A Novel Deep Learning Method for the Predictions of Current Forces on Bluff Bodies

2018 ◽  
Author(s):  
T. P. Miyanawala ◽  
Rajeev K. Jaiman
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Bluff Body ◽  
The Body ◽  
Stochastic Gradient Descent ◽  
Approximation Procedure ◽  
Bluff Bodies ◽  
Force Coefficients ◽  
Fluid Forces ◽  
Body Shapes

Unsteady separated flow behind a bluff body causes fluctuating drag and transverse forces on the body, which is of great significance in many offshore and marine engineering applications. While physical experimental and computational techniques provide valuable physics insight, they are generally time-consuming and expensive for design space exploration and flow control of such practical scenarios. We present an efficient Convolutional Neural Network (CNN) based deep-learning technique to predict the unsteady fluid forces for different bluff body shapes. The discrete convolution process with a non-linear rectification is employed to approximate the mapping between the bluff-body shape and the fluid forces. The deep neural network is fed by the Euclidean distance function as the input and the target data generated by the full-order Navier-Stokes computations for primitive bluff body shapes. The convolutional networks are iteratively trained using a stochastic gradient descent method to predict the fluid force coefficients of different geometries and the results are compared with the full-order computations. We have extended this CNN-based technique to predict the variation of force coefficients with the Reynolds number as well. Within the error threshold, the predictions based on our deep convolutional network have a speed-up nearly three orders of magnitude compared to the full-order results and consumes an insignificant fraction of computational resources. The deep CNN-based model can predict the hydrodynamic coefficients required for the well-known Lighthill’s force decomposition in almost real time which is extremely advantageous for offshore applications. Overall, the proposed CNN-based approximation procedure has a profound impact on the parametric design of bluff bodies and the feedback control of separated flows.

Data-Driven Computing With Convolutional Neural Networks for Two-Phase Flows: Application to Wave-Structure Interaction

2018 ◽  
Author(s):  
X. Mao ◽  
V. Joshi ◽  
T. P. Miyanawala ◽  
Rajeev K. Jaiman
Keyword(s):  
Neural Network ◽  
Free Surface ◽  
Bluff Body ◽  
Data Driven ◽  
Model Parameters ◽  
Wave Forces ◽  
Approximation Procedure ◽  
High Fidelity ◽  
Bluff Bodies ◽  
Fluid Forces

Fluctuating wave force on a bluff body is of great significance in many offshore and marine engineering applications. We present a Convolutional Neural Network (CNN) based data-driven computing to predict the unsteady wave forces on bluff bodies due to the free-surface wave motion. For the full-order modeling and high-fidelity data generation, the air-water interface for such wave-body problems must be captured accurately for a broad range of physical and geometric parameters. Originated from the thermodynamically consistent theories, the physically motivated Allen-Cahn phase-field method has many advantages over other interface capturing techniques such as level-set and volume-of-fluid methods. The Allen-Cahn equation is solved in the mass-conservative form by imposing a Lagrange multiplier technique. While a tremendous amount of wave-body interaction data is generated in offshore engineering via both CFD simulations and experiments, the results are generally underutilized. Design space exploration and flow control of such practical scenarios are still time-consuming and expensive. An alternative to semi-analytical modeling, CNN is a class of deep neural network for solving inverse problems which is efficient in parametric data-driven computation and can use the domain knowledge. It establishes a model with arbitrarily generated model parameters, makes predictions using the model and existing input parametric settings, and adjusts the model parameters according to the error between the predictions and existing results. The computational cost of this prediction process, compared with high-fidelity CFD simulation, is significantly reduced, which makes CNN an accessible tool in design and optimization problems. In this study, CNN-based data-driven computing is utilized to predict the wave forces on bluff bodies with different geometries and distances to the free surface. The discrete convolution process with a non-linear rectification is employed to approximate the mapping between the bluff-body shape, the distance to the free-surface and the fluid forces. The wave-induced fluid forces on bluff bodies of different shapes and submergences are predicted by the trained CNN. Finally, a convergence study is performed to identify the effective hyper-parameters of the CNN such as the convolution kernel size, the number of kernels and the learning rate. Overall, the proposed CNN-based approximation procedure has a profound impact on the parametric design of bluff bodies experiencing wave loads.

