Adaptive Neural Network Algorithm Research Articles

Solving polynomial systems using a fast adaptive back propagation-type neural network algorithm

This paper proposes a neural network architecture for solving systems of non-linear equations. A back propagation algorithm is applied to solve the problem, using an adaptive learning rate procedure, based on the minimization of the mean squared error function defined by the system, as well as the network activation function, which can be linear or non-linear. The results obtained are compared with some of the standard global optimization techniques that are used for solving non-linear equations systems. The method was tested with some well-known and difficult applications (such as Gauss–Legendre 2-point formula for numerical integration, chemical equilibrium application, kinematic application, neuropsychology application, combustion application and interval arithmetic benchmark) in order to evaluate the performance of the new approach. Empirical results reveal that the proposed method is characterized by fast convergence and is able to deal with high-dimensional equations systems.

Improving Stability and Convergence for Adaptive Radial Basis Function Neural Networks Algorithm. (On-Line Harmonics Estimation Application)

In this paper, an adaptive Radial Basis Function Neural Networks (RBFNN) algorithm is used to estimate the fundamental and harmonic components of nonlinear load current. The performance of the adaptive RBFNN is evaluated based on the difference between the original signal and the constructed signal (the summation between fundamental and harmonic components). Also, an extensive investigation is carried out to propose a systematic and optimal selection of the Adaptive RBFNN parameters. These parameters will ensure fast and stable convergence and minimum estimation error. The results show an improving for fundamental and harmonics estimation comparing to the conventional RBFNN. Also, the results show how to control the computational steps and how they are related to the estimation error. The methodology used in this paper facilitates the development and design of signal processing and control systems.Article History: Received Dec 15, 2016; Received in revised form Feb 2nd 2017; Accepted 13rd 2017; Available onlineHow to Cite This Article: Almaita, E.K and Shawawreh J.Al (2017) Improving Stability and Convergence for Adaptive Radial Basis Function Neural Networks Algorithm (On-Line Harmonics Estimation Application). International Journal of Renewable Energy Develeopment, 6(1), 9-17.http://dx.doi.org/10.14710/ijred.6.1.9-17

An adaptive linear neural network with least mean M-estimate weight updating rule employed for harmonics identification and power quality monitoring

This paper describes a combined adaptive linear neural network and least mean M-estimate (ADALINE-LMM) algorithm for estimating the amplitude and phase of the individual harmonic contained in a distorted power system current signal. The weight vector of the ADALINE is updated iteratively by LMM algorithm. A Hampel’s three parts redescending M-estimator function is incorporated in the instantaneous cost function to provide thresholds for identifying and eliminating the effect of temporary fluctuation owing to the presence of impulsive noise. This type of combined approach shows more accurate and faster tracking capability than the combined ADALINE and variable step size least mean square (ADALINE-VSLMS) algorithm. In addition to this, the proposed algorithm is suggested in shunt hybrid active power filter (SHAPF) for extracting the harmonics and reactive power components from the distorted load currents. Extensive time domain simulation is carried out to evaluate the performance of the SHAPF for maintaining the power quality of a system under various demanding situations. Moreover, an experimental setup is developed in the laboratory for verification of the proposed control technique in a real-time application using a Spartan 3A DSP processor.

Adaptive Neural Network Algorithm for Power Control in Nuclear Power Plants

The aim of this paper is to design, test and evaluate a prototype of an adaptive neural network algorithm for the power controlling system of a nuclear power plant. The task of power control in nuclear reactors is one of the fundamental tasks in this field. Therefore, researches are constantly conducted to ameliorate the power reactor control process. Currently, in the Department of Automation in the National Research Nuclear University (NRNU) MEPhI, numerous studies are utilizing various methodologies of artificial intelligence (expert systems, neural networks, fuzzy systems and genetic algorithms) to enhance the performance, safety, efficiency and reliability of nuclear power plants. In particular, a study of an adaptive artificial intelligent power regulator in the control systems of nuclear power reactors is being undertaken to enhance performance and to minimize the output error of the Automatic Power Controller (APC) on the grounds of a multifunctional computer analyzer (simulator) of the Water-Water Energetic Reactor known as Vodo-Vodyanoi Energetichesky Reaktor (VVER) in Russian. In this paper, a block diagram of an adaptive reactor power controller was built on the basis of an intelligent control algorithm. When implementing intelligent neural network principles, it is possible to improve the quality and dynamic of any control system in accordance with the principles of adaptive control. It is common knowledge that an adaptive control system permits adjusting the controller’s parameters according to the transitions in the characteristics of the control object or external disturbances. In this project, it is demonstrated that the propitious options for an automatic power controller in nuclear power plants is a control system constructed on intelligent neural network algorithms.

