Cascade-forward Back Propagation Neural Network Research Articles

Automatic Flood Detection in Multi-Temporal Sentinel-1 Synthetic Aperture Radar Imagery Using ANN Algorithms

Natural Calamities like floods cause wide-range of damage to human existence as well as substructures. For automatic extraction of flooded area in multi-temporal satellite imagery acquired by Sentinel-1 Synthetic Aperture Radar (SAR), this paper presents two neural network algorithms: Feed-Forward Neural Network, Cascade-forward back-propagation neural network. This work currently focuses on Uttar Pradesh in India, which was affected due to floods during August 2017. The two models are trained, validated and tested using MATLAB R2018b. The models are first trained using a variety of input data until the percentage of error with respect to water body detection is within an acceptable error limit. These models are then used to extract the water features effectively and to detect the flooded regions. Finally, flood area is calculated in sq. km in during flood and post-flood imagery using these algorithms. The results thus obtained are compared with that from the binary thresholding method from previous studies. The results show that the Feed- Forward Neural Network gives better accuracy than the Cascade-forward back propagation neural network. Based on the promising results, the proposed method may assist in our understanding of the role of machine learning in disaster detection.

Performance Analysis of Fuzzy Multilayer Support Vector Machine for Epileptic Seizure Disorder Classification using Auto Regression Features

Background:Around 1% of the total population in the world suffers from epilepsy, a central nervous system disorder. Epilepsy is the neurological disorder of the human brain which can affect people of all ages. Classification techniques and Signal processing are basic methods in the advancement of an algorithm for seizure detection. The primary procedures of a typical biomedical evaluation and processing framework are data acquisition, feature extraction, preprocessing, and classification. Based on this, seizure detection is performed by using the following two methods.Methods:This paper proposes a technique for the classification of EEG signals to detect the epileptic seizures by using Cascade Forward Backpropagation Neural Network (CFBNN) and Fuzzy Multilayer Support Vector Machine (FMSVM) methods.Results:Finally, the results of developed classifiers are identified with seizure disorder activities. This research concentrated on Parametric Features such as AR (Autoregressive) Burg, AR YuleWalker, AR Covariance, AR Modified Covariance, and Levinson Durbin Recursion. Linear Prediction Coefficient was analyzed with the EEG dataset gathered from Karunya University. The sensitivity, specificity, and accuracy were calculated for the proposed classifiers.Conclusion:The results of the proposed classifiers were computed with minimum and maximum accuracy and these results were compared with the previous results of the classifiers like FFNN, and PNN as shown in the tables. Based on the obtained outputs and calculated parametric functions, the results validated that the FMSVM classifier performed better in the detection of epileptic seizure disorder in terms of accuracy, sensitivity and specificity.

ANN model for predicting acrylonitrile wastewater degradation in supercritical water oxidation

The discharged acrylonitrile wastewater had aroused more and more attention due to the increasingly serious water pollution. Supercritical water oxidation (SCWO) was an effective and fast way to degrade it completely without secondary pollution. To better illustrate the performances of SCWO of acrylonitrile wastewater, the experimental research covered the effects of different operation conditions on TOC reduction, such as reduced temperature (T/Tc), reduced pressure (P/Pc), initial total organic carbon concentration (TOC0), stoichiometric ratio (SR) and residence time (t). For a more accurate prediction of the emissions, two kinds of artificial neural network (ANN) models were adopt to simulate the TOC reductions in the processes of SCWO of acrylonitrile wastewater, including the Cascade-forward back propagation neural network (CFBPNN) and Feed-forward back propagation neural network (FFBPNN). The input parameters of ANN models were T/Tc, P/Pc, TOC0, SR and t. The output parameter was TOC reduction (η). The mean square error (E2) and the coefficient of determination (R2) were used to evaluate the model performances, respectively. Both the model and the experiment results had shown the TOC reduction could be greatly improved by reduced temperature, reduced pressure, initial TOC concentration, stoichiometric ratio and residence time. The FFBPNN model with the hidden neurons numbers of 12 was shown much better performances than the CFBPNN model.

