Intelligent networks for crosswise stream nanofluidic model with Cu–H2O over porous stretching medium

The porous media transport theories are thoroughly operative to analyse transferral phenomenon in reducing the bio-convective flow instabilities and biological tissues. The present study is designed to investigate the heat transfer phenomena in nanofluidic system involving Cu ? H₂O over the stretched porous media with the strength of stochastic solver via Levenberg-Marquardt backpropagation networks. The mathematical model of physical phenomena is described in PDEs that are reduced to system of ODEs through scaling group transformations. The datasets are determined through explicit Runge-Kutta numerical method and used as a target parameter for the development of continuous neural networks mapping. The training, testing and validation processes are utilized in learning of neural network models based on backpropagation of Levenberg-Marquardt technique to determines the solution of different scenarios constructed on the various values of porosity parameter along with six different cases based on the stretching ratio values. Validation and verification of neural network model to find the solution of nanfluidic problem is endorsed on the assessment of achieved accuracy through mean squared error, error histograms and regression studies.     

Neuro-intelligent mappings of hybrid hydro-nanofluid Al2O3–Cu–H2O model in porous medium over rotating disk with viscous dissolution and Joule heating

The aim of study is to investigate the mass and heat transfer phenomena in hybrid hydro-nanofluidic system involving Al₂O₃–Cu–H₂O over the rotating disk in porous medium with viscous dissolution and Joule heating through the stochastic solver by way of Levenberg-Marquardt backpropagation neural networks. The mathematical model in system of PDEs describes the physical phenomena of the hybrid hydro-nanofluid flow problem are converted into set of ODEs by means of scaling group transformations. The datasets are constructed by utilizing the power of explicit Runge-Kutta numerical method that help to the develop a continuous neural networks mapping. The validation, training and testing processes are utilized to learn the neural network mapping to estimate the solution of various scenarios with cases that are constructed by varying different values of physical constraints such as porosity factor, inertia coefficient, Prandtl number, Brinkman number, radiation parameter, mgnetic parameter, concentration of nanoparticles on the velocities and temperature profiles. Determination, convergence, verification and stability of Levenberg-Marquardt backpropogation neural network mappings are validated on the assessment of achieved accuracy through regression based statistical analysis, mean squared error and error histograms for hybrid hydro-nanofluidic model.     

A Comparison of ODE Solution Methods for Spacecraft Thermal Problems

Abstract

In spacecraft thermal design and analysis practice, the lumped-parameter network formulation is used extensively to construct mathematical models. The models lake the form of a system of coupled, nonlinear, first-order ordinary differential equations (ODEs). The number of equations may vary from a few tens to a few thousands. It is necessary to solve these equations in an efficient and economical way. This article reviews various methods available for the numerical solution of ODEs. General-purpose codes available for this are discussed. Numerical experiments are conducted to investigate the efficiency of various methods. The results indicate that the Crank-Nicholson method with provision for automatic selection of step size to control the local truncation error is a very good choice for the solution of spacecraft thermal problems.     

Application of artificial neural network in performance prediction of PEM fuel cell

Investigations on using artificial neural networks to predict the performance of single proton exchange membrane fuel cell has been carried out. Two sets of polarization data obtained at different temperatures and flow rates are used to create and simulate the network. Cell temperature, humidification temperatures, H₂/air flow rates and current density have been used as inputs, and voltage is used as observed (output) value to train and simulate the network. This nonlinear data are batch trained, and artificial neural network has been constructed using feed forward backpropagation algorithm. Performance of the training has been improved by increasing the number of neurons to reduce the error. Simulation results are in agreement with experimental data, and the corresponding networks are used to predict the polarization behavior for unknown inputs. Copyright © 2011 John Wiley & Sons, Ltd.     

Artificial neural networks and neuro-fuzzy inference systems as virtual sensors for hydrogen safety prediction

Hydrogen is increasingly investigated as an alternative fuel to petroleum products in running internal combustion engines and as powering remote area power systems using generators. The safety issues related to hydrogen gas are further exasperated by expensive instrumentation required to measure the percentage of explosive limits, flow rates and production pressure. This paper investigates the use of model based virtual sensors (rather than expensive physical sensors) in connection with hydrogen production with a Hogen^®20 electrolyzer system. The virtual sensors are used to predict relevant hydrogen safety parameters, such as the percentage of lower explosive limit, hydrogen pressure and hydrogen flow rate as a function of different input conditions of power supplied (voltage and current), the feed of de-ionized water and Hogen^®20 electrolyzer system parameters. The virtual sensors are developed by means of the application of various Artificial Intelligent techniques. To train and appraise the neural network models as virtual sensors, the Hogen^®20 electrolyzer is instrumented with necessary sensors to gather experimental data which together with MATLAB neural networks toolbox and tailor made adaptive neuro-fuzzy inference systems (ANFIS) were used as predictive tools to estimate hydrogen safety parameters. It was shown that using the neural networks hydrogen safety parameters were predicted to less than 3% of percentage average root mean square error. The most accurate prediction was achieved by using ANFIS.     