Deep Learning Techniques for Effective Prediction of Aerodynamic Properties of Elliptical Bluff Bodies

2021 ◽  
Author(s):  
W. M. U. Weerasekara ◽  
H. M. C. D. B. Gunarathna ◽  
W. A. K. P. Wanigasooriya ◽  
T. P. Miyanawala
Keyword(s):  
Neural Network ◽  
Reynolds Number ◽  
Deep Learning ◽  
Aerodynamic Force ◽  
Flow Behavior ◽  
Reynolds Numbers ◽  
Bluff Bodies ◽  
Flow Conditions ◽  
Force Coefficients ◽  
Drag And Lift

Abstract Predicting aerodynamic forces on bluff bodies remains to be a challenging task due to the unpredictable flow behavior, specifically at higher Reynolds numbers. Experimental approaches to determine aerodynamic coefficients could be costly and time consuming. In the meantime, use of numerical techniques could also require a considerable computational cost and time depending on complexity of the flow behavior. The research focusses on developing an effective deep learning technique to predict aerodynamic force coefficients acting on elliptical bluff bodies for a given aspect ratio and given flow condition. Collecting data for drag and lift coefficients of several aspect ratios for flow conditions starting from onset of vortex shredding to verge of subcritical region is conducted by an accurate full order model. The specified region will provide a transient flow behavior and thus lift coefficient will be represented in terms of root mean square value and drag coefficient in terms of a mean value. With variations in flow behavior and vortex shredding frequencies, it requires to select an appropriate turbulence model, optimum discretization of fluid domain and time step to obtain an accurate result. Flow simulations are conducted primarily using Unsteady Reynolds Averaged Navier-Stokes Equations (URANS) model and Detached Eddy Simulations (DES) model. Effectiveness in using different turbulence models for specified flow regimes are also explored in comparison to available experimental results. At lower Reynolds numbers, aerodynamic force coefficients for a specified body will only depend on Reynolds number. But after a certain specific Reynolds number, aerodynamic forces are dependent on the Mach number in addition to Reynolds number. Therefore, for higher Reynolds numbers, aerodynamic force coefficients are recorded for multiple Mach numbers with same Reynolds number and will be fed to the neural network. With the development of the machine learning and neural network modelling, many of the fields have nourished and created effective and efficient technologies to ease complex functions and activities. Our goal is to ease the complexity in the computational fluid dynamic field with a deep neural network tool created to predict drag and lift coefficient of elliptical bluff bodies for a given aspect ratio with an acceptable accuracy level. Researchers have developed deep neural network tools to predict various flow conditions and have succeeded with sufficient accuracy and a satisfying reduction of computational cost. In our proposed deep learning neural network, we have chosen to model the network with inputs as the geometry setup and the flow conditions with validated drag and lift coefficients. The model will extract the necessary flow features into filters with the convolution operation performed on the inputs. Our main directive is to create a deep learned neural network tool to predict the target values within an acceptable range of accuracy while minimizing the computation cost.

Share Market Data Prediction Strategies using Deep Learning Algorithm

2019 ◽  
Vol 13 ◽  
Author(s):  
A John. ◽  
D. Praveen Dominic ◽  
M. Adimoolam ◽  
N. M. Balamurugan
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Stock Market ◽  
Predictive Analytics ◽  
Learning Algorithm ◽  
Market Price ◽  
Stochastic Gradient ◽  
Stochastic Gradient Descent ◽  
Mining Machine ◽  
Gradient Descent Algorithm

