Artificial Neural Network Procedure Research Articles

Liquid fuel production from syngas: Simulation and optimization using artificial neural network

This study aimed to increase liquid fuel (C5+) product selectivity from CO and H2 conversion over a Cu/Fe-based catalyst and followed by artificial neural network simulation to enable a profound visualization strategy for Fischer–Tropsch (FT) process. The experiments were performed in a fixed-bed reactor over temperatures (T) of 483–553 K, pressures (P) of 14.5–203 psig, space velocities (SV) of 1000–2900 hr−1, and H₂/CO feed ratios of 1–6.2. Mathematical modeling (MM) and artificial neural network (ANN) were utilized to simulate the nonlinear CO conversion and product selectivity of the FT process. Nine simultaneous reaction pathway was developed to evaluate the reactant and product behaviors under varying FT conditions. In addition, a multilayered parameter processed by a neural network was used to predict the performance of the CO hydrogenation process. A comparison between these two approaches demonstrated that ANN more closely mimicked the nonlinear conversion and selectivity of the FT process than MM, and the correlation coefficients for the predicted mole fractions and CO conversion were 0.991 and 0.986, respectively. A hybrid Genetic Algorithm–ANN procedure was constructed to predict the optimum operating conditions to increase liquid fuel over the catalyst. The optimum C5+ selectivity (75.1 %), which meets the FT process objective, was achieved under operating conditions of T = 483 K, P = 203 psig, SV = 1303 hr−1, and H₂/CO feed ratio = 1.01 for process design optimization.

A programmable digital metasurface structure designed using ANN technique

ABSTRACT In this paper, an 8-bit programmable digital metasurface is designed in the operating frequency range from 4 to 7 GHz. A monopole antenna operating at 5 GHz is characterised by using a metasurface structure as the ground plane. Various combinations of the metasurface structure are examined by altering the state of each unit-cell between ‘ON (1)’ and ‘OFF (0)’ in the coding matrix. The purpose of the 8-bit programmable digital metasurface is to control the electromagnetic wave effectively in the frequency range of interest. Dynamically controllable metasurface structure is utilised to change the unit cells’ configurations. In addition, the proposed adjustable metasurface is capable to control the monopole antenna directivity, gain and main lobe magnitude, efficiently. According to various via conditions, different metasurface-based antenna parameters such as return loss (S11), radiation pattern, directivity and surface current distribution are investigated by means of the commercial numerical simulation software, CST microwave studio. Employing the simulation results of the metasurface-based antennas, the artificial neural network (ANN) data set is obtained. 128 activation conditions are trained and tested by Levenberg Marquart learning algorithm. During the ANN procedure, MATLAB is used to obtain accurate results by changing the rate of test and training data set.

Prediction of discharge coefficients for broad-crested weirs using expert systems

ABSTRACT Broad crested weirs can be used to make discharge measurements in irrigation canals; the entrance of stepped weirs or chutes are sometimes designed as a broad crested weir structure. These structures are also sometimes used for the dam body. In this study, the Artificial Neural Network (ANN) and M5 model tree methods are used to predict discharge coefficients (Cd ) for broad crested weirs. The results from these two models are compared with nonlinear regression equations. Four series of data obtained from different rectangular broad crested weirs have been used and important dimensionless parameters have been defined. Results show that the ANN procedure is superior to the M5 model and regression approaches. The accuracy for ANN is quantified by R = 0.966 and RMSE = 0.038. All the three methods are able to provide a reasonable prediction for Cd ; the M5 model tree provides four linear equations that can be used to estimate Cd . The shape of the Cd contours shows that the effect of weir height (P) exceeds that of the weir length (L).

Connectionist based models for solving Diophantine equation

Diophantine equation is an important concept while studying number theory. But there is as such no standard method or procedure to find the solution of the Diophantine equation. In this article, the numerical solution of Diophantine equation has been discussed using the concept of Artificial Neural Network (ANN). A four layer ANN architecture has been constructed for solving Diophantine eqution. Detail ANN procedure with few example problems and their network architectures have been addressed. Convergence plots and/or convergence tables for different example problems have also been included to validate the proposed method.

