Multilayer Feedforward Neural Network Model Research Articles

Retraction Note to: Artificial neural networks for prediction Charpy impact energy of Al6061/SiCp-laminated nanocomposites

In this work, Charpy impact energy of Al6061-SiCp nanocomposites produced by mechanical alloying has been modeled by artificial neural networks in both crack divider and crack arrester configurations. Al6061 powders were mixed by 2, 3, and 5 vol % of SiC nanoparticles and ball-milled for 45 min. Afterward, the produced powders were hot-pressed in aluminum cans and then were extruded to produce a dense bulk. Charpy impact specimens were prepared from the produced samples in layered form with different adhesives and thickness of the layers. To build the model, training, validating, and testing was performed using 171 pair input-target data. The used data in the multilayer feed-forward neural networks models were arranged in a format of six input parameters including the thickness of layers, the number of layers, the adhesive type, the crack tip configuration, the content of SiC nanoparticles, and the test trial number. The output parameter was Charpy impact energy of the nanocomposites. The training, validating, and testing results in the neural network models have shown a strong potential for predicting Charpy impact energy of Al6061/SiCp nanocomposites in the considered range of input-target values.

A New Ore Grade Estimation Using Combine Machine Learning Algorithms

Accurate prediction of mineral grades is a fundamental step in mineral exploration and resource estimation, which plays a significant role in the economic evaluation of mining projects. Currently available methods are based either on geometrical approaches or geostatistical techniques that often considers the grade as a regionalised variable. In this paper, we propose a grade estimation technique that combines multilayer feed-forward neural network (NN) and k-nearest neighbour (kNN) models to estimate the grade distribution within a mineral deposit. The models were created by using the available geological information (lithology and alteration) as well as sample locations (easting, northing, and altitude) obtained from the drill hole data. The proposed approach explicitly maintains pattern recognition over the geological features and the chemical composition (mineral grade) of the data. Prior to the estimation of grades, rock types and alterations were predicted at unsampled locations using the kNN algorithm. The presented case study demonstrates that the proposed approach can predict the grades on a test dataset with a mean absolute error (MAE) of 0.507 and R2=0.528, whereas the traditional model, which only uses the coordinates of sample points as an input, yielded an MAE value of 0.862 and R2=0.112. The proposed approach is promising and could be an alternative way to estimates grades in a similar modelling tasks.

Perceived neighborhood: Preferences versus actualities

Housing recovery plays a key role in the overall restoration of a community. A multitude of factors affect housing recovery, many of which are associated with interactions of residents with their perceived neighborhoods. Targeting perceived neighborhoods rather than administratively defined measures of land helps with devising recovery plans that could better address social preferences of the residents. However, such measures are commonly subject to collection of information via expensive and time-consuming surveys. The current research aims to contribute to the domain by exploring the relationship between perception of households of their neighborhood anchors (perceived anchors) and the anchors that exist within perceived neighborhood boundaries (actual anchors). The goal is to propose a model for classifying households’ perceived anchors from publicly available data on actual anchors.Data were collected on households' attributes, perceived neighborhood boundaries, and perceived community anchors through an online survey of New York and Louisiana residents. Actual anchors were mined from the OpenStreetMap database. Correlation analysis revealed several significant associations between actual and perceived anchors. A multilayer feed-forward neural network model was also developed to predict the classification of households’ perceived anchors from actual anchors. Sensitivity analysis of the model disclosed that individuals whose perceived neighborhood comprised more categories of actual anchors were more likely to prioritize infrastructure to other neighborhood assets, a preference that was more dominant in high-density areas.

