Feedforward Multilayer Perceptron Research Articles

Comparative Analysis of Linear Models and Artificial Neural Networks for Sugar Price Prediction

Sugar is an important commodity that is used beyond the food industry. It can be produced from sugarcane and sugar beet, depending on the region. Prices worldwide differ due to high volatility, making it difficult to estimate their forecast. Thus, the present work aims to predict the prices of kilograms of sugar from four databases: the European Union, the United States, Brazil, and the world. To achieve this, linear methods from the Box and Jenkins family were employed, together with classic and new approaches of artificial neural networks: the feedforward Multilayer Perceptron and extreme learning machines, and the recurrent proposals Elman Network, Jordan Network, and Echo State Networks considering two reservoir designs. As performance metrics, the MAE and MSE were addressed. The results indicated that the neural models were more accurate than linear ones. In addition, the MLP and the Elman networks stood out as the winners.

Thermomechanical behavior of a novel hybrid epoxy/ZnO nanocomposite adhesive in structural bonding: Experimental analysis and ANN modeling

ObjectivesEpoxy adhesives are advanced materials, but suffer from high brittleness and low toughness due to their high crosslinking degree, which limits their service life in structural applications. The lack of appropriate thermal stability at high temperatures is another constraint of epoxy adhesives. The main aim of this research was to increase the thermal stability and toughness of epoxy adhesives using phenolic resin, zinc oxide (ZnO) nanoparticles, and poly (butyl acrylate-block-styrene) copolymer as a toughening agent. Furthermore, the mechanical properties of epoxy adhesives were predicted by designing two feed-forward multilayer perceptron (MLP) networks. MethodsEpoxy adhesives with different contents of phenolic resin (10, 20, and 30 phr), copolymer (1.25, 2.5, and 3.75 phr), and ZnO (0.5, 1, 2, and 5 phr) nanoparticles were synthesized under mechanical mixing by using xylene as solvent. Then, the epoxy adhesive samples for tensile and lap shear tests were cured at room temperature for 7 and 2 days, respectively. The mechanical properties of adhesive samples were measured by tensile and lap shear tests. The thermal stability of the epoxy adhesive samples was investigated by thermogravimetric analysis (TGA). The kinetics of the curing reaction of the optimum epoxy adhesive samples was also studied by differential scanning calorimetry (DSC). ResultsThe toughness of epoxy adhesive containing 10 phr phenolic, 2.5 phr block copolymer, and 2 phr ZnO nanoparticles increased by 20%, and shear strength by 99% compared to pure epoxy adhesive, notifying a significant synergistic effect. The TGA results showed that all three mentioned additives have increased the thermal stability of pure epoxy. The highest thermal stability was observed for epoxy adhesive containing 2.5 phr block copolymer and 2 phr ZnO nanoparticles. In addition, DSC analysis proved the positive effect of ZnO nanoparticles incorporation and block copolymer on the curing kinetics of epoxy adhesive. Moreover, the predicted tensile strength, tensile modulus, toughness, and shear strength of epoxy adhesives by MLP networks showed a very good consistency between the experimental data and the model predictions (e.g., MRE < 1.1 for lap shear). SignificanceIn the current study, the artificial neural networks (ANN) results showed that there was very good consistency between the ANN predictions and the experimental data.

Geometrical Interpretation and Design of Multilayer Perceptrons.

The multilayer perceptron (MLP) neural network is interpreted from the geometrical viewpoint in this work, that is, an MLP partition an input feature space into multiple nonoverlapping subspaces using a set of hyperplanes, where the great majority of samples in a subspace belongs to one object class. Based on this high-level idea, we propose a three-layer feedforward MLP (FF-MLP) architecture for its implementation. In the first layer, the input feature space is split into multiple subspaces by a set of partitioning hyperplanes and rectified linear unit (ReLU) activation, which is implemented by the classical two-class linear discriminant analysis (LDA). In the second layer, each neuron activates one of the subspaces formed by the partitioning hyperplanes with specially designed weights. In the third layer, all subspaces of the same class are connected to an output node that represents the object class. The proposed design determines all MLP parameters in a feedforward one-pass fashion analytically without backpropagation. Experiments are conducted to compare the performance of the traditional backpropagation-based MLP (BP-MLP) and the new FF-MLP. It is observed that the FF-MLP outperforms the BP-MLP in terms of design time, training time, and classification performance in several benchmarking datasets. Our source code is available at https://colab.research.google.com/drive/1Gz0L8AnT4ijrUchrhEXXsnaacrFdenn?usp = sharing.

