pereira, professor of computational methods at the Nuclear Engineering Institute (IEN) of CNEN.
The LMP team has already developed a second version of the SVQ, delivered in 2024. The technicians from Eletronuclear are currently being trained to use the new tool. According to physician Andressa Nicolau, manager of the SVQ project in the LMP, while the first model only provides temperature data up to 2020, the new version of the software will allow the engineers from Eletronuclear to retrain the algorithm with current data, collected by the installed sensors, enabling continuous reassessment of the operational conditions.
The control room of Angra 1, the first nuclear power plant in BrazilEletronuclear
The new version will also allow the prediction of future temperatures and the lifespan of equipment over the years, while automatically correcting the forecast based on the actual conditions the plant presents over time. “Engineers will also be able to set new assessment parameters and retrain the software according to new needs,” adds Nicolau.
The lifespan inspection process at Angra 1 also boasts its own radiation assessment system developed in the country. Radiation is the second most significant factor in the aging of equipment at a nuclear facility. The system uses an alanine/ESR dosimeter to measure the radiation. The dosimeter is a small box that contains alanine, one of the amino acids that living beings use to synthesize proteins. After exposure to radiation, the alanine is taken to a laboratory, where an electron spin resonance spectrometer (ESR) measures the dose of radiation received. Analog systems are employed to measure the radiation in complex radiotherapy treatments and industrial processes.
With funding from FAPESP, the team of physicist Oswaldo Baffa, from the Department of Physics at the School of Philosophy, Sciences and Languages of Ribeirão Preto of the University of São Paulo (FFCLRP-USP), has developed a system based on these devices to measure the radiation that affects the equipment at Angra 1. The device has been in operation since 2015.
According to Baffa, the alanine dosimeters present a significant advantage. “They are able to measure very low doses of radiation, less than 1 gray (Gy), up to high doses of thousands of Gy,” he explains—gray is the unit of measurement for absorbed radiation. Nuclear power plants traditionally use luminescent or thermoluminescent dosimeters, which do not operate with such a wide range of radiation levels, informs the researcher from USP.
Prior to their use by Angra 1, there was only one reference in the scientific literature about the use of alanine/ESR dosimeters in nuclear plants. It was a study from 2008 by South Korean scientists, but which was never put into practical use. In 2022, the team from the Department of Physics at USP published an article in the journal Radiation Physics and Chemistry assessing the radiation measurement in the equipment at Angra 1, performed using 85 alanine/ESR dosimeters. The authors of the study concluded that the data obtained in all the measurements show lower levels of radiation than expected for 40 years of operation, which is 400 Gy. The results were compared with the radiation levels defined in one of the plant’s safety analysis reports.
In addition to Angra 1, Eletronuclear’s Brazilian complex also boasts Angra 2, with a capacity of 1,350 MW, which began operating commercially in 2001. The most recent plant has temperature and radiation sensors. The expectation among technicians at Eletronuclear is that the SVQ and the alanine/ESR dosimeter also play an important role in its license renewal program, scheduled for 2040.
The story above was published with the title “To measure the past temperature” in issue 347 of january/2025.
Project
Advances in dosimetry, archaeological dating, and characterization of biomaterials by electron spin resonance (nº 07/06720-4); Grant Mechanism Regular Research Grant; Principal Investigator Oswaldo Baffa Filho (USP); Investment R$507,101.37.
Scientific articles
NICOLAU, A. S. et al. Deep neural networks for estimation of temperature values for thermal ageing evaluation of nuclear power plant equipment. Progress in Nuclear Energy Journal. feb. 2023.
BAFFA, O. et al. Alanine/electron spin resonance dosimetry for environmental qualification of electric equipment in a nuclear power plant. Radiation Physics and Chemistry. apr. 2022.
