Performance Of Artificial Neural Network Research Articles

Assessing the Potential of Artificial Intelligence (Artificial Neural Networks) in Predicting the Spatiotemporal Pattern of Wildfire-Generated PM2.5 Concentration

To evaluate the health effects of wildfire smoke, it is crucial to identify reliable models, at fine spatiotemporal resolution, of exposure to wildfire-generated PM2.5. To this end, satellite-drived aerosol optical depth (AOD) measurements are widely used in exposure models, providing long and short-term PM2.5 predictions. Multiple regression models, specifically land use regression (LUR), incorporating AOD images have shown good potential for estimating long-term PM2.5 exposure, but less so for short-term predictions. In this study, we developed artificial neural networks (ANNs) and, in particular, multilayer perceptron (MLP) by integrating ground-based PM2.5 measurements with AOD images and meteorological and spatial variables. Moreover, we used spatial- and temporal-ANNs to investigate and compare the ANNs’ ability to predict different PM2.5 concentration levels caused by abrupt spatial and temporal changes in fire smoke. The study herein analyzes and compares the viability of previously established neural network approaches in predicting short-term PM2.5 exposure during the 2014–2017 wildfire seasons in the province of Alberta, Canada. The performance of ANNs is also compared to classical models, including simple correlation (PM2.5 vs. AOD) and multiple linear regression (MLR) including meteorological and land-use predictors (MET_AOD_LUR). Our study shows that ANN achieved a 15% to 113% R2 increase compared to competing models.

Feature Extraction and Classification of Power Quality Events Based on Fast Fourier Transformation and Artificial Neural Network

This paper presents an effective method for detection and classification of Power Quality Events (PQE), based on Fast Fourier Transformation (FFT) for event identification and Artificial Neural Network (ANN) technique for classifying of these events. Firstly, synthetic data such as pure sine as a reference, voltage sag, voltage swell, flicker, transient, voltage with harmonics are created in MATLAB based on TS EN 50160 standard. Database with 480 PQE waveforms is generated with 80 samples for each of the 6 types of the waveform with randomly different event amplitude, beginning occurrence time, time duration, frequency component and angle according to a type of event. FFT is used to extract features of the events by decomposing the signal. Then, 16384×480 data are reduced to 480×480 data by applying Principal Component Analysis (PCA) that is prevent over-learning, obtain less runtime using less computing power and reduce data and storage space. Finally, a total of 480 PQE are classified by using ANN. 336 of these PQE are used for training cluster, 72 of PQE are used for verification and the remaining 72 are used for testing. Firstly, the ANN has been trained correctly. The classification performance of the ANN in PQE has been examined by inserting the test into ANN. The performance of ANN is 99.8% for these PQE. The purpose of this research is to provide an artificial intelligence assistant that can fast and accurately advise the power system operators for the networks, and the results also show that the goal has been achieved.

