Three-layer Feedforward Neural Network Research Articles

A BP Artificial Neural Network Model for Earthquake Magnitude Prediction in Himalayas, India

The aim of this study is to evaluate the performance of BP neural network techniques in predicting earthquakes occurring in the region of Himalayan belt (with the use of different types of input data). These parameters are extracted from Himalayan Earthquake catalogue comprised of all minor, major events and their aftershock sequences in the Himalayan basin for the past 128 years from 1887 to 2015. This data warehouse contains event data, event time with seconds, latitude, longitude, depth, standard deviation and magnitude. These field data are converted into eight mathematically computed parameters known as seismicity indicators. These seismicity indicators have been used to train the BP Neural Network for better decision making and predicting the magnitude of the pre-defined future time period. These mathematically computed indicators considered are the clustered based on every events above 2.5 magnitude, total number of events from past years to 2014, frequency-magnitude distribution b-values, Gutenberg-Richter inverse power law curve for the n events, the rate of square root of seismic energy released during the n events, energy released from the event, the mean square deviation about the regression line based on the Gutenberg-Richer inverse power law for the n events, coefficient of variation of mean time and average value of the magnitude for last n events. We propose a three-layer feed forward BP neural network model to identify factors, with the actual occurrence of the earthquake magnitude M and other seven mathematically computed parameters seismicity indicators as input and target vectors in Himalayan basin area. We infer through comparing curve as observed from seismometer in Himalayan Earthquake catalogue comprised of all events above magnitude 2.5 mg, their aftershock sequences in the Himalayan basin of year 2015 and BP neural network predicting earthquakes in 2015. The model yields good prediction result for the earthquakes of magnitude between 4.0 and 6.0.

Modeling and Optimizing Tensile Strength and Yield Point on a Steel Bar Using an Artificial Neural Network With Taguchi Particle Swarm Optimizer

A Taguchi particle swarm optimization (TPSO) with a three-layer feedforward artificial neural network (ANN) is used to model and optimize the chemical composition of a steel bar. The novel contribution of a TPSO is the use of a Taguchi method mechanism to exploit better solutions in the search space through iterations, the use of the conventional non-linear PSO to increase convergence speed, and the use of random movement for particle diversity. The exploration and exploitation capability of the TPSO were confirmed by performance comparisons with other PSO-based algorithms in solving high-dimensional global numerical optimization problems. Experiments in this paper showed that the TPSO provides higher computational efficiency and higher robustness when solving problems involving seven non-linear benchmark functions, including three unimodal functions, one multimodal functions, two rotated functions, and one shifted functions. The results for the computational experiments show that the TPSO outperforms other PSO-based algorithms reported in the literature. Finally, the results obtained by a TPSO-based ANN model of the chemical composition of the steel bar were consistent with the actual data. That is, the proposed TPSO with three-layer feedforward ANN can be used in practical applications.

Estimation of thermal conductivity of CNTs-water in low temperature by artificial neural network and correlation

An accurate artificial neural network (ANN) model and new correlation are developed to predict thermal conductivity of functionalized carbon nanotubes (MWNT-10nm in diameter)-water nanofluid based on experimental data. Experimental values of thermal conductivity are in six concentrations of nanoparticles from 0.005% up to 1.5%. The temperatures were changed within 10–60°C. In order to estimate the thermal conductivity, a feed-forward three-layer neural network is utilized. The obtained results exhibited that the new correlation and ANN model have a good agreement with the experimental data. The maximum values of deviation and mean square error of neural network outputs were 2% and 8.2E−05, respectively. The findings illustrated that the artificial neural network can estimate and model the thermal conductivity of CNTs-water nanofluid very efficiently and accurately.

Application of artificial neural network to study the performance of multi-gravity separator (MGS) treating iron ore fines

ABSTRACTThe present article describes an attempt made to study the possibility of beneficiating low-grade iron ore fines of Barbil Area of Orissa state, India, using multi-gravity separator (MGS) after grinding the −10 mm fines to < 75 micron size and prepare a pellet feed of 65% Fe content. For the performance analysis, an artificial neural network (ANN) mathematical modeling approach was attempted. A three-layer feedforward neural network with a backpropagation method has been adopted, considering the three significant parameters of MGS, mainly drum inclination, drum speed, and shake amplitude, were varied and the results were evaluated for grade, recovery, and separation efficiency. The results of beneficiation studies showed that good recovery of hematite is possible with simultaneous increase in Fe(T) grade from 50.74% to 65.26% with 71.25% recovery. The predicted value obtained by ANN shows good agreement with the experimental values.

