Artificial Neural Network Model Method Research Articles

Presentation of artificial neural network models based on optimum theories for predicting accident severity on rural roads in Iran

This study focuses on predicting the number and severity of rural accidents using nine accident-related variables by the multi-layer perceptron (MLP) approach as a type of artificial neural network modeling method. The models were developed using the input parameters of shoulder width, roadway width, roadside hazard, access density, passing zone ratio, speed limit, pavement condition index, shoulder rumble strips, and centerline rumble strips. MLP models were created based on accident data that occurred from 2019 to 2020 on the Tehran–Qom and Tehran-Saveh rural roads in Iran. In order to achieve the highest accuracy, twenty MLP models have been built with various structures. The MLP model, consisting of a multilayer feedforward network with hidden sigmoid and softmax output, has been used in this study. In this regard, an attempt was made to present an optimum theory to select the best model. The most accurate model was chosen based on the R-value and root mean square error (RMSE), mean absolute error (MAE), f, SD, and R2. Results indicated that the R-value obtained from the optimum model was 0.912, representing the accurate performance of the selected model. In addition, access density, roadside hazard, and roadway width were identified as the most significant variables.

Modeling of nonlinear characteristics of switched reluctance motor based on optimized RBF neuron network

The strong nonlinearity of switched reluctance motor (SRM) makes it difficult to establish an accurate mathematical model of its electromagnetic characteristics. In this paper, an artificial neural network modeling method is adopted to estimate the nonlinear flux linkage characteristics of a four-phase SRM. Based on the samples obtained by experimental measurement, a radial basis function neural network is trained and further optimized using the K-Means algorithm. The modeling accuracy of the optimized flux linkage model is analyzed in detail and also verified by dynamic simulation of a direct instantaneous torque control system of the motor.

Optimizing the material and printing parameters of the additively manufactured fiber-reinforced polymer composites using an artificial neural network model and artificial bee colony algorithm

Fiber-reinforced polymer composite (FRP) has always been attracting the attention of researchers due to its multiple applications as a primary material for manufacturing structural elements and parts or as a reinforcing material for the structural elements. Controlling the quality of the additively manufactured FRP is critical to guarantee an FRP material of high performance for different applications. Therefore, in this research, a machine learning model based on data collected from experimental studies was developed using the artificial neural network (ANN) modeling method to control the quality of FRP manufactured by fused deposition modeling (FDM) printing technology. ANN model predicts the tensile properties, particularly ultimate tensile strength (UTS), of the FRP as a function of 7 material and printing parameters, which are fiber tensile strength, polymer tensile strength, layer thickness, fiber volume fraction, fiber form, infill pattern, and fiber orientation. A parametric study was conducted to confirm the credibility of the developed ANN model in representing the pairwise relationship between each input parameter and the predicted UTS. The UTS of the FRP was maximized by optimizing the printing and material parameters by using the artificial bee colony (ABC) algorithm. Several case studies were conducted to prove the efficiency of the coupled ANN model and optimization algorithm. The ANN modeling method and ABC algorithm were codified and applied using MATLAB software. The results showed that the developed ANN model could accurately predict the UTS of FRP. The parametric study confirmed that the developed ANN model was valid for representing the pairwise relationships between the input parameters and the predicted UTS. Also, it was found that the fiber properties, fiber form, fiber orientation, and fiber volume fraction have the highest influence on the UTS of FRP manufactured by FDM compared with other parameters. Furthermore, the ABC optimization algorithm can design the input parameters (i.e., material and printing parameters) to obtain an FRP with maximum UTS. For future studies, it is recommended to include more input parameters, such as printing speed, nozzle temperature, and infill density, in addition to outputs, such as impact and flexural properties.