CAN EXCHANGE RATES BE PREDICTABLE? EXCHANGE RATE PREDICTION MODEL WITH ANN FOR TURKEY

Ponte Academic JournalMar 2021, Volume 77, Issue 3 CAN EXCHANGE RATES BE PREDICTABLE? EXCHANGE RATE PREDICTION MODEL WITH ANN FOR TURKEYAuthor(s): Fuat Sekmen ,Haşmet Gokırmak, Murat Kürkcü, Şükrü Apaydın, Hasan MemişJ. Ponte - Mar 2021 - Volume 77 - Issue 3 doi: 10.21506/j.ponte.2021.3.9 Abstract:The forecasting concept can be expressed as predicting a variable’s future values using various methods under certain assumptions. In the estimations made using the time series, the value of a variable in the current period and the past period values are used. In the time series analysis, past time values of the variable to be estimated constitute the model’s independent variables. Learning from data, generalizing, working with very large samples, etc. are essential Artificial Neural Networks (ANN) methodology features. ANN methodology, which provides significant advantages thanks to these features, has vast usage possibilities in prediction modeling as in other fields. ANN has become one of the time series estimations’ methods since the late 1980s. This study handles an artificial neural network learning paradigm to estimate the exchange rate for economic targets. This paradigm has not prior conditions such as normal distribution or stationary. Therefore it has been frequently preferred in time-series analyses. The present study predicts the Turkish Lira’s value against the U.S. Dollar by using an ANN methodology. Our study is built upon purchasing power parity and the different researches dealing with exchange rate predictions. Our primary purpose is to minimize the error between the expected output of the network and the output it produces. This study uses an Artificial Neural Network model that builds upon the MLP to estimate the exchange rates. We first define the MLP with one hidden layer for this model, using monthly data from 2000 to 2019. In contrast, the independent variables are interest rates, Gross Domestic Product, and Consumer Price Index data for Turkey and the United States of America. We prefer batch training to train the ANN in SPSS Neural Networks 17.0 software. Sixty percent of the data were used in training, 25 percent in the testing, and 10 percent in the holdout, of the artificial neural networks, respectively. According to research results, the performance of ANN is relatively higher. While the relative error in the training set is 1 percent, the test set’s relative error is below 1 percent (0,007). Download full text:Check if you have access through your login credentials or your institution Username Password

A review study of application of artificial intelligence in construction management and composite beams

This paper is aimed to review the use of artificial intelligence (AI) algorithms in diverse civil engineering applications such as predicting and evaluating the different parameters of composite beams and shear connectors and determining the compressive strength of concrete. Also, the application of AI methods especially artificial neural network (ANN) in construction engineering and management including prediction and estimation, decision-making, classification or selection, optimization and risk analysis and safety has been thoroughly discussed. Furthermore, the integration of Artificial Neural network (ANN) with other soft computing methods, such as Backpropagation (BP), imperialist competitive algorithm (ICA), support vector regression (SVR), back-propagation neural network (BPNN), Genetic Algorithms (GA) and Multilayer feed forward (MLFF) has been reviewed. It has been reported that the combination of ANN with other intelligence algorithms leads to providing more accurate results. Moreover, the performance of ANN with other soft computing techniques, such as BP, BPNN, SVR, GA, ICA, and MLFF in various fields has been compared and ANN in many cases had superiority over other models.

A Cartesian Genetic Programming Based Parallel Neuroevolutionary Model for Cloud Server’s CPU Usage Prediction

Cloud computing use is exponentially increasing with the advent of industrial revolution 4.0 technologies such as the Internet of Things, artificial intelligence, and digital transformations. These technologies require cloud data centers to process massive volumes of workloads. As a result, the data centers consume gigantic amounts of electrical energy, and a large portion of data center electrical energy comes from fossil fuels. It causes greenhouse gas emissions and thus ensuing in global warming. An adaptive resource utilization mechanism of cloud data center resources is vital to get by with this huge problem. The adaptive system will estimate the resource utilization and then adjust the resources accordingly. Cloud resource utilization estimation is a two-fold challenging task. First, the cloud workloads are sundry, and second, clients’ requests are uneven. In the literature, several machine learning models have estimated cloud resources, of which artificial neural networks (ANNs) have shown better performance. Conventional ANNs have a fixed topology and allow only to train their weights either by back-propagation or neuroevolution such as a genetic algorithm. In this paper, we propose Cartesian genetic programming (CGP) neural network (CGPNN). The CGPNN enhances the performance of conventional ANN by allowing training of both its parameters and topology, and it uses a built-in sliding window. We have trained CGPNN with parallel neuroevolution that searches for global optimum through numerous directions. The resource utilization traces of the Bitbrains data center is used for validation of the proposed CGPNN and compared results with machine learning models from the literature on the same data set. The proposed method has outstripped the machine learning models from the literature and resulted in 97% prediction accuracy.