Lie group analysis of hyperbolic tangent fluid flow in the presence of thermal radiation

The present study is devoted to the flow and heat transfer analysis of the hyperbolic tangent fluid through a stretching sheet by considering the effect of thermal radiation in addition to an applied transverse magnetic field, as well as thermal and velocity slip conditions. The Lie group analysis technique has been utilized for establishing similarity transformations, which effectively transform the governing equations to a system of nonlinear ordinary differential equations (ODEs). These ODEs are numerically solved by utilizing the shooting method. The heat transfer properties and flow features under the influence of various physical parameters are also studied. We noted that by increasing the thermal radiation parameter, the temperature profile increases and also the thermal boundary layer thickens. Furthermore, it is deduced that rising the thermal radiation parameter reduces the local Nusselt number. Moreover, the numerical results obtained are in agreement with the existing results in the literature.     

Design Methods and Strategies for Forward and Inverse Problems of Turbine Blades Based on Machine Learning

To study the feasibility of using machine learning technology to solve the forward problem(prediction of aerodynamic parameters)and the inverse problem(prediction of geometric parameters)of turbine blades,this paper built a forward problem model based on backpropagation artificial neural networks(BP-ANNs)and an inverse problem model based on radial basis function artificial neural networks(RBF-ANNs).The S2(a stream surface obtained by extending a radial curve in turbo blades)calculation program was used to generate the dataset for single-stage turbo blades,and the back propagation algorithm was used to train the model.The parameters of five blade sections in a single-stage turbine were selected as inputs of the forward problem model,including stagger angle,inlet geometric angle,outlet geometric angle,wedge angle of leading edge pressure side,wedge angle of leading edge suction side,wedge angle of trailing edge,rear bending angle,and leading edge diameter.The outputs are efficiency,power,mass flow,relative exit Mach number,absolute exit Mach number,relative exit flow angle,absolute exit flow angle and reaction degree,which are eight aerodynamic parameters.The inputs and outputs of the inverse problem model are the opposite of that of the forward problem model.The models can accurately predict the aerodynamic parameters and geometric parameters,and the mean square errors(MSEs)of the forward problem test set and the inverse problem test set are 0.001 and 0.00035,respectively.This study shows that machine learning technology based on neural networks can be flexibly applied to the design of forward and inverse problems of turbine blades,and the models built by this method have practical application value in regression prediction problems.     

Fault diagnosis for temperature,flow rate and pressure sensors in VAV systems using wavelet neural network

Wavelet neural network, the integration of wavelet analysis and neural network, is presented to diagnose the faults of sensors including temperature, flow rate and pressure in variable air volume (VAV) systems to ensure well capacity of energy conservation. Wavelet analysis is used to process the original data collected from the building automation first. With three-level wavelet decomposition, the series of characteristic information representing various operation conditions of the system are obtained. In addition, neural network is developed to diagnose the source of the fault. To improve the diagnosis efficiency, three data groups based on several physical models or balances are classified and constructed. Using the data decomposed by three-level wavelet, the neural network can be well trained and series of convergent networks are obtained. Finally, the new measurements to diagnose are similarly processed by wavelet. And the well-trained convergent neural networks are used to identify the operation condition and isolate the source of the fault.     

Improved artificial neural network method for predicting photovoltaic output performance

To ensure the safety and stability of power grids with photovoltaic (PV) generation integration, it is necessary to predict the output performance of PV modules under varying operating conditions. In this paper, an improved artificial neural network (ANN) method is proposed to predict the electrical characteristics of a PV module by combining several neural networks under different environmental conditions. To study the dependence of the output performance on the solar irradiance and temperature, the proposed neural network model is composed of four neural networks, it called multi- neural network (MANN). Each neural network consists of three layers, in which the input is solar radiation, and the module temperature and output are five physical parameters of the single diode model. The experimental data were divided into four groups and used for training the neural networks. The electrical properties of PV modules, including I–V curves, P– V curves, and normalized root mean square error, were obtained and discussed. The effectiveness and accuracy of this method is verified by the experimental data for different types of PV modules. Compared with the traditional single-ANN (SANN) method, the proposed method shows better accuracy under different operating conditions.     