A Scale and Rotation Invariant Urdu Nastalique Ligature Recognition Using Cascade Forward Backpropagation Neural Network

In the emerging age of technologies, machines are becoming more and more skilled and capable just like humans. Despite the fact that machines do not have their own intelligence, but still due to advancement in Artificial Intelligence (AI), machines are rapidly advancing. The area of Pattern Recognition (PR) deals with bringing enhancements to identify obscure patterns corresponding to specific classes. Optical Character Recognition (OCR) is a subfield of PR which deals with the recognition of characters. A great work has been done for Japanese, Hindi, Arabic and Chinese scripts, but only a diminutive work has been done for Urdu script. The Urdu language is highly cursive and is written in different calligraphic styles like Naskh, Nastalique, Kofi, Devani and Riqa. The Nastalique font is very calligraphic with aesthetic beauty. The ligature segmentation of Urdu Nastalique is also more difficult as compared to other languages. Urdu Nastalique has some characteristics like stacking of ligatures and cursiveness which makes its ligature segmentation a difficult task. Cursiveness means ligatures are joined together to form a new shape. It contains connected ligatures which makes it more complicated as compared to other languages. The ligature recognition of Urdu text by an OCR is a strenuous task due to variants of scaling, rotation, orientation and font style. In this study, a scale and rotation invariant classifier for Urdu Nastalique OCR is proposed. A combination of scale and location invariant moments is used for feature extraction and the classification is performed using Cascade Forward Backpropagation Neural Network. The model is validated through independent dataset testing and 5-fold cross-validation which gave 96.474% and 96.922% accuracy. The results depict the adaptability of the proposed model due to its high accuracy for recognition of Urdu Nastalique Ligature.

Nonlinear Modeling of the Flux Linkage in 2-D Plane for the Planar Switched Reluctance Motor

This paper proposes a nonlinear flux linkage model in 2-D plane for the planar switched reluctance motor (PSRM). The inputs of the proposed model are the 2-D positions and the current, and the output is the flux linkage. The proposed model is established via a cascade-forward backpropagation neural network (CFNN). The designed CFNN consists of four layers: one input layer, two hidden layers, and one output layer. The first hidden layer has 20 neurons with a tan-sigmoid transfer function, and the second hidden layer has 20 neurons with a log-sigmoid transfer function. The output layer is a pure linear layer. The sample set with 179 755 samples is obtained experimentally in a dSPACE-based PSRM system by applying the dc excitation method. The sample set is divided into three sets. 35% and 30% of the samples are randomly chosen as the training sample set and validation sample set, respectively, and the remaining samples are utilized as the test sample set to assess the generalization performance of the CFNN-based model. According to the results of the test sample set, the maximum relative error is 11.05% and the mean relative error is 0.42% when the current ranges from 1 to 9 A. The CFNN has the capability to build a multi-input nonlinear model. The CFNN-based model is capable of reflecting the variations of flux linkage in 2-D plane caused by manufacturing tolerances. The effectiveness of the CFNN-based model is finally verified.

Application of response surface methodology and artificial neural network modeling to assess non-thermal plasma efficiency in simultaneous removal of BTEX from waste gases: Effect of operating parameters and prediction performance

Abstract This study aimed to assess the prediction efficiencies of response surface methodology (RSM) and artificial neural network (ANN)-based models in terms of benzene, toluene, ethylbenzene, and xylenes (BTEX) removal from a polluted airstream using non-thermal plasma (NTP). The effect that key elements of the NTP process, including temperature, BTEX concentration, voltage and flow rate, had on the BTEX elimination efficiency was investigated using a central composite RSM design along with three ANN models including Feed-Forward Back Propagation Neural Network (FFBPNN), Cascade-Forward Back Propagation Neural Network (CFBPNN) and Elman-Forward Back Propagation Neural Network (EFBPNN) with the topology of 4-h-1. The RSM and ANN models were statistically compared using some indicators including Sum of Squared Errors (SSE), adjusted R2, determination coefficient (R2), Root Mean Squared Error (RMSE), Absolute Average Deviation (AAD). According to the RSM output, voltage was the most efficient variable with a coefficient proportion of 8.28. Besides, FFBPNN was the best model among the considered ANN models. Also, the R2 achieved for ANN (FFBPNN) and RSM models were 0.9736 and 0.9656 correspondingly. Therefore, it was concluded that the ANN (FFBPNN) represents a powerful tool for modeling the BTEX removal.