A Performance Comparison of Artificial Neural Networks and Spiking Neural Networks

The continual development of artificial neural networks (ANNs) on von-Neuman based hardware poses a threat to potential advances in artificial intelligence (AI). At current rates, power usage of such models will rival that of large cities or even supersede the amount of power produced globally. Therefore, it is imperative to make a transition to a more power efficient computing architecture to make way for future advances in AI. The development of neuromorphic hardware has paved a path for a new set of neural networks called spiking neural networks (SNNs) to act as a viable alternative for machine learning tasks on such brain-like hardware. The primary motive of the research is to contribute towards a search for viable applications of SNNs in their current infancy. The project wishes to address whether SNNs can outperform ANNs image classification of handwritten digits on the MNIST dataset. Despite training on non-optimal von Neumann hardware, statistically significant evidence suggests that the SNN outperformed the ANN in terms of the mean accuracy of the task. As a result of the findings, the paper suggests that further research should focus on translating older ANN models to SNN counterparts that outperform on their tasks and finding applications of SNNs on von Neumann hardware that outperform ANNs. Both prior focuses will narrow the gap between ANN and SNN performance, and in doing so makes a transition to more power efficient and uninhibited AI advancement a reality.

Microwave Imaging of the Neck by Means of Artificial Neural Networks for Tumor Detection

In this paper, a microwave imaging approach based on artificial neural networks (ANNs) for neck tumor detection is presented. The aim of this technique is to retrieve the geometric and dielectric properties of the neck to identify the possible presence of tumors, starting from scattered electric field data. A fully-connected neural network is developed to test the feasibility of the proposed approach. Moreover, a numerical model including the main features of a cross section of the neck is specifically designed in order to create a suitable training dataset. Subsequently, for the optimization of the ANN architecture and performance evaluation, a numerical analysis is conducted. A set of simulated cases, based on realistic neck phantoms, is tested to evaluate the robustness of the network. Preliminary results show the possibility to identify and locate neck tumors.

Highly Accurate Prediction Model for Daily Runoff in Semi-Arid Basin Exploiting Metaheuristic Learning Algorithms

Developing trustworthy rainfall-runoff (R-R) models can offer serviceable information for planning and managing water resources. Use of artificial neural network (ANN) in adopting such models and predicting changes in runoff has become popular among many hydrologists from a long time. However, since the optimization is the most significant phase in ANN training, researchers’ attentiveness has been attracted to the ANN’s biggest problem, i.e. its susceptibility of being blocked in local minima. Consequently, use of genetic algorithms (GA), particle swarm optimization (PSO), firefly algorithm (FFA) and improved particle swarm optimization (IPSO) approaches to increase the performance of ANN, have gained remarkable interest among distinct modern heuristic optimization approaches. In this paper, the capability of four improved ANN methods, hybrid GA-based ANN, PSO-based ANN, FFA-based ANN and IPSO-based ANN in modeling rainfall-runoff (R-R) is investigated. IPSO has been used in order to increase the ability of PSO, where the new positions of particles are dynamically adjusted using two procedures which is given form the velocity obtained by PSO and proposed velocity in IPSO. The random normal grated number with a dynamical scale factor is used to compute the new position of the best particles in proposed velocity. Daily R-R data from six stations distributed in the Seybouse watershed located in semi-arid region in Algeria were used in models’ development. The selection of the input data sets was carried out using the autocorrelation, partial autocorrelation and cross correlation functions. The results of the four hybrid models were compared via performance metrics, viz., Root Mean Square Error (RMSE), Pearson’s correlation coefficient (R), Nash Sutcliffe Efficiency coefficient (NSE), and via graphical analysis (scatter plots, time series and Taylor diagram). Outcomes of the analysis at all study stations disclosed that all the ANN models enhanced with IPSO overachieved the GA-based ANN, PSO-based ANN and FFA-based ANN models in estimating runoff for both training and testing periods. The outcomes of the study indicate that the IPSO hybrid metaheuristic algorithm is the best technique in improving ANN capability in modeling daily R-R.

