• International journal of advanced smart convergence
• /
• v.1 no.2
• /
• pp.30-33
• /
• 2012

This paper propose a new design of fuzzy neural networks based on hard partition to generate the rules of the networks. For this we use hard c-means (HCM) clustering algorithm. The premise part of the rules of the proposed networks is realized with the aid of the hard partition of input space generated by HCM clustering algorithm. The consequence part of the rule is represented by polynomial functions. And the coefficients of the polynomial functions are learned by BP algorithm. The number of the hard partition of input space equals the number of clusters and the individual partitioned spaces indicate the rules of the networks. Due to these characteristics, we may alleviate the problem of the curse of dimensionality. The proposed networks are evaluated with the use of numerical experimentation.

• Proceedings of the KIEE Conference
• /
• 2003.07d
• /
• pp.2633-2635
• /
• 2003

In this study, we propose a fuzzy polynomial neural networks (FPNN) and a genetically optimized fuzzy polynomial neural networks(GoFPNN) for identification of non-linear system. GoFPNN architecture is designed by a FPNN based on fuzzy set and its structure and parameters are optimized by genetic algorithms. A fuzzy neural networks(FNN) based on fuzzy set divide into two structures that is simplified inference structure and linear inference structure. The proposed FPNN is resulted from integration and extension of simplified and linear inference structure of FNN. The consequence structure of the FPNN consist of polynomials represented by networks using connection weights for rules. The networks comprehend simplified(Type 0), linear (Type 1), and quadratic(Type 3) inferences. The proposed FPNN can select polynomial type of consequence part for each rule. Therefore, proposed scheme can offer flexible structure design capability for a system characteristics. Moreover, GAs is applied to networks structure and parameters tuning of proposed FPNN, and its efficient application method is discussed, these subjects are result in GoFPNN that is optimal FPNN. To evaluate proposed model performance, a numerical experiment is carried out.

• Proceedings of the KIEE Conference
• /
• 2006.04a
• /
• pp.270-272
• /
• 2006

In this paper, we discuss optimal design of Fuzzy Polynomial Neural Networks by means of Genetic Algorithms(GAs) using symbolic coding for non-linear data. One of the major subject of genetic algorithms is representation of chromosomes. The proposed model optimized by the means genetic algorithms which used symbolic code to represent chromosomes. The proposed gFPNN used a triangle and a Gaussian-like membership function in premise part of rules and design the consequent structure by constant and regression polynomial (linear, quadratic and modified quadratic) function between input and output variables. The performance of the proposed model is quantified through experimentation that exploits standard data already used in fuzzy modeling. These results reveal superiority of the proposed networks over the existing fuzzy and neural models.

• Structural Engineering and Mechanics
• /
• v.56 no.5
• /
• pp.787-796
• /
• 2015

When the temperature of a structure varies, there is a tendency to produce changes in the shape of the structure. The resulting actions may be of considerable importance in the analysis of the structures having non-prismatic members. The computation of design forces for the non-prismatic beams having symmetrical parabolic haunches (NBSPH) is fairly difficult because of the parabolic change of the cross section. Due to their non-prismatic geometrical configuration, their assessment, particularly the computation of fixed-end horizontal forces and fixed-end moments becomes a complex problem. In this study, the efficiency of the Artificial Neural Networks (ANN) and Adaptive Neuro Fuzzy Inference Systems (ANFIS) in predicting the design forces and the design moments of the NBSPH due to temperature changes was investigated. Previously obtained finite element analyses results in the literature were used to train and test the ANN and ANFIS models. The performances of the different models were evaluated by comparing the corresponding values of mean squared errors (MSE) and decisive coefficients ($R^2$). In addition to this, the comparison of ANN and ANFIS with traditional methods was made by setting up Linear-regression (LR) model.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.11 no.3
• /
• pp.231-240
• /
• 2001