Background:: Predictive analytics has a multiplicity of statistical schemes from predictive modelling, data mining, machine learning. It scrutinizes present and chronological data to make predictions about expectations or if not unexplained measures. Most predictive models are used for business analytics to overcome loses and profit gaining. Predictive analytics is used to exploit the pattern in old and historical data. Objective: People used to follow some strategies for predicting stock value to invest in the more profit-gaining stocks and those strategies to search the stock market prices which are incorporated in some intelligent methods and tools. Such strategies will increase the investor’s profits and also minimize their risks. So prediction plays a vital role in stock market gaining and is also a very intricate and challenging process. Method: The proposed optimized strategies are the Deep Neural Network with Stochastic Gradient for stock prediction. The Neural Network is trained using Back-propagation neural networks algorithm and stochastic gradient descent algorithm as optimal strategies. Results: The experiment is conducted for stock market price prediction using python language with the visual package. In this experiment RELIANCE.NS, TATAMOTORS.NS, and TATAGLOBAL.NS dataset are taken as input dataset and it is downloaded from National Stock Exchange site. The artificial neural network component including Deep Learning model is most effective for more than 100,000 data points to train this model. This proposed model is developed on daily prices of stock market price to understand how to build model with better performance than existing national exchange method.

Deep learning for denoising

Geophysics ◽  
2019 ◽  
Vol 84 (6) ◽  
pp. V333-V350 ◽  
Author(s):  
Siwei Yu ◽  
Jianwei Ma ◽  
Wenlong Wang
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Graphics Processing Unit ◽  
Random Noise ◽  
Stochastic Gradient Descent ◽  
Processing Unit ◽  
Noise Attenuation ◽  
Optimal Parameters ◽  
Training Set ◽  
Unknown Variance

Compared with traditional seismic noise attenuation algorithms that depend on signal models and their corresponding prior assumptions, removing noise with a deep neural network is trained based on a large training set in which the inputs are the raw data sets and the corresponding outputs are the desired clean data. After the completion of training, the deep-learning (DL) method achieves adaptive denoising with no requirements of (1) accurate modelings of the signal and noise or (2) optimal parameters tuning. We call this intelligent denoising. We have used a convolutional neural network (CNN) as the basic tool for DL. In random and linear noise attenuation, the training set is generated with artificially added noise. In the multiple attenuation step, the training set is generated with the acoustic wave equation. The stochastic gradient descent is used to solve the optimal parameters for the CNN. The runtime of DL on a graphics processing unit for denoising has the same order as the [Formula: see text]-[Formula: see text] deconvolution method. Synthetic and field results indicate the potential applications of DL in automatic attenuation of random noise (with unknown variance), linear noise, and multiples.

scholarly journals Kill Line Model Cross Flow Inline Coupled Vortex-Induced Vibration

2017 ◽  
Author(s):  
Baiheng Wu ◽  
Jorlyn Le Garrec ◽  
Dixia Fan ◽  
Michael S. Triantafyllou
Keyword(s):  
Bluff Body ◽  
Oil And Gas ◽  
Cross Flow ◽  
The Body ◽  
Bluff Bodies ◽  
Structural Vibrations ◽  
Vortex Induced Vibrations ◽  
Series Of Experiments ◽  
Central Pipe ◽  
Semi Empirical

Currents and waves cause flow-structure interaction problems in systems installed in the ocean. Particularly for bluff bodies, vortices form in the body wake, which can cause strong structural vibrations (Vortex-Induced Vibrations, VIV). The magnitude and frequency content of VIV is determined by the shape, material properties, and size of the bluff body, and the nature and velocity of the oncoming flow. Riser systems are extensively used in the ocean to drill for oil wells, or produce oil and gas from the bottom of the ocean. Risers often consist of a central pipe, surrounded by several smaller cylinders, including the kill and choke lines. We present a series of experiments involving forced in-line and cross flow motions of short rigid sections of a riser containing 6 symmetrically arranged kill and choke lines. The experiments were carried out at the MIT Towing Tank. We present a systematic database of the hydrodynamic coefficients, consisting of the forces in phase with velocity and the added mass coefficients that are also suitable to be used with semi-empirical VIV predicting codes.