Using artificial neural networks to improve phosphorus indices

The phosphorus index (PI) was developed as a field-scale assessment tool used to identify critical source areas of phosphorus (P) loss, thus most US states have adopted the PI as their strategy for targeted management and conservation practices for effective mitigation of P loss from agricultural landscapes to surface waters. Recent studies have focused on evaluating and updating PI weighting factors (WFs) to ensure agreement between PI values and measured losses of P. Given that the WF of each site characteristic is usually determined individually without considering possible interactions, the goal of this study was to demonstrate how artificial neural networks (ANNs) that consider real-world interdependence can be used to determine WFs. Our specific objectives were to evaluate ANN performance for predicting soluble P (SP) concentrations in tile effluent using site characteristics as predictor variables, and to evaluate whether ANN-generated WFs can be used to improve PI performance. Garson’s algorithm was used to determine the relative importance of each site characteristic to SP loss. Data from a monitored in-field laboratory were used to evaluate the ability of a PI with no WFs (PI_NO), a PI with WFs as proposed in the original Lemunyon and Gilbert PI (PI_LG), and a PI with ANN-generated WFs (PI_ANN), to estimate SP loss potential in tile discharge. Simulation results showed that the ANN model provided reliable estimates of SP in tile effluent (<i>R²</i> = 0.99; RMSE = 0.0024). The relative importance analysis underscored the value of routine soil P testing for agronomic sufficiency for environmental stewardship, and highlighted the necessity of prioritizing both contemporary and legacy P sources during P loss risk assessments. Unlike the other PIs, PI_ANN was able to provide reasonable estimates of SP loss potential as illustrated with significant exponential relationships (<i>R²</i> = 0.60; <i>p</i> &lt; 0.001) between PI_ANN values and measured mean annual SP concentrations in tile effluent. These findings demonstrate that ANNs can be used to develop PIs with a strong correlation to measured SP.

Transesterification via Parametric Modelling and Optimization of Marula (Sclerocarya birrea) Seed Oil Methyl Ester Synthesis.

This study investigates Marula (Sclerocarya birrea) seed oil (SBSO) as a novel feedstock for biodiesel production through the transesterification process catalysed by heterogeneous bio-alkali derived from banana (Musa acuminata) peels. Response surface methodology (RSM) and artificial neural network (ANN) tools were used for the modelling and optimization of the process variables. The reaction process parameters considered were methanol/SBSO molar ratio, catalyst loading levels, reaction time and temperature. Central composite design (CCD) was espoused to generate 30 experimental conditions which were deployed in investigating the individual and synergetic effect of the process input variables on Sclerocarya birrea oil methyl ester (SBOME) yield. Appropriate statistical indices were adopted to investigate the predictive aptitude of the two models. Analysis shows that ANN model obtained for the transesterification process has a higher coefficient of determination (R2) of 0.9846 and lower absolute average deviation (AAD) of 0.07% compared to RSM model with R2 of 0.9482 and AAD of 0.12%. The process modelling outcome also confirmed ANN performance to be more precise than RSM. At methanol/SBSO ratio of 6:1, catalyst loading level of 2 wt%, process reaction time of 50 min and temperature of 55°C, the experimental maximum SBOME yield was observed to be 96.45 wt % following the ANN predicted yield of 96.45 wt % and RSM predicted yield of 96.65 wt % respectively. The analysed fuel properties of SBOME was found satisfactory within the biodiesel stipulated standard limit(s). The study establishes that SBSO is a good source for biodiesel production and its biodiesel methyl ester is a potential substitute for petroleum diesel and a bioenergy fuel.

