Feed-forward Back-propagation Artificial Neural Network Research Articles

Modeling and optimization of polymer enhanced ultrafiltration using hybrid neural-genetic algorithm based evolutionary approach

A stochastic genetic algorithm (GA) based strategy along with artificial neural network (ANN) was applied to optimize the retention of reactive red 120 (RR 120) dye from its aqueous solutions by way of polymer (polyethyleneimine (PEI)) enhanced ultrafiltration (PEUF). The optimal feed forward back propagation ANN (4-10-1) model network, trained initially through Levenberg–Marquardt (LM) algorithm, was suitably manoeuvred by the GA approach to predict the membrane performance index (PFI) response, evaluated as the product of dye rejection and permeation flux, for a randomly generated population of chromosomes. Each chromosome was constituted by four principal genes, namely, cross-flow rate, transmembrane pressure, polymer to dye ratio, and pH. The local exploitation capacity of the canonical GA was enhanced further by combining hill-climbing (HC) local search with the optimization levels of standard GA. The near-optimal and economically feasible factor levels were predicted by the hybrid ANN-GA-HC strategy, keeping PFI maximization and the constrained PEUF process dynamics in perspective; the optimal process factor settings experimentally yielded a pragmatic PFI of 143.8L/m2h, corresponding to high (99.9%) dye rejection, and a satisfactory permeation flux (144L/m2h).

Artificial neural network model using ultrasonic test results to predict compressive stress in concrete

This study focused on modeling the behavior of the compressive stress using the average strain and ultrasonic test results in concrete. Feed-forward backpropagation artificial neural network (ANN) models were used to compare four types of concrete mixtures with varying water cement ratio (WC), ordinary concrete (ORC) and concrete with short steel fiber-reinforcement (FRC). Sixteen (16) <TEX>$150mm{\times}150mm{\times}150mm$</TEX> concrete cubes were used; each contained eighteen (18) data sets. Ultrasonic test with pitch-catch configuration was conducted at each loading state to record linear and nonlinear test response with multiple step loads. Statistical Spearman's rank correlation was used to reduce the input parameters. Different types of concrete produced similar top five input parameters that had high correlation to compressive stress: average strain (<TEX>${\varepsilon}$</TEX>), fundamental harmonic amplitude (A1), <TEX>$2^{nd}$</TEX> harmonic amplitude (A2), <TEX>$3^{rd}$</TEX> harmonic amplitude (A3), and peak to peak amplitude (PPA). Twenty-eight ANN models were trained, validated and tested. A model was chosen for each WC with the highest Pearson correlation coefficient (R) in testing, and the soundness of the behavior for the input parameters in relation to the compressive stress. The ANN model showed increasing WC produced delayed response to stress at initial stages, abruptly responding after 40%. This was due to the presence of more voids for high water cement ratio that activated Contact Acoustic Nonlinearity (CAN) at the latter stage of the loading path. FRC showed slow response to stress than ORC, indicating the resistance of short steel fiber that delayed stress increase against the loading path.

Prediction of cyanobacterial blooms in the Dau Tieng Reservoir using an artificial neural network

An artificial neural network (ANN) model was used to predict the cyanobacteria bloom in the Dau Tieng Reservoir, Vietnam. Eight environmental parameters (pH, dissolved oxygen, temperature, total dissolved solids, total nitrogen (TN), total phosphorus, biochemical oxygen demand and chemical oxygen demand) were introduced as inputs, whereas the cell density of three cyanobacteria genera (Anabaena, Microcystis and Oscillatoria) with microcystin concentrations were introduced as outputs of the three-layer feed-forward back-propagation ANN. Eighty networks covering all combinations of four learning algorithms (Bayesian regularisation (BR), gradient descent with momentum and adaptive learning rate, Levenberg–Mardquart, scaled conjugate gradient) with two transfer functions (tansig, logsig) and 10 numbers of hidden neurons (6–16) were trained and validated to find the best configuration fitting the observed data. The result is a network using the BR learning algorithm, tansig transfer function and nine neurons in the hidden layer, which shows satisfactory predictions with the low values of error (root mean square error=0.108) and high correlation coefficient values (R=0.904) between experimental and predicted values. Sensitivity analysis on the developed ANN indicated that TN and temperature had the most positive and negative effects respectively on microcystin concentrations. These results indicate that ANN modelling can effectively predict the behaviour of the cyanobacteria bloom process.