A Bayesian regularized feed-forward neural network model for conductivity prediction of PS/MWCNT nanocomposite film coatings

In our present work, a multi-layered feed-forward neural network (FFNN) model was designed and developed to predict electrical conductivity of multi-walled carbon nanotube (MWCNT) doped polystyrene (PS) latex nanocomposite (PS/MWCNT) film coatings using data set gathered from several conductivity measurements. Surfactant concentrations (Cs), initiator concentrations (Ci), molecular weights (MPS) and particle sizes of PS latex (DPS) together with MWCNT concentrations (RMWCNT) were introduced as inputs while electrical conductivity (σ) was assigned as a single output in FFNN topology. Network training was carried out using a Bayesian regulation backpropagation algorithm. Optimal geometry of the hidden layer was first studied to search out the best FFNN topology providing the most accurate performance results. Mean squared error, MSE, mean absolute error, MAE, root-mean-squared error, RMSE, determination of coefficient, R2, variance accounted for, VAF, and regression analysis were employed as performance assessment parameters for proposed network model. Correlation coefficients (r) of each input variable together with relative importance-based sensitivity analysis results have shown that RMWCNT is the most significant input variable strongly affecting the σ value of PS/MWCNT nanocomposite film coatings and training performance of the neural network. Mathematical explicit function has been derived to model electrical conductivity by using weights and bias values at each neuron found in FFNN development. All predicted conductivity values are in a very good agreement with measured conductivity values, showing robustness and reliability of suggested FFNN model and it can be effectively used to predict electrical conductivity of PS/MWCNT nanocomposite film coatings.

Retraction Note to: Artificial neural networks for prediction compressive strength of geopolymers with seeded waste ashes

In the present work, compressive strength of inorganic polymers (geopolymers) made from seeded fly ash and rice husk bark ash has been predicted by artificial neural networks. Different specimens were subjected to compressive strength tests at 7 and 28 days of curing. One set of the specimens were cured at room temperature until reaching to 7 and 28 days, and the other sets were oven-cured for 36 h at the range of 40–90 °C and then room cured until 7 and 28 days. A model based on artificial neural networks for predicting the compressive strength of the specimens has been presented. To build the model, training and testing using experimental results from 120 specimens were conducted. The data used in the multilayer feed-forward neural networks models are arranged in a format of six input parameters that cover the percentage of fine fly ash in the ashes mixture, the percentage of coarse fly ash in the ashes mixture, the percentage of fine rice husk bark ash in the ashes mixture, the percentage of coarse rice husk bark ash in the ashes mixture, the temperature of curing, and the time of water curing. According to these input parameters, in the neural networks model, the compressive strength of each specimen was predicted. The training and testing results in the neural networks model have shown a strong potential for predicting the compressive strength of the geopolymer specimens.

Analysis of equilibrium CO 2 solubility in aqueous APDA and its potential blends with AMP / MDEA for postcombustion CO 2 capture

The utility of polyamine-based solvent-activators for the possible application in postcombustion CO2 capture technology has drawn considerable attention recently owing to its higher loading capacity as well as superior kinetics. The current work involves a comprehensive experimental cum theoretical investigation on the equilibrium solubility of CO2 pertaining to aqueous N-(3-aminopropyl)-1,3-propanediamine and its blends with N-methyldiethanolamine and 2-amino-2-methyl-1-propanol. The analysis was conducted within the operating temperature and CO2 partial pressure range of 303.2-323.2 K and 2-200 kPa, respectively. Two different mathematical models based on nonrigorous approaches such as equilibrium based modified Kent-Eisenberg (KE) model and a multilayer feedforward neural network model have been developed to correlate the CO2 solubility data over a wide range of experimental conditions. Both the model predictions are well-validated with the experimental results. The reaction scheme as well as the prevalence of important reaction products was further confirmed with qualitative 13C NMR as well as ATR-FTIR analysis. Apart from these some of the important thermally induced transport properties viz, density, viscosity, and surface tension of the aqueous single and blended systems were measured and correlated with various consistent empirical models such as Redlich-Kister and Grunberg-Nissan model while surface tension data are modeled using temperature-based multiple linear regression technique.