YAPAY SİNİR AĞLARI VE DERİN ÖĞRENME KULLANARAK TÜRKİYE''DEKİ COVID-19 VAKALARI İÇİN TAHMİN MODELLERİ

Governments face a dilemma between public health and the economy while making strategic decisions on health during a pandemic outbreak. It is of great importance to forecast the number of cases in terms of strategic decisions to be taken by governments especially in outbreak periods and to manage the dilemma mentioned. One of the important issues today is the Covid-19 outbreak for almost all countries. Unfortunately, no effective vaccine or treatment has been found for Covid-19 yet. At the time of this study, however, it was reported that the total number of reported cases by the World Health Organization worldwide was more than thirteen million. Various quarantine measures have been necessary to deal with such a large epidemic. Quarantine measures taken by governments bring countries to face to face with the economic crisis. This creates economic uncertainties and puts governments under tremendous pressure to make accurate and least harmful strategic decisions. For these reasons, governments prefer to make strategic decisions for Covid-19 step by step observing the situation rather than making a sudden decision. If the number of pandemic cases could be predicted before a predetermined time, it would be used as an important guide for governments to manage public health and economic dilemma more accurately. Therefore, this study provides artificial neural network (ANN) and deep learning models (long-short term memory, LSTM networks) to forecast Covid-19 cases before 7-day. The proposed models were tested on real data for Turkey. The results showed that LSTM models performed better than ANN models in both cumulative cases and new cases on the training data set. Comparing the performance of the proposed models over the whole data set, it was observed that the ANN and LSTM algorithms gave competitive results. In addition, the cumulative case forecast performances of both ANN and LSTM models were observed to be better than the new case forecast.

Fan-out and fan-in properties of superconducting neuromorphic circuits

Neuromorphic computing has the potential to further the success of software-based artificial neural networks (ANNs) by designing hardware from a different perspective. Current research in neuromorphic hardware targets dramatic improvements to ANN performance by increasing energy efficiency and speed of operation and even seeks to extend the utility of ANNs by natively adding functionality such as spiking operation. One promising neuromorphic hardware platform is based on superconductive electronics, which has the potential to incorporate all of these advantages at the device level in addition to offering the potential of near lossless communications both within the neuromorphic circuits and between disparate superconductive chips. Here, we explore one of the fundamental brain-inspired architecture components, the fan-in and fan-out as realized in superconductive circuits based on Josephson junctions. From our calculations and WRSPICE simulations, we find that the fan-out should be limited only by junction count and circuit size limitations, and we demonstrate results in simulation at a level of 1-to-10 000, similar to that of the human brain. We find that fan-in has more limitations, but a fan-in level on the order of a few 100-to-1 should be achievable based on current technology. We discuss our findings and the critical parameters that set the limits on fan-in and fan-out in the context of superconductive neuromorphic circuits.

Neural Network Based Pavement Condition Assessment with Hyperspectral Images

Hyperspectral image processing techniques, with their ability to provide information about the chemical compositions of materials, have great potential for pavement condition assessment. This study introduces a novel age-based pavement assessment method, employing an integrated algorithm with artificial neural network (ANN) and spectral angle mapping (SAM) on hyperspectral images. In the proposed method, the resulting ANN prediction outputs are used to make a new prediction along with the results from SAM scores. Tests are performed on hyperspectral images that have 360 spectral bands between 400 and 900 nm, collected by a specifically designed vehicular system for proximal image acquisition. The acquired images have eight classes, including three different pavement classes (good (5-year), medium (10-year), and poor (25-year)), yellow dye, white dye, soil, paving stone, and shadow. Several experiments are performed to evaluate the robustness of the followed methodology with limited learning data that include 5, 10, 25, and 50 samples per class, selected randomly from our independent spectral database. For a fair comparison, the individual ANN, SAM, support vector machine (SVM), and stacked auto-encoders (SAE) algorithms are also evaluated. The classification performances of individual ANN and SAM are significantly increased with their joint use, demonstrating a 1.2% to 21% classification accuracy improvement in relation to the training sample size. The study proves that the proposed approach is quite robust in cases wherein few training data are available, while SAE and standard ANN algorithms are more successful in cases wherein more learning data are present.

