Neuro-fuzzy Network Research Articles

THE USE OF NEURO-FUZZY MODELS IN EXPERT SUPPORT SYSTEMS FOR FORENSIC BUILDING-TECHNICAL EXPERTISE

The paper is focused on solving the problem of assessing the impact of repair-building works on the technical condition of objects near which these works were or are being carried out. Particular attention is paid to the analysis of the problems that accompany the creation of expert systems for supporting forensic building-technical expertise. The main aim of the work: conceptual modeling of an expert system for supporting forensic building-technical expertise. Object of research: the process of execution of forensic building-technical expertise and expert research. Solved problem: automation of a system capable of functioning in conditions of fuzzy uncertainty caused by the non-uniformity of the logic of the process of performing forensic building-technical expertise and the ambiguity and inconsistency of the information provided for research. Main scientific results: a model of a knowledge-based system is proposed and the use of neuro-fuzzy networks is justified to solve the problem of supporting the decision to assess the impact of repair-building works on the technical condition of the object, which has become the subject of expertise. Field of practical use of research results: forensic activities in the framework of building-technical expertise to determine the possible causes of deterioration in the technical condition of structural elements of buildings and their individual premises. Innovative technological product: a support system for forensic building-technical expertise based on knowledge and neuro-fuzzy models. Scope of application of an innovative technological product: forensic and investigative practice in resolving issues requiring the use of special knowledge in assessing the impact of repair-building works on the technical condition of nearby facilities.

Fuzzy neural networks and neuro-fuzzy networks: A review the main techniques and applications used in the literature

This paper presents a review of the central theories involved in hybrid models based on fuzzy systems and artificial neural networks, mainly focused on supervised methods for training hybrid models. The basic concepts regarding the history of hybrid models, from the first proposed model to the current advances, the composition and the functionalities in their architecture, the data treatment and the training methods of these intelligent models are presented to the reader so that the evolution of this category of intelligent systems can be evidenced. Finally, the features of the leading models and their applications are presented to the reader. We conclude that the fuzzy neural network models and their derivations are efficient in constructing a system with a high degree of accuracy and an appropriate level of interpretability working in a wide range of areas of economics and science.

Adaptive Backstepping Hybrid Fuzzy Sliding Mode Control for Uncertain Fractional-Order Nonlinear Systems Based on Finite-Time Scheme

A fractional-order integral fuzzy sliding mode control scheme is proposed for a class of uncertain fractional-order nonlinear systems subject to uncertainties and external disturbances. First, in each step, a neuro-fuzzy network system is developed to approximate the uncertain nonlinear function existing in fractional-subsystem and a fractional sliding mode surface is presented. Second, based on the fractional Lyapunov stability theory and the finite-time stability theory, a fractional adaptive backstepping neuro-fuzzy sliding mode controller is designed to drive the state trajectories of fractional-order systems to the prescribed sliding mode surface. Meanwhile, the finite-time stability of the fractional-order closed-loop system is proved. At last, three numerical examples are given to illustrate the effectiveness of the proposed control method.

An integrated approach for the identification of lithofacies and clay mineralogy through Neuro-Fuzzy, cross plot, and statistical analyses, from well log data

Today, researchers face multiple challenges identifying clay mineral types and lithofacies from well log data. This research paper hopes to offer new insight into this particular challenge. Formation evaluation characteristics play a significant role in the exploration and production of future and current oil and gas fields. The proposed methodology in this study uses an integrated approach that includes: (1) numerical equations, (2) Neuro-Fuzzy neural networks, (3) cross plots, and (4) statistical analyses. This proposed integrated approach is capable of dramatically improving the accuracy of the results. Well logging data provide valuable information for identifying lithofacies, clay mineralogy types, as well as other important hydrocarbon reservoir characteristics. Talhar Shale in the Southern Lower Indus Basin, Pakistan, is composed of interbedded shale, sand, and shaly-sand, intervals that have been identified via the lithological interpretation process of well logs. Talhar Shale contains montmorillonite type clay with minor amounts of illite, glauconite, and various micas that can be easily identified by natural gamma ray absorption profiles, as well as through ratio logs, bulk density log, and photoelectric absorption index log. These interpretations can be further confirmed via cross plots and other statistical analyses. This approach consists of a comprehensive study of well logging data and thus can lend itself to be a helpful component in characterizing the hydrocarbon structures of the Talhar Shale.