Using Artificial Neural Network for Estimation of Density and Viscosities of Biodiesel–Diesel Blends

In recent years, biodiesel has been considered as a good alternative of diesel fuels. Density and viscosity are two important properties of these fuels. In this study, density and kinematic viscosity of biodiesel-diesel blends were estimated by using artificial neural network (ANN). A three-layer feed forward neural network with Levenberg-Marquard (LM) algorithm was used for learning empirical data (previous studies data and this study empirical data). Input data for estimating density and kinematic viscosity includes components volume fraction, temperature and pure component properties (pure density at 293.15 K and pure kinematic viscosity at 313.15 K). Results of neural network simulation for density and kinematic viscosity showed a high accuracy (mean relative error for density and kinematic viscosity are 0.021% and 0.73%, respectively).

Permeability Prediction of Tight Sandstone Reservoirs Using Improved BP Neural Network

By analyzing the permeability controlling factors of tight sandstone reservoir in Wuhaozhuang Oil Field, the permeability is considered to be mainly controlled by porosity, clay content, irreducible water saturation and diagenetic coefficient. Because the conventional BP algorithm has its drawbacks such as slow convergence speed and easy falling into the local minimum value, an improved three-layer feed-forward BP neural network model is built by MATLAB neural network toolbox to predict permeability according to the four permeability controlling factors, while studying samples of model are selected based on the representative core analysis data. The simulation based on improved neural network model shows that the improved model has a faster convergence speed and better accuracy. The consistency between model prediction value and lab test value is good and the mean squared error is less. Therefore, the new model can meet the needs of the development geology research of oil field better in the future.

A novel word embedding learning model using the dissociation between nouns and verbs

In recent years, there have been researches on using semantic knowledge and global statistical features to guide the learning of word embeddings. Though the syntax knowledge also plays an very important role in natural language understanding, its effectiveness on the word embedding learning is still far from well investigated. Inspired by the principle of the dissociation between nouns and verbs (DNV) in language acquisition observed in neuropsychology, we propose a novel model for word embeddings learning using DNV (named Continuous Dissociation between Nouns and Verbs Model, CDNV). CDNV uses a three-layer feed forward neural network to integrate DNV generated by auto-tagged noun/verb information into the word embeddings learning process, which can still preserve the word order of local context. The advantage of the CDNV lies in that it is able to learn high-quality word embeddings with relatively low time complexity. Experimental results show that: (1) CDNV takes about 1.5h to learn word embeddings on a corpus of billions of words, which is comparable with CBOW and Skip-gram and more efficient than other models; (2) the nearest neighbors of some representative words derived from the word embeddings learnt by CDNV are more reasonable than other word embeddings; (3) the performance improvement on F1 measure from CDNV word embeddings is greater than other word embeddings on NER and Chunking.

Reconstructing meteorological time series to quantify the uncertainties of runoff simulation in the ungauged Qira River Basin using data from multiple stations

The existence and development of oases in arid plain areas depends mainly on the runoff generated from alpine regions. Quantifying the uncertainties of runoff simulation under climatic change is crucial for better utilization of water resources and management of oases in arid areas. In the present study, based on the ungauged Qira River Basin in Xinjiang, China, a modified version of the Delta statistical downscaling method was applied to reconstruct the monthly mean temperature (MMT), monthly accumulated precipitation (MAP), and monthly accumulated evaporation (MAE) of two target stations. Then, the uncertainty in runoff simulation, implemented using the Three-Layered Feedforward Neural Network model with the Back-Propagation learning algorithm, was quantified. The modified Delta method reproduced the MMT, MAP, and MAE time series of the two target stations very well during the calibrated periods, and the reconstructed uncertainty ranges were small among reconstructed datasets using data from 12 observation stations. The monthly accumulated runoff simulated by the reconstructed MMT, MAP, and MAE as input variables of the model possessed unpredictable uncertainty. Although the use of multi-data ensembles in model inputs are considered an effective way to minimize uncertainties, it could be concluded that, in this case, the efficiency of such an approach was limited because of errors in the meteorological data and the deficiency of the model’s structure. The uncertainty range in the runoff peak was unable to capture the actual monthly runoff. Nevertheless, this study represents a significant attempt to reproduce historical meteorological data and to evaluate the uncertainties in runoff simulation through multiple input ensembles in an ungauged basin. It can be used as reference meteorological data for researching long-term climate change and producing runoff forecasts for assessing the risk of droughts and/or floods, as well as the existence and management of plain oases in the Qira River Basin.