Modeling and predicting of water production by capacitive deionization method using artificial neural networks

Water desalination is a method to deal with water shortage that today considered as a way to meet the growing human demand. Obtaining freshwater at low cost guarantees the stability of the desalination method. Capacitive Deionization (CDI) method is a new technology for deionization of water with saline properties, the least energy, and environmental pollution. In this paper, an Artificial Neural Network (ANN) model was developed to estimate the amount of water produced by the CDI method based on experimental data. The backpropagation Multi-Layer Perceptron (MLP) model with Adam (Adaptive Moment Estimation) learning algorithm with two hidden layers was used. To evaluate performance, RMSE, MAE, MSE and R2 mathematical indices were used. The input layer had three variables that include the amount of initial solution concentration, the flow rate, and the amount of voltage applied to the cell. The output layer had a neuron as the amount of percentage of water desalination. For estimating the amount of salt removal percentage, the ANN modeling method was acceptable. The amount of correlation coefficient between experimental laboratory data and data estimated by ANN for training data was equal to 0.972. The overall correlation coefficient was 0.90 that for estimating the amount of salt removal percentage by CDI method is acceptable. The RMSE values for testing and training data were 0.008 and 0.003, respectively. In the statistical study of the effect of variables, it was found that by CDI method the initial concentration of salt has an inverse relationship with the amount of salt removed from water. Also, the amount of applied voltage has a direct relationship and the amount of feed rate has an inverse relationship with the amount of salt removed from water.

Automated Multilayer Neural Network Structure Adaptation Method With l₁ Regularization for Microwave Modeling

Artificial neural network (ANN) model development for microwave components principally includes two parts of work, i.e., data sampling and model structure adaptation. In existing various ANN modeling methods, the model structure adaptation process mainly focuses on adjusting the number of neurons within each hidden layer of ANN while keeping the number of layers unchanged. To make the ANN modeling process more flexible and efficient, an automated multilayer neural network structure adaptation method with <inline-formula> <tex-math notation="LaTeX">$ {l_{1}}$ </tex-math></inline-formula> regularization is proposed in this letter. We propose a new ANN model structure combining multilayer perceptron (MLP) and additional connections between the output layer and each hidden layer/input layer. A new training scheme with <inline-formula> <tex-math notation="LaTeX">$ {l_{1}}$ </tex-math></inline-formula> regularization is proposed to automatically determine the final model structure with user-desired model accuracy. Using the proposed model structure adaptation method, both the number of layers and the number of neurons within each layer of the final ANN model can be adaptively determined to address different needs for different microwave modeling problems. The proposed method is demonstrated by two microwave filter modeling examples in which the model development process achieves a time saving of at least 40&#x0025; over existing methods.

Study on HHV prediction of municipal solid wastes: A machine learning approach

In this work, artificial neural network (ANN) models and particle swarm optimization (PSO) models based on machine learning were built to predict the HHVs of MSW quickly. Four kinds of BP ANN models and two PSO models were built using proximate analysis and ultimate analysis of 33 MSW samples as input variables. As a comparison, three classical linear empirical models employed from publications were also used. The modeling results show that the input variables had significant influence on the prediction accuracy. The ANN model based on proximate analysis had lower precision for HHV estimation. The ultimate analysis had better prediction performance while the combination of ultimate analysis and proximate analysis (ANN-4 model) had the best accuracy. With regard to ANN-4 model, the largest relative deviation for all samples was lower than 10%. Due to the complex composition of MSW, the linear empirical model was not suitable for accurate prediction of the calorific value of MSW. Nonlinear empirical formulas obtained by PSO models improved the prediction performance for most samples. In general, the ANN modeling method could predict the thermochemical properties of MSW and provide rapid and effective guidance for the operation of MSW incineration process.

Modelling of multilinear gradient retention time of bio-sweetener rebaudioside A in HPLC analysis

Artificial neural network (ANN), as one of the artificial intelligence methods, has been widely using in HPLC studies for modelling purposes. Stevia rebaudiana is an important industrial plant due to its bio-sweetener molecule, rebaudioside-a, in its leaves. Although rebaudioside-a is up to 300-fold sweeter than sucrose, its calorie is almost zero. In this study, HPLC optimization of rebaudioside-a was studied and the optimization data based on multilinear gradient retention times were modelled by ANN. The input parameters were selected as concentrations, column temperatures, initial acetonitrile percentage for the first step of gradient elution, initial acetonitrile percentage for the second step of gradient elution, slope of acetonitrile, wavelengths, flow rates. The retention time was the output. Also, dried S. rebaudiana leaves were extracted and the concentrations were evaluated by HPLC. According to the ANN results, the most effective parameters on the prediction of non-linear gradient retention time for rebaudioside-a were found as flow rate and initial acetonitrile percentage for the second step of gradient. The best back propagation was selected as Levenberg-Marquardt algorithm. The highest rebaudioside-a level was found as 96.53 ± 6.36 μg mL-1. ANN modelling methods can be used in preparative HPLC applications to estimate the retention time of steviol glycosides.