Jelatin Çözeltilerinin Dinamik Viskozitesine Yapay Sinir Ağı (YSA) Yaklaşımı: Esnek Hesaplama Çalışması

In this research, we present a multi-layered feed-forward neural network (ANN) model developed for prediction of dynamic viscosity of aqueous gelatin solutions using experimental data collected from a number of measurements. In ANN architecture, shear stress, shear strain, torque of spindle, the angular velocity of spindle together with mass concentrations of gelatin solutions were introduced as input neurons, whereas dynamic viscosity of aqueous gelatin solutions was assigned as a single output neuron to be predicted. Developed ANN model was trained using backpropagation algorithm optimized with Bayesian regulation. Optimal geometry of the hidden layer was first studied to search out the ANN architecture which yields the most accurate performance results. Mean squared error (MSE), mean absolute error (MAE), root-mean-squared error (RMSE), determination of coefficient (R^2), the variance accounted for (VAF) and regression analyses were used as performance assessment parameters for suggested network models. Sensitivity analysis was carried out to investigate the most effective input neuron strongly influencing the performance of the developed ANN model. As a result, the use of 8 neurons in the hidden layer has shown excellent performance results yielding the least MSE and the highest R^2 values compared to other suggested ANN models. Upon sensitivity analysis, the shear rate was found to be the most effective input neuron significantly affecting network performance. ANN-based predicted dynamic viscosity values were found to be in excellent agreement with measured viscosity values, demonstrating the robustness as well as the accuracy of the developed ANN model. Developed ANN model can, therefore, be effectively used to predict the dynamic viscosity of aqueous polymer solutions using the same input and output parameters in specific data range reported in this paper with statistical details.

Application of neural network modelling for classifying hydrocarbon bearing zone, water bearing zone and shale with estimation of petrophysical parameters in Cauvery basin, India

This work has been developed to classify sand and shale from seven wells in the Cauvery basin using a multilayered feedforward neural network (MLFN) model. Seven wells distributed over 5100 km2 of this basin have been utilized for analysis of conventional well logs and reservoir characterization. Hydrocarbon bearing sediments of Andimadam, Bhuvanagiri, Nannilam and Niravi formations of the Cauvery basin are evaluated in terms of shaliness, cementation factor, porosity, water saturation, and permeability. Pickett plot has been applied to investigate the cementation factor, formation water resistivity, permeability. The cementation factor (m) varies from 1.31 to 1.86 in these formations, whereas permeability varies from 0.01 to 400 md. Very good quality reservoir exists in the Bhuvanagiri formation with high permeability 300–400 md, whereas a good quality reservoir is occurring in Niravi, Nannilam and Andimadam formations with hydrocarbon saturation 60–70%.

Modeling traffic noise in a mountainous city using artificial neural networks and gradient correction

Different regions have established traffic noise prediction models to adapt to their particular environmental characteristics. This paper aimed to develop a traffic noise prediction model for mountainous cities. In China, the traffic noise prediction model HJ 2.4-2009, which itself is based on the sound pressure level corrected for roadway gradients (RGs), has been receiving widespread acceptance. On the basis of the model in HJ 2.4-2009, the RG correction coefficient was proposed to modify the original model and a per-vehicle noise prediction model was built using a multilayer feedforward artificial neural network (ANN) model. The data collected from a municipal road of a hilly city, Chongqing, was used to train and validate the ANN model. The predictor variables comprised the per-vehicle noise value, vehicle type, vehicle velocity, and roadway gradient. The results showed that the modified HJ 2.4-2009 model incorporating the gradient correction coefficient achieved a significantly higher R2 for mountainous cities than the original model. Besides, the ANN-based noise prediction model achieved considerable accuracy improvement over the empirical predictive equations.