Optimisation of ANN topology for predicting the rehydrated apple cubes colour change using RSM and GA

In this study, an efficient optimisation method by combining response surface methodology (RSM) and genetic algorithm (GA) is introduced to find the optimal topology of artificial neural networks (ANNs) for predicting colour changes in rehydrated apple cubes. A multi-layered feed-forward backpropagation ANN model of algorithms was developed to correlate one output (colour change) to four input variables (drying air temperature, drying air velocity, temperature of distilled water and rehydration time). A predictive model for ANN topology in terms of the best mean squared error (MSE) performance on validation samples was created using RSM. RSM model was integrated with an effective GA to find the optimum topology of ANN. The optimum ANN had minimum MSE when the number of hidden neurons, learning rate, momentum constant, number of epochs and number of training runs were 13, 0.33, 0.89, 3869 and 3, respectively. MSE of optimal ANN topology on validation samples was 0.0072095. It turned out that the optimal ANN topology can be considered as more precise for predicting colour change in the rehydrated apple cubes. Mean absolute error and regression coefficient (R) of the optimal ANN topology were determined as 0.0259 and 0.96475 for training, 0.0399 and 0.95243 for testing and 0.0264 and 0.95151 for validation data sets. The results of the testing model on new samples showed excellent agreement between the actual and predicted data with coefficient of determination R2 = 0.97.

Estimation of biogas and methane yields in an UASB treating potato starch processing wastewater with backpropagation artificial neural network

Three-layered feedforward backpropagation (BP) artificial neural networks (ANN) and multiple nonlinear regression (MnLR) models were developed to estimate biogas and methane yield in an upflow anaerobic sludge blanket (UASB) reactor treating potato starch processing wastewater (PSPW). Anaerobic process parameters were optimized to identify their importance on methanation. pH, total chemical oxygen demand, ammonium, alkalinity, total Kjeldahl nitrogen, total phosphorus, volatile fatty acids and hydraulic retention time selected based on principal component analysis were used as input variables, whiles biogas and methane yield were employed as target variables. Quasi-Newton method and conjugate gradient backpropagation algorithms were best among eleven training algorithms. Coefficient of determination (R2) of the BP-ANN reached 98.72% and 97.93% whiles MnLR model attained 93.9% and 91.08% for biogas and methane yield, respectively. Compared with the MnLR model, BP-ANN model demonstrated significant performance, suggesting possible control of the anaerobic digestion process with the BP-ANN model.

Forecasting the BIST 100 Index Using Artificial Neural Networks with Consideration of the Economic Calendar

Artificial Neural Networks (ANN) is an analysis method that mimics the operating principle of the human brain. The problem-solving skills and the high rate of success in solving complex problems of ANN, relative to the other traditional methods has made it a preference as well in the fields of finance and economics. I used Feed-forward back-propagation ANN (BPN-ANN) similar to related studies to forecast the indices, and employed trial-and-error method for the parameters like number of layers and neurons at layers to reach the optimal ANN structure. Extant literature has used important macroeconomic variables like inflation, interest rates and money supply to shape the future expected value of the dependent variables of the index. This study is followed a new and unique perspective from established literature to make use of the advantages of ANN to forecast on Turkish stock market index. Further, the study used the daily data between 29th July and 15th November of 2015 for BIST 100, including codified economic calendar events and major parities, dollar index and indices as variables. Due to the expected technical superiority of the system as a result of the nature of the variables to be used, a higher forecast performance is expected, even with the occurrence of events outside the scope, like important political developments. The expected results of this study, other than being an immense contribution to literature will be developed into an important tool that can be utilized by investors. Among several models, 18-20-1 structured MLP has best explanatory level with 0.893 R 2 and 0.207 MSE values. This was followed the 18-16-1 structured MLP which had the minimum MSE as 0.025 and 0.88 R 2 . These are models 1 and 2 respectively. It is also observed that ECE (Economic Calendar Events) and ‘Other’ variables have notable effects that explain on the fluctuation of the index. Similarly, the two variables have shown their significant in other models as well. Prediction of opening is more successful than closing. ECE has greater success forecasting open prices.

