The safe and reliable operation of nuclear power plants (NPPs) depends on advanced monitoring and predictive systems to anticipate critical failures, such as reactor TRIPs, which can cause unplanned shutdowns and safety risks. This study presents a deep learning-based approach for the early identification of reactor TRIPs and the prediction of event sequences during operational transients in Pressurized Water Reactors (PWRs). The methodology employs Long Short-Term Memory (LSTM) and Transformer neural networks, trained on ten years of historical data from the Alarm Processing System (APS) of the Angra 2 NPP. By analyzing alarm activations and binary variable transitions, the developed models are able to identify patterns in event sequences and deviations from expected operational behaviors. In this scope, LSTMs are employed for their ability to learn long-term temporal dependencies, whereas Transformers are adopted for their attention-based capacity to capture relationships across complex event patterns. The best-performing model achieved an average of 99.94% accuracy in TRIP classification and 99.82% in next-event forecasting, surpassing conventional machine-learning baselines trained on the same dataset in terms of precision and robustness. The findings demonstrate the potential of deep learning for nuclear safety applications and can be extended to other critical industrial systems requiring high reliability.

This study proposes a Master-Slave Parallel Particle Swarm Optimization (MSPPSO) approach to optimize a NuScale-like seed-blanket reactor core, addressing challenges in fuel utilization, neutron economy, and reactor sustainability. By leveraging parallel computing, the approach runs multiple SERPENT simulations simultaneously, achieving a computational speed-up of approximately 16x compared to a purely sequential execution, making large-scale parameter searches and in-depth burnup evaluations feasible within practical time limits. In this preliminary investigation, the optimization was performed only for the initial fuel cycle, as the main objective is to provide a proof-of-concept of a PSO-based approach for NuScale-like seed-blanket reactor core design optimization. Therefore, the optimization process refined fuel rod distribution, uranium and thorium enrichment levels, and burnable poison management, ensuring balanced reactivity control. The developed model successfully reproduced NuScale benchmark results, with keff deviations within 100 pcm, confirming its accuracy. The optimized core configurations sustained operation for nearly 550 days, demonstrating enhanced fuel cycle efficiency while balancing U-233 production and keff stability, validating the feasibility of thorium-based fuel cycles. Moreover, the objective optimization framework effectively minimized Pu-239 accumulation, ensuring favorable neutron economy and reduced proliferation risks. These results confirm that the MSP-PSO represents a valid proof-of-concept for the proposed optimization approach.

In the oil industry, accurate flow rate determination and control in pipelines are critical for ensuring operational efficiency. However, most conventional flowmeters require direct contact with the transported fluids, which necessitates periodic maintenance and may cause system shutdowns, thereby increasing operational costs. To address these limitations, this study proposes a minimally intrusive methodology for predicting fluid volume fractions and calculating superficial velocities aimed at flow rate determination in two-phase systems using the radioactive particle tracking technique. This approach employs a sealed radiation source (137Cs, emitting 662 keV gamma-rays) inserted into the pipeline to obtain volume fraction data. The simulated setup consisted of a polyvinyl chloride (PVC) pipe, five NaI(Tl) scintillation detectors, and the radiation source, configured for a stratified saltwater-oil flow regime. Simulations were performed using the MCNP6 Monte Carlo code. Volume fractions were predicted using deep neural networks, while time delays for superficial velocity calculations were derived from the cross-correlation function applied to the oil-phase signals. The proposed method achieved a maximum mean absolute percentage error (MAPE) of 2.24% for the oil flow rate when compared with theoretical values.

