Artificial neural network modelling of the thermal performance of a compact heat exchanger

This paper reports the use of artificial neural network models to simulate the thermal performance of a compact, fin-tube heat exchanger with air and water/ethylene glycol anti-freeze mixtures as the working fluids. The model predictions were compared with experimental data over a range of flow rates and inlet temperatures and with various ethylene glycol concentrations. In addition, the inlet air flow was distorted by obstructing part of the inlet ducting near the front face of the exchanger. The artificial neural networks were able to predict the overall rate of heat transfer in the exchanger with a high degree of accuracy and in this respect were found to be superior over conventional non-linear regression models in capturing the underlying non-linearity in the data. Moreover the detailed spatial variations in outlet air temperature were also adequately predicted. The results indicate that appropriately trained neural networks can simulate both the overall and “local” characteristics of the compact heat exchanger. In addition the paper demonstrates how an alternative type of neural network, the so-called Self-Organising-Map (SOM), can be employed for heat exchanger condition monitoring by identifying and classifying the deterioration in exchanger performance which, in this case, was associated with different levels of inlet obstruction.     

Inverse neural network based control strategy for absorption chillers

This paper proposes a novel, model-based control strategy for absorption cooling systems. First, a small-scale absorption chiller was modelled using artificial neural networks (ANNs). This model takes into account inlet and outlet temperatures as well as the flow rates of the external water circuits. The configuration 9-6-2 (9 inputs, 6 hidden and 2 output neurons) showed excellent agreement between the prediction and the experimental data (R² > 0.99 and RMSE < 0.05%). This type of ANN model is used to explain the behaviour of the system when operating conditions are measured and these measurements are available. A control strategy was also developed by using the inverse artificial neural network (ANNi) method. For a particular output (cooling load) the ANNi calculates the optimal unknown parameter(s) (controlling temperatures and flow rates). An optimization method was used to fit the unknown parameters of the ANNi method. The very low percentage of error and short computing time make this methodology suitable for the on-line control of absorption cooling systems.     

基于NARX模型的岩溶地下河日流量预测

针对传统的统计学方法难以精确刻画岩溶地下河日流量变化的非线性动态特性,引入有源自回归神经网络(NARX)技术,建立了基于NARX模型的岩溶地下河日流量预测模型,基于寨底地下河2013年1月15日~2014年6月30日的降雨量和流量数据,利用该模型对寨底地下河日流量进行了短期预测。结果表明,该模型预测效果较好,能够很好地预测岩溶地下河流量的变化趋势和极值等动态特性,另外该模型神经元个数越多,延迟阶数越大,神经网络对数据的学习能力和灵活性越强,但该模型不宜进行归一化处理。     

Convective heat transfer and pressure drop study on nanofluids in double-walled reactor by developing an optimal multilayer perceptron artificial neural network

The effect of nanofluid on the cooling performance and pressure drop of a jacked reactor has experimentally been investigated. Aqueous nanofluids of Al₂O₃ and CuO was used as the cool ant inside the cooling jacket of the reactor. The application of the artificial neural networks (ANNs) to predict the performance of a double-walled reactor has been studied. Different architectures of artificial neural networks were developed to predict the convective heat transfer and pressure drop of nanofluids. The experimental results are used for training and testing the ANNs based on two optimal models via feed-forward back-propagation multilayer perceptron (MLP). The comparison of statistical criteria of different network shows that the optimal structure for predicting the convective heat transfer coefficient is the MLP network with one hidden layer and 10 neurons, which has been trained with Levenberg–Marquardt (LM) algorithm. The predicted pressure drop values by the MLP network with two hidden layers and 6 neurons in the each layer has been used from LM training algorithm, which showed a reasonable agreement with the experimental results.     