An adaptive power system stabilizer using on-line trained neuralnetworks

This paper presents an approach to the design of an adaptive power system stabilizer (PSS) based on on-line trained neural networks. Only the inputs and outputs of the generator are measured and there is no need to determine the states of the generator. The proposed neural adaptive PSS (NAPSS) consists of an adaptive neuro-identifier (ANI), which tracks the dynamic characteristics of the plant, and an adaptive neuro-controller (ANC) to damp the low frequency oscillations. These two subnetworks are trained in an on-line mode utilizing the backpropagation method. The use of a single-element error vector along with a small network simplifies the learning algorithm in terms of computation time. The improvement of the dynamic performance of the system is demonstrated by simulation studies for a variety of operating conditions and disturbances     

Investigation of flow parameters of Reiner–Philippoff nanofluid flow with higher-order slip properties,activation energy,and bioconvection by artificial neural networks

In this study, the flow parameters of Reiner–Philippoff nanofluid flow with high-order slip properties, activation energy, and bioconvection have been analyzed using artificial neural networks (ANNs). Local Nusselt number (LNN), local Sherwood number (LSN), and motile density number (MDN) are considered as flow parameters. Numerical values have been obtained by numerical methods using flow equations. To estimate the flow parameters, three different ANN models have been designed. The Levenberg–Marquardt training algorithm is used in multilayer perceptron network models with 10 neurons in the hidden layers. In all, 70% of the data set has been used for training the models, 15% for validation, and 15% for testing. The performance analysis of the network models has been made by calculating the determined performance parameters. The R values for the LNN, LSN, and MDN parameters have been calculated as 0.99261, 0.98769, and 0.99102, respectively, and the average deviation values are −0.65%, 0.06%, and −0.11%, respectively. The attained outcomes showed that the ANNs can predict the LNN, LSN, and MDN, which are the flow parameters of the Reiner–Philippoff nanofluid flow, with high accuracy.     

A comparison of empirically based steady-state models for vapor-compression liquid chillers

This paper presents a comprehensive comparison of empirically based models for steady-state modeling of vapor-compression liquid chillers. Next to the considered models already proposed in the open literature, i.e. regression, thermodynamic, and a radial basis function neural network model, a multilayer perceptron neural network model is introduced. The models predict the coefficient of performance by only using input variables that are readily known to the operating engineer. They are applied to two different chillers operating at the University of Auckland, New Zealand. The comparison demonstrates that neural networks show higher generalization abilities and at least equal forecast results compared to the regression models. Procedures are presented that make models without any physical meaning in the parameters possible to be used in fault detection and diagnosis. It is inferred that black-box models, in particular the radial basis function neural network model, may be preferred for predicting a chiller’s performance in these purposes.     

基于小波神经网络替代模型的抽水井涌水量计算方法

影响抽水井涌水量的水文地质因素多且存在不确定性,采用确定性模型模拟会导致较大的误差。以某研究区抽水井涌水量为依据,建立该地区地下水不确定性数值模型。采用参数敏感性分析确定对抽水井涌水量影响较大的水文地质参数,随后用改进的随机进化算法在参数取值范围内抽样,最后将抽样参数输入地下水数值模型中计算抽水井涌水量。根据输入、输出数据集建立小波神经网络模型代替地下水数值模型。研究结果表明,不同工况下基于小波神经网络替代模型计算结果与数值模拟结果相差不大,两者误差在10%以内,表明用小波神经网络替代模型满足精度要求,且避免了传统数值模型的反复试算,计算简便、效率高,且因其可公式化,扩大了在推求地下水抽水井涌水量方面的应用。     

The effect of Joule heating and viscous dissipation on the boundary layer flow of a magnetohydrodynamics micropolar-nanofluid over a stretching vertical Riga plate

Joule heating and viscous dissipation effects on the behavior of the boundary layer flow of a micropolar nanofluid over a stretching vertical Riga plate (electro magnetize plate) are considered. The flow is disturbed by an external electric magnetic field. The problem is formulated mathematically by nonlinear system of partial differential equations (PDEs). By using suitable variables transformations, this system is transformed onto a system of nonlinear ordinary differential equations (ODEs). The Parametric NDsolve package of the commercial software Mathematica is used to solve the obtained ODEs as well as the considered numerical results for different physical parameters with appropriate boundary conditions. Novel results are obtained by studying the stream lines flow around the plate in two and three dimensions. Moreover, the effects of the pertinent parameters on the skin friction coefficient, couple stress, local Nusselt, and Sherwood number are discussed. Special cases of the obtained results show excellent agreements with previous works. The results showed that as the magnetic field parameter increases the velocity of the boundary layer adjacent to the stretching sheet decreases. Also, for a productive chemical reaction near the sheet surface, the angular velocity decreases but opposite trend is observed far from the sheet surface. The importance of this study comes from its significant applications in many scientific fields, such as nuclear reactors, industry, medicine, and geophysics.     