Analysis of EMG Signals during Stance and Swing Phases for Controlling Magnetorheological Brake applications

Abstract The development of ankle foot orthoses (AFO) for lower limb rehabilitation have received significant attention over the past decades. Recently, passive AFO equipped with magnetorheological brake had been developed based on ankle angle and electromyography (EMG) signals. Nonetheless, the EMG signals were categorized in stance and swing phases through visual observation as the signals are stochastic. Therefore, this study aims to classify the pattern of EMG signals during stance and swing phases. Seven-time domains features will be extracted and fed into artificial neural network (ANN) as a classifier. Two different training algorithms of ANN namely Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) will be applied. As number of inputs will affect the classification performance of ANN, different number of input features will be employed. In this study, three participants were recruited and walk on the treadmills for 60 seconds by constant the speed. The ANN model was designed with 2, 10, 12, and 14 inputs features with LM and SCG training algorithms. Then, the ANN was trained ten times and the performances of each inputs features were measured using classification rate of training, testing, validation and overall. This study found that all the inputs with LM training algorithm gained more than 2% average classification rate than SCG training algorithm. On the other hand, classification accuracy of 10, 12 and 14 inputs were 5% higher than 2 inputs. It can be concluded that LM training algorithm of ANN was performed better than SCG algorithm with at least 10 inputs.

Modeling phosphorus adsorption onto polyaluminium chloride water treatment residuals

Abstract Beneficial reuse and appropriate disposal of water treatment residuals (WTRs) are of great concern for sustainable drinking water treatment. Using WTRs to remove phosphorus (P) is widely regarded as a feasible approach. However, the information is still limited on air-dried WTRs containing polyaluminium chloride (PAC) and anionic polyacrylamide (APAM) used to adsorb P. The objectives of this study were to construct artificial neural network (ANN) models for P adsorption onto WTRs from distilled de-ionized (DDI) water solution and stormwater, to investigate the performance of ANN in predicting phosphorous adsorption, and to model isotherm adsorption, kinetics, and thermodynamics by using the index of model performance. Batch experiments were performed with different WTRs dosage, pH, initial P concentration, temperature, and time. ANN models accurately predicted the P concentration at equilibrium. Non-linearized Langmuir model fitted the isotherm data best. Pseudo second-order kinetic model provided a better fit to experimental data. The adsorption process may be at least simultaneously controlled by surface adsorption and intraparticle diffusion. The P adsorption is a homogenous monolayer adsorption that is spontaneous, endothermic, and entropy production process. WTRs were found to be favorable and effective in removing P, but the P removals had significant differences in both solutions.

Performance of Artificial Neural Networks (ANN) At Transportation Planning Model

This study based on Artificial Neural Networks (ANN) in modeling trip production, which represent the first stage at transportation planning. The study analyzed the performance of ANN based on data that collected for central business district (CBD) of the city of Nasiriyah by calculating the all trips (y) that produced from this sector. As results, The ANN model gave a high performance comparative with Statistical Techniques in predicting trips. The coefficient of determination R2 for the number of all trips (y) is (0.948 ) and (0.871) for ANN and Statistical Techniques approaches respectively which is considered as an acceptable relationship obtained; also the ANN prediction model is more accurate for prediction process.

Artificial neural network classification of foliage targets from spectrograms of sequential echoes using a biomimetic audible sonar

Classifying foliage targets using echolocation is important for recognizing landmarks by bats using ultrasonic emissions and blind human echolocators (BEs) using palatal clicks. Previous attempts to classify foliage used ultrasonic frequencies and single sensor (monaural) detection. Motivated by the echolocation capabilities of BEs, a biomimetic sonar emitting audible clicks acquired 5600 binaural echoes from five sequential emissions that probed two foliage targets at aspect angles separated by 18°. Echo spectrograms formed feature vector inputs to artificial neural networks (ANNs) for classifying two targets, Ficus benjamina and Schefflera arboricola, with leaf areas that differ by a factor of four. Classification performances of ANNs without and with hidden layers were analyzed using tenfold cross-validation. Performance improved with input feature size, with binaural echo classification outperforming that using monaural echoes for the same number of emissions and for the same number of echoes. Linear classification accuracy was comparable to that using nonlinear classification with both achieving fewer than 1% errors with binaural spectrogram features from five sequential emissions. This result was better by a factor of 20 compared to previous classification of these targets using only the time envelopes of the same echoes.