Performance predictions of laminar and turbulent heat transfer and fluid flow of heat exchangers having large tube-diameter and large tube-row by artificial neural networks

In this work an artificial neural network (ANN) is used to correlate experimentally determined and numerically computed Nusselt numbers and friction factors of three kinds of fin-and-tube heat exchangers having plain fins, slit fins and fins with longitudinal delta-winglet vortex generators with large tube-diameter and large the number of tube rows. First the experimental data for training the network was picked up from the database of nine samples with tube outside diameter of 18 mm, number of tube rows of six, nine, twelve, and Reynolds number between 4000 and 10,000. The artificial neural network configuration under consideration has twelve inputs of geometrical parameters and two outputs of heat transfer Nusselt number and fluid flow friction factor. The commonly-implemented feed-forward back propagation algorithm was used to train the neural network and modify weights. Different networks with various numbers of hidden neurons and layers were assessed to find the best architecture for predicting heat transfer and flow friction. The deviation between the predictions and experimental data was less than 4%. Compared to correlations for prediction, the performance of the ANN-based prediction exhibits ANN superiority. Then the ANN training database was expanded to include experimental data and numerical data of other similar geometries by computational fluid dynamics (CFD) for turbulent and laminar cases with the Reynolds number of 1000–10,000. This in turn indicated the prediction has a good agreement with the combined database. The satisfactory results suggest that the developed ANN model is generalized to predict the turbulent or/and laminar heat transfer and fluid flow of such three kinds of heat exchangers with large tube-diameter and large number of tube rows. Also in this paper the weights and biases corresponding to the neural network architecture are provided so that future research can be carried out. It is recommended that ANNs might be used to predict the performances of thermal systems in engineering applications, especially to model heat exchangers for heat transfer analysis.     

Artificial neural network estimation of global solar radiation using air temperature and relative humidity

Measured air temperature and relative humidity values between 1998 and 2002 for Abha city in Saudi Arabia were used for the estimation of global solar radiation (GSR) in future time domain using artificial neural network method. The estimations of GSR were made using three combinations of data sets namely: (i) day of the year and daily maximum air temperature as inputs and GSR as output, (ii) day of the year and daily mean air temperature as inputs and GSR as output and (iii) time day of the year, daily mean air temperature and relative humidity as inputs and GSR as output. The measured data between 1998 and 2001 were used for training the neural networks while the remaining 240 days’ data from 2002 as testing data. The testing data were not used in training the neural networks. Obtained results show that neural networks are well capable of estimating GSR from temperature and relative humidity. This can be used for estimating GSR for locations where only temperature and humidity data are available.     

Process modeling and optimization of PECVD silicon nitride coated on silicon solar cell using neural networks

The process of plasma enhanced chemical vapor deposition silicon nitride films coated on silicon solar cells as antireflection layers is modeled and optimized using neural networks. This neural network model is built based on the robust design technique with process input–output experimental data. The input parameters selected are as substrate temperature, SiH₄ and NH₃ flow rates, and RF power; while the output parameters are deposition rate, refractive index, and short circuit current. This model can then be applied to predict the input–output relationships of the process. Optimal operating conditions of this process can be determined using this model.     

Study of flow field,heat transfer,and entropy generation of nanofluid turbulent natural convection in an enclosure utilizing the computational fluid dynamics‐artificial neural network hybrid method

In this study, the turbulent natural convection of Ag‐water nanofluid in a tall, inclined enclosure has been investigated. The main objective of this study is finding the optimized angle of the enclosure with operational boundary condition in cooling from ceiling utilizing the computational fluid dynamics‐artificial neural network (CFD‐ANN) hybrid method, which has not been noticed in previous studies. To achieve this, we proposed two approaches. First, the simulations have been done with a deviation angle of 0 to 90° by using water and Ag‐water nanofluid. And second, a new prediction approach is proposed based on radial basis function artificial neural networks (RBF‐ANN) to predict the mean Nusselt number and entropy generation with the variation of Rayleigh numbers, deviation angles, and volume fractions as inputs. The results from the first approach indicate that the Rayleigh number has a considerable function in the determination of optimized angle. The results from the second approach, which used the first approach simulation results as training data set, could predict the mean Nusselt number and entropy generation with 1.4577e^?022 and 1.552e^?015 mean square error, respectively. Moreover, a new set of data for Rayleigh numbers, deviation angles, and volume fractions were used to test the performance of the prediction model, which shows promising and superior prospects for RBF‐ANN.     