Multi-objective optimum design of an alpha type Stirling engine using meta-models and co-simulation approach

An alpha type Stirling engine was optimized using meta-models considering uninterrupted electric power supply concurrently with natural gas combi boilers at homes during electricity interruptions. To predict and optimize the power and efficiency of the designed Stirling engine, an artificial neural network (ANN) model was trained as a meta-model. The ANN modeling method was used in solving a multi-objective Pareto optimization problem under some constraints to determine the optimum engine parameters. The design parameters were swept volume, hot and cold cylinder temperatures, gas constant, charge pressure and engine operation speed. Feed forward and Levenberg–Marquardt back propagation algorithms were evaluated to determine the best resulting network architecture that was found as 6–12–8–1. Subsequently, the fraction of variance (Rf) value was calculated close to 1 and the absolute mean error percentage (AMEP) was calculated as 6.07%. Trained ANN model was used in solving the multi-objective optimization problem. Using the optimum design parameters, the meta-model predicted the power as 73.3 W and efficiency as 32.2%. Co-simulation approach was followed to verify the optimization results, and the nominal power output and corresponding efficiency were calculated using the Schmidt theory and the calibrated 1-D model created by the GT-Suite software that yield respectively, 144.6 W and 85.8 W for the power and 35% and 35.1% for the cycle efficiency. Consequently, the use of an ANN model in solving the associated optimization problem proved itself as a fast, accurate enough and powerful method to find the optimum design parameters and predict the engine performance.

Artificial neural network modeling method incorporating spatial structure

Artificial neural network modeling method incorporating spatial structure

Comparison of Deformation Models of Flexible Manipulator Joints for Use in DEMO

The joint deformation contributes significantly to final end-effector displacement of a manipulator, especially when the manipulator is in the form a serial kinematics with long links. When the manipulators employed in the DEMO are under the heavy payload, the deformation of manipulator is inevitable and the magnitude is significant. In order to maneuver the large object through several via points with high positioning accuracy in the remote handling process of DEMO, the real-time computation of manipulator deformation has to be conducted, which is crucial to the DEMO adaptive position control system for the displacement compensation. Three computation-effective deformation modeling methods are proposed in this paper, which are parametric modeling method, nonparametric deterministic artificial neural network (ANN) modeling method, and nonparametric Bayesian ANN modeling method, respectively. A specific joint in a boom equipped in a telescopic articulated remote mast is taken as the study object in this paper. A nodal deformation in the joint is investigated by three modeling methods, respectively. The parametric deformation model is derived by using the structural mechanics, whose parameters are identified by using the Markov chain Monte Carlo (MCMC) method; the deformation model of deterministic ANN is trained by using the Levenberg–Marquardt method; and the deformation model of the Bayesian ANN is trained by using the MCMC method. The results show that the parametric model from the structural mechanics is linear and is incompetent in the deformation modeling when the nonlinearity presents; both the deterministic and Bayesian ANNs are capable of model the nodal deformation of joint. The performance of both the deterministic network and Bayesian network cannot rival for one another in the application scenario of this paper. The training of the Bayesian network can provide the criterions for estimation of possible ranges of the modeling outputs from its probabilistic distribution curves, and the judgment of proper size of network.

