scholarly journals Predicting atmospheric optical properties for radiative transfer computations using neural networks

2020 ◽  
Author(s):  
Menno Veerman ◽  
Robert Pincus ◽  
Caspar van Leeuwen ◽  
Damian Podareanu ◽  
Robin Stoffer ◽  
...  
Keyword(s):  
Neural Networks ◽  
Optical Properties ◽  
Radiative Transfer ◽  
Grid Cell ◽  
High Accuracy ◽  
Average Error ◽  
Multilayer Perceptrons ◽  
Shallow Convection ◽  
Wide Range ◽  
The Neural Networks

<p>A fast and accurate treatment of radiation in meteorological models is essential for high quality simulations of the atmosphere. Despite our good understanding of the processes governing the transfer of radiation, full radiative transfer solvers are computationally extremely expensive. In this study, we use machine learning to accelerate the optical properties calculations of the Rapid Radiative Transfer Models for General circulation model applications - Parallel (RRTMGP). These optical properties control the absorption, scattering and emission of radiation within each grid cell. We train multiple neural networks that get as input the pressure, temperature and concentrations of water vapour and ozone of each grid cell and together predict all 224 or 256 quadrature points of each optical property. All networks are multilayer perceptrons and we test various network sizes to assess the trade-off between the accuracy of a neural network and its computational costs. We train two different sets of neural networks. The first set (generic) is trained for a wide range of atmospheric conditions, based on the profiles chosen by the Radiative Forcing Model Intercomparison Project (RFMIP). The second set (case-specific) is trained only for the range in temperature, pressure and moisture found in one large-eddy simulation based on a case with shallow convection over a vegetated surface. This case-specific set is used to explore the possible performance gains of case-specific tuning.</p><p>Most neural networks are able to predict the optical properties with high accuracy. Using a network with 2 hidden layers of 64 neurons, predicted optical depths in the longwave spectrum are highly accurate (R<sup>2 </sup>> 0.99). Similar accuracies are achieved for the other optical properties. Subsequently, we take a set of 100 atmospheric profiles and calculate profiles of longwave and shortwave radiative fluxes based on the optical properties predicted by the neural networks. Compared to fluxes based on the optical properties computed by RRTMGP, the downwelling longwave fluxes have errors within 0.5 W m<sup>-2</sup> (<1%) and an average error of -0.011 W m<sup>-2</sup> at the surface. The downwelling shortwave fluxes have an average error of -0.0013 W m<sup>-2</sup> at the surface. Using Intel’s Math Kernel Library’s (MKL) BLAS routines to accelerate matrix multiplications, our implementation of the neural networks in RRTMGP is about 4 times faster than the original optical properties calculations. It can thus be concluded that neural networks are able to emulate the calculation of optical properties with high accuracy and computational speed.</p>

scholarly journals Predicting atmospheric optical properties for radiative transfer computations using neural networks

2021 ◽  
Vol 379 (2194) ◽  
pp. 20200095
Author(s):  
Menno A. Veerman ◽  
Robert Pincus ◽  
Robin Stoffer ◽  
Caspar M. van Leeuwen ◽  
Damian Podareanu ◽  
...  
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Optical Properties ◽  
Radiative Transfer ◽  
Training Data ◽  
Machine Learning Techniques ◽  
Climate Modelling ◽  
Atmospheric Conditions ◽  
Promising Tool ◽  
The Neural Networks

The radiative transfer equations are well known, but radiation parametrizations in atmospheric models are computationally expensive. A promising tool for accelerating parametrizations is the use of machine learning techniques. In this study, we develop a machine learning-based parametrization for the gaseous optical properties by training neural networks to emulate a modern radiation parametrization (RRTMGP). To minimize computa- tional costs, we reduce the range of atmospheric conditions for which the neural networks are applicable and use machine-specific optimized BLAS functions to accelerate matrix computations. To generate training data, we use a set of randomly perturbed atmospheric profiles and calculate optical properties using RRTMGP. Predicted optical properties are highly accurate and the resulting radiative fluxes have average errors within 0.5 W m −2 compared to RRTMGP. Our neural network-based gas optics parametrization is up to four times faster than RRTMGP, depending on the size of the neural networks. We further test the trade-off between speed and accuracy by training neural networks for the narrow range of atmospheric conditions of a single large-eddy simulation, so smaller and therefore faster networks can achieve a desired accuracy. We conclude that our machine learning-based parametrization can speed-up radiative transfer computations while retaining high accuracy. This article is part of the theme issue ‘Machine learning for weather and climate modelling’.