Some Aspects of the Modelling of Dried Red Beets Rehydration Process

Some dehydrated products must be rehydrated before consumption or further industry processing. Optimization of the rehydration process needs mathematical models of the process. Despite the widespread use of computers and their associated software, empirical equations are still widely used in view of their simplicity and ease of computation. The mathematical description of the kinetics of mass gain, volume increase, dry matter loss, and moisture content increase and changes of rehydration indices during the rehydration of dried red beets was investigated. The effects of drying air temperature (Td), drying air velocity (vd), characteristic dimension (L), and rehydration temperature (Tr) on model constants were also examined. Red beets cubes (10 mm) and slices (5 and 10 mm) were dried in natural convection (vd = 0.01 m/s), forced convection (vd = 2 m/s), and fluidization (vd = 6 m/s) at Td = 50, 60, and 70 °C. The rehydration was conducted in distilled water at Tr = 20, 45, and 70 °C. The kinetics of rehydrating dried red beets was modelled applying five empirical models: Peleg, Lewis (Newton), Henderson–Pabis, Page, and modified Page. Equations were developed to make the model constants dependent on Td, vd, L, and Tr. Artificial neural networks (ANNs) (feedforward multilayer perceptron) were adopted to condition the rehydration indices on Td, vd, L, and Tr. The following models can be recommended as the most acceptable: (1) the modified Page model for mass gain (RMSE = 0.0236–0.0897) and for volume increase (RMSE = 0.0213–0.0972), (2) the Peleg model for dry mass loss (RMSE = 0.0161–0.610), and (3) the Henderson–Pabis model for moisture content increase (RMSE = 0.0350–0.1062). The ANNs performed the rehydration indices in an acceptable way (RMSE = 0.0528–0.2285). Both the rehydration indices and model constants depended (but to a different degree) on the investigated drying and rehydration conditions.

Subspace learning machine (SLM): Methodology and performance evaluation

Inspired by the feedforward multilayer perceptron (FF-MLP), decision tree (DT) and extreme learning machine (ELM), a new classification model, called the subspace learning machine (SLM), is proposed in this work. SLM first identifies a discriminant subspace, S0, by examining the discriminant power of each input feature. Then, it uses probabilistic projections of features in S0 to yield 1D subspaces and finds the optimal partition for each of them. This is equivalent to partitioning S0 with hyperplanes. A criterion is developed to choose the best q partitions that yield 2q partitioned subspaces among them. We assign S0 to the root node of a decision tree and the intersections of 2q subspaces to its child nodes of depth one. The partitioning process is recursively applied at each child node to build an SLM tree. When the samples at a child node are sufficiently pure, the partitioning process stops and each leaf node makes a prediction. The idea can be generalized to regression, leading to the subspace learning regressor (SLR). Furthermore, ensembles of SLM/SLR trees can yield a stronger predictor. Extensive experiments are conducted for performance benchmarking among SLM/SLR trees, ensembles and classical classifiers/regressors.

Verification study on how macrofungal fruitbody formation can be predicted by artificial neural network

The occurrence and regularity of macrofungal fruitbody formation are influenced by meteorological conditions; however, there is a scarcity of data about the use of machine-learning techniques to estimate their occurrence based on meteorological indicators. Therefore, we employed an artificial neural network (ANN) to forecast fruitbody occurrence in mycorrhizal species of Russula and Amanita, utilizing meteorological factors and validating the accuracy of the forecast of fruitbody formation. Fungal data were collected from two locations in Western Hungary between 2015 and 2020. The ANN was the commonly used algorithm for classification problems: feed-forward multilayer perceptrons with a backpropagation algorithm to estimate the binary (Yes/No) classification of fruitbody appearance in natural and undisturbed forests. The verification indices resulted in two outcomes: however, development is most often studied by genus level, we established a more successful, new model per species. Furthermore, the algorithm is able to successfully estimate fruitbody formations with medium to high accuracy (60–80%). Therefore, this work was the first to reliably utilise the ANN approach of estimating fruitbody occurrence based on meteorological parameters of mycorrhizal specified with an extended vegetation period. These findings can assist in field mycological investigations that utilize sporocarp occurrences to ascertain species abundance.