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.     

Prediction of dynamic viscosity of a hybrid nano-lubricant by an optimal artificial neural network

In this paper, at first, a new correlation was proposed to predict the relative viscosity of MWCNTs-SiO₂/AE40 nano-lubricant using experimental data. Then, considering minimum prediction error, an optimal artificial neural network was designed to predict the relative viscosity of the nano-lubricant. Forty-eight experimental data were used to feed the model. The data set was derived to training, validation and test sets which contained 70%, 15% and 15% of data points, respectively. The correlation outputs showed that there is a deviation margin of 4%. The results obtained from optimal artificial neural network presented a deviation margin of 1.5%. It can be found from comparisons that the optimal artificial neural network model is more accurate compared to empirical correlation.     

Applications of feedforward multilayer perceptron artificial neural networks and empirical correlation for prediction of thermal conductivity of Mg(OH)2–EG using experimental data

This paper presents an investigation on the thermal conductivity of nanofluids using experimental data, neural networks, and correlation for modeling thermal conductivity. The thermal conductivity of Mg(OH)₂ nanoparticles with mean diameter of 10 nm dispersed in ethylene glycol was determined by using a KD2-pro thermal analyzer. Based on the experimental data at different solid volume fractions and temperatures, an experimental correlation is proposed in terms of volume fraction and temperature. Then, the model of relative thermal conductivity as a function of volume fraction and temperature was developed via neural network based on the measured data. A network with two hidden layers and 5 neurons in each layer has the lowest error and highest fitting coefficient. By comparing the performance of the neural network model and the correlation derived from empirical data, it was revealed that the neural network can more accurately predict the Mg(OH)₂–EG nanofluids' thermal conductivity.     

Development of a neural network-based fault diagnostic system for solar thermal applications

The objective of this work is to present the development of an automatic solar water heater (SWH) fault diagnosis system (FDS). The FDS system consists of a prediction module, a residual calculator and the diagnosis module. A data acquisition system measures the temperatures at four locations of the SWH system and the mean storage tank temperature. In the prediction module a number of artificial neural networks (ANN) are used, trained with values obtained from a TRNSYS model of a fault-free system operated with the typical meteorological year (TMY) for Nicosia, Cyprus and Paris, France. Thus, the neural networks are able to predict the fault-free temperatures under different environmental conditions. The input data to the ANNs are various weather parameters, the incidence angle, flow condition and one input temperature. The residual calculator receives both the current measurement data from the data acquisition system and the fault-free predictions from the prediction module. The system can predict three types of faults; collector faults and faults in insulation of the pipes connecting the collector with the storage tank and these are indicated with suitable labels. The system was validated by using input values representing various faults of the system.     

Application of neural networks for the prediction of hourly mean surface temperatures in Saudi Arabia

This paper utilizes artificial neural networks for the prediction of hourly mean values of ambient temperature 24 h in advance. Full year hourly values of ambient temperature are used to train a neural network model for a coastal location — Jeddah, Saudi Arabia. This neural network is trained off-line using back propagation and a batch learning scheme. The trained neural network is successfully tested on temperatures for years other than the one used for training. It requires only one temperature value as input to predict the temperature for the following day for the same hour. The predicted hourly temperature values are compared with the corresponding measured values. The mean percent deviation between the predicted and measured values is found to be 3.16, 4.17 and 2.83 for three different years. These results testify that the neural network can be a valuable tool for hourly temperature prediction in particular and other meteorological predictions in general.     