scholarly journals A Novel Ultrasonic TOF Ranging System Using AlN Based PMUTs

Micromachines ◽  
2021 ◽  
Vol 12 (3) ◽  
pp. 284
Author(s):  
Yihsiang Chiu ◽  
Chen Wang ◽  
Dan Gong ◽  
Nan Li ◽  
Shenglin Ma ◽  
...  
Keyword(s):  
Clock Cycle ◽  
High Accuracy ◽  
Ultrasonic Waves ◽  
Oxide Semiconductor ◽  
Average Error ◽  
Cmos Process ◽  
Clock Frequency ◽  
Time Frequency ◽  
Range Finding ◽  
Wide Range

This paper presents a high-accuracy complementary metal oxide semiconductor (CMOS) driven ultrasonic ranging system based on air coupled aluminum nitride (AlN) based piezoelectric micromachined ultrasonic transducers (PMUTs) using time of flight (TOF). The mode shape and the time-frequency characteristics of PMUTs are simulated and analyzed. Two pieces of PMUTs with a frequency of 97 kHz and 96 kHz are applied. One is used to transmit and the other is used to receive ultrasonic waves. The Time to Digital Converter circuit (TDC), correlating the clock frequency with sound velocity, is utilized for range finding via TOF calculated from the system clock cycle. An application specific integrated circuit (ASIC) chip is designed and fabricated on a 0.18 μm CMOS process to acquire data from the PMUT. Compared to state of the art, the developed ranging system features a wide range and high accuracy, which allows to measure the range of 50 cm with an average error of 0.63 mm. AlN based PMUT is a promising candidate for an integrated portable ranging system.

Intelligent Regime Recognition in Upward Vertical Gas-Liquid Two Phase Flow Using Neural Network Techniques

2010 ◽  
Author(s):  
Soheil Ghanbarzadeh ◽  
Pedram Hanafizadeh ◽  
Mohammad Hassan Saidi ◽  
Ramin Bozorgmehry Boozarjomehry
Keyword(s):  
Neural Networks ◽  
Flow Regime ◽  
Two Phase Flow ◽  
Data Bank ◽  
Hydrodynamic Characteristics ◽  
Multilayer Perceptrons ◽  
Phase Flow ◽  
Two Phase ◽  
Operational Conditions ◽  
Wide Range

In order to safe design and optimize performance of some industrial systems, it’s often needed to categorize two-phase flow into different regimes. In each flow regime, flow conditions have similar geometric and hydrodynamic characteristics. Traditionally, flow regime identification was carried out by flow visualization or instrumental indicators. In this research3 kind of neural networks have been used to predict system characteristic and flow regime, and results of them were compared: radial basis function neural networks, self organized and Multilayer perceptrons (supervised) neural networks. The data bank contains experimental pressure signalfor a wide range of operational conditions in which upward two phase air/water flows pass to through a vertical pipe of 5cm diameter under adiabatic condition. Two methods of signal processing were applied to these pressure signals, one is FFT (Fast Fourier Transform) analysis and the other is PDF (Probability Density Function) joint with wavelet denoising. In this work, from signals of 15 fast response pressure transducers, 2 have been selected to be used as feed of neural networks. The results show that obtained flow regimes are in good agreement with experimental data and observation.

scholarly journals Learning the architectural features that predict functional similarity of neural networks

2020 ◽  
Author(s):  
Adam Haber ◽  
Elad Schneidman
Keyword(s):  
Neural Networks ◽  
Architectural Design ◽  
Neural Circuits ◽  
High Accuracy ◽  
Structure And Function ◽  
Functional Similarity ◽  
Architectural Features ◽  
Wide Range ◽  
And Function ◽  
Sparse Set

ABSTRACTThe mapping of the wiring diagrams of neural circuits promises to allow us to link structure and function of neural networks. Current approaches to analyzing connectomes rely mainly on graph-theoretical tools, but these may downplay the complex nonlinear dynamics of single neurons and networks, and the way networks respond to their inputs. Here, we measure the functional similarity of simulated networks of neurons, by quantifying the similitude of their spiking patterns in response to the same stimuli. We find that common graph theory metrics convey little information about the similarity of networks’ responses. Instead, we learn a functional metric between networks based on their synaptic differences, and show that it accurately predicts the similarity of novel networks, for a wide range of stimuli. We then show that a sparse set of architectural features - the sum of synaptic inputs that each neuron receives and the sum of each neuron’s synaptic outputs - predicts the functional similarity of networks of up to 100 cells, with high accuracy. We thus suggest new architectural design principles that shape the function of neural networks, which conform with experimental evidence of homeostatic mechanisms.

