scholarly journals Artificial Neural Networks as Surrogate Models for Uncertainty Quantification and Data Assimilation in 2-D/3-D Fuel Performance Studies

2020 ◽  
Vol 1 (1) ◽  
pp. 54-62
Author(s):  
Carlo Fiorina ◽  
Alessandro Scolaro ◽  
Daniel Siefman ◽  
Mathieu Hursin ◽  
Andreas Pautz
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Data Assimilation ◽  
Uncertainty Quantification ◽  
Monte Carlo Sampling ◽  
Fuel Performance ◽  
Fast Running ◽  
Use Of Data ◽  
Artificial Neural ◽  
Trained Neural Network

This paper preliminarily investigates the use of data-driven surrogates for fuel performance codes. The objective is to develop fast-running models that can be used in the frame of uncertainty quantification and data assimilation studies. In particular, data assimilation techniques based on Monte Carlo sampling often require running several thousand, or tens of thousands of calculations. In these cases, the computational requirements can quickly become prohibitive, notably for 2-D and 3-D codes. The paper analyses the capability of artificial neural networks to model the steady-state thermal-mechanics of the nuclear fuel, assuming given released fission gases, swelling, densification and creep. An optimized and trained neural network is then employed on a data assimilation case based on the end of the first ramp of the IFPE Instrumented Fuel Assemblies 432.

An Optimal Design Method for Artificial Neural Networks by Using the Design of Experiments

2007 ◽  
Vol 11 (6) ◽  
pp. 593-599 ◽  
Author(s):  
Eiichi Inohira ◽  
◽  
Hirokazu Yokoi
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Optimal Design ◽  
Design Method ◽  
Feedforward Neural Networks ◽  
Learning Rate ◽  
Design Parameters ◽  
Approximation Accuracy ◽  
Artificial Neural ◽  
Trained Neural Network

This paper presents a method to optimally design artificial neural networks with many design parameters using the Design of Experiment (DOE), whose features are efficient experiments using an orthogonal array and quantitative analysis by analysis of variance. Neural networks can approximate arbitrary nonlinear functions. The accuracy of a trained neural network at a certain number of learning cycles depends on both weights and biases and its structure and learning rate. Design methods such as trial-and-error, brute-force approaches, network construction, and pruning, cannot deal with many design parameters such as the number of elements in a layer and a learning rate. Our design method realizes efficient optimization using DOE, and obtains confidence of optimal design through statistical analysis even though trained neural networks very due to randomness in initial weights. We apply our design method three-layer and five-layer feedforward neural networks in a preliminary study and show that approximation accuracy of multilayer neural networks is increased by picking up many more parameters.

scholarly journals Data assimilation: Particle filter and artificial neural networks

2008 ◽  
Vol 135 ◽  
pp. 012073 ◽  
Author(s):  
Helaine Cristina Morais Furtado ◽  
Haroldo Fraga de Campos Velho ◽  
Elbert Einstein Nehrer Macau
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Data Assimilation ◽  
Particle Filter ◽  
Artificial Neural

scholarly journals Incorporation of artificial neural networks and data assimilation techniques into a third-generation wind–wave model for wave forecasting

2006 ◽  
Vol 8 (1) ◽  
pp. 65-76 ◽  
Author(s):  
Zhixu Zhang ◽  
Chi-Wai Li ◽  
Yok-Sheung Li ◽  
Yiquan Qi
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Data Assimilation ◽  
Wave Model ◽  
Altimeter Data ◽  
Northwestern Pacific ◽  
Third Generation ◽  
Model Output ◽  
Artificial Neural ◽  
Wave Forecasting

Although the third-generation formulation of the ocean wave model describes the wave generation, dissipation and nonlinear interaction processes explicitly, many empirical parameters exist in the model which have to be determined experimentally. With the advance in oceanographic remote-sensing techniques, information on oceanic parameters including significant wave height (SWH) can be obtained daily by satellite altimeters. The assimilation of these data into the wave model provides a way of improving the hindcasting results. However, for wave forecasting, no altimeter data exist during the forecasting period, by definition. To improve the forecasting accuracy of the wave model, Artificial Neural Networks (ANN) are introduced to mimic the errors introduced by the wave model. This is achieved by training the ANN using the wave model output as input, and the results after data assimilation as the targeted output. The trained ANN is then used as a post-processor of the output from the wave model. The proposed method has been applied in wave simulation in the northwestern Pacific Ocean. The statistical interpolation method is used to assimilate the altimeter data into the wave model output and a back-propagation ANN is used to mimic the relation between the wave model outputs with or without data assimilation. The results show that an apparent improvement in the accuracy of forecasting can be obtained.

scholarly journals Using artificial neural networks to determine the location of wind farms. Miedzna district case study

2016 ◽  
Vol 30 (1) ◽  
pp. 101-111 ◽  
Author(s):  
Krzysztof Pokonieczny
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Wind Farm ◽  
Digital Terrain Model ◽  
Wind Farms ◽  
Shuttle Radar Topography Mission ◽  
Point Of View ◽  
Artificial Neural ◽  
Trained Neural Network ◽  
Area Classification

Abstract The article concerns issues pertaining to of selecting suitable areas for wind farms. The basic assumption of the study was to take into account both criteria related to the profitability of this type of power plant, as well as public interest, which means the harmonious and not burdensome functioning of these installations for local communities. The problem of wind farm localization may be solved by the application of artificial neural networks (ANN), which are a computational intelligence element. In the conducted analysis, the possibility of wind farm localization was considered for the primary grid field with dimensions of 100 by 100 m. To prepare the training set, topographic vector data from the VMap L2 and SRTM (Shuttle Radar Topography Mission) digital terrain model were used. For each 100-meter × 100-meter grid, the input data was prepared, consisting of the factors which are important from the point of view of wind farm localization (forests, rivers, built-up areas etc.). Studies show that a properly trained neural network (using a representative number of samples and for the appropriate architecture), allows to process automation area classification in terms of placement on the wind turbines.

Sign in / Sign up

Export Citation Format

Share Document