Performance comparison of feedforward neural network training algorithms in modelling microwave pyrolysis of oil palm fibre for hydrogen and biochar production

AbstractMicrowave pyrolysis of oil palm fibre (OPF) was conducted to study the effect of microwave power, temperature, and nitrogen (N2) flow rate on the production of hydrogen and biochar. The effect of microwave power ranging from 400 to 900 W, reaction temperatures ranging from 450°C to 700°C, and N2 flow rates ranging 200 to 1,200 cm3 min−1 were investigated. The microwave power, reaction temperature, and N2 flow rates are directly proportional to hydrogen and inversely proportional to biochar yield. The characteristics of the biochar were analysed using CHNOS, scanning electron microscopy, and Brunauer–Emmett–Teller results. Central composite design was used to optimize the number of experiment. The optimized experimental datasets were used as inputs for developing a model that can predict the weight of hydrogen and biochar from microwave pyrolysis of OPF. A multilayer feedforward neural network model was developed by comparing the performance of 11 different types of backpropagation training algorithms. The 11 training algorithms belong to six classes, which are (a) additive momentum, (b) self‐adaptive learning rate, (c) resilient backpropagation, (d) conjugate gradient, (e) quasi‐Newton, and (f) Bayesian regulation (BR). Best performing training algorithm was selected based on the lowest error values computed. The objective of this research is to identify the most suitable training algorithm for this process. Levenberg–Marquardt and BR exhibited very good performance. BR exhibited best performance with lowest error values in predicting weight of hydrogen and biochar. The testing data error values for hydrogen weight is 0.216 root mean square error (RMSE), 0.144 mean absolute error (MAE), and 0.020 mean absolute percentage error (MAPE), and biochar is 0.886 RMSE, 0.489 MAE, and 0.019 MAPE. The optimum number of hidden neuron for this algorithm is 19 with mean square error 0.23. The biochar exhibits porous structure suggesting possible land application.

MoDTRAP: Improved heart rate tracking and preprocessing of motion-corrupted photoplethysmographic data for personalized healthcare

ObjectiveA new joint framework for heart rate (HR) tracking and preprocessing of motion artifact (MA) corrupted photoplethysmogram (PPG) data is presented. MethodsTwo popular signal decomposition techniques, viz., variational mode decomposition (VMD) to track HR, and ensemble empirical mode decomposition (EEMD) in conjunction with an artificial neural network (NN) model was utilized for preprocessing. The HR frequency was estimated from the signal strengths of VMD decomposed mode functions and tracked with an indigenous algorithm utilizing the fact that the PPG signal normally comprises base HR frequency along with its harmonics. For MA reduction, The HR synchronous signal component was extracted from the appropriate mode functions of VMD decomposition and reconstructed to minimize MA using a template matching technique utilizing an autoencoder and pre-trained multilayer feedforward neural network model. A combination of EEMD followed by a LSTM binary classifier was used to generate and update the reference PPG template beat for the above purpose. ResultsWith 2015 IEEE signal processing cup challenge database, an average absolute error and percent absolute error of 1.02 and 0.86 respectively compared to the ground truth HR were obtained for all 22 subjects. For MA reduction, we used wrist PPG data collected from 30 human subjects (normal and cardiovascular patients), achieving an average root mean squared error (RMSE) of 0.31 and SNR improvement of 21.23 dB. ConclusionProposed MoDTRAP technique provides noticeable improvements, separately, over existing methods on HR tracking and MA reduction. SignificanceThe proposed technique can be utilized for ambulatory healthcare monitoring.

Investigation on the Potential to Integrate Different Artificial Intelligence Models with Metaheuristic Algorithms for Improving River Suspended Sediment Predictions