Computing the pull-in voltage of fixed–fixed micro-actuators by artificial neural network

In this study, a feed-forward backpropagation (FFBP) artificial neural network (ANN) model based on multilayer perceptron (MLP) for computing the pull-in voltage value of fixed–fixed micro-actuators is presented. The proposed ANN model is trained using the numerous pull-in voltage value results of clamped–clamped actuators, which have various physical properties, simulated by a software that employs the finite element method (FEM). As the presented ANN model has been trained for the simulation data, it implicitly takes into account the fringing field, mid-plane stretching and size effects. The accuracy and robustness of the results obtained by the proposed ANN model are verified by comparing with those of the theoretical ones through the simulation and experimental studies previously presented in the literature. The average percentage errors for training and testing data are found to be 0.30 and 0.40 %, respectively, while the maximum percentage error for literature data is calculated as 1.96 %. These results show that the main advantage of the presented ANN model gives satisfying pull-in voltage results over solution space of the simulations with effortless way.

Alzheimer Disease Detection in MRI Using Curvelet Transform with K-NN

Identification of Alzheimer disease at a preliminary stage from Brain MR Image is a major concern in the field of neurosciences. This paper proposes the novel approach for the recognition of Alzheimer disease from brain MR images. We applied method of feature extraction and pattern recognition technique for the diagnosis of MR image. The proposed algorithm consists of three steps: feature extraction, feature reduction, and classification. In the first step, curvelet transform is used for the extraction of the feature from MR images. In the second step, Principal Component Analysis (PCA) is applied and the dimension of the extracted feature is reduced. In the final step, the PCA output acts as an input to k-nearest neighbor (K-NN) classifier and Feed-forward Backpropagation Artificial Neural Network. The structural Magnetic Resonance Image (SMRI) data are selected from the Open Access series of imaging studies (OASIS) database. The experimental result shows that feature extracted with curvelet transformper forms better than 2D discrete wavelet transform (DWT) with both K-NN and ANN and separates Alzheimer disease (AD) and Healthy controls (HC) with high performance.

Modeling and simulation of xylitol production in bioreactor by Debaryomyces nepalensis NCYC 3413 using unstructured and artificial neural network models

This study examines the use of unstructured kinetic model and artificial neural networks as predictive tools for xylitol production by Debaryomyces nepalensis NCYC 3413 in bioreactor. An unstructured kinetic model was proposed in order to assess the influence of pH (4, 5 and 6), temperature (25°C, 30°C and 35°C) and volumetric oxygen transfer coefficient kLa (0.14h−1, 0.28h−1 and 0.56h−1) on growth and xylitol production. A feed-forward back-propagation artificial neural network (ANN) has been developed to investigate the effect of process condition on xylitol production. ANN configuration of 6-10-3 layers was selected and trained with 339 experimental data points from bioreactor studies. Results showed that simulation and prediction accuracy of ANN was apparently higher when compared to unstructured mechanistic model under varying operational conditions. ANN was found to be an efficient data-driven tool to predict the optimal harvest time in xylitol production.