A direct normal irradiation forecasting model based on artificial neural networks

We investigate the forecasting of the hourly Direct Normal Irradiation (DNI) using Artificial Neural Networks (ANN). The data used are hourly satellite data for the region of Ouarzazate in the South West Mediterranean basin region. The database is a total of 19 years. A feed-forward multilayer Perceptron is used with ten input parameters. The results show a very good accuracy for the four seasons of the years forecasted.

Void fraction prediction using prompt gamma neutron activation analysis and artificial intelligence

This paper presents the simulation of a water-gas biphasic system in annular and stratified flow regimes using a neutron beam from a241Am–Be radiation source. The objective is to evaluate the potential of prompt gamma neutron activation analysis (PGNAA) and artificial intelligence-based algorithms in flow measurements. A measurement geometry consisted of a neutron flat source with parallel emission and a spherical detector was developed using the MCNP6 code. Prompt gamma rays spectra were used as input data of a feed-forward multilayer perceptron (MLP) artificial neural network aiming to predict void fraction and to identify the investigated flow regimes. The results showed that the MLP was able to predict void fraction, for both flow regimes, with mean absolute percentage error within 1.58%. In addition, the same MLP achieved 100% of success in flow regime identification. Therefore, it is possible to combine PGNAA with MLP in order to investigate flow characteristics.

The Use of Forest Biomass for Energy Purposes in Selected European Countries

The utilization of primary and secondary woody biomass resources, despite controversies, is being promoted to reduce dependence on fossil fuels and due to the need to diversify energy sources and ensure energy security in European Union countries. Forest biomass is one of the renewable and sustainable energy sources that can be used for electricity, heat, and biofuel production. In the context of the ongoing energy crisis in Europe, an attempt was made to analyze the production and consumption of woody biomass for energy purposes (fuel wood, chips, and pellets). Specifically, an analysis of similarities between European countries in terms of biomass utilization was conducted. The analysis was complemented by a forecast of primary biomass production in selected European countries. The similarity analysis was conducted using the Ward method. Artificial neural networks (ANNs), including multi-layer feedforward perceptron (MLP) and radial basis function (RBF) models, were used to predict fuelwood extraction. The study showed that woody biomass remains an important source of bioenergy in Europe, and its significance as a strategic resource guaranteeing energy security is likely to increase. Fuel wood harvesting in Europe generally shows an upward trend, particularly in the Czech Republic, Germany, Estonia, Denmark, and the UK. A decreasing trend was observed in France, Spain, Greece, and Cyprus. The analysis revealed differences between countries in terms of woody biomass consumption. The ANN-based forecasts of fuelwood supply generally showed an increase in primary biomass harvesting.

ARTIFICIAL INTELLIGENCE APPLICATION IN PHOTOACOUSTIC OF GASES

Photoacoustic spectroscopy is a powerful, non-destructive, ultrasensitive technique that covers a wide range of applications including atmospheric monitoring, industrial, environmental, and biomedical practice. In this paper, our attention is focused on the area of artificial intelligence implementation in photoacoustic spectroscopy of gases. Artificial intelligence has been proven as a very successful, effective, and promising method for the accurate and real-time determination of photoacoustic signal parameters, related to relaxation, thermal and other physical properties of various media (i.e., for solving the inverse photoacoustic problem). To improve the sensitivity and selectivity of the photoacoustic method feedforward multilayer perceptron network is applied for real-time simultaneous determination of photoacoustic signal parameters: vibrational-to-translational relaxation time, and radius of the laser beam. Also, to solve the problem of finding optimal values of these photoacoustic parameters, metaheuristic algorithms, genetic algorithms and simulated annealing are used. The performance of artificial intelligence methods has been tested on a set of experimental signals generated in the (SF6+Ar) gas. The potential advantages and disadvantages of those methods are discussed.

