Neuro-fuzzy Network Research Articles

Detection and classification of faults in a microgrid using wavelet neural network

It is necessary to detect the fault disturbances as quick as possible to improve performance of microgrid. Keeping an eye to the above issue this paper introduces a novel technique for the detection and classification of different faults in microgrid consisting of as Wind Turbine (WT), diesel generator, Solid Oxide Fuel Cell (SOFC) and micro-turbine. Wavelet transform (WT) and Wavelet Packet Transform (WPT) are used for detection and feature extraction to characterize the various faulted signals by using multi-resolution technique. Further, taking the input feature information of all fault disturbances, artificial neural network (ANN), neuro-fuzzy (NF) and Wavelet Neural Network (WNN) are implemented to accurately classify various faults. Two practically relevant 3-bus system and 14-bus microgrid system comprised with various types of distribution generations are considered for the protection analysis which is simulated using MATLAB/SIMULINK environment. Validation of the proposed technique has been done and compared with other two well proven and extensively used methods like ANN and NF under different operating scenarios.

EEG classification using recurrent adaptive neuro-fuzzy network based on time-series prediction

Brain–computer interface (BCI) is a system that provides a way for brain and computer to communicate with each other directly. Electroencephalogram (EEG) is an important process in a BCI that can be used to determine whether the subject is doing action and/or imagination. This paper presents a motor imagery (MI) classification for BCI systems using recurrent adaptive neuro-fuzzy interface system (ANFIS). The classification system is based on time-series prediction where features are exploited from the EEG signals recorded from subjects imagining of the right hand, left hand, tongue, and foot movement. The classification system contains some recurrent ANFISes. Each recurrent ANFIS is trained on MI signals of one class and specializes in recognizing the signals of the same class from the signals of other categories. Recurrent ANFISes are employed to predict one step ahead for the EEG time-series data, and then, the classification is performed by mean square error (MSE) of the predicted signals. This approach is carried out on twelve subjects MI signals of four classes in online mode. Average prediction MSE of 0.0302 and average classification accuracy of 85.52% are obtained as results.

A neuro-fuzzy Kohonen network for data stream possibilistic clustering and its online self-learning procedure

A task of Data Stream Fuzzy Clustering is considered when data is processed sequentially under a priori uncertainty conditions about both a number of clusters and a degree of clusters’ overlapping. A modified two-layer neuro-fuzzy Kohonen network is used for solving the possibilistic fuzzy clustering tasks. This system tunes centers’ coordinates and membership levels of every pattern to clusters during the self-learning procedure and automatically increases a number of neurons during data processing. A distinguishing feature of the proposed approach is its computational simplicity due to the fact that a recurrent modification of the possibilistic fuzzy clustering procedure is used for tuning the network’s parameters. The proposed neuro-fuzzy system is based on the concepts of evolving systems of Computational Intelligence, the recurrent optimization, the possibilistic fuzzy clustering, and Data Stream Mining.

Output power levelling for DFIG wind turbine system using intelligent pitch angle control

ABSTRACTBlade pitch angle control, as an indispensable part of wind turbine, plays a part in getting the desired power. In this regard, several pitch angle control methods have been proposed in order to limit aerodynamic power gained from the wind turbine system (WTS) in the high-wind-speed regions. In this paper, intelligent control methods are applied to control the blade pitch angle of doubly-fed induction generator (DFIG) WTS. Conventional fuzzy logic and neuro-fuzzy-particle swarm optimization controllers are used to get the appropriate wind power, where fuzzy inference system is based on fuzzy c-means clustering algorithm. It reduces the extra repetitive rules in fuzzy structure which in turn would reduce the complexity in neuro-fuzzy network with maximizing efficiently. In comparing the controllers at any given wind speed, adaptive neuro-fuzzy inference systems controller involving both mechanical power and rotor speed revealed better performance to maintain the aerodynamic power and rotor speed at the rated value. The effectiveness of the proposed method is verified by simulation results for a 9 MW DFIG WTS.