Design of inverse multiquadric radial basis neural networks for the dynamical analysis of MHD casson nanofluid flow along a nonlinear stretchable porous surface with multiple slip conditions

The present research, a numerical approach to examine magnetohydrodynamics (MHD) Casson nanofluid flow in a porous medium along a stretchable surface with different slips using artificial neural networks (ANNs) by taking inverse multiquadric (IMQ) radial basis function (RBF) as an activation function. i.e. ANNs-IMQ-RBF. The hybridization of genetic algorithms (GAs) and sequential quadratic programming (SQP) is used for learning in ANNs-IMQ-RBF. The PDEs representing the fluid flow are converted into a nonlinear system of dimensionless form of ODEs through an appropriate transformation while effects of variation in the values of Casson parameter (β), Brownian motion parameter (Nb), Prandtl number (Pr), stretching parameter (n), porosity parameter (P), Lewis number (Le) along with temperature slip parameter (λ₂) on velocity, temperature and nanofluid concentration are depicted through graphs. The effectiveness, convergence and accuracy of the proposed solver are validated evidently through boxplot analysis, histograms and cumulative distribution function (CDF) plots.     

一种前向神经网络自适应学习算法

针对时变和(或)非线性输入的前向神经网络提出了一种感知自适应算法,其本质是要求输出的实际值和期望值的误差满足一个渐稳定的差分方程,而不是用后向传播方法使误差函数极小化。通过适当排列扩张输出可以避免算法的奇异性。     

A robust surrogate model of a solid oxide cell based on an adaptive polynomial approximation method

Surrogate models that predict the behaviors of solid oxide cells (SOCs) accurately at low computational cost are crucial to the control and optimization of SOC plants. Lumped physical models of SOCs, while widely used in such applications, lack accuracy because of neglected physical details. Data-driven models are the other options of surrogate models, which are proved to be more accurate because these models are identified directly from experiments or numerical simulations. However, due to the time cost of experiments and numerical simulations of SOCs, it is hoped that data-driven models can be constructed from a minimum amount of data. Also, the trained data-driven models should be robust, in other words, insensitive to the data set as well as the initial settings. These requirements are hard to be achieved by existing data-driven models of SOCs, such as lookup tables and artificial neural networks (ANNs). Aiming to preserve robustness and reduce the required amount of data, this paper introduces an adaptive polynomial approximation (APA) method, which is derived from the latest findings of numerical computation science, to the surrogate modeling of SOCs. The obtained models by the APA method are validated by both experiments and simulations. By analyzing the models, the coupling relationship among operating parameters of SOCs is revealed. The physical interpretability makes the APA method distinctive from common data-driven modeling methods. Performance comparison shows that the APA method is more accurate and robust than the existing ones with similar sampling costs. Additionally, the APA method can control the accuracy of the model by setting an error criterion in the algorithm iteration, endowing the APA method with an error control ability as per different accuracy requirements for SOC modeling.     

Bayesian adaptive combination of short-term wind speed forecasts from neural network models

Short-term wind speed forecasting is of great importance for wind farm operations and the integration of wind energy into the power grid system. Adaptive and reliable methods and techniques of wind speed forecasts are urgently needed in view of the stochastic nature of wind resource varying from time to time and from site to site. This paper presents a robust two-step methodology for accurate wind speed forecasting based on Bayesian combination algorithm, and three neural network models, namely, adaptive linear element network (ADALINE), backpropagation (BP) network, and radial basis function (RBF) network. The hourly average wind speed data from two North Dakota sites are used to demonstrate the effectiveness of the proposed approach. The results indicate that, while the performances of the neural networks are not consistent in forecasting 1-h-ahead wind speed for the two sites or under different evaluation metrics, the Bayesian combination method can always provide adaptive, reliable and comparatively accurate forecast results. The proposed methodology provides a unified approach to tackle the challenging model selection issue in wind speed forecasting.     

On comparing three artificial neural networks for wind speed forecasting

Wind speed forecasting is critical for wind energy conversion systems since it greatly influences the issues such as the scheduling of a power system, and the dynamic control of the wind turbine. In this paper, we present a comprehensive comparison study on the application of different artificial neural networks in 1-h-ahead wind speed forecasting. Three types of typical neural networks, namely, adaptive linear element, back propagation, and radial basis function, are investigated. The wind data used are the hourly mean wind speed collected at two observation sites in North Dakota. The performance is evaluated based on three metrics, namely, mean absolute error, root mean square error, and mean absolute percentage error. The results show that even for the same wind dataset, no single neural network model outperforms others universally in terms of all evaluation metrics. Moreover, the selection of the type of neural networks for best performance is also dependent upon the data sources. Among the optimal models obtained, the relative difference in terms of one particular evaluation metric can be as much as 20%. This indicates the need of generating a single robust and reliable forecast by applying a post-processing method.