Using artificial neural network to predict velocity of sound in liquid water as a function of ambient temperature, electrical and magnetic fields

One of the main thermophysical properties of liquid water is velocity of sound. However, the effect of different externalities on velocity of sound in liquid water is not well known. Therefore, in current study, by designing an artificial neural network (ANN) velocity of sound in liquid water under different externalities is predicted. Selected externalities are ambient temperature from 272.65K to 348.43K, different electrical fields in range of 0V/m to 4.03E + 9V/m and magnetic fields in range of 0–10.0594T. To prepare of reference dataset for entry to ANN, numerical and experimental data as macroscopic reference data are extracted from microscopic characteristic of water HB strength. In order to achieve an appropriate ANN, ANN architecture sensitivity analysis is conducted by using an iterative algorithm. Learning procedure in the selected feed-forward back propagation ANN is done by hyperbolic transfer functions. Also, Levenberg–Marquardt algorithm is utilized for the optimization process. ANNs output showed that the maximum MSE in prediction of velocity of sound is 0.00066. Also, the minimum of correlation coefficient in prediction of velocity of sound is 0.99131. Based on the ANNs outputs, weights and bias, an equation to predict of velocity of sound in liquid water under intended externalities is proposed. Also, according to weight sensitivity analysis input of electrical fields with 63% relative importance percentage has a grater impression on the response variable of velocity of sound in liquid water.

Quantitative Structure-Property Relationship Study to Predict the Retention Times of Some Volatile Compounds in Rosé Wines

In this work, quantitative structure-property relationship (QSPR) models for prediction of retention time (RT) of 47 volatile compounds in rosé wines on the basis of their molecular structures were developed by applying multiple linear regression (MLR) and artificial neural network (ANN). A Levenberg- Marquardt algorithm trained feed-forward back-propagation artificial neural network (ANN) was employed. The data were randomly divided into 29 training and 9 validation sets. For comparison purpose, multiple linear regression (MLR) model of the same data was developed. Cross-validation was used to validate the QSPR models. Also, the application of the models for prediction of external set’s retention times of compounds without any contribution to model development steps was another validation process. The MLR model yielded marginally acceptable statistics with test correlation R2 = 0.993 and mean squared error (MSE) = 0.0075. Not surprisingly, the ANN model was significantly more accurate with test correlation R2 = 0.995 and mean squared error (MSE) = 0.00013.

Modeling the Effects of Cu Content and Deformation Variables on the High-Temperature Flow Behavior of Dilute Al-Fe-Si Alloys Using an Artificial Neural Network.

The hot deformation behavior of Al-0.12Fe-0.1Si alloys with varied amounts of Cu (0.002–0.31 wt %) was investigated by uniaxial compression tests conducted at different temperatures (400 °C–550 °C) and strain rates (0.01–10 s−1). The results demonstrated that flow stress decreased with increasing deformation temperature and decreasing strain rate, while flow stress increased with increasing Cu content for all deformation conditions studied due to the solute drag effect. Based on the experimental data, an artificial neural network (ANN) model was developed to study the relationship between chemical composition, deformation variables and high-temperature flow behavior. A three-layer feed-forward back-propagation artificial neural network with 20 neurons in a hidden layer was established in this study. The input parameters were Cu content, temperature, strain rate and strain, while the flow stress was the output. The performance of the proposed model was evaluated using the K-fold cross-validation method. The results showed excellent generalization capability of the developed model. Sensitivity analysis indicated that the strain rate is the most important parameter, while the Cu content exhibited a modest but significant influence on the flow stress.

Flow time and product cost estimation by using an artificial neural network (ANN): A case study for transformer orders

ABSTRACTIn electromechanical industrial corporations, determining the production cost of the orders according to the technical specifications demanded by the customer has great importance in giving an accurate price offer. Labor cost is one of the important and most variable cost components that must be estimated in order to give an accurate price offer. In this study, a feed-forward back-propagation artificial neural network (FF-BPN) is used to predict the flow times of power transformer orders of a transformer producer according to the technical specifications given by the customer. The results of this study show that the prediction capability of an artificial neural network is very good for this type of problem and results in better cost estimation than current company practice. A case study is carried out for a manufacturer of electrical transformers in Turkey.