Comparison of Kinetic-based and Artificial Neural Network Modeling Methods for a Pilot Scale Vacuum Gas Oil Hydrocracking Reactor

An artificial neural network (ANN) and kinetic-based models for a pilot scale vacuum gas oil (VGO) hydrocracking plant are presented in this paper. Reported experimental data in the literature were used to develop, train, and check these models. The proposed models are capable of predicting the yield of all main hydrocracking products including dry gas, light naphtha, heavy naphtha, kerosene, diesel, and unconverted VGO (residue). Results showed that kinetic-based and artificial neural models have specific capabilities to predict yield of hydrocracking products. The former is able to accurately predict the yield of lighter products, i.e. light naphtha, heavy naphtha and kerosene. However, ANN model is capable of predicting yields of diesel and residue with higher precision. The comparison shows that the ANN model is superior to the kinetic-base models. © 2013 by Authors, Published by BCREC Group. This is an open access article under the CC BY-SA License (https://creativecommons.org/licenses/by-sa/4.0)

Estimating High Spatial Resolution Clear-Sky Land Surface Upwelling Longwave Radiation From MODIS Data

Surface upwelling longwave radiation (LWUP) is an important component in the surface radiation budget. Existing satellite-derived LWUP data sets are too coarse to support high-resolution numerical models, and their accuracy needs to be improved. In this paper, we evaluate three methods for estimating clear-sky land LWUP from the Moderate Resolution Imaging Spectroradiometer (MODIS) data at 1-km spatial resolution. The three methods are as follows: (1) the temperature-emissivity method; (2) the linear model method; and (3) the artificial neural network (ANN) model method. Methods 2 and 3 are new methods based on extensive radiative transfer simulations and statistical analysis. We explicitly considered surface emissivity effects by incorporating the University of California Santa Barbara emissivity library in the radiative transfer simulation. The three methods were evaluated using ground-measured LWUP from six SURFRAD sites. Although methods 2 and 3 were developed using MODIS Terra atmospheric profiles, they were applied to both Terra and Aqua data because the designs of the two sensors are similar. The root mean squared errors (rmses) of the ANN model method are smaller than that of the other two methods at all sites. The averaged rmses of the ANN model method are 15.89 W/m2 (Terra) and 14.57 W/m>2 (Aqua); the averaged biases are -8.67 W/m2 (Terra) and -7.21 W/m2 (Aqua). The biases and rmses for Aqua are ~1.3 W/m2 smaller than that of Terra. The biases and rmses of the ANN model method are ~5 W/m2 smaller than that of the temperature-emissivity method and ~2.5 W/m2 smaller than that of the linear model method.

Probing Norwalk-like virus presence in shellfish, using artificial neural networks

A database was examined using artificial neural network (ANN) models to investigate the efficacy of predicting PCR-identified Norwalk-like virus presence and absence in shellfish. The relative importance of variables in the model and the predictive power obtained by application of ANN modelling methods were compared with previously developed logistic regression models. In addition, two country-specific datasets were analysed separately with ANN models to determine if the relative importance of the input variables was similar for geographically diverse regions. The results of this analysis found that ANN models predicted Norwalk-like virus presence and absence in shellfish with equivalent, and better, precision than logistic regression models. For overall classification performance, ANN modelling had a rate of 93%, vs 75% for the logistic regression. ANN models were able to illuminate the site-specific relationships between indicators and pathogens.

Utilisation of non-supervised neural networks and principal component analysis to study fish assemblages

Kohonen self-organizing maps (SOM) belong to the non-supervised artificial neural network modelling methods. It typically displays a high dimensional data set in a lower dimensional space. In this way, that method can be considered as a non-linear surrogate to the principal component analysis (PCA). In order to test the efficiency of SOM on complex ecological data gathered in the natural environment, we made a comparison between PCA and SOM capabilities to analyse the spatial occupancy of several European freshwater fish species in the littoral zone of a large French lake. The same data matrix consisting of 710 samples and 15 species was analysed using PCA and SOM. Both methods provided insights on the major trends in fish spatial occupancy. However, a more detailed analysis showed that only SOM was able to reliably visualise the entire fish assemblage in a two dimensional space (i.e. both dominant and scarce species). On the contrary PCA provided irrelevant ecological information for some species. These drawbacks were afforded to data heterogeneity, scarce species being poorly represented on the PCA plane. These results led us to conclude that SOM constitute a more reliable data representation method than PCA when complex ecological data sets are used.