Discrimination of quarry blasts from earthquakes using artificial neural networks

2021 ◽  
Author(s):  
Deniz Ertuncay ◽  
Andrea De Lorenzo ◽  
Giovanni Costa
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Spectral Characteristics ◽  
High Accuracy ◽  
Anthropogenic Sources ◽  
North East ◽  
The Neural Networks ◽  
Artificial Neural ◽  
Seismic Networks ◽  
Intensive Work

<p>Seismic networks record vibrations that are captured by their stations. These vibrations can be coming from various sources, such as tectonic tremors, quarry blasts and anthropogenic sources. Separation of earthquakes from other sources may require human intervention and it can be a labor-intensive work. In case of lack of such a separation, seismic hazard may be miscalculated. Our goal is to discriminate earthquakes from quarry blasts by using artificial neural networks. To do that, we used two different databases for earthquake signals and quarry blasts. Neither of them have data from our study of interest, which is North-East of Italy. We used three channel signals from the stations as an input for the neural networks. Signals are used as both time series and their spectral characteristics and are fed to the neural networks with this information. We then separated earthquakes from quarry blasts in North-East Italy by using our algorithm. We conclude that our algorithm is able to discriminate earthquakes from quarry blasts with high accuracy and the database can be used in different regions with different earthquake and quarry blast sources in a large variety of distances.</p>

scholarly journals Magnetic Resonance Spectroscopy Frequency and Phase Correction Using Convolutional Neural Networks

2021 ◽  
Author(s):  
David J. Ma ◽  
Hortense A-M. Le ◽  
Yuming Ye ◽  
Andrew F. Laine ◽  
Jeffery A. Lieberman ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Frequency Offset ◽  
Phase Correction ◽  
Superior Performance ◽  
Multilayer Perceptrons ◽  
Resonance Spectroscopy ◽  
Phase Offset ◽  
Wide Range

We introduce DeepSPEC, a novel convolutional neural network (CNN) -based approach for frequency-and-phase correction (FPC) of MRS spectra to achieve fast and accurate FPC of single-voxel PRESS MRS and MEGA-PRESS data. In DeepSPEC, two neural networks, including one for frequency correction and one for phase correction were trained and validated using published simulated and in vivo PRESS and MEGA-PRESS MRS dataset with wide-range artificial frequency and phase offsets applied. DeepSPEC was subsequently tested and compared to the current deep learning solution - a ″vanilla″ neural network approach using multilayer perceptrons (MLP). Furthermore, random noise was added to the original simulated dataset to further investigate the model performance with noise at varied signal-to-noise (SNR) levels (i.e., 6 dB, 3 dB, and 1.5 dB). The testing showed that DeepSPEC is more robust to noise compared to the MLP-based approach due to having a smaller absolute error in both frequency and phase offset prediction. The DeepSPEC framework was capable of correcting frequency offset with 0.01±0.01 Hz and phase offset with 0.12±0.09° absolute errors on average for unseen simulated data at a high SNR (12 dB) and correcting frequency offset with 0.01±0.02 Hz and phase offset within -0.07±0.44° absolute errors on average at very low SNR (1.5 dB). Furthermore, additional frequency and phase offsets (i.e., small, moderate, large) were applied to the in vivo dataset, and DeepSPEC demonstrated better performance for FPC when compared to the MLP-based approach. Results also show DeepSPEC has superior performance than the model-based SR implementation (mSR) in FPC by having higher accuracy in a wider range of additional offsets. These results represent a proof of concept for the use of CNNs for preprocessing MRS data and demonstrate that DeepSPEC accurately predicts frequency and phase offsets at varying noise levels with state-of-the-art performance.

