Momentum Method Research Articles

Prospective scenarios analysis Impact on demand for oil and its derivatives after the COVID-19 pandemic

Goal: This paper aims to illustrate one of the strategies found in the literature to deal with uncertainties: prospecting scenarios in the demand for oil derivatives after the COVID-19 pandemic (identifying the desirable scenario, the undesirable and the trend, at a national level in the Brazilian market) and proposing a corresponding strategy plan for each prospection in order to minimize the resumption impacts.
 Design / Methodology / Approach: This study was carried out with relevant indicators of the oil and gas industry (downstream). Commercialization data (from 2017 to 2020) were compared in order to verify the pandemic impacts. After collecting the data, we used the Momentum approach to prospect the scenarios, suggesting a strategic action plan for each prospection.
 Results: The results reveal that, regardless the post-pandemic scenario that we will encounter, the industries belonging to the oil and gas sector will have to define strategies and implement solid action plans to remain competitive in the market. It was also possible to note that the action plan for the desirable and trend scenario has synergy.
 Limitations of the investigation: The results obtained in this research refer only to the Brazilian scenario, more specifically to the downstream sector of the oil and gas industry and may differ from other sectors or regions.
 Practical implications: This research provides ways for oil and gas companies to be prepared for the post-pandemic future, thus remaining competitive in the market. In addition, as a result of applying the Momentum method, it is possible to see how the sector's indicators relate to each other through the cross-impact matrix.
 Originality / Value: This is an unprecedented study that allows the downstream oil and gas companies to be better prepared to the post-COVID-19 scenario. It adds value not only to the oil and gas industry but also to other industries, since it is possible to replicate the method presented here in other contexts with different databases.

Underwater Wireless Sensor Networks: An Energy-Efficient Clustering Routing Protocol Based on Data Fusion and Genetic Algorithms

Due to the limited battery energy of underwater wireless sensor nodes and the difficulty in replacing or recharging the battery underwater, it is of great significance to improve the energy efficiency of underwater wireless sensor networks (UWSNs). We propose a novel energy-efficient clustering routing protocol based on data fusion and genetic algorithms (GAs) for UWSNs. In the clustering routing protocol, the cluster head node (CHN) gathers the data from cluster member nodes (CMNs), aggregates the data through an improved back propagation neural network (BPNN), and transmits the aggregated data to a sink node (SN) through a multi-hop scheme. The effective multi-hop transmission path between the CHN and the SN is determined through the enhanced GA, thereby improving transmission efficiency and reducing energy consumption. This paper presents the GA based on a specific encoding scheme, a particular crossover operation, and an enhanced mutation operation. Additionally, the BPNN employed for data fusion is improved by adopting an optimized momentum method, which can reduce energy consumption through the elimination of data redundancy and the decrease of the amount of transferred data. Moreover, we introduce an optimized CHN selecting scheme considering residual energy and positions of nodes. The experiments demonstrate that our proposed protocol outperforms its competitors in terms of the energy expenditure, the network lifespan, and the packet loss rate.

Variational posterior approximation using stochastic gradient ascent with adaptive stepsize

Scalable algorithms of variational posterior approximation allow Bayesian nonparametrics such as Dirichlet process mixture to scale up to larger dataset at fractional cost. Recent algorithms, notably the stochastic variational inference performs local learning from minibatch. The main problem with stochastic variational inference is that it relies on closed form solution. Stochastic gradient ascent is a modern approach to machine learning and is widely deployed in the training of deep neural networks. In this work, we explore using stochastic gradient ascent as a fast algorithm for the posterior approximation of Dirichlet process mixture. However, stochastic gradient ascent alone is not optimal for learning. In order to achieve both speed and performance, we turn our focus to stepsize optimization in stochastic gradient ascent. As as intermediate approach, we first optimize stepsize using the momentum method. Finally, we introduce Fisher information to allow adaptive stepsize in our posterior approximation. In the experiments, we justify that our approach using stochastic gradient ascent do not sacrifice performance for speed when compared to closed form coordinate ascent learning on these datasets. Lastly, our approach is also compatible with deep ConvNet features as well as scalable to large class datasets such as Caltech256 and SUN397.