In Nuclear Power Plants (NPPs), most monitoring and diagnostic systems operate based on the principle of Detection and Response (D&R), in which operator actions are triggered only after an anomaly is detected. While effective for real-time monitoring, this approach lacks predictive capability, which is critical for anticipating the evolution of accidents and enhancing operational safety. To address this limitation, this study investigates the use of Deep Learning models for multi-horizon forecasting the temporal behavior of key state variables during normal operation and postulated accident scenarios in nuclear reactors. Two datasets were employed: the LABIHS dataset, composed of simulated time series from a Pressurized Water Reactor (PWR) under a Loss-ofCoolant Accident (LOCA), and the SICA dataset, which contains real operational data from the Angra 1 nuclear power plant. The methodology included data preprocessing and data augmentation using instrumentation noise. Four deep learning architectures were evaluated: Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), Time-series Dense Encoder (TiDE), and Neural Hierarchical Interpolation for Time Series (N-HiTS). These models were trained using a sliding window approach and evaluated across multiple forecasting horizons. Comparative results showed that TCN outperformed LSTM among the classical models, while TiDE and N-HiTS achieved the best overall accuracy and stability across all forecasting horizons. With average MAE values of 1.01 +/- 2.39 (LABIHS) and 1.45 +/- 1.33 (SICA), these findings confirm the effectiveness of modern Deep Learning architectures for predictive monitoring in nuclear power plant operations.

This work proposes a fuel assembly design based on the Seed-Blanket Unit (SBU) concept, characterized by two distinct regions: Uranium plates in the central region (seed) to enhance neutron leakage and the surrounding region (blanket) consists of mixed oxide fuel rods composed of thorium and uranium, as originally proposed by Radkowsky. This configuration leverages the advantages of enhanced safety, improved efficiency, and greater resistance to nuclear proliferation. An innovation introduced in this study is the integration of two different geometries within a single fuel assembly: plate-type elements in the seed region-favoring increased neutron leakage-and cylindrical rods in the blanket region-minimizing neutron leakage. This hybrid geometry aims to optimize the neutron economy by enhancing fissile material breeding while controlling neutron losses. The main objective of this research is to investigate the viability and neutronic performance of a novel multigeometry fuel assembly-featuring a thorium-based blanket and a metallic uranium seed-by using Particle Swarm Optimization (PSO) as a computational tool to determine optimal design parameters. The goal is to maximize the in-core breeding of U-233, thereby extending fuel cycle longevity, reducing reliance on enriched uranium, minimizing plutonium generation, and lowering the radiotoxicity and decay heat of spent fuel. Neutronic simulations were carried out using the SERPENT 2.1.30 Monte Carlo code. To couple SERPENT with the PSO algorithm, a Pythonbased interface was developed to automatically update input parameters and generate simulation files within predefined value ranges. The optimization process was successful, yielding a fuel assembly with a 25.40 % higher conversion factor compared to a reference configuration at the end of the burnup cycle. Additionally, the optimized assembly achieved the production of 2.08 kg of U-233 over a burnup period exceeding 550 effective full-power days (EFPDs).

A predição precisa da Dose de Entrada na Pele (DEP) é essencial para a garantia de qualidade e segurança em exames de radiografia de tórax. Este estudo investigou a aplicação de algoritmos de inteligência artificial para prever a DEP a partir de fatores técnicos como mAs, kV, distância foco-filme (DFF) e espessura do paciente. Foram analisados dados de 150 radiografias de tórax com dosimetria física. Modelos de aprendizado supervisionado, incluindo Rede Neural Artificial (RNA) e Random Forest Regressor (RFR), foram treinados e avaliados quanto ao desempenho preditivo. A RNA apresentou erro absoluto médio (Mean Absolute Error - MAE) de 0,009 mGy e coeficiente de determinação R² = 0,990, enquanto o modelo RFR obteve MAE de 0,020 mGy e R² = 0,827. A média da dose física medida foi de 0,284 mGy, comparada às médias preditas de 0,274 mGy pela RNA e 0,282 mGy pelo RFR. Ambos os modelos apresentaram bom desempenho preditivo, com destaque para a RNA quanto à precisão. Esses achados reforçam o potencial da inteligência artificial como ferramenta de apoio às práticas de radioproteção e otimização dos parâmetros técnicos em ambientes clínicos.