Numerical and Experimental Evaluation of Lift Forces in Flows Around Surface-Mounted Bluff Bodies

2002 ◽  
Author(s):  
T. Stengel ◽  
F. Ebert ◽  
M. Fallen
Keyword(s):  
Velocity Field ◽  
Bluff Body ◽  
Recirculation Zone ◽  
Separation Zone ◽  
Laser Doppler Anemometry ◽  
Turbulence Models ◽  
Reynolds Numbers ◽  
The Body ◽  
Bluff Bodies ◽  
Large Eddy

The flow around a surface-mounted bluff body with cuboid shape is investigated. Therefore, the velocity field including the distribution of the turbulent kinetic energy is computed and compared with experimental Laser Doppler Anemometry data. Several different turbulence models, namely the standard k-ε model, the Wolfshtein two-layer k-ε model and a Large-Eddy approach are validated. Since the Large-Eddy model remains the only model representing the flow accurate, it is chosen for further investigations. The pressure distribution on the body and on the carrying surface around the body is analysed. The lift coefficients are computed for Reynolds numbers, ranging from 1.1 × 104 up to 4.4 × 104. The lengths of the separation zone above and the recirculation zone downstream the body are evaluated.

CFD Study of the Pressure Distribution on the Back of a 3-D Bluff Body (CUBE) of Air Flow in Different Reynolds Numbers

2009 ◽  
Author(s):  
Mohammad Javad Izadi ◽  
Pegah Asghari ◽  
Malihe Kamkar Delakeh
Keyword(s):  
Reynolds Number ◽  
Bluff Body ◽  
Free Stream ◽  
Fluid Flows ◽  
Reynolds Numbers ◽  
The Body ◽  
Stream Velocity ◽  
Bluff Bodies ◽  
Unsteady Forces ◽  
Flow Round

The study of flow around bluff bodies is important, and has many applications in industry. Up to now, a few numerical studies have been done in this field. In this research a turbulent unsteady flow round a cube is simulated numerically. The LES method is used to simulate the turbulent flow around the cube since this method is more accurate to model time-depended flows than other numerical methods. When the air as an ideal fluid flows over the cube, flow separate from the back of the body and unsteady vortices appears, causing a large wake behind the cube. The Near-Wake (wake close to the body) plays an important role in determining the steady and unsteady forces on the body. In this study, to see the effect of the free stream velocity on the surface pressure behind the body, the Reynolds number is varied from one to four million and the pressure on the back of the cube is calculated numerically. From the results of this study, it can be seen that as the velocity or the Reynolds number increased, the pressure on the surface behind the cube decreased, but the rate of this decrease, increased as the free stream flow velocity increased. For high free stream velocities the base pressure did not change as much and therefore the base drag coefficient stayed constant (around 1.0).

scholarly journals A Precision Agricultural Application: Manggis Fruit Classification Using Hybrid Deep Learning

2021 ◽  
Vol 35 (5) ◽  
pp. 375-381
Author(s):  
Putra Sumari ◽  
Wan Muhammad Azimuddin Wan Ahmad ◽  
Faris Hadi ◽  
Muhammad Mazlan ◽  
Nur Anis Liyana ◽  
...  
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Stochastic Gradient Descent ◽  
Traditional Methods ◽  
High Performing ◽  
The Neural Network ◽  
Different Types ◽  
Filter Size ◽  
Fully Connected ◽  
Convolution Filters