English

English

Characteristics of atmospheric single particles during haze periods in a typical urban area of Beijing: A case study in October, 2014

To investigate the composition and possible sources of particles, especially during heavy haze pollution, a single particle aerosol mass spectrometer (SPAMS) was deployed to measure the changes of single particle species and sizes during October of 2014, in Beijing. A total of 2,871,431 particles with both positive and negative spectra were collected and characterized in combination with the adaptive resonance theory neural network algorithm (ART-2a). Eight types of particles were classified: dust particles (dust, 8.1%), elemental carbon (EC, 29.0%), organic carbon (OC, 18.0%), EC and OC combined particles (ECOC, 9.5%), Na-K containing particles (NaK, 7.9%), K-containing particles (K, 21.8%), organic nitrogen and potassium containing particles (KCN, 2.3%), and metal-containing particles (metal, 3.6%). Three haze pollution events (P1, P2, P3) and one clean period (clean) were analyzed, based on the mass and number concentration of PM2.5 and the back trajectory results from the hybrid single particle Lagrangian integrated trajectory model (Hysplit-4 model). Results showed that EC, OC and K were the major components of single particles during the three haze pollution periods, which showed clearly increased ratios compared with those in the clean period. Results from the mixing state of secondary species of different types of particles showed that sulfate and nitrate were more readily mixed with carbon-containing particles during haze pollution episodes than in clean periods.

Application of Adaptive Neural Network Algorithms for Determination of Salt Composition of Waters Using Laser Spectroscopy

Application of Adaptive Neural Network Algorithms for Determination of Salt Composition of Waters Using Laser Spectroscopy

A large ratio data compression method for pipeline inspection based on EMAT-generated guided wave

In order to meet the engineering requirement of gas pipeline inline inspection for cracking defects, a new ultrasonic guided wave (GW) method based on electromagnetic acoustic transducer (EMAT) has emerged. In consideration of the large amount of original inspection data and the limited capacity of the storing device, the research on rapid and large ratio data compression is needed. This paper proposes a new large ratio data compression method based on subsection adaptive method (SAM) and wavelet neural network (WNN) algorithm. The influences of parameter selection on the compression quality, speed and ratio and the method of enhancing the compression speed of this method is studied. The experiment results indicate that compared with the traditional WWN compression algorithm, the proposed method has higher compression quality and speed, and a large compression ratio of 110:1 is achieved by the proposed method. It is thus suitable for gas pipeline crack inline inspection based on EMAT-generated GW.

Hysteresis Analysis of High Precision PZT Actuator in Dual-Stage Hard Disk Drive

In Hard disk drive (HDD) industry, the demand for faster HDD with higher storage capacities had led to the use of a dual-stage servo system with a high precision PZT actuator. However, the PZT actuator is made of piezoelectric material which exhibits hysteresis. Hysteresis causes inaccuracies and oscillations in the system responses. In this paper, the hysteresis behavior is examined for a new high resolution (a few nanometers) PZT actuator fabricated for dual stage actuator. The new PZT actuator has the first resonance at a frequency twenty times higher than the first resonant mode of PZT actuator used at present in HDD. A generalized Prandtl-Ishlinskii (GPI) model is obtained by using curve-fitting with the examined hysteresis behavior. Then the inverse of the GPI model is extracted which can be used as the hysteresis compensator. Finally a model-free hysteresis compensator using adaptive Neural Network algorithm is being proposed for future research.