Conventional expert systems has been criticized due to its lack of capability to adapt to the changing decision-making environments. In literature, many methods have been proposed to make expert systems more environment-adaptive by incorporating fuzzy logic and neural networks. The objective of this paper is to propose a new approach to building a self-evolving expert system inference mechanism by integrating fuzzy neural network and fuzzy rule extraction technique. The main recipe of our proposed approach is to fuzzify the training data, train them by a fuzzy neural network, extract a set of fuzzy rules from the trained network, organize a knowledge base, and refine the fuzzy rules by applying a pruning algorithm when the decision-making environments are detected to be changed significantly. To prove the validity, we tested our proposed self-evolving expert systems inference mechanism by using the bankruptcy data, and compared its results with the conventional neural network. Non-parametric statistical analysis of the experimental results showed that our proposed approach is valid significantly.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.4 no.2
• /
• pp.182-186
• /
• 2004

This paper describes an advanced compensation for non-linear functions designed to remove steering aberrations from phased array antennas. This system alters the steering command applied to the antenna in a way that the appropriate angle commands are given to the array steering software for the antenna to point to the desired position instead of squinting. Artificial neural networks are used to develop the inverse function necessary to correct the aberration. Also a straightforward antenna steering function is implemented with neural networks for the 9-term polynomials of forward steering function. In all cases the aberration is removed resulting in small RMS angular errors across the operational angle space when the actual antenna position is compared with the desired position. The use of neural network model provides a method of producing a non-linear system that can correct antenna performance and demonstrates the feasibility of generating an inverse steering algorithm.

• Computers and Concrete
• /
• v.16 no.6
• /
• pp.919-932
• /
• 2015

This paper presents a new method for assessing the three-point-bending (3PB) strength of mortar beams in a non-destructive manner, based on neural network (NN) models. The models are based on the radial basis function (RBF) architecture and the fuzzy means algorithm is employed for training, in order to boost the prediction accuracy. Data for training the models were collected based on a series of experiments, where the cement mortar beams were subjected to various bending mechanical loads and the resulting pressure stimulated currents (PSCs) were recorded. The input variables to the NN models were then calculated by describing the PSC relaxation process through a generalization of Boltzmannn-Gibbs statistical physics, known as non-extensive statistical physics (NESP). The NN predictions were evaluated using k-fold cross-validation and new data that were kept independent from training; it can be seen that the proposed method can successfully form the basis of a non-destructive tool for assessing the bending strength. A comparison with a different NN architecture confirms the superiority of the proposed approach.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.12
• /
• pp.78-86
• /
• 1999

The improvement of the water level control is important since it will prevent the steam generator trip so that improve the reliability and credibility of operation system. In this paper, the closed loop system identification is performed which can be used for the system monitoring and prediction of the system response. The model also can be used for the prediction control. Irving model is used as a steam generator model. The plant is an open loop unstable and non-minimum phase system. Fuzzy controller stabilize the system and the stable controller stabilize the system and the stable closed loop system is identified using neural networks. The obtained neural network model is validated using the untrained input and output. The results of computer simulation show the obtained Neural Network model represents the closed loop system well.

• Journal of Korean Institute of Industrial Engineers
• /
• v.23 no.1
• /
• pp.39-54
• /
• 1997

This paper presents four process models for machining processes : 1) an economical mathematical model of machining process, 2) a prediction model for surface roughness, 3) a decision model for fuzzy cutting conditions, and 4) a judgment model of machinability with automatic selection of cutting conditions. Each model was developed the economic machining, and these models were applied to theories widely studied in industrial engineering which are nonlinear programming, computer simulation, fuzzy theory, and neural networks. The results of this paper emphasize the human oriented domain of a nonlinear programming problem. From a viewpoint of the decision maker, fuzzy nonlinear programming modeling seems to be apparently more flexible, more acceptable, and more reliable for uncertain, ill-defined, and vague problem situations.

• Journal of Internet Computing and Services
• /
• v.9 no.1
• /
• pp.129-135
• /
• 2008

This paper presents a methodology to forecast KOSPI index by extracting fuzzy rules based on the neural network with weighted fuzzy membership functions (NEWFM) and the minimized number of input features using the distributed non-overlap area measurement method. NEWFM classifies upward and downward cases of KOSPI using the recent 32 days of CPPn,m (Current Price Position of day n for n-1 to n-m days) of KOSPI. The five most important input features among CPPn,m and 38 wavelet transformed coefficients produced by the recent 32 days of CPPn,m are selected by the non-overlap area distribution measurement method. For the data sets, from 1991 to 1998, the proposed method shows that the average of forecast rate is 67.62%.