Potential ANN prediction model for multiperformances WEDM on Inconel 718

This paper proposes a machining performance prediction approach on multiple performances of wire electrical discharge machining (WEDM) on Inconel 718. Artificial neural network (ANN) is emphasized to predict the machining performances. Many efforts have been made to model the performances of WEDM using ANN. However, to obtain the best ANN model, generally the network parameters are not consistent and so far, the selection has been made in a random manner and resulted in an excessive experimental trial. Taguchi design orthogonal array L256 is implemented in the process of network parameter selection to search for the potential machining model. This approach, prescribed as OrthoANN, is simplified to avoid as much as unnecessary experimentations. Material removal rate, surface roughness (Ra), cutting speed (Vc) and sparking gap (Sg) are the machining performances considered in this study. Five machining parameters considered; pulse on time, pulse off time, peak current, servo voltage and flushing pressure. Cascade forward back-propagation neural network (CFNN) is found to be the best network type of the selected data set. One hidden layer 5–14–4 CFNN showed the most precise and generalized network architecture with very good prediction accuracy. An average of 5.16% error is generated which seems to be in superior concurrence with the actual experimental results. Confirmation test is carried out to verify the machining performances suggested by this approach.

Modelling of dust removal in rotating packed bed using artificial neural networks (ANN)

Artificial neural network (ANN) models, including the Cascade-forward back propagation neural network (CFBPNN), feed-forward back propagation neural network (FFBPNN) and Elman-forward back propagation neural network (EFBPNN), were proposed to predict the dust removal efficiency in rotating packed bed (RPB) to speed up its development. Total 326 data sets for separation grade efficiency had been collected from literatures for training and verifying the model. Gas Reynolds number (ReG), liquid Reynolds number (ReL), rotational Reynolds number (Reω), M (d02ρL/dp2/ρp) and Csi/ρG were used as input data. While, the variable η (separation grade efficiency) was taken as output data for each model. Various of hidden neurons were compared based on the mean square error (E2), coefficient of determination (R2) and residual for each model. The separation grade efficiency in RPB was also compared with other existed dust removal equipments.

Performance comparison of artificial neural network models for daily rainfall prediction

With an aim to predict rainfall one-day in advance, this paper adopted different neural network models such as feed forward back propagation neural network (BPN), cascade-forward back propagation neural network (CBPN), distributed time delay neural network (DTDNN) and nonlinear autoregressive exogenous network (NARX), and compared their forecasting capabilities. The study deals with two data sets, one containing daily rainfall, temperature and humidity data of Nilgiris and the other containing only daily rainfall data from 14 rain gauge stations located in and around Coonoor (a taluk of Nilgiris). Based on the performance analysis, NARX network outperformed all the other networks. Though there is no major difference in the performances of BPN, CBPN and DTDNN, yet BPN performed considerably well confirming its prediction capabilities. Levenberg Marquardt proved to be the most effective weight updating technique when compared to different gradient descent approaches. Sensitivity analysis was instrumental in identifying the key predictors.

Modelling of adsorption in rotating packed bed using artificial neural networks (ANN)

Rotating packed bed (RPB) exhibits good performance for absorption, distillation and extraction. However, the development of RPB was restricted by complicated liquid flow pattern on activated carbon in adsorption. Therefore, a model of artificial neural network (ANN) was adopted to predict the adsorption in RPB. The experimental data were classified into two groups, and they were training ones for establishing the model and testing ones for validating the model. Different types of ANN models, including Cascade-forward back propagation neural network (CFBPNN), Elman-forward back propagation neural network (EFBPNN) and Feed-forward back propagation neural network (FFBPNN), were investigated in this study. High gravity factor, liquid Reynolds number, adsorption time to the maximum adsorption time and packing density to liquid concentration were used as input data. While, the adsorption amount to the maximum adsorption amount was taken as output data for each model. Optimal hidden neurons for FFBPNN, EBPNN, CFBPNN were 9, 12 and 8, respectively. Experimental adsorption amount data were in good agreement with the predicted data indicating that the ANN models had a superior performance. Besides, the ANN models exhibited a more accurate prediction and had more generalization ability than multiple nonlinear regression model (MNR). The proposed FFBPNN model of absorption in RPB will provide an optimization tool to maximize the adsorption amount.

Side-Channel Information Characterisation Based on Cascade-Forward Back-Propagation Neural Network

Traditional cryptanalysis assumes that an adversary only has access to input and output pairs, but has no knowledge about internal states of the device. However, the advent of side-channel analysis showed that a cryptographic device can leak critical information. In this circumstance, Machine learning is known as a powerful and promising method of analysing of side-channel information. In this paper, an experimental investigation on a FPGA implementation of elliptic curve cryptography (ECC) was conducted to explore the efficiency of side-channel information characterisation based on machine learning techniques. In this work, machine learning is used in terms of principal component analysis (PCA) for the preprocessing stage and a Cascade-Forward Back-Propagation Neural Network (CFBP) as a multi-class classifier. The experimental results show that CFBP can be a promising approach in characterisation of side-channel information.