A New Approach to Short-term Price Forecast Strategy with an Artificial Neural Network Approach: Application to the Nord Pool

In new deregulated electricity market, short-term price forecasting is key information for all market players. A better forecast of market-clearing price (MCP) helps market participants to strategically set up their bidding strategies for energy markets in the short-term. This paper presents a new prediction strategy to improve the need for more accurate short-term price forecasting tool at spot market using an artificial neural networks (ANNs). To build the forecasting ANN model, a three-layered feedforward neural network trained by the improved Levenberg-marquardt (LM) algorithm is used to forecast the locational marginal prices (LMPs). To accurately predict LMPs, actual power generation and load are considered as the input sets, and then the difference is used to predict price differences in the spot market. The proposed ANN model generalizes the relationship between the LMP in each area and the unconstrained MCP during the same period of time. The LMP calculation is iterated so that the capacity between the areas is maximized and the mechanism itself helps to relieve grid congestion. The addition of flow between the areas gives the LMPs a new equilibrium point, which is balanced when taking the transfer capacity into account, LMP forecasting is then possible. The proposed forecasting strategy is tested on the spot market of the Nord Pool. The validity, the efficiency, and effectiveness of the proposed approach are shown by comparing with time-series models

Forecasting crude oil price with ensemble neural networks based on different feature subsets method

In this study, an ensemble neural network is proposed based on different feature subsets method in order to forecast the world crude oil spot price. To this end, a number of experts in database gathering and appropriate time delays were interviewed to forecast 1-step ahead of the crude oil spot price. Subsequently, different features subsets were generated randomly, each of which was then used for each of the basic classifiers. Then, three-layered feed-forward neural network models were used to model each of the basic classifiers. Finally, the prediction results of all basic classifiers were combined with a single layer perceptron neural network to formulate an ensemble output for the original crude oil price series. In order to verify and evaluate the presented method, one of the main crude oil price series, i.e. WTI crude oil spot price, was used to test the effectiveness of the proposed method. Empirical results provided evidence for the effectiveness of the proposed ensemble learning method compared to linear and nonlinear models.

Process simulation and dynamic control for marine oily wastewater treatment using UV irradiation

UV irradiation and advanced oxidation processes have been recently regarded as promising solutions in removing polycyclic aromatic hydrocarbons (PAHs) from marine oily wastewater. However, such treatment methods are generally not sufficiently understood in terms of reaction mechanisms, process simulation and process control. These deficiencies can drastically hinder their application in shipping and offshore petroleum industries which produce bilge/ballast water and produced water as the main streams of marine oily wastewater. In this study, the factorial design of experiment was carried out to investigate the degradation mechanism of a typical PAH, namely naphthalene, under UV irradiation in seawater. Based on the experimental results, a three-layer feed-forward artificial neural network simulation model was developed to simulate the treatment process and to forecast the removal performance. A simulation-based dynamic mixed integer nonlinear programming (SDMINP) approach was then proposed to intelligently control the treatment process by integrating the developed simulation model, genetic algorithm and multi-stage programming. The applicability and effectiveness of the developed approach were further tested though a case study. The experimental results showed that the influences of fluence rate and temperature on the removal of naphthalene were greater than those of salinity and initial concentration. The developed simulation model could well predict the UV-induced removal process under varying conditions. The case study suggested that the SDMINP approach, with the aid of the multi-stage control strategy, was able to significantly reduce treatment cost when comparing to the traditional single-stage process optimization. The developed approach and its concept/framework have high potential of applicability in other environmental fields where a treatment process is involved and experimentation and modeling are used for process simulation and control.

THE APPLIANCE EFFICIENCY ESTIMATION OF PID-REGULATOR PARAMETERS OF NEURAL OPTIMIZER FOR SOLVING OF CONTROL PROBLEM OF METALLURGICAL HEATING PLANTS

The article considers the questions of control scheme with neural optimizer implementation. That optimizer is used for PID-regulator parameters of on-line tuning according to current situation. Three-layer feed-forward neural network was chosen for neural optimizer implementation. That network was on-line trained with the help of back propagation algorithm. The algorithm was modified with several conditions in order to give the neural network the opportunity to consider features of control of heating plants. Energy efficiency estimation of neural optimizer usage for PID-regulator parameters tuning was made during the experiment of cast steel heating in laboratory furnace. Electrical energy saving estimation during the mentioned process was made for control schemes with conventional PID-regulator, PID-regulator with neural optimizer and adaptive PID-regulator, made by Siemens.