Industrial Energy Load Profile Forecasting under Enhanced Time of Use Tariff (ETOU) using Artificial Neural Network

The demand response program involves consumers to mitigate peak demand and reducing global CO2 emission. In sustaining this effort, energy provider such as Tenaga Nasional Berhad (TNB) in Peninsular Malaysia has introduced Enhance Time of Use (ETOU) tariff. However, since 2015, small numbers join the ETOU program due to less confidence in managing their energy consumption profile. Thus, this study provides an optimum forecasting load profile model for TOU and ETOU tariffs using Artificial Neural Network (ANN). An industry's average energy profile has been used as a case study, while the forecasting technique has been conducted to find the optimum energy load profile congruently. The load shifting technique has been adopted under ETOU tariff price while integrating to the ANN procedure. A significant comparison in terms of cost reduction between TOU and ETOU electricity tariffs has been made. In contrast, ANN performance results in searching for the best-shifted load profile have been analyzed accordingly. From the proposed method, the total electricity cost saving has been founded to be saved for about 7.9% monthly. It is hoped that this work will benefit the energy authority and consumers in future action, respectively.

Neural network approach for solving nonlinear eigenvalue problems of structural dynamics

This article introduces a novel connectionist approach for dynamic analysis of structural problem. In general, dynamic analysis of structures leads to eigenvalue problems. Sometimes these eigenvalue problems may be nonlinear eigenvalue problem, which may be difficult to address by traditional methods. As such, we have proposed here an artificial neural network (ANN)-based method to handle nonlinear eigenvalue problems. A four-layer ANN architecture has been constructed for handling the eigenvalue problems, and detailed ANN procedure has been included for clear understanding. Two example problems of overdamped spring mass system have been addressed to show the efficacy of the proposed method. Further, convergence plots and tables for different eigenvalues have also been included to validate the proposed ANN procedure.

Reliable and accurate point-based prediction of cumulative infiltration using soil readily available characteristics: A comparison between GMDH, ANN, and MLR

Developing accurate and reliable pedo-transfer functions (PTFs) to predict soil non-readily available characteristics is one of the most concerned topic in soil science and selecting more appropriate predictors is a crucial factor in PTFs’ development. Group method of data handling (GMDH), which finds an approximate relationship between a set of input and output variables, not only provide an explicit procedure to select the most essential PTF input variables, but also results in more accurate and reliable estimates than other mostly applied methodologies. Therefore, the current research was aimed to apply GMDH in comparison with multivariate linear regression (MLR) and artificial neural network (ANN) to develop several PTFs to predict soil cumulative infiltration point-basely at specific time intervals (0.5–45 min) using soil readily available characteristics (RACs). In this regard, soil infiltration curves as well as several soil RACs including soil primary particles (clay (CC), silt (Si), and sand (Sa)), saturated hydraulic conductivity (Ks), bulk (Db) and particle (Dp) densities, organic carbon (OC), wet-aggregate stability (WAS), electrical conductivity (EC), and soil antecedent (θi) and field saturated (θfs) water contents were measured at 134 different points in Lighvan watershed, northwest of Iran. Then, applying GMDH, MLR, and ANN methodologies, several PTFs have been developed to predict cumulative infiltrations using two sets of selected soil RACs including and excluding Ks. According to the test data, results showed that developed PTFs by GMDH and MLR procedures using all soil RACs including Ks resulted in more accurate (with E values of 0.673–0.963) and reliable (with CV values lower than 11 percent) predictions of cumulative infiltrations at different specific time steps. In contrast, ANN procedure had lower accuracy (with E values of 0.356–0.890) and reliability (with CV values up to 50 percent) compared to GMDH and MLR. The results also revealed that Ks exclusion from input variables list caused around 30 percent decrease in PTFs accuracy for all applied procedures. However, it seems that Ks exclusion resulted in more practical PTFs especially in the case of GMDH network applying input variables which are less time consuming than Ks. In general, it is concluded that GMDH provides more accurate and reliable estimates of cumulative infiltration (a non-readily available characteristic of soil) with a minimum set of input variables (2–4 input variables) and can be promising strategy to model soil infiltration combining the advantages of ANN and MLR methodologies.