Comparison between fuzzy, neuro-fuzzy and neural network models to estimate the expansion of coffee berry borer in colombian coffee crops PDF XML Nychol Bazurto, Carlos Martinez, Helbert Eduardo Espitia 19-24 Comparative Analysis between a Discrete Spiral Chamber and a Continuous Spiral Chamber via ANSYS

Hydraulic turbines have been elements of utmost importance to meet the energy needs of communities worldwide. These are used with elements that improve the flow behavior of water to the turbine so that its energy benefits are improved. Spiral chambers are circular elements where their cross-section is constantly changing and are used in reaction turbines, such as Francis turbines. Their main function is to distribute evenly the fluid into the impeller inlet. This paper sought to use engineering simulation tools (ANSYS - CFX) to compare behavior and performance from a discreet continuous spiral chamber in a virtual environment under the same physical operating conditions, providing a comparative view of the performance present between the two manufacturing methods of spiral chambers

Development of Neurofuzzy Architectures for Electricity Price Forecasting

In 20th century, many countries have liberalized their electricity market. This power markets liberalization has directed generation companies as well as wholesale buyers to undertake a greater intense risk exposure compared to the old centralized framework. In this framework, electricity price prediction has become crucial for any market player in their decision-making process as well as strategic planning. In this study, a prototype asymmetric-based neuro-fuzzy network (AGFINN) architecture has been implemented for short-term electricity prices forecasting for ISO New England market. AGFINN framework has been designed through two different defuzzification schemes. Fuzzy clustering has been explored as an initial step for defining the fuzzy rules while an asymmetric Gaussian membership function has been utilized in the fuzzification part of the model. Results related to the minimum and maximum electricity prices for ISO New England, emphasize the superiority of the proposed model over well-established learning-based models.

Regression-Based Neuro-Fuzzy Network Trained by ABC Algorithm for High-Density Impulse Noise Elimination

Salt and pepper (SAP) noise elimination is a crucial step for further image processing and pattern recognition applications. The main aim of this article is to propose a novel SAP noise elimination method which employs a regression-based neuro-fuzzy network for highly corrupted gray scale and color images. In the proposed method, multiple neuro-fuzzy filters trained with artificial bee colony algorithm is combined with a decision tree algorithm. The performance of the proposed filter is compared with a number of well known methods with respect to popular metrics including, structural similarity index, peak signal-to-noise ratio, and correlation on well known test images. The results reveal that the proposed filter has superior performance in terms of all comparison metrics.

A Chebyshev polynomial feedforward neural network trained by differential evolution and its application in environmental case studies

This paper introduces a polynomial feedforward neural network based on Chebyshev polynomials able to effectively model non-linear and highly complex environmental data. The data sets were cautiously selected from the fields of biology, ecology, climate, and environmental management, and economics as to represent a scientifically meaningful and consistent corpus of disparate domains of intensive focus and interest in current ecological/environmental research, covering issues related to body growth/age, biomass production, energy efficiency/consumption, and ecology/geographic extension. The proposed network uses a number of layers to estimate the output in terms of a weighted sum of truncated Chebyshev series expansions applied to linear combinations of the input variables, and it is trained by the differential evolution algorithm. Its performance was compared to three neural networks. First, a polynomial feedforward network that uses Hermite polynomials as activation function in the hidden nodes; second, a radial basis function neural network; third, a Takagi-Sugeno-Kang neuro-fuzzy network. All the above networks were trained by evolutionary optimization algorithms. The comparison was carried out by standard criteria such as the root mean square error and the mean absolute error. Moreover, a non-parametric Kruskal-Wallis statistical test used to compare the median values of the root mean square errors between methods. The main experimental outcomes are: (a) the network's efficiency improves for higher polynomial orders, (b) the statistical analysis suggests that the proposed network appears to be very competitive to the other three networks.

Adaptive Command Filtered Neuro-Fuzzy Control Design for Fractional-Order Nonlinear Systems With Unknown Control Directions and Input Quantization

This article studies the adaptive backstepping control problem for a class of fractional-order (FO) nonlinear systems subject to input quantization and unknown control directions by combining with an indirect FO Lyapunov stability method, and a command filter-based FO dynamic surface control (FODSC) technique. First, a modified FODSC method is utilized to reduce the computational complexity existing in the conventional recursive procedure in which an FO command filter is designed to obtain the command signals and their FO derivatives. Furthermore, the Nussbaum function and neuro-fuzzy networks (NFNs) are adopted to deal with the problem of the unknown control directions and unknown nonlinear functions existing in the system. Moreover, by introducing the compensation and prediction mechanism to controller design, the adaptive controllers, and adaptive laws are constructed to ensure that all the signals of the controlled systems are bounded. Finally, two examples are given to show the validity of the developed control method.