Music Emotions Recognition by Machine Learning With Cognitive Classification Methodologies

Music emotions recognition (MER) is a challenging field of studies addressed in multiple disciplines such as musicology, cognitive science, physiology, psychology, arts and affective computing. In this article, music emotions are classified into four types known as those of pleasing, angry, sad and relaxing. MER is formulated as a classification problem in cognitive computing where 548 dimensions of music features are extracted and modeled. A set of classifications and machine learning algorithms are explored and comparatively studied for MER, which includes Support Vector Machine (SVM), k-Nearest Neighbors (KNN), Neuro-Fuzzy Networks Classification (NFNC), Fuzzy KNN (FKNN), Bayes classifier and Linear Discriminant Analysis (LDA). Experimental results show that the SVM, FKNN and LDA algorithms are the most effective methodologies that obtain more than 80% accuracy for MER.

A Comparative Study of Various Hybrid Wavelet Feedforward Neural Network Models for Runoff Forecasting

Considering network topologies and structures of the artificial neural network (ANN) used in the field of hydrology, one can categorize them into two different generic types: feedforward and feedback (recurrent) networks. Different types of feedforward and recurrent ANNs are available, but multilayer perceptron type of feedforward ANN is most commonly used in hydrology for the development of wavelet coupled neural network (WNN) models. This study is conducted to compare performance of the various wavelet based feedforward artificial neural network (ANN) models. The feedforward ANN types used in the study include the multilayer perceptron neural network (MLPNN), generalized feedforward neural network (GFFNN), radial basis function neural network (RBFNN), modular neural network (MNN) and neuro-fuzzy neural network (NFNN) models. The rainfall-runoff data of four catchments located in different hydro-climatic regions of the world is used in the study. The discrete wavelet transformation (DWT) is used in the present study to decompose input rainfall data using db8 wavelet function. A total of 220 models are developed in this study to evaluate the performance of various feedforward neural network models. Performance of the developed WNN models is compared with their counterpart simple models developed without applying wavelet transformation (WT). The results of the study are further compared with - multiple linear regression (MLR) model which suggest that the WNN models outperformed their counterpart simple models. The hybrid wavelet models developed using MLPNN, the GFFNN and the MNN models performed best among the six selected data driven models explored in the study. Moreover, performance of the three best models is found to be similar and thus the hybrid wavelet GFFNN and the MNN models can be considered as an alternative to the most commonly used hybrid WNN models developed using MLPNN. The study further reveals that the wavelet coupled models outperformed their counterpart simple models only with the parsimonious input vector.

Predicting potential of controlled blasting-induced liquefaction using neural networks and neuro -fuzzy system

In recent years, controlled blasting has turned into an efficient method for evaluation of soil liquefaction in real scale and evaluation of ground improvement techniques. Predicting blast-induced soil liquefaction by using collected information can be an effective step in the[a1] study of blast-induced liquefaction. In this study, to estimate residual pore pressure ratio, first, multi- layer perceptron neural network is used in which error (RMS) for the network was calculated as 0.105. Next, neuro-fuzzy network, ANFIS was used for modeling. Different ANFIS models are created using Grid partitioning (GP), Subtractive Clustering (SCM), and Fuzzy C-means Clustering (FCM). Minimum error is obtained using by FCM at about 0.081. Finally, radial basis function (RBF) network is used. Error of this method was about 0.06. Accordingly, RBF network has better performance. Variables including fine-content, relative density, effective overburden pressure and SPT value are considered as input components and the Ru, residual[a2] pore pressure ratio was used as the only output component for designing prediction models. In the next stage the network output is compared with the results of a regression analysis. Finally, sensitivity analysis for RBF network is tested, its results reveal that and SPT are the most effective factors in determining Ru.

Fault detection and identification using combination of EKF and neuro-fuzzy network applied to a chemical process (CSTR)

In this paper, a new algorithm is proposed for fault detection and identification (FDI) in a class of nonlinear systems by combining the extended Kalman filter (EKF) and neuro-fuzzy networks (NFNs). There is an abundance of the literature on fault diagnosis ranging from model-based methods to data-driven approaches that have advantages and drawbacks. One may employ the advantages of different approaches to develop a high-efficient method for fault diagnosis. Initially, an EKF is designed to estimate the system output and to generate accurate residuals by a mathematical model of the process. Then, an NFN is designed for making decision using the mean value of the residuals. The network assigns a locally linear model to each faulty condition of the system. The validity of the models is determined based on the fuzzy rules. Combining the introduced EKF and the introduced NFN causes the proposed method to be independent of pre-designing a bank of observers in the model-based methods. Moreover, there is no need for extracting the features from the signals without any physical insight as well as computational complexity in the data-driven techniques. The effectiveness of the proposed FDI scheme is verified by applying it to a chemical plant as the case study, namely, continuous stirred tank reactor process. Simulation results show that the proposed methodology is very effective to detect and identify the faults of the system in different faulty modes.

