Periodic Reporting for period 3 - CORTEX (Core monitoring techniques and experimental validation and demonstration)
Okres sprawozdawczy: 2020-09-01 do 2021-08-31
Being able to monitor the state of reactors while they are running at nominal conditions is extremely advantageous. The early detection of anomalies gives the possibility for the utilities to take proper actions before such problems lead to safety concerns or impact plant availability. The analysis of measured fluctuations of process parameters (primarily the neutron flux) around their mean values provides non-intrusive on-line core monitoring capabilities. These fluctuations, often referred to as noise, primarily arise either from the turbulent character of the flow in the core, from coolant boiling (in the case of two-phase systems), or from mechanical vibrations of reactor internals. Because such fluctuations carry valuable information concerning the dynamics of the reactor core, one can infer some information about the system state under certain conditions and detect possible anomalous behaviour of the system. The main goal of the project was to develop a monitoring technique, based on pre-computed reactor responses, allowing backtracking the nature and spatial distribution of possible anomalies, which give rise to the neutron flux fluctuations recorded by the in-core and ex-core instrumentation of the reactor. The project also contributes to a lowering of the CO2 footprint to the environment and to a higher availability of cheap base-load electricity to the consumers.
Detailed models for fluid-induced vibrations of reactor components were generated and a reduced order model was developed. Capabilities were built and tested to simulate reactor neutron noise arising from fuel assembly vibrations with the codes SIMULATE-3, SIMULATE-3K, PARCS, KMACS, FEMFFUSION and CORE SIM+. Neutron noise solvers using higher order approximations in energy, space and angle, were developed, using either probabilistic or deterministic methods. A statistical methodology for uncertainty and sensitivity analysis specific to neutron noise was developed.
The first measurement campaigns at AKR-2 and CROCUS were carried out successfully in 2018, demonstrating the availability of vibrating absorber and absorber of variable strength experimental setups. The data acquisition systems were furthermore qualified against industry-grade equipment. Additional measurements were carried out in 2019 and 2021 for CROCUS and in 2020 and 2021 for AKR-2, based on the feedback received from the code modellers and the experience gained from the early measurement campaigns. A very tight collaboration between the experimentalists and code developers resulted in the validation activities progressing as planned, with a better understanding of the discrepancies between calculations and measurements and in the identification of specific points to be addressed in future measurements.
A large number of simulated data, in frequency and time domains, in related scenarios, were generated to develop and test signal processing methods and machine learning algorithms and architectures. Advanced signal processing methods, based on wavelet and frequency transformations were developed for trend and noise removal and visualization purposes. Novel machine and deep learning methods were successfully applied to the data for predicting perturbation type/location. Very promising results were obtained and published. Extensions to real data were thereafter successfully carried out, with the required modifications of the methods and algorithms.
The simulation tools and signal processing/machine learning methods were then properly merged and applied to actual plant data. For that purpose, four measurement campaigns for additional neutron noise data were prepared and executed at Gösgen NPP. A collection of already available data of KWU, US PWR and VVER reactors was prepared and distributed within the project. Also, core data for steady state calculations were distributed, so that neutron noise simulations for a very large set of postulated scenarios could be executed. Based on those simulation data, the measurement data were pre-processed and fed to the developed machine learning architectures to detect anomalies, classify them, and localize them when relevant. This represented a world-premiere, thus demonstrating the capabilities of the neutron noise-based method developed in CORTEX. An assessment of the effect of uncertainties onto the simulation data and thereafter on the accuracy of the machine-learning based diagnostics was carried out, demonstrating the robustness of the method.
Eight short courses/workshops were arranged on the following topics: signal processing and noise analysis, fundamentals of reactor kinetics and theory of small space-time dependent fluctuations, reactor dynamics, advanced signal processing methods, simulation of neutron noise in power reactors, uncertainty and sensitivity analysis applied to neutron noise calculations, neutron noise measurements and their modelling. The publication records are as follows: journal papers: 15, conference papers: 35, presentations: 23, posters: 5. The following communications channels were developed and regularly updated: website, LinkedIn page, leaflet. A popular science video was also released.
The first measurement campaigns at AKR-2 and CROCUS were carried out successfully in 2018, demonstrating the availability of vibrating absorber and absorber of variable strength experimental setups. The data acquisition systems were furthermore qualified against industry-grade equipment. Additional measurements were carried out in 2019 and 2021 for CROCUS and in 2020 and 2021 for AKR-2, based on the feedback received from the code modellers and the experience gained from the early measurement campaigns. A very tight collaboration between the experimentalists and code developers resulted in the validation activities progressing as planned, with a better understanding of the discrepancies between calculations and measurements and in the identification of specific points to be addressed in future measurements.