The turbulent wave boundary layer under a free stream circular orbital motion

This paper presents a simple model of the turbulent wave boundary layer for the case where the orbital motion formed by the waves just outside the boundary layer is circular. The analysis is based on the integral momentum method. It turns out that the wall friction under such a flow is larger than the similar 2D-case. It is demonstrated that even though the motion is quite different, the wall friction predictions are not significantly different from the 2D-case, where the orbital motion follows a straight line: the bed friction increases by 15–40% for the circular orbit case, depending on roughness and orbital radius. The analysis includes the case of a weak current superposed the waves. For this case, the flow resistance for the current increases by around 30–55% for a circular orbit compared to a straight motion back and forth.

Characterizing and mitigating ice-induced vibration of monopile offshore wind turbines

Offshore wind turbines (OWTs) are vulnerable to ice-induced vibrations (IIV) when they are located in ice-prone regions with drifting sea ice. The ice-induced frequency lock-in (FLI) phenomenon will cause excessive vibration and fatigue damage of OWTs. The present study characterizes the ice-induced FLI of OWTs under various ice conditions through coupled analysis and mitigates the significant FLI response using a three-dimensional pendulum tuned mass damper (3D-PTMD). Based on the National Renewable Energy Laboratory (NREL) 5 MW baseline monopile OWT model, a numerical model of the OWT with a 3D-PTMD is established wherein the interaction between the blades and the tower is modeled and soil effect is considered. Aerodynamic loading is computed using the Blade Element Momentum (BEM) method where the Prandtl's tip loss factor and the Glauert correction are considered. A coupled simulation program based on Määttänen–Blenkarn model is developed to simulate ice-structure interactions. Different ice velocities, thicknesses, and attack angles in the Bohai Sea are chosen to study the ice-structure interactions and characterize the FLI. Research results show that the FLI of the monopile OWT occurs when the ice velocity is between 0.01 m/s to 0.06 m/s and the ice thickness is 7 cm, and 0.03 m/s to 0.09 m/s when the ice thickness is 32 cm. When the FLI happens, the 3D-PTMD has significant effectiveness in mitigating the root mean square and peak response of the nacelle/tower and the foundation in fore-aft and side-side directions. The present study enhances the understanding of ice-induced FLI and provides a feasible solution for vibration control of OWTs in ice-prone areas.

Momentum Acceleration in the Individual Convergence of Nonsmooth Convex Optimization With Constraints.

Momentum technique has recently emerged as an effective strategy in accelerating convergence of gradient descent (GD) methods and exhibits improved performance in deep learning as well as regularized learning. Typical momentum examples include Nesterov's accelerated gradient (NAG) and heavy-ball (HB) methods. However, so far, almost all the acceleration analyses are only limited to NAG, and a few investigations about the acceleration of HB are reported. In this article, we address the convergence about the last iterate of HB in nonsmooth optimizations with constraints, which we name individual convergence. This question is significant in machine learning, where the constraints are required to impose on the learning structure and the individual output is needed to effectively guarantee this structure while keeping an optimal rate of convergence. Specifically, we prove that HB achieves an individual convergence rate of O(1/√t) , where t is the number of iterations. This indicates that both of the two momentum methods can accelerate the individual convergence of basic GD to be optimal. Even for the convergence of averaged iterates, our result avoids the disadvantages of the previous work in restricting the optimization problem to be unconstrained as well as limiting the performed number of iterations to be predefined. The novelty of convergence analysis presented in this article provides a clear understanding of how the HB momentum can accelerate the individual convergence and reveals more insights about the similarities and differences in getting the averaging and individual convergence rates. The derived optimal individual convergence is extended to regularized and stochastic settings, in which an individual solution can be produced by the projection-based operation. In contrast to the averaged output, the sparsity can be reduced remarkably without sacrificing the theoretical optimal rates. Several real experiments demonstrate the performance of HB momentum strategy.

An Artificial intelligence approach for predicting compressive strength of eco-friendly concrete containing waste tire rubber

Using waste tire rubber as a aggregate replacement in the production of concrete can be considered as an effective way for environment and economies. This study presents an approach based on a prediction model using Artificial Neural Networks (ANN) to predict compressive strength of eco-friendly concrete containing waste tire rubber (RC). A data set with nine influencing features including water, cement, supplementary cementitious materials, coarse aggregate, coarse rubber aggregate, fine aggregate, fine rubber aggregate, superplasticizer, age using for training and validating models have been collected from the literature. The output was compressive strength of RC. The combination of root mean square propagation and stochastic gradient descent with momentum method is employed to train the ANN. Using various validation criteria such as coefficient of determination (R2), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE), the ANN model was validated and compared with two machine learning (ML) techniques Random Forest (RF) and Multilayer Perceptron (MLP). A Sensitivity analysis also was carried out to validate the robustness and stability of these models. The experimental results showed that the ANN model outperformed in comparing with other models and therefore it can be used as a suitable approach to predict compressive strength of eco-friendly rubber concrete.