Extending the operational life of the Angra 1 Nuclear Power Plant (NPP) requires an accurate estimation of historical temperature exposure for equipment within the containment area to assess the aging and degradation of critical components. This assessment is essential for extending the plant's license by 20 years. Since Mobile Temperature Sensors (MTSs) were installed only in 2015, this study employs Deep Neural Networks (DNNs), including Deep Rectifier Neural Networks (DRNNs), Convolutional Neural Networks (CNNs), and Long ShortTerm Memory Networks (LSTMs), to infer historical temperature data before MTS deployment. The DNNs utilize time series data from Plant Fixed Sensors (PFS), monitored by the Angra 1 Integrated Computer System (SICA), as inputs. Feature importance and outlier detection methods are investigated to enhance DNN performance. Feature importance techniques, such as XGBoost, Random Forest, Principal Component Analysis (PCA), and outlier detection methods, including autoencoders, DBSCAN, and isolation forest, are evaluated. Results indicate that preprocessing significantly improves model accuracy. For instance, PCA without outlier detection combined with a CNN achieved a Mean Absolute Error (MAE) of 3.194, whereas the integration of Random Forest and XGBoost for feature importance with DBSCAN for outlier detection and a CNN reduced the MAE to 0.497.

The safe management of radioactive waste is a critical challenge in nuclear engineering, particularly in ensuring occupational exposure remains within the ALARA principle. This work proposes a methodology for 3D dose rate mapping in radioactive waste repositories by integrating MCNP6 radiation transport simulations with Deep Rectified Neural Networks (DRNNs). MCNP6 simulations were used to model a hypothetical repository, incorporating geometrical features and Cs-137 volumetric sources. The DRNN was trained on a coarse dose grid generated by MCNP, augmented with additive white noise to increase the effective size of the training dataset, improving generalization while reducing computational demand. The trained network achieved high predictive accuracy, with R2 = 0.9863 and limited error growth for an independent test set. The methodology successfully predicted and refined spatial dose distributions, capturing nonlinear radiation field behavior in geometrically complex environments. Overall, the proposed MCNP-DRNN methodology proved capable of providing high-resolution dose predictions across the repository volume, supporting dose-optimized task planning and enhanced radiological safety.

This article proposes a Light Water Small Modular Reactor (LW-SMR) based on the seed-blanket concept with thorium fuel, utilizing Artificial Intelligence (AI) for optimization. The objectives include defining SMR parameters and developing parallel multi-objective AI models using Genetic Algorithm (GA) and Particle Swarm Optimization PSO. A GA has been developed and applied to a simplified fuel assembly, while a PSO algorithm has been used together with the SERPENT code to perform precise calculations for SMR simulation. The results obtained demonstrated a great improvement when compared to classical parametric studies, demonstrating the possibility of designing a fuel assembly using artificial intelligence.

This study presents an application of an Artificial Neural Network (ANN) to detect fluids in an annular flow regime using Prompt-Gamma Neutron Activation Analysis (PGNAA). The ANN was trained using gamma-ray spectra resulting from neutron interactions with chemical elements found in fluids typical of multiphase flow in oil exploration. These spectra were generated through mathematical simulation using the MCNP6 Monte Carlo computer code to model nuclear particle transport. A 241Am-Be polyenergetic neutron source was simulated for these calculations. Several combinations of fluid fractions were developed to create a dataset used for both training and evaluation of the ANN. The ANN demonstrated robust generalization capabilities by accurately predicting the volume fraction of the three investigated fluids (saltwater, oil, and gas), even for cases not included in the training phase. The combination of ANN and PGNAA proved effective for analyzing multiphase systems, with over 92% of all showing errors of less than 5%.