Condition-Based Maintenance for medium speed diesel engines used in vessels in operation

Condition-Based Maintenance for diesel engines has contributed to reliability, energy-efficiency, and cost reduction. Both, the modelling of engine performance and fault detection require large amounts of data; usually, these are obtained on a test bench. In contrast, in operative engines, provoking faults onboard is not a viable proposition. Condition-Based Maintenance, fault detection and diagnosis need to be solved on engines installed in commercial vessels: the present contribution answers this need. A medium-speed diesel engine was monitored using thermocouples, pressure sensors, a propeller shaft torque meter and fuel oil flow-meters, during more than 10,000 running hours. Monitored data were used to train a three-layer feed-forward neural network, to generate the engine performance model; thus, determine the engine's fuel consumption and faulty conditions. The faulty conditions considered were: (1) a polluted turbine; (2) a dirty air filter/compressor; (3) a dirty air cooler; (4) and bad fuel injection, i.e. bad combustion. The sensor's precision and the experience gained by monitoring the engine served as a baseline to define the fault threshold values. The results proved the feasibility of installing a Condition-Based Maintenance, for vessels in operation, by monitoring engine performance and analysing the data with the aid of artificial neural networks.

Degradation of Rhodamine B by an electrochemical ozone generating system consist of a Ti anode coated with nanocomposite of Sn–Sb–Ni oxide

An ozonation process was performed using a recycled electrochemical ozone generator system. A titanium based electrode, coated with nanocomposite of Sn–Sb–Ni was applied as anode in a laboratory-made electrochemical reactor. A constant flow rate of 192mg/h of generated ozone was entered to an ozonation reactor to contact with a typical target pollutant, i.e., Rhodamine B (Rh.B) molecules in aqueous solution. Four operational parameters such as: initial dye concentration, pH, temperature and the contact time were evaluated for the ozonation process. Experimental findings revealed that for a solution of 8mg/L of the dye, the degradation efficiency could reach to 99.5% after 30min at pH 3.7 and temperature of 45°C as the optimum conditions. Kinetic studies showed that a second order equation can describe the ozonation adequately well under different temperatures. Also, considering to the importance of process simulation, a three-layered feed forward back propagation artificial neural network model was developed. Sensitivity analysis indicated order of the operational parameter's relative importance on the model output as: time≫pH>Rh.B initial concentration>temperature.

Artificial neural network modeling of photocatalytic removal of a disperse dye using synthesized of ZnO nanoparticles on montmorillonite

In this study, the photocatalytic ability of ZnO/Montmorilonite (ZnO/MMT) nanocomposite under UV-A, UV-B and UV-C radiation was investigated. ZnO nanoparticles were synthesized on the surface of MMT and used as photocatalyst in decolorization of Disperse Red 54 (DR54) solution. Synthesized nanocomposite was characterized by X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM) techniques and nitrogen adsorption/desorption isotherms curves. The average width of synthesized ZnO particles is in the range of 30–45nm. Effect of UV light regions, initial dye concentration, initial dosage of nanocomposite, and reusability of catalyst was studied on decolorization efficiency. The highest decolorization efficiency was achieved under UV-C radiation. A three-layered feed forward back propagation artificial neural network model was developed to predict the photocatalysis of DR54 under UV-C radiation. According to ANN model the ZnO/MMT dosage with a relative importance of 49.21% is the most influential parameter in the photocatalytic decolorization process.

Novel method to predict body weight in children based on age and morphological facial features.

A new and novel approach of predicting the body weight of children based on age and morphological facial features using a three-layer feed-forward artificial neural network (ANN) model is reported. The model takes in four parameters, including age-based CDC-inferred median body weight and three facial feature distances measured from digital facial images. In this study, thirty-nine volunteer subjects with age ranging from 6-18 years old and BW ranging from 18.6-96.4 kg were used for model development and validation. The final model has a mean prediction error of 0.48, a mean squared error of 18.43, and a coefficient of correlation of 0.94. The model shows significant improvement in prediction accuracy over several age-based body weight prediction methods. Combining with a facial recognition algorithm that can detect, extract and measure the facial features used in this study, mobile applications that incorporate this body weight prediction method may be developed for clinical investigations where access to scales is limited.