Numerical analysis of unsteady aerodynamic performance of floating offshore wind turbine under platform surge and pitch motions

The aerodynamic performance of floating offshore wind turbines is extremely complex due to the motions of the floating platform. The surge and pitch motions are the most influential motions among the six degrees-of-freedom motions. In view of this, the aim of this study is to investigate the unsteady aerodynamic characteristics of a floating offshore wind turbine under single (surge or pitch) and combined motions using computational fluid dynamics simulations, In addition, the coupling technique of dynamic mesh and sliding mesh is employed, as well as the unsteady Reynolds averaged Navier-Stokes method. The numerical simulation method in this paper is first validated by comparison to the results of the blade element momentum method and the vortex method. Then, the aerodynamic characteristics of the floating offshore wind turbine under harmonic platform motions with different periods and amplitudes are investigated. The results show that the increase of amplitude and frequency aggravate the fluctuation of the overall aerodynamic performance of the wind turbine. In addition, the combined surge-pitch motion reduces the average power generation indicating that complex platform motions adversely affect the power generation of floating offshore wind turbines.

Data classification based on fractional order gradient descent with momentum for RBF neural network

ABSTRACT The weight-updating methods have played an important role in improving the performance of neural networks. To ameliorate the oscillating phenomenon in training radial basis function (RBF) neural network, a fractional order gradient descent with momentum method for updating the weights of RBF neural network (FOGDM-RBF) is proposed for data classification. Its convergence is proved. In order to speed up the convergence process, an adaptive learning rate is used to adjust the training process. The Iris data set and MNIST data set are used to test the proposed algorithm. The results verify the theoretical results of the proposed algorithm such as its monotonicity and convergence. Some non-parametric statistical tests such as Friedman test and Quade test are taken for the comparison of the proposed algorithm with other algorithms. The influence of fractional order, learning rate and batch size is analysed and compared. Error analysis shows that the algorithm can effectively accelerate the convergence speed of gradient descent method and improve its performance with high accuracy and validity.

An engineering model for the induction of crosswind kite power systems

The present paper introduces a new, physically consistent definition of effective induction that should be used in engineering models for power kite performance that use aerodynamic coefficients for the wing. It is argued that in such cases it is physically inconsistent to use disc-based induction models – like momentum models – and thus a new, physically consistent induction model using vortex theory methodology is derived. Simulation results using the new induction model are compared to the previously often used momentum method and Actuator Line (AL) CFD simulations. The comparison shows that the new vortex based model is in much better agreement with the AL results than the momentum method. The new model is as computationally light as the momentum induction method.

Performance of a wind turbine blade in sandstorms using a CFD-BEM based neural network

In arid regions, such as the North African desert, sandstorms impose considerable restrictions on horizontal axis wind turbines (HAWTs), which have not been thoroughly investigated. This paper examines the effects of debris flow on the power generation of the HAWT. Computational Fluid Dynamics (CFD) models were established and validated to provide novel insight into the effects of debris on the aerodynamic characteristics of NACA 63415. To account for the change in the chord length and Reynolds number along the span of the blade and the 3D flow patterns, the power curves for a wind turbine were obtained using the Blade Element Momentum (BEM) method. We present a novel coupled application of the neural network, CFD, and BEM to investigate the erosion rates of the blade due to different sandstorm conditions. The proposed model can be scaled and developed to assist in monitoring and prediction of HAWT blade conditions. This work shows that HAWT performance can be significantly diminished due to the aerodynamic losses under sandstorm conditions. The power generated under debris flow conditions can decrease from 10 to 30% compared to clean conditions.