A dynamic two-level artificial neural network for estimation of parameters in combined mode conduction-radiation heat transfer in porous medium: An application to handle huge dataset with noise

A dynamic two-level artificial neural network (DTLANN) approach is used for the estimation of parameters in combined mode conduction–radiation heat transfer in a porous medium. Four commonly used neural networks: feed forward, cascade forward, fitnet, and radial basis are used in mapping artificial neural network (ANN), and their performance is compared under noisy big data (10,302 × 1300 matrix size). Governing equations for heat transfer in the porous medium through conduction and radiation modes are solved by finite volume method and discrete transfer method. This numerical model is called a direct model. A large amount of data is generated by using the direct model for different values of extinction coefficient β and convective coupling P₂. These data were divided into different groups (class) based on the temperature difference between the gas and solid phase. In the inverse analysis, a new pair of temperature profiles for the solid and gas phase is taken as input and classified with the help of a pattern net artificial neural network model. On the basis of classification, data from that particular class and its neighbor class are used for training the mapping ANN model. After the training of the mapping ANN model, corresponding values of β and P₂ are obtained as output for any new input. This DTLANN model has a high regression coefficient (R) of .99131 and can predict highly accurate values of parameters under a huge dataset with noise, within much less CPU time.     

Direct and inverse neural networks modelling applied to study the influence of the gas diffusion layer properties on PBI-based PEM fuel cells

This article shows the application of a very useful mathematical tool, artificial neural networks, to predict the fuel cells results (the value of the tortuosity and the cell voltage, at a given current density, and therefore, the power) on the basis of several properties that define a Gas Diffusion Layer: Teflon content, air permeability, porosity, mean pore size, hydrophobia level. Four neural networks types (multilayer perceptron, generalized feedforward network, modular neural network, and Jordan-Elman neural network) have been applied, with a good fitting between the predicted and the experimental values in the polarization curves. A simple feedforward neural network with one hidden layer proved to be an accurate model with good generalization capability (error about 1% in the validation phase). A procedure based on inverse neural network modelling was able to determine, with small errors, the initial conditions leading to imposed values for characteristics of the fuel cell. In addition, the use of this tool has been proved to be very attractive in order to predict the cell performance, and more interestingly, the influence of the properties of the gas diffusion layer on the cell performance, allowing possible enhancements of this material by changing some of its properties.     

Modelling fuel cell performance using artificial intelligence

Over the last few years, fuel cell technology has been increasing promisingly its share in the generation of stationary power. Numerous pilot projects are operating worldwide, continuously increasing the amount of operating hours either as stand-alone devices or as part of gas turbine combined cycles. An essential tool for the adequate and dynamic analysis of such systems is a software model that enables the user to assess a large number of alternative options in the least possible time. On the other hand, the sphere of application of artificial neural networks has widened covering such endeavours of life such as medicine, finance and unsurprisingly engineering (diagnostics of faults in machines). Artificial neural networks have been described as diagrammatic representation of a mathematical equation that receives values (inputs) and gives out results (outputs). Artificial neural networks systems have the capacity to recognise and associate patterns and because of their inherent design features, they can be applied to linear and non-linear problem domains. In this paper, the performance of the fuel cell is modelled using artificial neural networks. The inputs to the network are variables that are critical to the performance of the fuel cell while the outputs are the result of changes in any one or all of the fuel cell design variables, on its performance. Critical parameters for the cell include the geometrical configuration as well as the operating conditions. For the neural network, various network design parameters such as the network size, training algorithm, activation functions and their causes on the effectiveness of the performance modelling are discussed. Results from the analysis as well as the limitations of the approach are presented and discussed.     