Modeling And Design Of Nano Scale Cmos Controller Using Adaptive Neuro-Fuzzy For Speed Control Of Dc Shunt Motor

The main aim of this work is to control the speed of the DC shunt motor by using a nano scale Complimentary Metallic oxide Field Effect Transistor (CMOS) inverter with Adaptive Neuro-Fuzzy based method for optimal parameter selection. The speed of the DC shunt motor can be varied by varying armature voltage with field current constant and varying the field current with armature voltage constant. It is found that the speed is said to increase with increase in armature voltage for which the graph is a linear in case of armature controlled method. Whereas for the field control method the speed is said to increase with decrease in field current with armature voltage to be constant. The stated condition is cross verified by plotting the graphs for conventional and proposed nano scale CMOS inverter with ANFIS optimization scheme. A novel technique for modeling and devising a nano scale CMOS circuit using adaptive neuro-fuzzy network for controlling and tuning the switching characteristics of the circuit so that it is symmetric with equal values of rise time and fall time with equal output for propagation delay during low to high and high to low outputs. This control over the CMOS switching characteristics using ANFIS controller offers an effective speed control for DC shunt motors. The designed scheme for 45nm process technology is simulated using MATLAB. The results for nano scale CMOS inverter with ANFIS optimization scheme proved to be accurate with an accuracy of 97.5% to 100% which is compared with experimentation results obtained in laboratory using conventional speed control techniques like field and armature control methods.

A new method for force prediction in an accelerometer force balance system using support vector regression

The accurate prediction of force is very important in the present scenario of aerodynamic force measurement. The high accuracy of force prediction during calibration facilitates a better accuracy of force measurement in aerodynamic facilities like shock tunnels and wind tunnels. The present study describes the force prediction in an accelerometer force balance system using support vector regression (SVR). The comparison of SVR with the existing force prediction techniques namely, adaptive neuro-fuzzy inference system (ANFIS) and artificial neural network (ANN) has also been carried out. The accelerometer force balance used in the current experimentation consists of a tri-axial accelerometer to measure the response on an aluminium hemispherical model on the application of force. The impulse forces were applied along the axial, normal and azimuthal directions. The forces were predicted using the accelerations obtained from the tri-axial accelerometer. SVR method was able to predict the forces quite accurately as compared to ANFIS and ANN. However, SVR has the advantage over ANFIS and ANN in that it is independent of the magnitude of the training and testing data. It is capable of an accurate prediction of forces with any magnitude of training and testing data, unlike ANFIS and ANN.

Determining the pile bearing capacity use of pda test results and neural networks

The dynamic loading or PDA test is one of the on-site experiments to estimate the bearing capacity of a pile. This test is based on the wave-propagation theory and can provide a good estimate of the bearing capacity of a pile as well as an assessment of the health of the pile. In this research, using the results of 100 dynamic loading tests carried out with different piles and projects, three types of artificial neural network (ANN) have been used to estimate the load. Initially, the perceptron multi-layer neural network was one of the most used neural networks. Subsequently, the neuro-fuzzy network is used in a combination of neuro-fuzzy networks and, at the end of the radial-based neural network, a successful network was used for non-linear problems. Between the different models of the neural network used by researchers, the multi-layered perceptron network has a better performance. However, other networks used have also proven successful. Finally, different models of the neural networks were compared and the network that has the best performance was identified in both phases. The models based on neural networks, unlike conventional behavioral models, do not explain how the input parameters affect the output. In this research, by analyzing the sensitivity to the optimal structure of the introduced models in each step, we have tried to partly answer this question. Also, the extraction and presentation of the relations governing a neural network model to the user is more reliable in the use of such models, which facilitates the application of such models in engineering works. In this research, four first indicators were used to evaluate and compare the models and structures.

Neuro-Fuzzy Networks and Their Applications in Medical Fields.

The fusion of artificial neural networks and fuzzy logic systems allows researchers to model real-world problems through the development of intelligent and adaptive systems. Artificial neural networks are able to adapt and learn by adjusting the interconnections between layers, while fuzzy logic inference systems provide a computing framework based on the concept of fuzzy set theory, fuzzy if-then rules, and fuzzy reasoning. The combined use of those adaptive structures is known as "neuro-fuzzy" systems. In this paper, the basic elements of both approaches are analyzed, noticing that this blending could be applied for pattern recognition in medical applications.

Crops reclamation management based on hybrid neuro-fuzzy systems

The article deals the technique of application of neuro-fuzzy systems in problems of management of melioration of agricultural crops is considered. In particular, an adaptive network based on the fuzzy inference system is developed, a method for predicting data flows based on the neuro-fuzzy network is proposed, which is implemented in the MATLAB ANFIS module.