Data-driven and physics-based approach for wave downscaling: A comparative study

The physics-based approach is the most common framework for modelling ocean engineering processes; however, in the last years data-driven models are becoming more and more popular. In the case of wave downscaling, the number of studies involving data-driven models is scarce. In this paper, both approaches are developed and compared on a case study in the Mediterranean Sea. Two Feed Forward Multilayer Perceptron Neural Networks are trained with the Levenberg-Marquardt algorithm and compared with a SWAN model. The behaviour of both models turns out to be similar with respect to the wave height. However, with respect to the peak period, neural networks outperform the SWAN model, which slightly overestimates the peak period. Neural networks show a correlation coefficient and error of 0.96 and 1.00 s, respectively, versus 0.72 and 6.2 s for the SWAN model. Regarding the computational cost, the data-driven approach clearly outperforms the physics-based approach. Therefore, it may be concluded that at those sites where a minimum of 25,000 data samples are available, data-driven models are of interest – not least in semi-enclosed basins governed by local patterns, where large-scale or global physics-based models typically struggle.

Recovering shear stiffness degradation curves from classification data with a neural network approach

Shear stiffness is critical in assessing the stress–strain response of geotechnical infrastructure, and is a complex, nonlinear parameter. Existing methods characterise stiffness degradation as a function of strain and require either bespoke laboratory element tests, or adoption of a curve fitting approach, based on an existing data set of laboratory element tests. If practitioners lack the required soil classification parameters, they are unable to use these curve fitting functions. Within this study, we examine the ability and versatility of an artificial neural network (ANN), in this case a feedforward multilayer perceptron, to predict strain-based stiffness degradation on the data set of element test results and soil classification data that underpins current curve fitting functions. It is shown that the ANN gives similar or better results to the existing curve fitting method when the same parameters are used, but also that the ANN approach enables curves to be recovered with ‘any’ subset of the considered soil classification parameters, providing practitioners with a great versatility to derive a stiffness degradation curve. A user-friendly and freely available graphical calculation app that implements the proposed methodology is also presented.

An Analysis of the Influence of Surface Roughness and Clearance on the Dynamic Behavior of Deep Groove Ball Bearings Using Artificial Neural Networks.

The deep groove ball bearing is one of the most important components of the rotary motion system and is the research subject in this paper. After factory assembly, new ball bearings need to pass quality control. The conventional approach relies on measuring the vibration amplitudes for each unit and sorting them into classes according to the vibration level. In this paper, based on experimental research, models are created to predict the vibration class and analyze the dynamic behavior of new ball bearings. The models are based on artificial neural networks. A feedforward multilayer perceptron (MLP) was applied, and a backpropagation learning algorithm was used. A specific method of training groups of artificial neural networks was applied, where each network provided an answer to the input within the group, and the final answer was the mean value of the answers of all networks in the group. The models achieved a prediction accuracy of over 90%. The main aim of the research was to construct models that are able to predict the vibration class of a new ball bearing based on the geometric parameters of the bearing rings. The models are also applied to analyze the influence of surface roughness of the raceways and the internal radial clearance on bearing vibrations.

Analysis of Cleaner Production Performance in Manufacturing Companies Employing Artificial Neural Networks

Cleaner production has emerged as a comprehensive paradigm, aiming to reduce, or even avoid, the environmental impact in the production stage, in a broad variety of fields. However, the great number of interacting factors makes the assessment of efficiency and the identification of critical factors pose significant challenges to researchers and companies. Artificial intelligence and, particularly, artificial neural networks have proven their suitability to lead with diverse multi-variable problems, but have not yet been applied to model production systems. In this work, we employ dimensionality reduction in combination with a fully connected feed-forward multi-layer perceptron to model the relation between the input (cleaner production techniques) and output variables (cleaner production performance) and, subsequently, quantify the sensibility of the different output variables on the input variables. In particular, we consider Product Design, Production Processes, and Reuse as the input latent variables, whereas the Environmental Performance of Product, Environmental Performance of Processes, and Economic Performance comprises the output variables of our model. The results, employing data collected from a direct survey of 205 Brazilian companies, reveal that the best configuration for the ANN uses eight neurons in the hidden layer. Regarding sensitivity, the obtained results show that improving practices with poor marks leads to a higher enhancement of output figures. In particular, since reuse presents mainly low marks, it can be identified as an area for improvement, in order to increase overall performance.