Performance enhancement of respiratory tumor motion prediction using adaptive support vector regression: Comparison with adaptive neural network method

ABSTRACTThe breathing motion moves internal organs and targeted regions determined by radiation therapy planning. For the radiation therapy, accurate prediction for breathing motion is of great interest as the outer targeted region treatment could endanger sensitive tissue. In this study, the use of a prediction algorithm with adaptive support vector regression (aSVR) was proposed and compared with the adaptive neural network (ANN) algorithm considering the prediction accuracy and training and predicting time. Respiration data from 87 patients treated by radiation therapy, were acquired with an optical marker at 30 Hz. Five types of prediction filters with the ANN or aSVR filters, were implemented and their performance was compared according to the size of the sliding window (2.5 and 5.0 sec), and the prediction latencies (100, 200, 300, 400, and 500 msec). Training and testing of the prediction algorithms with aSVR and ANN were performed. The root mean square error (RMSE) was used as the accuracy metric. The aSVR with an RBF kernel outperformed other prediction filters, including not only various types of ANN filters but also the aSVR with a linear kernel. A sliding window of 2.5 sec significantly and independently enhanced the overall accuracy. Otherwise, the training and prediction testing times were significantly prolonged in case of aSVR with an RBF kernel. The aSVR filter with the RBF kernel is in all cases superior to other filters regarding its accuracy; it also shows clinically applicable results from the viewpoint of training and predicting time, which may be effective for predicting patient breathing motion and thus enhancing the efficacy of radiation therapy.

An adaptive PID neural network for complex nonlinear system control

Usually it is difficult to solve the control problem of a complex nonlinear system. In this paper, we present an effective control method based on adaptive PID neural network and particle swarm optimization (PSO) algorithm. PSO algorithm is introduced to initialize the neural network for improving the convergent speed and preventing weights trapping into local optima. To adapt the initially uncertain and varying parameters in the control system, we introduce an improved gradient descent method to adjust the network parameters. The stability of our controller is analyzed according to the Lyapunov method. The simulation of complex nonlinear multiple-input and multiple-output (MIMO) system is presented with strong coupling. Empirical results illustrate that the proposed controller can obtain good precision with shorter time compared with the other considered methods.

Adaptive Dynamic Clone Selection Neural Network Algorithm for Motor Fault Diagnosis

Abstract A fault diagnosis method based on adaptive dynamic clone selection neural network (ADCSNN) is proposed in this paper. In this method the weights of neural network is encoded as the antibody, and the network error is considered as the antigen. The algorithm is then applied to fault detection of motor equipment. The experiments results show that the fault diagnosis method based on ADCS neural network has the capability in escaping local minimum and improving the algorithm speed, this gives better performance.

The mahalanobis taguchi system—adaptive resonance theory neural network algorithm for dynamic product designs

This work presents a novel algorithm, which combines the Mahalanobis Taguchi system (MTS) with the adaptive resonance theory neural network (ARTN) method for parameter selections in a dynamic product design system (DPDS). The utility of the algorithm is assessed in two dimensions: the MTS shows how individual product parameter dimensions are selected, and the ARTN links parameter selection decisions across two different timelines and can be used to focus on DPDS and to identify product architecture dimensions that are critical for a DPDS. The MTS can easily solve product parameter selection problems and has proved to be computationally efficient in previous studies. Additionally, the ARTN algorithm provides a simple and efficient means of constructing a DPDS, which is verified by the ARTN algorithm and is presented in this study. From the results of this study, we conclude that the MTS-ARTN algorithm can be applied successfully in a DPDS environment.

Integration of ANFIS, NN and GA to determine core porosity and permeability from conventional well log data

Routine core analysis provides useful information for petrophysical study of the hydrocarbon reservoirs. Effective porosity and fluid conductivity (permeability) could be obtained from core analysis in laboratory. Coring hydrocarbon bearing intervals and analysis of obtained cores in laboratory is expensive and time consuming. In this study an improved method to make a quantitative correlation between porosity and permeability obtained from core and conventional well log data by integration of different artificial intelligent systems is proposed. The proposed method combines the results of adaptive neuro-fuzzy inference system (ANFIS) and neural network (NN) algorithms for overall estimation of core data from conventional well log data. These methods multiply the output of each algorithm with a weight factor. Simple averaging and weighted averaging were used for determining the weight factors. In the weighted averaging method the genetic algorithm (GA) is used to determine the weight factors. The overall algorithm was applied in one of SW Iran?s oil fields with two cored wells. One-third of all data were used as the test dataset and the rest of them were used for training the networks. Results show that the output of the GA averaging method provided the best mean square error and also the best correlation coefficient with real core data.

Adaptive Neural Network for Image Edge Detection

This paper presents a neural network adaptive image edge detection method, and from neural network theory, this paper gives the formula of adaptive neural network algorithm; quantitative given the momentum factor and error, momentum factor and error on the weight vector of norm of the gradient of the quantitative relationship; and gives the algorithm flow diagram. Through experiment we get the conclusion: by using this adaptive neural network for image edge detection is feasible, and it has good generalization ability.