Enhanced Adsorption of Hexavalent Chromium onto Magnetic Calcium Ferrite Nanoparticles: Kinetic, Isotherm, and Neural Network Modeling

Calcium ferrite nanoparticles with super-paramagnetic behavior were synthesized via simple chemical precipitation method for effective removal of hexavalent chromium from aqueous media. The properties of synthesized nanoparticles were studied by X-ray diffraction (XRD), field emission scanning electron microscope (FESEM), Fourier transform infrared (FTIR) spectroscopy, Brunauer-Emmett-Teller (BET), and vibrating sample magnetometer (VSM) measurements. The ferrite nanoparticles have shown polycrystalline nature and high BET specific surface area (229.83 m2/g) with active functional groups on the surface. The adsorption process follows second-order kinetics with the involvement of intra-particle diffusion and adsorption capacity as much as 124.11 mg/g was determined from the Langmuir isotherm. The thermodynamic analysis revealed that the adsorption process was feasible, spontaneous, and exothermic in nature. A three-layer feed-forward back-propagation artificial neural network (ANN) model was employed to predict the removal (%) of Cr(VI) ions as output. Optimal ANN network (4:8:1) shows the minimum mean squared error (MSE) of 0.00161 and maximum coefficient of determination (R2) of 0.984. The adsorption process is mostly influenced by solution pH and followed by adsorbent dosage, initial Cr(VI) concentration, and contact time as illustrated by sensitivity analysis. With small size and high surface area, biocompatibility, ecofriendly nature, easy magnetic separation, and enhanced adsorption capacity towards Cr(VI), calcium ferrite nanoparticles will find its potential application in wastewater remediation.

A Comparative Study of ANN and CFD Modelling for Pressure Drop Prediction in a Fluidized Bed with Internals

In the present study, experiments on fluidized beds were conducted by varying hydrodynamic parameters such as air velocity, bed height and internal spacing and these data were used in the development of a model using artificial neural network (ANN) and simulation using computational fluid dynamics (CFD). Staggered rod internals of width 25% and 50% of the column diameter were used. A two-layered feed-forward back propagation ANN model was developed to predict the pressure drop with 15 neurons and it gave a high R2 value of 0.9966. The model prediction and experimental data are in good agreement. The results from CFD studies are also in good agreement with the experimental data. Pressure drop was found to reduce in the presence of internals and it reduced further with a decrease in internal spacing due to continuous splitting of bubbles. The average error in CFD predicted pressure drop is 9.56% and ANN predicted pressure drop is 0.95%.

Neurocomputational Models for Parameter Estimation of Circular Microstrip Patch Antennas

Neurocomputational models eliminates the complex, lengthy and time consuming mathematical procedures for design, analysis and calculating performance parameters of Microstrip antenna. No single ANN based model has been proposed till date for calculating all parameters of circular microstrip antennas simultaneously. This paper presents a Neuro-Computational (NC) approach for estimation of all performance parameters such as Return Loss (RL), Voltage Standing Wave Ratio (VSWR), resonant frequency (fr), Band-Width (BW), Gain(G), Directivity(D) and antenna efficiency(η) of Circular Microstrip Patch Antenna (CMPA) simultaneously. The difficulty in calculating the parameters of these antennas lies due to the involvement of a large number of physical parameters including their associated optimal values. It is indeed very difficult to formulate an exact numerical solution merely on practical observations based empirical studies. In order to circumvent this problem, an alternative solution is achieved using artificial neural network (ANN). Feed-Forward Back-Propagation Artificial Neural Network (FFBP-ANN) trained with Levenberg-Marquardt algorithm is used for estimation of different performance parameters of CMPA. The results of NC estimation are in very agreement with simulated, measured and theoretical results.