Statistical study of squeezing for soft rocks based on factor and regression analyses of effective parameters

The time-dependent deformation of rocks due to stress released by excavation is referred to as squeezing. Accurate evaluation of the squeezing at the design stage can dramatically reduce technical problems and the financial costs of underground structures. Although various methods are presented to predict tunnel squeezing at the preliminary stage, being site-specific and incorporating incomplete databases are deficiencies of the available procedures. In this study, based on a comprehensive literature review, we prepared a database of tunnel squeezing for soft rocks, including possible effective parameters. Statistical processing methods such as univariate, reduction, and cleaning were employed to improve the statistical quality of the database. The statistically-processed datasets were also validated based on various scales such as accuracy, convergence, and usefulness. Significant predictors of squeezing are recognized as the ratio of strength to stress and the rock mass classification system. New squeezing criteria were developed using binary and multi-class regression methods to predict the squeezing occurrence and intensity of soft rocks. The results are confirmed by a Multilayer Perceptron Feed-Forward Neural Network and are compared to well-known empirical equations. The developed equations are more accurate comparing the empirical equations used to predict the squeezing of soft rocks. This methodology can be utilized at the design stage for another database to predict squeezing rocks for topographic-stress and tectonic-stress-based cases.

Application of machine learning to determine the shear stress and filtration loss properties of nano-based drilling fluid

A detailed understanding of the drilling fluid rheology and filtration properties is essential to assuring reduced fluid loss during the transport process. As per literature review, silica nanoparticle is an exceptional additive to enhance drilling fluid rheology and filtration properties enhancement. However, a correlation based on nano-SiO2-water-based drilling fluid that can quantify the rheology and filtration properties of nanofluids is not available. Thus, two data-driven machine learning approaches are proposed for prediction, i.e. artificial-neural-network and least-square-support-vector-machine (LSSVM). Parameters involved for the prediction of shear stress are SiO2 concentration, temperature, and shear rate, whereas SiO2 nanoparticle concentration, temperature, and time are the inputs to simulate filtration volume. A feed-forward multilayer perceptron is constructed and optimised using the Levenberg–Marquardt learning algorithm. The parameters for the LSSVM are optimised using Couple Simulated Annealing. The performance of each model is evaluated based on several statistical parameters. The predicted results achieved R2 (coefficient of determination) value higher than 0.99 and MAE (mean absolute error) and MAPE (mean absolute percentage error) value below 7% for both the models. The developed models are further validated with experimental data that reveals an excellent agreement between predicted and experimental data.

An artificial neural network optimized by grey wolf optimizer for prediction of hourly wind speed in Tamil Nadu, India

• The hourly wind speed prediction is an important problem. • A feed-forward (FF) multi-layer perceptron (MLP) artificial neural network (ANN) is used for prediction. • The FF MLP ANN is optimized by 9 state-of-art metaheuristic algorithms. • In the experiments, 38 years of hourly wind data belonging to 5 cities were used. • Grey Wolf Optimizer produced best results in terms of mean square error (MSE). The growing population has tremendously increased the daily energy demand all around the world. India is the second-most crowded nation in the world with approximately 1.4 billion people. New and renewable energy is on the agenda of India and in 2021 India possesses the fourth-largest installed capacity of wind power. Accurate prediction of wind speed is vital in wind farm design and operation. In this work, an hourly wind speed prediction with an artificial neural network optimized by a metaheuristics approach is conducted. A feed-forward (FF) multi-layer perceptron (MLP) artificial neural network (ANN) is used for the prediction of the hourly wind speed. In this study, 38 years of hourly wind data belonging to 5 cities (Ambur, Hosur, Kumbakonam, Nagapattinam, and Pudukottai) were used. These cities have different specific properties such as latitude, longitude, and altitude. The FF MLP ANN is optimized by 9 state-of-art metaheuristic algorithms. In this work, Artificial Bee Colony (ABC), Ant Colony Optimization (ACO), Biogeography Based Optimization (BBO), Evolutionary Strategy (ES), Genetic Algorithm (GA), Grey-Wolf-Optimizer (GWO), Population-Based Incremental Learning (PBIL), Particle Swarm Optimization (PSO), Tree-Seed Algorithm (TSA) have been used to optimize the weights of the ANN. GWO outperforms other metaheuristic algorithms in the prediction of wind speed with a FF MLP ANN model, with a success percentage rate of approximately 3% to 10,000%.