Prediction of Heat Transfer Rates for Shell-and-Tube Heat Exchangers by Artificial Neural Networks Approach

This work used artificial neural network(ANN)to predict the heat transfer rates of shell-and-tube heatexchangers with segmental baffles or continuous helical baffles,based on limited experimental data.The BackPropagation (BP) algorithm was used in training the networks.Different network configurations were alsostudied.The deviation between the predicted results and experimental data was less than 2%.Comparison withcorrelation for prediction shows ANN superiority.It is recommended that ANN can be easily used to predict theperformances of thermal systems in engineering applications,especially to model heat exchangers for heattransfer analysis.     

Air gasification of high-ash sewage sludge for hydrogen production: Experimental,sensitivity and predictive analysis

In this work, air gasification of sewage sludge was conducted in a lab-scale bubbling fluidized bed gasifier. Further, the gasification process was modeled using artificial neural networks for the product gas composition with varying temperatures and equivalence ratios. Neural network-based prediction will help to predict the hydrogen production from product gas composition at various temperatures and equivalence ratios. The gasification efficiency and lower heating values were also established as a function of temperatures and equivalence ratios. The maximum H₂ and CO was recorded as 16.26 vol% and 33.55 vol%. Intraileally at ER 0.2 gas composition H₂, CO, and CH₄ show high concentrations of 20.56 vol%, 45.91 vol%, and 13.32 vol%, respectively. At the same time, CO₂ was lower as 20.20 vol% at ER 0.2. Therefore, optimum values are suggested for maximum H₂ and CO yield and lower concentration of CO₂ at ER 0.25 and temperature of 850 °C. A predictive model based on an Artificial Neural network is also developed to predict the hydrogen production from product gas composition at various temperatures and equivalence ratios. The network has been trained with different topologies to find the optimal structure for temperature and equivalence ratio. The obtained results showed that the regression coefficients for training, validation, and testing are 0.99999, 0.99998, and 0.99992, respectively, which clearly identifies the training efficiency of the trained model.     

Artificial neural network modeling of nanofluid flow in a microchannel heat sink using experimental data

The present paper deals with the artificial neural network modeling (ANN) of heat transfer coefficient and Nusselt number in TiO₂/water nanofluid flow in a microchannel heat sink. The microchannel comprises of 40 channels; each channel has a length of 4 cm, a width of 500 μm, and a height of 800 μm. In the ANN modeling of heat transfer coefficient and Nusselt number 23 and 72 datasets have been used, respectively. The experimental Nusselt number has been calculated based on three different thermal conductivity models, four volume fractions of 0, 0.5, 1, and 2%, two values of Reynolds number i.e. 400 and 1200 and three different heating rates including 50.6, 60.7, and 69.1 W. Therefore, the inputs that are introduced to the neural network are volume fraction of nanoparticles, Reynolds number, heating rate, and model number while the output of network is the Nusselt number. It is elucidated that an appropriately trained network can act as a good alternative for costly and time-consuming experiments on the nanofluid flow in microchannels. The average relative errors in the prediction of Nusselt number and heat transfer coefficients were 0.3% and 0.2%, respectively.     

Comparison of the different artificial neural networks in prediction of biomass gasification products

In this study, artificial neural networks (ANNs) and a nonlinear autoregressive exogenous (NARX) neural network model were employed in order to model a fixed bed downdraft gasification. The relation between the feature group and the regression performance was investigated. First, feature group consists of the equivalence ratio (ER), air flow rate (AF), and temperature distribution (T0‐T5) obtained from the fixed bed downdraft gasifiers, while the second group includes ultimate and proximate values of biomasses, ER, AF, and the reduction temperature (T0). Models constructed to predict the syngas composition (H₂, CO₂, CO, CH₄) and calorific value. Experimental gasification data that involve 3831 data samples that belong to pinecone and wood pellet were used for training the ANNs. Different ANN architecture and NARX time series model have been constructed to examine the prediction accuracy of the models. The results of the ANN models were consistent with the experimental data (R² > 0.99). The overall score of NARX time series networks is found to be higher than other architecture types. A successful method is proposed to reduce the number of features, and the effect of the features on the prediction capability was examined by calculating the relative importance index using the Garson's equation.