A Computational Fluid Dynamics Investigation on the Axial Induction Factor of a Small Horizontal Axis Wind Turbine

Abstract The evolution of wind and hydrokinetic turbines stimulated the development of several tools to evaluate and to predict horizontal axis rotor behavior. From this perspective, the blade element momentum methods stand out as one of the most common approaches due to its reliability and computing speed. In the classical blade element momentum, the axial induction factor is a crucial variable to compute correctly the turbine parameters. Usually, the axial induction is determined by an interactive process that balances the forces at blade sections with momentum equations. The forces are computed based on the airfoil polars evaluated at each blade section with local inlet velocity. This procedure assumes that the swirl terms are linearized, where the lateral pressure forces is neglected. In order to evaluate these tri-dimensional effects on the blade element momentum method, the present work introduces a different methodology to determine the axial induction factor employing computational fluid dynamics simulations. The method was applied for a full-scale horizontal axis rotor with three blades and 1 m of diameter, with wind tunnel experiments for validation. The axial induction factor obtained with the new technique was compared to the classical blade element momentum method. The results show axial induction factor variations along the radial and axial coordinates. An analogy with Glauert power coefficient limit was made, finding a specific limit curve for the tested turbine, and, moreover, a correlation between turbine firing speed and induction factor.

INVESTIGATION OF THE BLOOD FLOW IN DEFORMABLE VESSELS USING STABILIZED FINITE ELEMENT METHOD

The study and simulation of blood flow in the cardiovascular system have many applications such as pathologies studies, surgical planning, and design of medical devices. Several works within this area consider the problem using a rigid wall assumption, while blood velocity and pressure in large arteries are greatly influenced by vessel wall dynamics. The Coupled Momentum Method (CMM) is based on a strong coupling of degrees-of-freedom of the fluid and the solid domains. This coupling is made by considering that the deformation of the wall, in a variational level, become a boundary condition for the fluid domain. As an advantage, description of motion (Eulerian) may be kept the same and a fixed mesh. In this study, blood is considered a Newtonian fluid and the wall a thin-walled linear elastic. This work is focused on using fluid-structure interaction in 3D geometries to evaluate the blood and vessel dynamics in contrast to the rigid wall formulation what is considered only a fluid dynamic problem. For realistic and physiological parameters, CMM has demonstrated to be a good alternative for the simulation of blood flow in large arteries by virtue of representing more realistic phenomena than a rigid wall model. The results were obtained using FEniCS and Python, and are in agreement with the theoretical and numerical solutions from the literature.

OPTIMASI CONJUGATE GRADIENT PADA BACKPROPAGATION NEURAL NETWORK UNTUK PREDIKSI HASIL TANGKAP IKAN

The need for fish catch by a company or fisherman in Rembang Regency affects market process and also welfare. The catch made by the fishermen is not on target, due to the weather and type of fishing gear. An accurate method is needed in making predictions and a correlation between catch and weather so that fisherman can get maximum predictions results, so that price adjustment can be made. The research was conducted using an experimental method, to determine the accuracy of the effect of the Conjugate Gradient on the Back Propagation Neural Network in obtaining the best value. Based on the results of the Cycle training test with the Conjugate Gradient Backpropagation Neural Network method, the smallest average value is obtained at the 400th Epoch compared to the Epoch Gradient Descent With Momentum method at Epoch 800.Thus it is proven that using the Conjugate Gradient Backpropagation Neural Network method is better with an average value of- MSE average 0.2223 in three stages of testing Training Cycle, Learning Rate and Hidden Layer.

Ship Traffic Flow Prediction Based on Fractional Order Gradient Descent with Momentum for RBF Neural Network

To prevent the maritime traffic accidents and make scientific decision, scientific and accurate prediction of the traffic flow is useful, which has often been made by neural network. The weight updating methods have played an important role in improving the performance of neural networks. To ameliorate the oscillating phenomenon in training radial basis function (RBF) neural network, a fractional order gradient descent (GD) with momentum method for updating the weights of RBF neural network (FOGDM-RBF) is proposed. Its convergence is proved. The new algorithm is used to predict vessel traffic flow at Xiamen Port. It performs stable and converges to zero as the iteration increases. The results verify the theoretical results of the proposed algorithm such as its monotonicity and convergence. The descending curve of error values by fractional order GDM is smoother than the GD and GDM method. Error analysis shows that the algorithm can effectively accelerate the convergence speed of the GD method and improve its performance with high accuracy and validity. The influence of fractional order, number of hidden layer neurons, tide peak hours, and ship size is analyzed and compared. 1. Introduction As the world shipping becomes more and more busy, the large ship traffic flow leads to frequent maritime traffic accidents, resulting in huge economic losses. Ship traffic flow is a basic system in marine traffic engineering and an important index to measure the construction of marine traffic infrastructure. Its prediction results can provide basis for formulating scientific Port management planning and ship navigation management. Therefore, to ensure the accuracy and rationality of ship traffic flow forecasting is of great significance for improving port infrastructure construction and formulating scientific port management strategies. Many advanced artificial intelligence optimization algorithms have been used for ship traffic flow forecasting, such as artificial neural network (Zhai 2013; Zhang 2015). Neural network can deal with complex nonlinear problems and has achieved some results. However, the neural network itself has some shortcomings, such as slow learning speed, easy to fall into the local extremum, learning and memory instability, etc.