Daily Estimation of Global Solar Irradiation and Temperatures Using Artificial Neural Networks through the Virtual Weather Station Concept in Castilla and León, Spain.

In this article, the interpolation of daily data of global solar irradiation, and the maximum, average, and minimum temperatures were measured. These measurements were carried out in the agrometeorological stations belonging to the Agro-climatic Information System for Irrigation (SIAR, in Spanish) of the Region of Castilla and León, in Spain, through the concept of Virtual Weather Station (VWS), which is implemented with Artificial Neural Networks (ANNs). This is serving to estimate data in every point of the territory, according to their geographic coordinates (i.e., longitude and latitude). The ANNs of the Multilayer Feed-Forward Perceptron (MLP) used are daily trained, along with data recorded in 53 agro-meteorological stations, and where the validation of the results is conducted in the station of Tordesillas (Valladolid). The ANN models for daily interpolation were tested with one, two, three, and four neurons in the hidden layer, over a period of 15 days (from 1 to 15 June 2020), with a root mean square error (RMSE, MJ/m2) of 1.23, 1.38, 1.31, and 1.04, respectively, regarding the daily global solar irradiation. The interpolation of ambient temperature also performed well when applying the VWS concept, with an RMSE (°C) of 0.68 for the maximum temperature with an ANN of four hidden neurons, 0.58 for the average temperature with three hidden neurons, and 0.83 for the minimum temperature with four hidden neurons.

Hilbert-Huang Transform-Based Seismic Intensity Parameters for Performance-Based Design of RC-Framed Structures

This study aims to develop the optimal artificial neural networks (ANNs) capable of estimating the seismic damage of reinforced concrete (RC)-framed structures by considering several seismic intensity parameters based on the Hilbert–Huang Transform (HHT) analysis. The selected architecture of ANN is the multi-layer feedforward perceptron (MFP) network. The values of the HHT-based parameters were calculated for a set of seismic excitations, and a combination of five to twenty parameters was performed to develop input datasets. The output data were the structural damage expressed by the Park and Ang overall damage index (DIPA,global). The potential contribution of nine training algorithms to developing the most effective MFP was also investigated. The results confirm that the evolved MFP networks, utilizing the employed parameters, provide an accurate estimation of the target output of DIPA,global. As a result, the developed MFPs can constitute a reliable computational intelligence approach for determining the seismic damage induced on structures and, thus, a powerful tool for the scientific community for the performance-based design of buildings.

Implementation of radial basis function network-based maximum power point tracking for a PV-fed high step-up converter

This research offers a maximum power point tracking (MPPT) algorithm for photovoltaic (PV) systems based on neural networks (NNs) and a rapid step-up converter configuration. An improved variable step size-radial basis function network (RBFN) in the NN algorithm is accomplished in the proposed system to track the maximum power point (MPP) with high convergence speed and obtain maximum power with reduced oscillations. Under various irradiance and temperature conditions, the performance of the recommended algorithm was compared to that of particle swarm optimization (PSO), modified perturb and observe (P&amp;amp;O) MPPT technique, artificial neural network (ANN), and multilayer perceptron feed-forward (MPF) NN-based MPPT method. In this system, a new interleaved non-isolated large step-up converter with the coupled inductor technique is suggested to compensate for the discord in PV devices to enable a continuous and independent power flow. The proposed PV-fed converter system is validated under partial shading conditions (PSCs) and uniform solar PV, and the results are experimentally verified with the use of a programmable direct current (DC) source. The obtained results indicate that the proposed converter produces output with high gain, continuous input current, low voltage stress on switches, minimal ripple, high power density, and extensive input and output operations. Finally, a prototype has been implemented to verify the functionality of the presented converter in continuous conduction mode operation with an input voltage range of 20 V and an output voltage of 200 V.