Suspended sediment load (SLL) prediction is a significant field in hydrology and hydraulic sciences, as sedimentation processes change the soil quality. Although the adaptive neuro fuzzy system (ANFIS) and multilayer feed-forward neural network (MFNN) have been widely used to simulate hydrological variables, improving the accuracy of the above models is an important issue for hydrologists. In this article, the ANFIS and MFNN models were improved by the bat algorithm (BA) and weed algorithm (WA). Thus, the current paper introduces improved ANFIS and MFNN models: ANFIS–BA, ANFIS–WA, MFNN–BA, and MFNN–WA. The models were validated by applying river discharge, rainfall, and monthly suspended sediment load (SSL) for the Atrek basin in Iran. In addition, seven input groups were used to predict monthly SSL. The best models were identified through root-mean-square error (RMSE), Nash–Sutcliff efficiency (NSE), standard deviation ratio (RSR), percent bias (PBIAS) indices, and uncertainty analysis. For the ANFIS–BA model, RMSE and RSR varied from 1.5 to 2.5 ton/d and from 5% to 25%, respectively. In addition, a variation range of NSE was between very good and good performance (0. 75 to 0.85 and 0.85 to 1). The uncertainty analysis showed that the ANFIS–BA had more reliable performance compared to other models. Thus, the ANFIS–BA model has high potential for predicting SSL.

Detecting Orbital Angular Momentum Modes of Vortex Beams Using Feed-Forward Neural Network

Detecting orbital angular momentum (OAM) modes with high-speed and -accuracy is always a challenge in the practical applications of vortex beams (VBs). Here, we design a multilayer feed-forward neural network (FNN) model to rapidly identify OAM modes. After trained with loads of learning samples influenced by the transmission distance, beam waist, and atmospheric turbulence, the model shows a good generalization ability in quickly identifying the OAM mode of VBs ranging from −25 to +25. The classification accuracy reaches 99.55% under the influence of moderate turbulence and achieves 68.53% even under strong turbulence. Furthermore, we construct a communication link to deliver a 100 × 100 pixels Lena gray image by encoding OAM modes. After free-space transmission, the OAM-SK signals are demodulated by the FNN model, and the Lena gray image is well reconstituted. With an i5-6500 CPU, the identification process takes only 7.5 × 10−5 s per OAM mode. These numerical simulation results demonstrate that the proposed FNN model provides a feasible way to rapidly identify the OAM modes of VBs and may show great potential in the optical OAM communication, etc.

MLP Modeling and Prediction of IP Subnet Packets Forwarding Performance

In IP networks, packets forwarding performance can be improved by adding more nodes and dividing the network into smaller segments. Being able to measure and predict traffic flows to direct to a given segment can be crucial in respecting traffic shaping, scheduling and QoS. This paper proposes to model network packets forwarding performance for optimization and prediction purposes by using multi-layer feed-forward neural network model that uses sigmoid functions to activate the hidden nodes. Gradient descent technique has been considered to optimize and enhance the MLP accuracy. Simulations of MPL neurons training stages pointed out a relative improvement of the forwarding process when network posses a larger density of neurons. Numerical results validated our theoretical analysis and confirmed that to enhance the forwarding process, it is necessary to divide the network into small segments by optimizing resources allocation.

A GA-PSO Hybrid Algorithm Based Neural Network Modeling Technique for Short-term Wind Power Forecasting

This article proposes a hybrid neural network modeling technique for forecasting of wind power generation based on an integrated algorithm combining genetic algorithm (GA) and particle swarm optimization (PSO). The share of wind energy in electric power generation keeps growing supported by favorable environmental policies aiming at achieving low-emission targets. However, due to the intermittent and uncertain nature of wind flow, integration of wind power into electric power systems brings operational challenges to address. Accurate wind power generation forecasting tools play a key role to address the challenges. A multi-layered feed-forward artificial neural network model optimized by a combination of genetic algorithm and particle swarm optimization algorithm is developed in this work for wind power generation forecasting. The proposed technique is tested based on practical information obtained from Goldwind Smart Microgrid in Beijing. The performance of the proposed method is superior to neural network models optimized using GA and PSO separately, as well as the benchmark persistence approach.