Optical Properties of HVEM Accelerators

1975 ◽  
Vol 33 ◽  
pp. 180-181
Author(s):  
A. Strojnik ◽  
J.W. Scholl ◽  
V. Bevc
Keyword(s):  
Optical Properties ◽  
Electron Accelerator ◽  
Systematic Investigation ◽  
Electron Source ◽  
High Brightness ◽  
The Past ◽  
Wide Range ◽  
Optical Behavior

The electron accelerator, as inserted between the electron source (injector) and the imaging column of the HVEM, is usually a strong lens and should be optimized in order to ensure high brightness over a wide range of accelerating voltages and illuminating conditions. This is especially true in the case of the STEM where the brightness directly determines the highest resolution attainable. In the past, the optical behavior of accelerators was usually determined for a particular configuration. During the development of the accelerator for the Arizona 1 MEV STEM, systematic investigation was made of the major optical properties for a variety of electrode configurations, number of stages N, accelerating voltages, 1 and 10 MEV, and a range of injection voltages ϕ0 = 1, 3, 10, 30, 100, 300 kV).

Estimating Pigment Concentrations from Spectral Images Using an Encoder‐Decoder Neural Network

2020 ◽  
Vol 64 (3) ◽  
pp. 30502-1-30502-15
Author(s):  
Kensuke Fukumoto ◽  
Norimichi Tsumura ◽  
Roy Berns
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Absorption Coefficient ◽  
Spectral Data ◽  
High Accuracy ◽  
Pigment Concentration ◽  
Scattering Coefficient ◽  
A Value ◽  
Input And Output ◽  
Pigment Concentrations

Abstract A method is proposed to estimate the concentration of pigments mixed in a painting, using the encoder‐decoder model of neural networks. The model is trained to output a value that is the same as its input, and its middle output extracts a certain feature as compressed information about the input. In this instance, the input and output are spectral data of a painting. The model is trained with pigment concentration as the middle output. A dataset containing the scattering coefficient and absorption coefficient of each of 19 pigments was used. The Kubelka‐Munk theory was applied to the coefficients to obtain many patterns of synthetic spectral data, which were used for training. The proposed method was tested using spectral images of 33 paintings, which showed that the method estimates, with high accuracy, the concentrations that have a similar spectrum of the target pigments.

Ab initio calculation on the Optical Properties of AA- Stacked two dimensional Graphene, Silicene, Germanene, and Stanene

2018 ◽  
Vol 1 (1) ◽  
pp. 46-50
Author(s):  
Rita John ◽  
Benita Merlin
Keyword(s):  
Optical Properties ◽  
Band Structure ◽  
Density Functional ◽  
Electronic Band Structure ◽  
Complex Refractive Index ◽  
Single Layer ◽  
Generalized Gradient ◽  
Electronic Band ◽  
Wide Range ◽  
Optical Region

In this study, we have analyzed the electronic band structure and optical properties of AA-stacked bilayer graphene and its 2D analogues and compared the results with single layers. The calculations have been done using Density Functional Theory with Generalized Gradient Approximation as exchange correlation potential as in CASTEP. The study on electronic band structure shows the splitting of valence and conduction bands. A band gap of 0.342eV in graphene and an infinitesimally small gap in other 2D materials are generated. Similar to a single layer, AA-stacked bilayer materials also exhibit excellent optical properties throughout the optical region from infrared to ultraviolet. Optical properties are studied along both parallel (||) and perpendicular ( ) polarization directions. The complex dielectric function (ε) and the complex refractive index (N) are calculated. The calculated values of ε and N enable us to analyze optical absorption, reflectivity, conductivity, and the electron loss function. Inferences from the study of optical properties are presented. In general the optical properties are found to be enhanced compared to its corresponding single layer. The further study brings out greater inferences towards their direct application in the optical industry through a wide range of the optical spectrum.

Squeak and rattle noise classification using radial basis function neural networks

2020 ◽  
Vol 68 (4) ◽  
pp. 283-293
Author(s):  
Oleksandr Pogorilyi ◽  
Mohammad Fard ◽  
John Davy ◽  
Mechanical and Automotive Engineering, School ◽  
Mechanical and Automotive Engineering, School ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
High Accuracy ◽  
Training Method ◽  
Vehicle Interior ◽  
Trained Classifier ◽  
Different Types ◽  
Noise Classification ◽  
Automatic Tool ◽  
Multi Class Classification

In this article, an artificial neural network is proposed to classify short audio sequences of squeak and rattle (S&R) noises. The aim of the classification is to see how accurately the trained classifier can recognize different types of S&R sounds. Having a high accuracy model that can recognize audible S&R noises could help to build an automatic tool able to identify unpleasant vehicle interior sounds in a matter of seconds from a short audio recording of the sounds. In this article, the training method of the classifier is proposed, and the results show that the trained model can identify various classes of S&R noises: simple (binary clas- sification) and complex ones (multi class classification).

Sign in / Sign up

Export Citation Format

Share Document