Hybrid artificial neural network and cooperation search algorithm for nonlinear river flow time series forecasting in humid and semi-humid regions

Accurate river flow forecasting is of great importance for the scientific management of water resources system. With the advantages of easy implementation and high flexibility, artificial neural network (ANN) has been widely employed to address the complex hydrological forecasting problem. However, the conventional ANN method often suffers from some defects in practice, like slow convergence and local minimum. In order to enhance the ANN performance, this study proposes a hybrid river flow forecasting method by integrating the novel cooperation search algorithm (CSA) into the learning process of ANN. In other words, the computational parameters of the ANN network (like threshold and linking weights) are iteratively optimized by the CSA method in the feasible state space. The proposed method is applied to the river flow data collected from two real-world hydrological stations in China. Several Quantitative indexes are chosen to compare the performance of the developed models, while the comprehensive analysis between the simulated and observed flow data are conducted. The experimental results show that in different scenarios, the hybrid method based on ANN and CSA always outperforms the control models and yields superior forecasting results during both training and testing phases. In Three Gorges Project, the presented method makes 11.10% and 5.42% improvements in the Nash–Sutcliffe efficiency and Coefficient correlation values of the standard ANN method in the testing phase. Thus, this interesting finding shows that the performance of the artificial intelligence models in river flow time series forecasting can be effectively improved by metaheuristic algorithm with outstanding global search ability.

Prediction of Critical Flashover Voltage of High Voltage Insulators Leveraging Bootstrap Neural Network

Understanding the flashover performance of the outdoor high voltage insulator has been in the interest of many researchers recently. Various studies have been performed to investigate the critical flashover voltage of outdoor high voltage insulators analytically and in the laboratory. However, laboratory experiments are expensive and time-consuming. On the other hand, mathematical models are based on certain assumptions which compromise on the accuracy of results. This paper presents an intelligent system based on Artificial Neural Networks (ANN) to predict the critical flashover voltage of High-Temperature Vulcanized (HTV) silicone rubber in polluted and humid conditions. Various types of learning algorithms are used, such as Gradient Descent (GD), Levenberg-Marquardt (LM), Conjugate Gradient (CG), Quasi-Newton (QN), Resilient Backpropagation (RBP), and Bayesian Regularization Backpropagation (BRBP) to train the ANN. The number of neurons in the hidden layers along with the learning rate was varied to understand the effect of these parameters on the performance of ANN. The proposed ANN was trained using experimental data obtained from extensive experimentation in the laboratory under controlled environmental conditions. The proposed model demonstrates promising results and can be used to monitor outdoor high voltage insulators. It was observed from obtained results that changing of the number of neurons, learning rates, and learning algorithms of ANN significantly change the performance of the proposed algorithm.

Control of a Buck DC/DC Converter Using Approximate Dynamic Programming and Artificial Neural Networks

This paper proposes a novel artificial neural network (ANN) based control method for a dc/dc buck converter. The ANN is trained to implement optimal control based on approximate dynamic programming (ADP). Special characteristics of the proposed ANN control include: 1) The inputs to the ANN contain error signals and integrals of the error signals, enabling the ANN to have PI control ability; 2) The ANN receives voltage feedback signals from the dc/dc converter, making the combined system equivalent to a recurrent neural network; 3) The ANN is trained to minimize a cost function over a long time horizon, making the ANN have a stronger predictive control ability than a conventional predictive controller; 4) The ANN is trained offline, preventing the instability of the network caused by weight adjustments of an on-line training algorithm. The ANN performance is evaluated through simulation and hardware experiments and compared with conventional control methods, which shows that the ANN controller has a strong ability to track rapidly changing reference commands, maintain stable output voltage for a variable load, and manage maximum duty-ratio and current constraints properly.