Real and predicted mortality under health spending constraints in Italy: a time trend analysis through artificial neural networks

BackgroundAfter 2008 global economic crisis, Italian governments progressively reduced public healthcare financing. Describing the time trend of health outcomes and health expenditure may be helpful for policy makers during the resources’ allocation decision making process. The aim of this paper is to analyze the trend of mortality and health spending in Italy and to investigate their correlation in consideration of the funding constraints experienced by the Italian national health system (SSN).MethodsWe conducted a 20-year time-series study. Secondary data has been extracted from a national, institution based and publicly accessible retrospective database periodically released by the Italian Institute of Statistics. Age standardized all-cause mortality rate (MR) and health spending (Directly Provided Services - DPS, Agreed-Upon Services - TAUS, and private expenditure) were reviewed. Time trend analysis (1995–2014) through OLS and Multilayer Feed-forward Neural Networks (MFNN) models to forecast mortality and spending trend was performed. The association between healthcare expenditure and MR was analyzed through a fixed effect regression model. We then repeated MFNN time trend forecasting analyses on mortality by adding the spending item resulted significantly related with MR in the fixed effect analyses.ResultsDPS and TAUS decreased since 2011. There was a mismatch in mortality rates between real and predicted values. DPS resulted significantly associated to mortality (p < 0.05). In repeated mortality forecasting analysis, predicted MR was found to be lower when considering the pre-constraints health spending trend.ConclusionsBetween 2011 and 2014, Italian public health spending items showed a reduction when compared to prior years. Spending on services directly provided free of charge appears to be the financial driving force of the Italian public health system. The overall mortality was found to be higher than the predicted trend and this scenario may be partially attributable to the healthcare funding constraints experienced by the SSN.

Performance comparison of feedforward neural network training algorithms in modeling for synthesis of polycaprolactone via biopolymerization

This paper reports the biopolymerization of e-caprolactone, using lipase Novozyme 435 catalyst at varied impeller speeds and reactor temperatures. A multilayer feedforward neural network (FFNN) model with 11 different training algorithms is developed for the multivariable nonlinear biopolymerization of polycaprolactone (PCL). In previous works, biopolymerization carried out in scaled-up bioreactors is modeled through FFNN. No review discussed the role of different training algorithms in artificial neural network on the estimation of biopolymerization performance. This paper compares mean absolute error, mean square error, and mean absolute percentage error (MAPE) in the PCL biopolymerization process for 11 different training algorithms that belong to six classes, namely (1) additive momentum, (2) self-adaptive learning rate, (3) resilient backpropagation, (4) conjugate gradient backpropagation, (5) quasi-Newton, and (6) Bayesian regulation propagation. This paper aims to identify the most effective training method for biopolymerization. Results show that the quasi-Newton-based and Levenberg–Marquardt algorithms have the best performance with MAPE values of 4.512, 5.31, and 3.21% for the number of average molecular weight, weight average molecular weight, and polydispersity index, respectively.

A Deep Learning Approach for Predicting Antidepressant Response in Major Depression Using Clinical and Genetic Biomarkers.