Impulsive noise reduction for transient Earth voltage‐based partial discharge using Wavelet‐entropy

Partial discharge (PD) is caused by the localised electrical field intensification in insulating materials. Early detection and accurate measurement of PD are very important for preventing premature failure of the insulating material. Detection of PDs in metal-clad apparatus through the transient Earth voltage method is a promising approach in non-intrusive on-line tests. However, the electrical interference from background environment remains the major barrier to improving its measurement accuracy. In this study, a wavelet-entropy-based PD de-noising method has been proposed. The unique features of PD are characterised by combining wavelet analysis that reveals the local features and entropy that measures the disorder. With such features, a feed-forward back-propagation artificial neural network is adopted to recognise the actual PDs from noisy background. Comparing with other methods such as the energy-based method and the similarity-comparing method, the proposed wavelet-entropy-based method is more effective in PD signal de-noising.

Tensile strength prediction of dissimilar friction stir-welded AA6351–AA5083 using artificial neural network technique

Friction stir welding provides an alternative method of joining aluminum in a reliable way. Anticipation of the joint efficiency is then a necessary step to optimize the process of the welding operation. In the light of this, artificial neural network (ANN) technique may then be applied as a reliable method for simulating and predicting the durability of the joints for different process parameters. In the present work, an ANN model is presented that predicts the ultimate tensile strength of friction stir-welded dissimilar aluminum alloy joints. Four parameters were considered including tool pin profile (straight square, tapered square, straight hexagon, straight octagon and tapered octagon), rotational speed, welding speed and axial force. Experimental tests were conducted according to a four-parameter five level central composite design. A feed-forward back propagation ANN with a single hidden layer comprising 20 neurons was employed to simulate the ultimate tensile strength (UTS) of the joints. The neural network was trained using the data obtained from the experimental work. A comparison between the experimental and simulated data showed that the ANN model reliably predicted the UTS of dissimilar aluminum alloy friction stir-welded joints. The models developed were capable of predicting values with less than 5 % error. Furthermore, the effect of different process parameters on the tensile behavior of dissimilar joints was also investigated and reported upon.

Artificial neural network approach for moiré fringe center determination

The moire effect has been used in high-accuracy positioning and alignment systems for decades. Various methods have been proposed to identify and locate moire fringes in order to relate the pattern information to dimensional and displacement measurement. These methods can be broadly categorized into manual interpretation based on human knowledge and image processing based on computational algorithms. An artificial neural network (ANN) is proposed to locate moire fringe centers within circular grating moire patterns. This ANN approach aims to mimic human decision making by eliminating complex mathematical computations or time-consuming image processing algorithms in moire fringe recognition. A feed-forward backpropagation ANN architecture was adopted in this work. Parametric studies were performed to optimize the ANN architecture. The finalized ANN approach was able to determine the location of the fringe centers with average deviations of 3.167 pixels out of 200 pixels ( ≈1.6% ) and 6.166 pixels out of 200 pixels ( ≈3.1% ) for real moire patterns that lie within and outside the training intervals, respectively. In addition, a reduction of 43.4% in the computational time was reported using the ANN approach. Finally, the applicability of the ANN approach for moire fringe center determination was confirmed.

Formulation development, modeling and optimization of emulsification process using evolving RSM coupled hybrid ANN-GA framework

In the present work, a mathematical and statistical approach was adopted to study the formation and stability of oil-in-water emulsion with an integrated hybrid genetic algorithm (GA) coupled with feed-forward back-propagation artificial neural network (BPANN) and response surface methodology (RSM) based on Box–Behnken design (BBD). The input parameters were oil concentration (10–50%, v/v), surfactant concentration (0.1–2%, w/v), stirring speed (2000–6000rpm) and stirring time (5–20min). The output parameter was relative emulsion volume expressed as emulsion stability index (ESI24). In the proposed hybrid GA model, outputs of BP-ANN model were used as initial population settings and RSM-BBD generated model equation was used as a fitness function. Error analysis was performed on the model fit to the experimental data using sum of square error (SSE), mean square error (MSE) and relative percent error (RPD). The optimum condition predicted by the hybrid GA was 0.913 of ESI24, with 4.70% error under 50% (v/v) oil concentrations, 2% (w/v) surfactant concentrations, 5691rpm and 5min stirring time. The proposed hybrid GA model was found to be useful for the optimization of process parameters for emulsion formation and stability analysis.