Momentum-Incorporated Symmetric Non-Negative Latent Factor Models

Symmetric high-dimensional and sparse (SHiDS) networks are frequently found in various industrial applications. A symmetric non-negative latent factor (SNLF) model can acquire essential features from them precisely, yet it suffers from slow convergence. To address this issue, this article integrates a generalized momentum method into a symmetric, single latent factor-dependent, non-negative and multiplicative update (S <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> LF-NMU) algorithm, thereby achieving a <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</u> omentum-incorporated, <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</u> ymmetric, <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">s</u> ingle-latent-factor-dependent <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</u> on-negative-multiplicative-update (MS <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> N) algorithm. Based on an MS <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> N algorithm, momentum-incorporated symmetric non-negative latent factor (MSNLF) models are proposed for an SHiDS network, which ensures fast convergence as well as high representative learning ability. Empirical studies on four SHiDS networks from industrial applications demonstrate that compared with state-of-the-art models, the proposed MSNLF models have significantly higher computational efficiency and representative learning ability.

A Unified Momentum Method with Triple-Parameters and Its Optimal Convergence Rate

A Unified Momentum Method with Triple-Parameters and Its Optimal Convergence Rate

Geometrically exact planar Euler-Bernoulli beam and time integration procedure for multibody dynamics

A new formulation of geometrically exact planar Euler-Bernoulli beam in multi-body dynamics is proposed. For many applications, the use of the Euler-Bernoulli model is sufficient and has the advantage of being a nodal displacement-only formulation avoiding the integration of rotational degrees of freedom. In this paper, an energy momentum method is proposed for the nonlinear in-plane dynamics of flexible multi-body systems, including the effects of revolute joints with or without torsional springs. Large rotational angles of the joints are accurately calculated. Several numerical examples demonstrate the accuracy and the capabilities of the new formulation.

Improved gradient descent algorithms for time-delay rational state-space systems: intelligent search method and momentum method

This study proposes two improved gradient descent parameter estimation algorithms for rational state-space models with time-delay. These two algorithms, based on intelligent search method and momentum method, can simultaneously estimate the time-delay and parameters without the matrix eigenvalue calculation in each iteration. Compared with the traditional gradient descent algorithm, the improved algorithms come with two advantages: having quicker convergence rates and less computational efforts, particularly meaningful for those large-scale systems. A simulated example is selected to illustrate the efficiency of the proposed algorithms.

Validation and accommodation of vortex wake codes for wind turbine design load calculations

Abstract. The computational effort for wind turbine design load calculations is more extreme than it is for other applications (e.g., aerospace), which necessitates the use of efficient but low-fidelity models. Traditionally the blade element momentum (BEM) method is used to resolve the rotor aerodynamic loads for this purpose, as this method is fast and robust. With the current trend of increasing rotor size, and consequently large and flexible blades, a need has risen for a more accurate prediction of rotor aerodynamics. Previous work has demonstrated large improvement potential in terms of fatigue load predictions using vortex wake models together with a manageable penalty in computational effort. The present publication has contributed towards making vortex wake models ready for application to certification load calculations. The observed reduction in flapwise blade root moment fatigue loading using vortex wake models instead of the blade element momentum (BEM) method from previous publications has been verified using computational fluid dynamics (CFD) simulations. A validation effort against a long-term field measurement campaign featuring 2.5 MW turbines has also confirmed the improved prediction of unsteady load characteristics by vortex wake models against BEM-based models in terms of fatigue loading. New light has been shed on the cause for the observed differences and several model improvements have been developed, both to reduce the computational effort of vortex wake simulations and to make BEM models more accurate. Scoping analyses for an entire fatigue load set have revealed the overall fatigue reduction may be up to 5 % for the AVATAR 10 MW rotor using a vortex wake rather than a BEM-based code.