Adaptive Lifting Transform for Classification of Hyperspectral Signatures

Supervised classification of hyperspectral images is a challenging task because of the higher dimensionality of a pixel signature. The conventional classifiers require large training data set; however, practically limited numbers of labeled pixels are available due to complexity and cost involved in sample collection. It is essential to have a method that can reduce such higher dimensional data into lower dimensional feature space without the loss of useful information. For classification purpose, it will be useful if such a method takes into account the nature of the underlying signal when extracting lower dimensional feature vector. The lifting framework provides the required flexibility. This article proposes the adaptive lifting wavelet transform to extract the lower dimensional feature vectors for the classification of hyperspectral signatures. The proposed adaptive update step allows the decomposition filter to adapt to the input signal so as to retain the desired characteristics of the signal. A three-layer feed forward neural network is used as a supervised classifier to classify the extracted features. The effectiveness of the proposed method is tested on two hyperspectral data sets (HYDICE & ROSIS sensors). The performance of the proposed method is compared with first generation discrete wavelet transform (DWT) based feature extraction method and previous studies that use the same data. The indices used for measuring performance are overall classification accuracy and Kappa value. The experimental results show that the proposed adaptive lifting scheme (ALS) has excellent results with a small size training set.

Artificial neural network modelling for removal of chromium (VI) from wastewater using physisorption onto powdered activated carbon

A three-layered feed-forward artificial neural network (ANN) model has been designed to predict the adsorption efficiency and adsorption capacity for the adsorptive removal of chromium (VI) from synthetic wastewater. The adsorbent dose, wastewater pH, initial pollutant concentration and contact time were used to develop the network. The data used to train and test the model were obtained from several batch experiments. Various algorithms and transfer functions for hidden layer were tested to find the most reliable network. Broyden–Fletcher–Goldfarb–Shanno (BFGS) quasi-Newton backpropagation algorithm gave the most satisfactory results for adsorption efficiency. Resilient and BFGS quasi-Newton backpropagation were the most suitable algorithm for adsorption capacity. The best combination of training algorithm and transfer function for adsorption efficiency was found to be trainrp and poslin, while poslin produced simulated results within 10% deviation for adsorption capacity. Eight to eleven neurons were found to be optimum using trial-and-error method. The ANN predicted and experimentally measured values were compared to test the accuracy of the model.

Neural Network Approach for Safety Climate Prediction

AbstractSafety climate is a snapshot of the safety culture of any organization and it is dynamic in nature. It reflects the employees’ perceptions and attitude towards the existing safety practices in construction projects. The main goal of this paper is to evaluate and differentiate construction projects based on their safety climate and to determine the significant constructs of a safety climate. This paper develops a model to predict safety climate in a construction project based on an artificial neural network (ANN). The important constructs for a safety climate have been determined through literature review. The constructs of safety climate are used as inputs and the safety climate of the project is used as an output for the ANN algorithm. This study collected a total of 250 responses through a questionnaire survey across the country. A three-layer feed-forward back-propagation neural network (10-18-1) has been utilized for the analysis. The developed model predicts the safety climate of a constructi...

Optimal load sharing strategy for a wind/diesel/battery hybrid power system based on imperialist competitive neural network algorithm

In this study, optimal load sharing strategy for a stand-alone hybrid power generation system that consists of wind turbine, diesel generator and battery banks is presented. The diesel generator is used to complement the intermittent output of the wind source whereas the battery is used to compensate for part of the temporary peak demand, which the wind and diesel generator cannot meet thus avoiding oversizing of the diesel generator. To optimise the performance of the system, imperialist competitive algorithm (ICA), ant colony optimisation (ACO) and particle swarm optimisation (PSO) are used to optimal load sharing. These algorithms are used to select the best available energy source so that the system has the best performance.To verify the system performance simulation studies have been carried out using forecasted data (load demand and wind speed). Accordingly, ICA, ACO and PSO are used to train a three-layer feed forward neural network. This trained artificial neural network is applied to short-term wind speed and load demand forecasting on a specific day in the Qazvin. The results show that the proposed control methods can reduce fuel consumption and increase the battery lifetime and battery ability to respond to real-time load turbulences simultaneously.