This study tackles the Nuclear Accident Identification Problem (NAIP) in Nuclear Power Plants (NPPs), focusing on employing the Cuckoo Optimization Algorithm (COA). The methodology involves classifying anomalous events using data from normal operational conditions and three design basis accidents within the simulated plant state dataset of the Brazilian NPP Angra 2. The classification process is enhanced through the use of Voronoi Vectors, delineating regions of influence for each plant state and facilitating the generation of a "don't know" response. A notable feature of this approach is the integration of Principal Component Analysis (PCA) for selecting process variables, effectively reducing the dimensionality of the problem. The proposed approach achieved nearly 100 % accuracy across all classifications, even in the presence of 1 % and 2 % noise in the data.

Mining serves as the initial link in the supply chain for metals in manufacturing, and comminution − the process of reducing ore size through crushing and milling during mineral processing − is notably energy-intensive, accounting for up to 4 % of global power consumption, and its efficient management hinges on the variability within the mineral body, relying heavily on a comprehensive understanding of the ore’s strength and grinding resistance. The assessment of these parameters typically involves comminution tests such as the Geopyörä Breakage Test, the SMC test, and the Bond Ball Mill Grindability test. Additionally, analytical techniques like the X-ray Diffraction (XRD) and the Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) prove indispensable for the determination of crystalline structures and chemical composition analysis within different segments of the mineral body. This article delves into the application of Deep Neural Network (DNN) and Particle Swarm Optimization (PSO) to predict rock strength and grinding resistance parameters: Drop weight index (DWI) and Bond work index (BWI) of mineral samples using ICP-AES and XRD data in order to develop models capable of unravelling the intricate relationships among material properties, chemical and mineralogical composition, and their collective influence on ore strength and grindability. The current article utilizes a dataset from a real comminution project to evaluate the performance of the DNN and PSO models. The results obtained with both DNN and PSO models are promising and remarkably similar, with the DNN achieving a coefficient of determination (R2) of 99.5 % for DWI and 99.8 % for BWI. Similarly, the PSO model achieving an R2 of 99.4 % for DWI and 99.7 % for BWI when compared directly with reference results, surpassing previous works that employed methodologies based on conventional regression models. Thus, the results found from this study suggest that the use of PSO and DNN is promising in solving this type of problem.

The In-Core Fuel Management or Loading Pattern Optimization (LPO) is the problem of finding an optimal configuration of nuclear fuel assemblies in a reactor core. Over the years Optimization Metaheuristics based on populations have been used successfully for solving the LPO. However, for such methods, thousands of evaluations of candidate loading patterns with reactor physics codes are required. In addition, more precise calculations are also desirable, which in principle may conflict with faster evaluations. For coping with both criteria simultaneously, a code based on the Galerkin Finite Element Method (GFEM) for solving the neutron diffusion equation designed for Graphics Processing Unit (GPU) is proposed in the present work. The resulting program GFEM_GPU has been implemented in Python language with a library for GPU-accelerated computing. A comparison between GFEM_GPU and the CPU version of GFEM regarding three nuclear reactor benchmarks (IAEA-2D, BIBLIS-2D, and ZION-2D) is provided and the accuracy of the results is compared to the Nodal Expansion Method (NEM). The implementation of the GFEM solve-phase with GFEM_GPU provides iterative solutions up to 50% faster than the CPU version depending on the discretization of the problem. This approach presents several advantages, such as a high-level implementation, faster prototyping, and application to LPO problems, while keeping the high accuracy provided by fine mesh methods.

Radioactive particle tracking is a nuclear technique that tracks a sealed radioactive particle inside a volume through a mathematical location algorithm, which is widely applied in many fields such as chemical and civil engineering in hydrodynamics flows. It is possible to reconstruct the trajectory of the radioactive particle using a traditional mathematical algorithm or artificial intelligence methods. In this paper, the traditional algorithm is based on solving a minimization problem between the simulated events and a calibration dataset, and it was written using C++ language. The artificial intelligence method is represented by a deep neural network, in which hyperparameters were defined using a Python optimization library called Optuna. This paper aims to compare the potentiality of both methods to evaluate the accuracy of the radioactive particle tracking technique. This study proposes a simplified model of a concrete mixer, six NaI(Tl) detectors, and a137Cs sealed radioactive particle. The simulated measurement geometry and the dataset (3615 patterns) were developed using the MCNPX code, which is a mathematical code based on the Monte Carlo Method. The results show a mean absolute percentage error (MAPE) of 20.81%, 10.33%, and 16.84% for x, y and z coordinates, respectively, for the traditional algorithm. For the deep neural network, MAPE is 6.87%, 2.70%, and 22.79% respectively for x, y and z coordinates. In addition, an investigation is carried out to analyze whether the size of the calibration dataset influences the performance of both methods.