Parametric Studies of a Simple Direct Expansion Solar Assisted Heat Pump Using ANN and GA

Abstract This paper reports the use of artificial neural network (ANN) integrated with genetic algorithm (GA) to predict the performance of direct expansion solar assisted heat pump (DX-SAHP), in Calicut (Latitude of 11.15 0 N, longitude of 75.49 0 E), India. The performance parameters such as power consumption, heating capacity, energy performance ratio and compressor discharge temperature of DX-SAHP obtained from the experimentation at different solar intensities and ambient temperatures are used as training data for the network. The back propagation learning algorithm with variants Lavenberg–Marguardt (LM) with 10 neurons in the hidden layer and logistic sigmoid transfer function were used in the network to predict the performance of DX-SAHP. Then those values obtained from the analysis using ANN are optimized further by integrating the ANN procedure with GA. The resulting computations confirmed that the use of ANN integrated with GA gives better optimized values compared to the value obtained from ANN for the performance prediction of DX-SAHP.

Evaluation of the effect of aggregate on concrete permeability using grey correlation analysis and ANN

In this study, the influence of coarse aggregate size and type on chloride penetration of concrete was investigated, and the grey correlation analysis was applied to find the key influencing factor. Furthermore, the proposed 6-10-1 artificial neural network (ANN) model was constructed, and performed under the MATLAB program. Training, testing and validation of the model stages were performed using 81 experiment data sets. The results show that the aggregate type has less effect on the concrete permeability, compared with the size effect. For concrete with a lower w/b, the coarse aggregate with a larger particle size should be chose, however, for concrete with a higher w/c, the aggregate with a grading of 5-20 mm is preferred, too large or too small aggregates are adverse to concrete chloride diffusivity. A new idea for the optimum selection of aggregate to prepare concrete with a low penetration is provided. Moreover, the ANN model predicted values are compared with actual test results, and the average relative error of prediction is found to be 5.62%. ANN procedure provides guidelines to select appropriate coarse aggregate for required chloride penetration of concrete and will reduce number of trial and error, save cost and time.

Influence mechanism of lightweight aggregate on concrete impermeability: prediction by ANN

The influence of lightweight aggregate (LWA) on concrete impermeability is complex, varying according to LWA type, water–cement ratio and curing age. A convenient, quick method to predict impermeability of lightweight concrete is needed. To investigate the influencing mechanism, the microstructure of LWA and the interfacial transition zone are observed, and the water absorbing/releasing character of LWA in cement paste is studied. Chloride diffusion coefficients of concrete with different LWAs and water–cement ratios are also tested at different ages. LWA is found to have both an advantageous and disadvantageous influence on concrete impermeability. Based on 108 experimental datasets collected, an artificial neural network (ANN) model is applied to predict the influence of LWA on chloride diffusivity of concrete. The model is developed, trained and tested using a multi-layer back-propagation method. The predicted values are then compared with actual test results; the average relative error of prediction is found to be 4·14% and coefficient of determination (R2) is 0·97. These results indicate that the developed model has successfully learned the relationship between the different input and output parameters, and can predict the chloride diffusivity with accuracy adequate for practical design purposes. The ANN procedure provides guidelines to select the appropriate LWA for the required chloride diffusivity of concrete, reducing trial and error attempts, and saving both cost and time.

Design Optimization of Reinforced Concrete Frames

In this paper, a design optimization for the reinforced concrete plane frame structure has been done in order to minimize the cost of the concrete and steel for beams and columns by adopting the Artificial Neural Network (ANN) computational model trough the NeuroShell-2 software program. The design procedure conforms to the ACI-318-08 Code. The variables used for design optimization are the width, depth and the area of reinforced steel, including longitudinal reinforcement and shear reinforcement. A three-bay two-story RC frame is modeled with selecting different span length and different load cases. Acceptable design results are obtained from more than 50 examples which are subjected to all the constraints of the ACI Code, using different cross-section sizes and these results are used to train the NeuroShell-2 program. The results obtained demonstrate the efficiency of the ANN procedure for the multi story RC frame design.