Short time prediction of cloud server round-trip time using a hybrid neuro-fuzzy network

The paper presents a cloud server roundtrip time prediction approach for cloud datacenters using neuro-fuzzy network with eight probability distribution functions (Normal, Rayleigh, Weibull, Gamma, Birnbaum-Saunders, Extreme Value, and Generalized Pareto) used for fuzzification and defuzzification. We predict the Round-Trip Time (RTT), i.e., the time for a network packet to travel from a client to a server and back. The proposed approach can achieve significant reduction in the short-time RTT prediction error, achieving an accuracy of 79.36%. The approach could be useful for increasing the efficiency of client-cloud systems, for example, when taking effective decisions for computational offloading, and contribute to the development of smart cloud computing.

AUTOMATIC SEED CLASSIFICATION BY MULTI-LAYER NEURAL NETWORK WITH SPATIAL-FEATURE EXTRACTION

Now a day’s research works on agriculture field have been widely implemented and it is the department that shows rapid growth. This improved growth has shaken its hands to technology for extreme growth. Man, free system for food processing unit like classification based on variety, quality and other aspects. This paper expresses about seed classification based on multiple feature extraction and minimum distance classifier. Feature extraction is associated with spatial, color, shape, texture and statistical features. We have used rice, corn and wheat for this classification process. Too many features have been extracted. Optimum morphological features are extracted for database creation. Training and classification by Means of Multi-Layer Perceptron neural network and Neuro-fuzzy neural network. Matlab tool have been used for entire process.

Motion Prediction Model Using Adaptive Neuro Fuzzy Network (ANFN) and Probabilistic Neural Network (PNN) Algorithm in 4-Dimensional Computed Tomography (4DCT) Images

Motion Prediction Model Using Adaptive Neuro Fuzzy Network (ANFN) and Probabilistic Neural Network (PNN) Algorithm in 4-Dimensional Computed Tomography (4DCT) Images

Parameter Selection and Optimization of an Intelligent Ultrasonic-Assisted Grinding System for SiC Ceramics

Ceramic grinding processing technology continues to advance rapidly. Recent studies are increasingly adopting ultrasonic-assisted grinding (UAG) for obtaining improved surface quality and reduced surface roughness. Because different grinding parameters could lead to different machining quality, operators may find it difficult to select grinding parameters on the basis of the expected machining quality alone. In this article, an intelligent UAG system (IUAGS) that provides suitable grinding parameters for the operator is proposed. This IUAGS employs a proposed one-dimensional convolutional neuro-fuzzy network (1DCNFN) to establish a surface roughness prediction model, and then a particle swarm optimization algorithm is used to optimize the grinding parameters. The experimental results demonstrate that our proposed 1DCNFN has a lower mean absolute percentage error (MAPE) in surface roughness prediction than other methods do. Moreover, our IUAGS can provide appropriate UAG parameters on the basis of the specific requirements of the operator.

A Batch Variable Learning Rate Gradient Descent Algorithm With the SmoothingL1/2Regularization for Takagi-Sugeno Models

A batch variable learning rate gradient descent algorithm is proposed to efficiently train a neuro-fuzzy network of zero-order Takagi-Sugeno inference systems. By using the advantages of regularization, the smoothing L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1/2</sub> regularization is utilized to find more appropriate sparse network. Combining the second-order information of the smoothing error function, a variable learning rate is chosen along the steep descent direction, which avoids line search procedure and may reduce the cost of computation. In order to appropriately adjust the Lipschitz constant of the smoothing error function in the learning rate, a new scheme is proposed by introducing a hyper-parameter. Also the article applies the modified secant equation for estimating the Lipschitz constant, which makes the algorithm greatly reduce the oscillating phenomenon and improve the robustness. Under appropriate assumptions, a convergent result of the proposed algorithm is also given. Simulation results for two identification and classification problems show that the proposed algorithm has better numerical performance and promotes the sparsity capability of the network, compared with the common batch gradient descent algorithm and a variable learning rate gradient-based algorithm.

Multi-Inputs and Multi-Outputs equivalent model based on data driven controller for a robotic system

Abstract This paper proposes the control of a data driven model for an experimental robotic system. The components of the robotic system are a redundant robot and a motion capture system considered them as a Multi-Inputs and Mulit-Outputs system. The Pseudo Jacobian Matrix computes the equivalent model of the robotic system taking into account the input and output signals. Besides, we design the adaptive gains for a proportional controller using an artificial neuro-fuzzy network for the robot’s end-effector control. The experimental results validate the proposed control scheme for a regulation control. We provided a Lyapunov analysis to guarantee convergence parameters of the controller.