A large number of simulated data, in frequency and time domains, in related scenarios, were generated to develop and test signal processing methods and machine learning algorithms and architectures. Advanced signal processing methods, based on wavelet and frequency transformations were developed for trend and noise removal and visualization purposes. Novel machine and deep learning methods were successfully applied to the data for predicting perturbation type/location. Very promising results were obtained and published. Extensions to real data were thereafter successfully carried out, with the required modifications of the methods and algorithms.
The simulation tools and signal processing/machine learning methods were then properly merged and applied to actual plant data. For that purpose, four measurement campaigns for additional neutron noise data were prepared and executed at Gösgen NPP. A collection of already available data of KWU, US PWR and VVER reactors was prepared and distributed within the project. Also, core data for steady state calculations were distributed, so that neutron noise simulations for a very large set of postulated scenarios could be executed. Based on those simulation data, the measurement data were pre-processed and fed to the developed machine learning architectures to detect anomalies, classify them, and localize them when relevant. This represented a world-premiere, thus demonstrating the capabilities of the neutron noise-based method developed in CORTEX. An assessment of the effect of uncertainties onto the simulation data and thereafter on the accuracy of the machine-learning based diagnostics was carried out, demonstrating the robustness of the method.
Eight short courses/workshops were arranged on the following topics: signal processing and noise analysis, fundamentals of reactor kinetics and theory of small space-time dependent fluctuations, reactor dynamics, advanced signal processing methods, simulation of neutron noise in power reactors, uncertainty and sensitivity analysis applied to neutron noise calculations, neutron noise measurements and their modelling. The publication records are as follows: journal papers: 15, conference papers: 35, presentations: 23, posters: 5. The following communications channels were developed and regularly updated: website, LinkedIn page, leaflet. A popular science video was also released.
WP1: development of high-fidelity simulation tools specifically targeted at the modelling of the neutron noise
WP2: establishement of first-of-a-kind noise specific measurements in research reactors for validating the simulation tools
WP3: applications of machine learning-based techniques for retrieving anomalies in nuclear reactors, using simulations as training data sets
WP4: demonstration of the technique on actual plant data and identification of anomalies
Economic impact
As the fleet of nuclear reactors in Europe is becoming older, operational problems will become more frequent. Concerning the availability of new units, lower plant availability is also expected in the starting phase of the renewal of the units. In addition, new plant technologies are being introduced whereas operational experience is limited for such technologies. This will also impact the availability of the units. As a consequence, the impact of core monitoring on plant availability and on the profitability of the plants as proposed in CORTEX is significant. The capabilities to automatize and correctly diagnose reactor internal anomalies in nuclear power plants plays a role in supporting the extension of the operating licenses for the existing reactors and when building and operating new reactors. Moreover, this justifies the use of nuclear power as a safe and efficient baseload power source. The proposed technology will also create new business opportunities for companies servicing the nuclear industry.
Societal impact
The increased availability of the reactor units resulting from the proposed technique allows maintaining an as large as possible fraction of electricity coming from nuclear power. With its low levelized cost compared to the other forms of electricity generation in Europe, the new method thus contributes to a lowering of the cost of electricity to the European consumers. Moreover, by making the plants more available, the technique further reduces CO2 emissions in the atmosphere. The project finally contributes to make the exploitation of nuclear reactors safer, thus leading to a better acceptance of nuclear power throughout Europe.
WP2: establishement of first-of-a-kind noise specific measurements in research reactors for validating the simulation tools
WP3: applications of machine learning-based techniques for retrieving anomalies in nuclear reactors, using simulations as training data sets
WP4: demonstration of the technique on actual plant data and identification of anomalies
Economic impact
As the fleet of nuclear reactors in Europe is becoming older, operational problems will become more frequent. Concerning the availability of new units, lower plant availability is also expected in the starting phase of the renewal of the units. In addition, new plant technologies are being introduced whereas operational experience is limited for such technologies. This will also impact the availability of the units. As a consequence, the impact of core monitoring on plant availability and on the profitability of the plants as proposed in CORTEX is significant. The capabilities to automatize and correctly diagnose reactor internal anomalies in nuclear power plants plays a role in supporting the extension of the operating licenses for the existing reactors and when building and operating new reactors. Moreover, this justifies the use of nuclear power as a safe and efficient baseload power source. The proposed technology will also create new business opportunities for companies servicing the nuclear industry.
Societal impact
The increased availability of the reactor units resulting from the proposed technique allows maintaining an as large as possible fraction of electricity coming from nuclear power. With its low levelized cost compared to the other forms of electricity generation in Europe, the new method thus contributes to a lowering of the cost of electricity to the European consumers. Moreover, by making the plants more available, the technique further reduces CO2 emissions in the atmosphere. The project finally contributes to make the exploitation of nuclear reactors safer, thus leading to a better acceptance of nuclear power throughout Europe.