Comparison of neuro-fuzzy network and response surface methodology pertaining to the viscosity of polymer solutions

This study has utilized the response surface methodology (RSM) and adaptive neuro-fuzzy inference system (ANFIS) approaches for the modeling of polymer solution viscosity. In the absence of reports in the previous study on applying these two approaches, the main objective of this study has been to compare the performance of these methods toward the viscosity modeling of a polymer solution. By utilizing RSM technique, the effects of three independent parameters including shear rate, polymer concentration, and sodium chloride concentration on viscosity of polymer solution were examined. The RSM results showed that all the parameters were not equally important in the polymer solution viscosity. Moreover, analysis of variance (ANOVA) was also carried out and indicated that there was no evidence of lack of fit in the RSM model. As a second approach for polymer solution viscosity modeling, ANFIS was utilized with two rules constructed based on the first-order Sugeno fuzzy approach and trained by back propagation neural networks algorithm. High coefficient of determination (R2) values ( >99%) showed that the prediction ability of both the ANFIS and RSM models was good enough for the response when the interpolation ability of the models was considered. In order to evaluate the extrapolation abilities of the two developed models, two data sets lying beyond the originally considered data were also taken into account. The results showed that their extrapolation predictive ability was poor. The reason could simply be the inherent behavior of the polymeric solution, i.e., the correlational structure seen in the sample used in the training step did not continue outside the sample space.

Nonlinear aeroacoustic characterization of Helmholtz resonators with a local-linear neuro-fuzzy network model

The nonlinear acoustic behavior of Helmholtz resonators is characterized by a data-based reduced-order model, which is obtained by a combination of high-resolution CFD simulation and system identification. It is shown that even in the nonlinear regime, a linear model is capable of describing the reflection behavior at a particular amplitude with quantitative accuracy. This observation motivates to choose a local-linear model structure for this study, which consists of a network of parallel linear submodels. A so-called fuzzy-neuron layer distributes the input signal over the linear submodels, depending on the root mean square of the particle velocity at the resonator surface. The resulting model structure is referred to as an local-linear neuro-fuzzy network. System identification techniques are used to estimate the free parameters of this model from training data. The training data are generated by CFD simulations of the resonator, with persistent acoustic excitation over a wide range of frequencies and sound pressure levels. The estimated nonlinear, reduced-order models show good agreement with CFD and experimental data over a wide range of amplitudes for several test cases.

Scour Depth Prediction around Bridge Piers Using Neuro-Fuzzy and Neural Network Approaches

Scour Depth Prediction around Bridge Piers Using Neuro-Fuzzy and Neural Network Approaches

Evaluation of the Performance of Telecommunication Systems by Approach of Hybrid Stochastic Automata Combined With Neuro-Fuzzy Networks

This paper presents a functional and dysfunctional behavioral study of a telecommunication system, with the aim to evaluate the performance of its constituent units. It is question of taking advantage offered by artificial intelligence in order to evaluate by modeling and simulation in system reliability. The methodological approach consists in combining ANFIS neuro-fuzzy networks with hybrid stochastic automata. The Neuro-Fuzzy ANFIS networks provide a prediction for the passage from nominal mode to degraded mode, by controlling the occurrence of malfunctions at transient levels. This allows to anticipate the occurrence of events degrading system performance, such as failures and disturbances. The objective is to maintain the system in nominal operating mode and prevent its tipping in degraded mode. The results are implanted around a demonstrator based on Scilab, and implemented on Matlab / Simulink.

Comparison between fuzzy, neuro-fuzzy and neural network models to estimate the expansion of coffee berry borer in colombian coffee crops

In this article, the comparison among different techniques applied to simulation and representation of the expansion of the coffee berry borer in the Colombian coffee crops is carried out. First, a brief description of supervised techniques (models focused on data adjustment): Neural networks, neuro-fuzzy and fuzzy logic. Then, implementing these models is described. Finally, the comparison among these models is shown, highlighting performance of neuronal networks.