A new hybrid quadratic regression and cascade forward backpropagation neural network

In this study, a quadratic regression model (QRM) and a cascade forward backpropagation neural network (CFBN) are jointly integrated together to form a hybrid model called the new hybrid quadratic regression method and cascade forward backpropagation neural (QRM-CFBN) network method. The new hybrid method was tested on a daily time series data obtained from the UCI repository data link and the data set was collected from a combined cycle power plant. The joint integration was made possible by the Bayesian model averaging technique, which was used to obtain a combined forecast from the two separate methods. The model resulting from the joint integration was applied on the log difference series of the original time series data. The results obtained from the new hybrid QRM-CFBN were compared with the results obtained from the hybrid ARIMA-RNN, standalone cascade forward backpropagation neural (CFBN) network and layered recurrent neural network (LRNN) after being tested on the same sample time series data respectively. The comparison indicates that the results emerging from the new hybrid QRM-CFBN method on the average, generally results in better performance when compared with the hybrid ARIMA-RNN, the standalone CFBN and the standalone LRNN for 1day, 3days as well as 5days prediction mean absolute error (MAE) and root mean square error (RMSE) for varying data samples of 50, 100, 200, 400 and 800days respectively. The RMSEs and the MAEs were applied to ascertain the assertion that the new jointly integrated forecast has better forecasting performance greater than the standalone CFBN and LRNN forecasts as well as ARIMA-RNN forecast. The analysis for this study was simulated using MATLAB software, version 8.03.

English

In this study, artificial neural networks are suggested as a model that can be ‘trained’ to yield qualitative results out of a huge amount of categorical data. It can be said that this is a new approach applied in educational qualitative data analysis. In this direction, a cascade-forward back-propagation neural network (CFBPN) model was developed to analyze categorical data for determine students’ attitudes. The data were collected using a conceptual understanding test which includes open-ended questions. The results of this study indicate that using CFBPN model in analyzing data from educational research examining attitudes, behaviors, or beliefs may help us obtain more detailed information about the data analyzed and hence about the characteristics of the participants involved. Key words: artificial neural networks; qualitative data analysis; education; biology

Face Recognition System Based on Different Artificial Neural Networks Models and Training Algorithms

Face recognition is one of the biometric methods that is used to identify any given face image using the main features of this face. In this research, a face recognition system was suggested based on four Artificial Neural Network (ANN) models separately: feed forward backpropagation neural network (FFBPNN), cascade forward backpropagation neural network (CFBPNN), function fitting neural network (FitNet) and pattern recognition neural network (PatternNet). Each model was constructed separately with 7 layers (input layer, 5 hidden layers each with 15 hidden units and output layer). Six ANN training algorithms (TRAINLM, TRAINBFG, TRAINBR, TRAINCGF, TRAINGD, and TRAINGD) were used to train each model separately. Many experiments were conducted for each one of the four models based on 6 different training algorithms. The performance results of these models were compared according to mean square error and recognition rate to identify the best ANN model. The results showed that the PatternNet model was the best model used. Finally, comparisons between the used training algorithms were performed. Comparison results showed that TrainLM was the best training algorithm for the face recognition system.

Wire electrical discharge machining of gamma titanium aluminide for optimum process criteria yield in single pass cutting operation

Selection of optimum machining parameter combinations for obtaining higher cutting efficiency and accuracy is a challenging task in WEDM due to the presence of large number of process variables and complicated stochastic process mechanism. In the present research, single pass wire electrical discharge machining (WEDM) of γ TiAl alloy has been studied. This paper attempts to develop an appropriate machining strategy for maximum process criteria yield in WEDM. A cascade-forward back-propagation neural network based on Bayesian regularisation is developed to model the machining process. The three most important parameters, cutting speed, surface roughness and wire offset, have been considered as measure of process performance. The model is capable of predicting the response parameters (cutting speed, surface roughness and wire offset) as a function of different control parameters (pulse on time, pulse off time, peak current, wire tension, dielectric flow rate and servo reference voltage). Verification experiments have been carried out to check the validity of the developed model and then optimal parametric combinations were searched out using an advanced optimisation strategy.