In the wake of recent advances in scientific research, personalized medicine using deep learning techniques represents a new paradigm. In this work, our goal was to establish deep learning models which distinguish responders from non-responders, and also to predict possible antidepressant treatment outcomes in major depressive disorder (MDD). To uncover relationships between the responsiveness of antidepressant treatment and biomarkers, we developed a deep learning prediction approach resulting from the analysis of genetic and clinical factors such as single nucleotide polymorphisms (SNPs), age, sex, baseline Hamilton Rating Scale for Depression score, depressive episodes, marital status, and suicide attempt status of MDD patients. The cohort consisted of 455 patients who were treated with selective serotonin reuptake inhibitors (treatment-response rate = 61.0%; remission rate = 33.0%). By using the SNP dataset that was original to a genome-wide association study, we selected 10 SNPs (including ABCA13 rs4917029, BNIP3 rs9419139, CACNA1E rs704329, EXOC4 rs6978272, GRIN2B rs7954376, LHFPL3 rs4352778, NELL1 rs2139423, NUAK1 rs2956406, PREX1 rs4810894, and SLIT3 rs139863958) which were associated with antidepressant treatment response. Furthermore, we pinpointed 10 SNPs (including ARNTL rs11022778, CAMK1D rs2724812, GABRB3 rs12904459, GRM8 rs35864549, NAALADL2 rs9878985, NCALD rs483986, PLA2G4A rs12046378, PROK2 rs73103153, RBFOX1 rs17134927, and ZNF536 rs77554113) in relation to remission. Then, we employed multilayer feedforward neural networks (MFNNs) containing 1–3 hidden layers and compared MFNN models with logistic regression models. Our analysis results revealed that the MFNN model with 2 hidden layers (area under the receiver operating characteristic curve (AUC) = 0.8228 ± 0.0571; sensitivity = 0.7546 ± 0.0619; specificity = 0.6922 ± 0.0765) performed maximally among predictive models to infer the complex relationship between antidepressant treatment response and biomarkers. In addition, the MFNN model with 3 hidden layers (AUC = 0.8060 ± 0.0722; sensitivity = 0.7732 ± 0.0583; specificity = 0.6623 ± 0.0853) achieved best among predictive models to predict remission. Our study indicates that the deep MFNN framework may provide a suitable method to establish a tool for distinguishing treatment responders from non-responders prior to antidepressant therapy.

Prediction of the Tensile Response of Carbon Black Filled Rubber Blends by Artificial Neural Network.

The precise experimental estimation of mechanical properties of rubber blends can be a very costly and time-consuming process. The present work explores the possibilities of increasing its efficiency by using artificial neural networks to study the mechanical behavior of these widely used materials. A multilayer feed-forward back-propagation artificial neural network model, with a strain and the carbon black content as input parameters and stress as an output parameter, has been developed to predict the uniaxial tensile response of vulcanized natural rubber blends with different contents of carbon black in the form of engineering stress-strain curves. A novel procedure has been created for the simulation of the optimized artificial neural network model with input datasets generated by a regression model of an experimental dependence of tensile strain-at-break on the carbon black content in the investigated blends. Errors of the prediction of experimental stress-strain curves, as well as of tensile strain-at-break, tensile stress-at-break and M100 tensile modulus were estimated for all simulated stress-strain curves. The present study demonstrated that the performance of a developed neural network model to predict the stress-strain curves of rubber blends with different contents of carbon black is also exceptionally high in the case of a network that had never learned the input data, which makes it a suitable tool for extensive use in practice.

Hyperspectral signature analysis using neural network for grade estimation of copper ore

The ever-increasing demand for the different metal and mineral resources from the earth’s subsurface has brought tremendous pressure on the geochemical laboratory for the growing countries. The success of any mining industry relies on the estimated values of ore grade in the mineral deposit. Hence, rapid assessment of ore grade is critical in daily schedule in mines operations. Commonly the assay value is determined by chemical analysis or X-Ray Fluorescence (XRF), which is one of the constrained by real-time grade estimation, duration of sample preparation and processing. Several researches carried out in exploration and revealed that hyperspectral technique is a promising tool for mineral identification and mapping. The goal of the present study is to determine the effectiveness of narrow band spectroscopy in Cu grade estimation. To achieve this, a multilayer feed-forward neural network model has been developed to establish a functional link between hyperspectral signature derived features with the copper grade. Altogether eight different types of features including absorption depth, band depth center, the area under the absorption curve, full width at half maxima were extracted from continuum removed spectra along with derivative reflectance features, e.g. band depth ratio, 1st and 2nd slopes from the hyperspectral profile. The dimensionality was reduced by applying Principal Component Analysis onto the extracted features. The first seven PCAs are then used as input vector of the ANN model. A five-fold cross-validation exercise is carried out for model performance. The high degree of correlation reveals that the PCA generated feature from hyperspectral data coupled with ANN model could be an alternative approach to predict the copper grade for the copper mine.