Implementation of hybrid ANN–PSO algorithm on FPGA for harmonic estimation

Harmonic estimation is the main process in active filters for harmonic reduction. A hybrid Adaptive Neural Network–Particle Swarm Optimization (ANN–PSO) algorithm is being proposed for harmonic isolation. Originally Fourier Transformation is used to analyze a distorted wave. In order to improve the convergence rate and processing speed an Adaptive Neural Network Algorithm called Adaline has then been used. A further improvement has been provided to reduce the error and increase the fineness of harmonic isolation by combining PSO algorithm with Adaline algorithm. The inertia weight factor of PSO is combined along with the weight factor of Adaline and trained in Neural Network environment for better results. ANN–PSO provides uniform convergence with the convergence rate comparable that of Adaline algorithm. The proposed ANN–PSO algorithm is implemented on an FPGA. To validate the performance of ANN–PSO; results are compared with Adaline algorithm and presented herein.

Fast and Robust Method for Dynamic Gesture Recognition Using Hermite Neural Network

Due to its shortcomings such as slow convergence rate and low recognition accuracy, the traditional BP neural networks perform poorly in dynamic gesture recognition, especially for online gesture training. In this paper, a novel adaptive Hermite neural networks algorithm for dynamic gesture recognition was proposed. At first, a three-layer feed-forward neural network, of which hidden layer neurons are activated by a group of Hermite orthogonal polynomial functions was constructed. Based on its special structure, a method to determine the network weights directly was introduced, and a novel algorithm to determine the optimal number of hidden nodes adaptively was proposed. Then a rapid method of fingertips tracking was put forward to get the trajectory of dynamic gesture. At last, dynamic gesture was recognized through the trained Hermite neural network. Experiment results show that Hermite neural network can enhance the speed and precision of network training, improve the learning speed and recognition accuracy, and has good robustness and generalization ability.

Improved Neural Network Algorithms with Time-Varying Widrow-Hoff Learning Rule for Harmonic Estimation

Estimation of harmonics is the main process in harmonic reduction. Algorithms are very well developed for estimation of harmonic components. In power electronic applications, objectives like fast response of the system are of primary importance. Effective methods for estimation of instantaneous harmonic components are presented in this paper. Originally, Fourier transformation is used to analyse a distorted waveform. In order to improve the convergence rate and processing speed an adaptive neural network algorithm, called ADALINE, has been used. Further improvements in the existing ADALINE algorithm have been made and two new algorithms, namely Modified ADALINE and Constricted ADALINE, are proposed by incorporating time-varying learning factor into the weight updating rule of ADALINE. The proposed methods stay effective as they converge to a minimum error and bring out finer estimation. The proposed methods are validated through harmonic estimation of PC load current. Then the results are compared with the existing ADALINE algorithm to illustrate their effectiveness.

An adaptive neural network algorithm for assessment and improvement of job satisfaction with respect to HSE and ergonomics program: The case of a gas refinery

Researchers have been continuously trying to improve human performance with respect to Health, Safety and Environment (HSE) and ergonomics (hence HSEE). This study proposes an adaptive neural network (ANN) algorithm for measuring and improving job satisfaction among operators with respect to HSEE in a gas refinery. To achieve the objectives of this study, standard questionnaires with respect to HSEE are completed by operators. The average results for each category of HSEE are used as inputs and job satisfaction is used as output for the ANN algorithm. Moreover, ANN is used to rank operators performance with respect to HSEE and job satisfaction. Finally, Normal probability technique is used to identify outlier operators. Moreover, operators with inadequate job satisfaction with respect to HSEE are identified. This would help managers to see if operators are satisfied with their jobs in the context of HSEE. This is the first study that introduces an integrated ANN algorithm for assessment and improvement of human job satisfaction with respect to HSEE program in complex systems.

Application of Fuzzy Neural Network in Gas-Solid Reaction Kinetics in Porous Media

Based on the gas-solid reactions in porous media model kinetic experiments, using the fuzzy neural network and adaptive fuzzy neural network algorithm, establishs the model of fuzzy neural network in gas-solid reactions in porous media. through software algorithms, and some experimental data input system for data fitting to prove the validity and correctness of the system.