Fruits come in different variants and subspecies. While some subspecies of fruits can be easily differentiated, others may require an expertness to differentiate them. Although farmers rely on the traditional methods to identify and classify fruit types, the methods are prone to so many challenges. Training a machine to identify and classify fruit types in place of traditional methods can ensure precision fruit classification. By taking advantage of the state-of-the-art image recognition techniques, we approach fruits classification from another perspective by proposing a high performing hybrid deep learning which could ensure precision mangosteen fruit classification. This involves a proposed optimized Convolutional Neural Network (CNN) model compared to other optimized models such as Xception, VGG16, and ResNet50 using Adam, RMSprop, Adagrad, and Stochastic Gradient Descent (SGD) optimizers on specified dense layers and filters numbers. The proposed CNN model has three types of layers that make up its model, they are: 1) the convolutional layers, 2) the pooling layers, and 3) the fully connected (FC) layers. The first convolution layer uses convolution filters with a filter size of 3x3 used for initializing the neural network with some weights prior to updating to a better value for each iteration. The CNN architecture is formed from stacking these layers. Our self-acquired dataset which is composed of four different types of Malaysian mangosteen fruit, namely Manggis Hutan, Manggis Mesta, Manggis Putih and Manggis Ungu was employed for the training and testing of the proposed CNN model. The proposed CNN model achieved 94.99% classification accuracy higher than the optimized Xception model which achieved 90.62% accuracy in the second position.

scholarly journals Automated Heart Dysfunctionality Identification Based on Iris using Deep Learning

2020 ◽  
Vol 9 (5) ◽  
pp. 528-531
Keyword(s):  
Neural Network ◽  
Heart Disease ◽  
Deep Learning ◽  
Early Stage ◽  
Region Of Interest ◽  
Histogram Equalization ◽  
The Body ◽  
Iris Localization ◽  
Human Iris ◽  
The World

One of the most deadly diseases in the world is Heart Disease. The dysfunctionality of the heart at the early stage can be detected using iridology. The study of iridology describes the structure of the human iris as an observation of the condition of organs in the body. In this article, we explore the heart condition through a series of stages such as iris localization, segmentation, extraction of region of interest, histogram equalization and classification using convolutional neural network. The results are evaluated using various quality metrics such as precision, recall, f-score & accuracy.

scholarly journals PEMODELAN PREDIKSI KUAT TEKAN BETON UMUR MUDA MENGGUNAKAN H2O'S DEEP LEARNING

2020 ◽  
Vol 25 (1) ◽  
pp. 40
Author(s):  
Stefanus Santosa ◽  
Suroso Suroso ◽  
Marchus Budi Utomo ◽  
Martono Martono ◽  
Mawardi Mawardi
Keyword(s):  
Neural Network ◽  
Machine Learning ◽  
Artificial Neural Network ◽  
Compressive Strength ◽  
Deep Learning ◽  
Information Needs ◽  
Chemical Properties ◽  
Stochastic Gradient Descent ◽  
Design Data ◽  
Artificial Neural

Artificial Neural Network (ANN) is a Machine Learning (ML) algorithm which learn by itself and organize its thinking to solve problems. Although the learning process involves many hidden layers (Deep Learning) this algorithm still has weaknesses when faced with high noise data. Concrete mixture design data has a high enough noise caused by many unidentified / measurable aspects such as planning, design, manufacture of test specimens, maintenance, testing, diversity of physical and chemical properties, mixed formulas, mixed design errors, environmental conditions, and testing process. Information needs about the compressive strength of early age concrete (under 28 days) are often needed while the construction process is still ongoing. ANN has been tried to predict the compressive strength of concrete, but the results are less than optimal. This study aims to improve the ANN prediction model using an H2O’s Deep Learning based on a multi-layer feedforward artificial neural network that is trained with stochastic gradient descent using backpropagation. The H2O’s Deep Learning best model is achieved by 2 hidden layers- 50 hidden neurons and ReLU activation function with a RMSE value of 6,801. This Machine Learning model can be used as an alternative/ substitute for conventional mix designs, which are environmentally friendly, economical, and accurate. Future work with regard to the concrete industry, this model can be applied to create an intelligent Batching and Mixing Plants.

Sign in / Sign up

Export Citation Format

Share Document