Low and medium activity radioactive waste, although they contain only a small fraction of the total activity produced on a world scale, represent more than 90% of the total volume of radioactive waste generated in the use of nuclear energy. Due to the potential risk to human health and the environment, all radioactive material is subject to strict control regarding its use and subsequent disposal. For this reason, all waste must be collected, treated and stored properly in specific facilities for this purpose. Therefore, methods and techniques that keep these wastes under control are of great importance for an efficient management of these wastes. Planning and constant training on waste management procedures help control these materials. One of the possible ways to carry out this planning and previous training of the activities is using computational simulators, these simulators allow a preview of the activities and a good estimate of the radiation that would be received, by the operators, in the real activity. In this context, the present work presents a methodology for a simulation system of dynamic reallocation of nuclear waste and the visualization of the generated radioactive mesh, in real time.

This work presents an approach for predicting break size and location in nuclear power plants (NPP) loss of coolant accidents (LOCA) with or without emergency shutdown of a nuclear reactor (SCRAM). To accomplish that, a multi-tasking deep neural network (MTDNN) model is proposed. The training and validation data sets were generated on a PWR simulator at the Human-Systems Interface Laboratory (LABIHS) at Instituto de Engenharia Nuclear (IEN), Brazil. For each simulation, the time behaviors of more than a hundred state variables were recorded and a random forest (RF) algorithm was applied to select the most relevant variables. Using the set of 15 variables selected, the MTDNN model was trained and applied to a test set, in which accuracy of 97.26 % for location and relative error of about 3.77 were obtained for break sizes between 0 and 1200 cm2.

The prediction of atmospheric dispersion of nuclear material is crucial to support emergency response during nuclear accidents, due to the potential severity and far-reaching consequences of such accidents. This importance is particularly evident in the context of severe accidents, like the incident that occurred at the Fukushima Daiichi Nuclear Power Plant (NPP). NPP are commonly supported by atmospheric dispersion systems (ADS), which estimate the spatial dose rates distribution, referred to as the dose rate map (DRM), by means of physical model simulations, including source term (ST), wind fields, atmospheric transport and diffusion, and dose calculations. However, it is well-known that a real accident scenario may differ from those modeled especially under severe conditions, and this fact may lead to erroneous predictions resulting in poor or catastrophic decision making. Aiming to minimize this problem, several efforts have been made to improve predictions based on field mea-surements, such as inverse problems and other methods to correct/adjust ST parameters. Such methods, how-ever, complement the physical model calculations. This paper proposes a novel approach to predict DRM at ground level, based only upon field measurements (avoiding execution of physical model simulations). The idea is to dynamically adapt a set of mobile sensors (for example monitoring drones swarm) to follow the radioactive plume. To achieve this, an active machine learning (AML) approach was developed based on Gaussian process regression (GRP) and swarm intelligence. To test the approach, computational experiments using realistic simulated data (generated on the ADS simulator of a Brazilian NPP) has been used. As result, DRMs have been reproduced with good qualitative (visual) representation and satisfactory quantitative metrics. The average correlation coefficient between predicted and real DRMs was approximately 0.72 (ranging from 0.64 to 0.97 during plume evolution), demonstrating to be a promising approach to predict DRMs.

Anais da Semana Nacional de Engenharia Nuclear e da Energia e Ciências das Radiações (978-85-5722-483-4) - Aplicação De Técnica Nuclear Na Indústria De Petróleo Com Uso De Modelagem Matemática E Inteligência Artificial