THE NEURO-FUZZY DIAGNOSTIC MODEL SYNTHESIS WITH HASHED TRANSFORMATION IN THE SEQUENCE AND PARALLEL MODE

Context. The urgent task of improving the speed of neuro-fuzzy model construction by the precedents has been solved. Objective is a creation of a neuro-fuzzy network synthesis method with high speed of computations and allowing to realize the synthesis of neuro-fuzzy networks in parallel mode. Method. The method of neuro-fuzzy model constructing by precedents, which reduces the dimension of the input data by hashing transformation to the one-dimensional axis saving local cluster topology in a feature space, estimates the significance of the features and instances on the basis of selected clusters, and also forms a partition of the original feature space in an automatic mode, synthesizes structure and adjusts parameters of the neuro-fuzzy model automatically, excluding from the training process of the neuro-fuzzy model the uninformative data, thus simplifying the structure of the obtained model, allows to perform most computationally costly operations in parallel mode, that allows to automate the process of neuro-fuzzy model synthesis by precedents, as well as to increase the speed of neuro-fuzzy model construction both in sequential and in parallel implementation of computations. Results. The software implementing proposed method have been developed and used in computational experiments investigating the properties of the method. The experiments confirmed the efficiency of the proposed method and software. Conclusions. The experiments also allow to recommend them for use in practice to solve the problems of diagnosis and automatic classification by the features.

A new method based on adaptive neuro-fuzzy inference system for determination of acid molarity using Compton scattered photons

The simplicity and accuracy in determining the exact concentration of a particular substance in a solution, is one of the major issues in industrial chemistry. In this paper, we introduce a new technique based on Compton scattering of gamma photons and artificial intelligence to determine the concentration of a solute in a solution. We used a 137Cs gamma ray source with few millicurie activity and a NaI(Tl) scintillator detector to determine the acid sulfuric molarity. We also applied Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) as artificial intelligence techniques to estimate the molarity of the samples. Monte Carlo simulations also support the present experimental results. The results of the proposed methods are in excellent agreement with the measured molarities.

Fukunaga-Koontz feature transformation for statistical structural damage detection and hierarchical neuro-fuzzy damage localisation

Considering jointly damage sensitive features (DSFs) of signals recorded by multiple sensors, applying advanced transformations to these DSFs and assessing systematically their contribution to damage detectability and localisation can significantly enhance the performance of structural health monitoring systems. This philosophy is explored here for partial autocorrelation coefficients (PACCs) of acceleration responses. They are interrogated with the help of the linear discriminant analysis based on the Fukunaga-Koontz transformation using datasets of the healthy and selected reference damage states. Then, a simple but efficient fast forward selection procedure is applied to rank the DSF components with respect to statistical distance measures specialised for either damage detection or localisation. For the damage detection task, the optimal feature subsets are identified based on the statistical hypothesis testing. For damage localisation, a hierarchical neuro-fuzzy tool is developed that uses the DSF ranking to establish its own optimal architecture. The proposed approaches are evaluated experimentally on data from non-destructively simulated damage in a laboratory scale wind turbine blade. The results support our claim of being able to enhance damage detectability and localisation performance by transforming and optimally selecting DSFs. It is demonstrated that the optimally selected PACCs from multiple sensors or their Fukunaga-Koontz transformed versions can not only improve the detectability of damage via statistical hypothesis testing but also increase the accuracy of damage localisation when used as inputs into a hierarchical neuro-fuzzy network. Furthermore, the computational effort of employing these advanced soft computing models for damage localisation can be significantly reduced by using transformed DSFs.