Mix proportions for HPC incorporating multi-cementitious composites using artificial neural networks

High performance concrete (HPC) is defined in terms of both strength and durability performance under anticipated environmental conditions. HPC can be manufactured involving up to 10 different ingredients whilst having to consider durability properties in addition to strength. The number of ingredients and the number of properties of HPC, which needs to be considered in its design, are more than those for ordinary concrete. Therefore, it is difficult to predict the mix proportions and other properties of this type of concrete using statistical empirical relationship. An alternative approach is to use an artificial neural network (ANN). Based on the experimentally obtained results, ANN has been used to establish its applicability to the prediction and optimization of mix proportioning for HPC. It was demonstrated that mix proportioning for HPC can be predicted using ANN. However, some trial mixes are necessary for better performance and elimination of material variability factors from place to place. ANN procedure provides guidelines to select appropriate material proportions for required strength and rheology of concrete mixes and will reduce the number of trial mixes.

A new approach to the analysis of alpha spectra based on neural network techniques

The analysis of alpha spectra requires good radiochemical procedures in order to obtain well differentiated alpha peaks in the spectrum, and the easiest way to analyze them is by directly summing the counts obtained in the Regions of Interest (ROIs). However, the low-energy tails of the alpha peaks frequently make this simple approach unworkable because some peaks partially overlap. Many fitting procedures have been proposed to solve this problem, most of them based on semi-empirical mathematical functions that emulate the shape of a theoretical alpha peak. The main drawback of these methods is that the great number of fitting parameters used means that their physical meaning is obscure or completely lacking. We propose another approach—the application of an artificial neural network. Instead of fitting the experimental data to a mathematical function, the fit is carried out by an artificial neural network (ANN) that has previously been trained to model the shape of an alpha peak using as training patterns several polonium spectra obtained from actual samples analyzed in our laboratory. In this sense, the ANN is able to learn the shape of an actual alpha peak. We have designed such an ANN as a feed-forward multi-layer perceptron with supervised training based on a back-propagation algorithm. The fitting procedure is based on the experimental observables that are characteristic of alpha peaks—the number of counts of the maximum and several peak widths at different heights. Polonium isotope spectra were selected because the alpha peaks corresponding to 208Po, 209Po, and 210Po are monoenergetic and well separated. The uncertainties introduced by this fitting procedure were less than the counting uncertainties. This new approach was applied to the problem of resolving overlapping peaks. Firstly, a theoretical study was carried out by artificially overlapping alpha peaks from actual samples in order to test the ability of the ANN to resolve each peak. Then, the ANN procedure was checked by determining the activity levels of different spectra obtained from certified samples for which one knows a priori the radioactive content, and its results were compared with those of other methods.

Explanation of the observed structure of functional feeding groups of aquatic macro-invertebrates by an ecological model and the maximum exergy principle

The maximum exergy principle presumes that an ecosystem with its species composition tends to move as far away from thermodynamic equilibrium as possible under the prevailing conditions. If this principle is valid, the species composition (or rather their properties) found in any ecosystem represent the best ‘solution’ among the possible ones for the ecosystem to move away from thermodynamic equilibrium. The application of structurally biogeochemical models (Integration of Ecosystem Theories: A Pattern (2002)) with exergy as goal function could be used to explain the biological structure observed—because it would be the structure giving the highest exergy of the system under the prevailing conditions. Five cases have been examined by this method combining modeling and the maximum exergy principle. In all five cases, the observed biological structure (sizes and biomass concentrations) gave the highest exergy. In principle, this method with a high causality could replace the application of artificial neural network (ANN) or any other predictive approach to find a relationship between dominant species and water quality in an aquatic ecosystem; but the method is unfortunately too time consuming to be applicable on a medium to large data-set covering many sites. Therefore, it is recommended to develop a method to incorporate the maximum exergy principle in the ANN procedure.