Robust adaptive self-organizing neuro-fuzzy tracking control of UUV with system uncertainties and unknown dead-zone nonlinearity

In this paper, a robust adaptive self-organizing neuro-fuzzy control (RASNFC) scheme for tracking of unmanned underwater vehicle with uncertainties and the unknown dead-zone nonlinearity is proposed. The proposed RASNFC scheme comprises an estimation-based adaptive controller (EBAC) using a self-organizing neuro-fuzzy network (SNFN) and a robust controller. The EBAC controller is constructed with a novel sliding mode reaching law control framework, and the unknown dynamic function is identified by the SNFN approximator which is able to online self-construct a neuro-fuzzy network with dynamic structure by generating and pruning fuzzy rule. The robust controller is employed to provide the finite $$L_{2}$$ -gain property to cope with reconstruction errors such that the robustness of the entire closed-loop control system is enhanced. Theoretical analysis shows that tracking errors and their derivatives are asymptotically stable and all signals in the closed-loop system are bounded. Comparative simulation results demonstrate the effectiveness and superiority of the proposed RASNFC scheme.

Fuzzy Group-Based Intersection Control via Vehicular Networks for Smart Transportations

Vehicular network has been recently used to achieve high efficient and flexible traffic scheduling at intersection roads for smart transportation systems. Different from existing works, where traffic signal is used to schedule waiting vehicles at each lane, we propose to divide vehicles in the same lane into small groups and schedule vehicle groups via wireless communication rather than traffic lights. Such direct scheduling of vehicles can reduce waiting time and improve fairness, especially when the traffic volume in different lanes is imbalanced. The key challenge in such a design lies in determining appropriate size of groups with respect to real-time traffic conditions. To cope with this issue, we propose a neuro-fuzzy network-based grouping mechanism, where the network is trained using reinforcement learning technique. Also, vehicle groups are scheduled via a neuro-fuzzy network. Simulations using ns3 are conducted to evaluate the performance of our algorithm and compare it with similar works. The results show that our algorithm can reduce waiting time and at the same time improve fairness in various cases, and the advantage against traffic light algorithms can be up to 40%.

Control Dynamics and Simulation of Inclined Cart and Pendulum System

This paper applies various soft-computing control strategies for offline mode control of highly non-linear cart and pendulum system moving on an inclined surface. The surface is considered at inclination of 12° from horizontal. The study compares performance of four different control techniques namely Proportional-integral-derivative (PID), Fuzzy logic, Adaptive neuro fuzzy inference system (ANFIS) and Neural networks for control of proposed system. A Matlab-Simulink model of system has been developed from mathematical equations derived using Newton's second law. The cart and pendulum system has been initially controlled using PID controllers and results were further used to train ANFIS and neural controllers. The ANFIS and fuzzy controllers were designed using three and nine gbell shape membership functions (MFs) respectively. The controllers were further compared in terms of settling time, overshoot and undershoot.

An adaptive neuro-fuzzy inference system for assessment of risks to an organization’s information security

Sergey A. Glushenko - Senior Lecturer, Department of Information Systems and Applied Computer Science, Rostov State University of EconomicsAddress: 69, Bolshaya Sadovaya Street, Rostov-on-Don, 344002, Russian FederationE-mail: gs-gears@yandex.ru This article explains the importance of applying risk assessment in the implementation of information security systems. It is considered the most common risk assessment procedure and entails application of fuzzy logic theory for this purpose. The paper describes the proposed fuzzy production model (FPM), which defines seven input linguistic variables describing risk factors, four output linguistic variables that characterize different areas of information security risks, as well as four base rules. It is noted that the FPM is the first approach to the subject area and requires optimization to minimize the model’s output errors. The most common methods of optimization of fuzzy models parameters are examined, and the advantages of applying methods based on neuro-fuzzy networks (NFN) are justified. The article describes the process of converting fuzzy model elements, such as unit fuzzification, rule base unit and unit defuzzification, into fragments of the neural network. The result of this process is a neuro-fuzzy network corresponding to the fuzzy model. Formation of the developed NFN is based on an adaptive neuro-fuzzy inference system (ANFIS), using the specialized Neuro-Fuzzy Designer package of MATLAB software. The model was trained by a hybrid method which represents a combination of the methods of least squares and backpropagation. The result of this process is optimization (setting) the parameters of membership functions of input linguistic variables. Application of neuro-fuzzy modeling made it possible to obtain a more appropriate fuzzy production model which is able to conduct linguistic analysis of the risks of an organization’s information security. The information obtained with its help allows IT managers to determine risk priorities and to develop effective action plans to reduce the impact of the most dangerous threats.[1] This research has been carried out with financial support of RFBR within the framework of scientific project No. 16-31-00285 “Fuzzy logic methods and models in risk management decision support systems”