Mohammad Fazle Azeem

Generalization of adaptive neuro-fuzzy inference systems

The paper aims at several objectives. The adaptive network-based fuzzy inference systems (ANFIS) of Jang is extended to the generalized ANFIS (GANFIS) by proposing a generalized fuzzy model (GFM) and considering a generalized radial basis function (GRBF) network. The GFM encompasses both the Takagi-Sugeno (TS)-model and the compositional rule of inference (CRI)-model. A local model, a property of TS-model, and the index of fuzziness, a property of CRI-model define the consequent part of a rule of GFM. The conditions by which the proposed GFM converts to TS-model or the CRI-model are presented. The basis function in GRBF is a generalized Gaussian function of three parameters. The architecture of the GRBF network is devised to learn the parameters of GFM, since it has been proved in this paper that GRBF network and GFM are functionally equivalent. It is shown that GRBF network can be reduced to either the standard RBF or the Hunts RBF network. The issue of the normalized versus the nonnormalized GRBF networks is investigated in the context of GANFIS. An interesting property of symmetry on the error surface of GRBF network is investigated in the present work. The proposed GANFIS is applied for the modeling of a multivariable system like stock market.

Artificial wavelet neural network and its application in neuro-fuzzy models

In the proposed work, two types of artificial neural networks are proposed by using well-known advantages and valuable features of wavelets and sigmoidal activation functions. Two neurons are derived by adding and multiplying the outputs of the wavelet and the sigmoidal activation functions. These neurons in a feed-forward single hidden layer network result summation wavelet neural network (SWNN) and multiplication wavelet neural network (MWNN). An algorithm is introduced for structure determination of the proposed networks. Approximation properties of SWNN and MWNN have been evaluated with different wavelet functions. The above networks in the consequent part of the neuro-fuzzy model result summation wavelet neuro-fuzzy (SWNF) and multiplication wavelet neuro-fuzzy (MWNF) models. Different types of wavelet function are tested with the proposed networks and fuzzy models on four different dynamical examples. Convergence of the learning process is also guaranteed by adaptive learning rate and performing stability analysis using Lyapunov function.

Modified queen bee evolution based genetic algorithm for tuning of scaling factors of fuzzy knowledge base controller

There are wide ranges of combination for genetic algorithm (GA) operators exist in the literature. Most of them have been applied on different type of tuning application for fuzzy knowledge base controller (FKBC). In this paper authors proposed a modification to the Sungs GA. The proposed GA utilizes the weighted crossover operator. A fitness function, which guides the evolution process, which is defined as inverse of integral absolute time error (IATE). The proposed method is applied, for the tuning of input and output scaling factors of FKBC, for two complex non-linear systems. The simulation results are encouraging.

Parameter identification of TSK neuro-fuzzy models

Abstract In the framework of the TSK neuro-fuzzy model a combination of the two well-known identification methods are employed for parameter estimation of the neuro-fuzzy inference system, namely the series–parallel and the parallel configurations. The presented paper proposes two new possible configurations for identifying the parameters of the TSK neuro-fuzzy model using the combinations of these two existing configurations. One of the proposed configurations constitutes the series–parallel configuration to the premise part and the parallel configuration to the consequent part of the neuro-fuzzy model, termed as PS-P configuration. The second one is composed of the series–parallel configuration to the consequent part and the parallel configuration to the premise part of the neuro-fuzzy model, termed as CS-P configuration. The presented work mainly deals with a comparative study of the proposed configurations and the existing configurations in the context of parameter identification of the TSK neuro-fuzzy model on three different benchmark examples. Moreover, it investigates upper bound of the learning rates, using the Lyapunov stability theorem, to assure the stability and the convergence of the model learning process. Implementation of the modified mountain clustering (MMC) and the cluster validity function yields initial models. To restrict the upper bound during the learning process it also presents a two-phase adaptive learning rate.

Parameter determination for a generalized fuzzy model

This paper underlines a way to evolve a generalized fuzzy model (GFM), using the interpolation of CRI and TS models in their consequent parts of fuzzy rules. The GFM possesses the index of fuzziness of CRI model and the local model of the TS model. The parameters of the GFM are estimated by a two-step process. The consequent part of fuzzy rules is reformulated to suit the LSE framework for estimating the associated parameters. By assuming Generalized Gaussian membership function for the premise parts, Gradient descent technique is used to update its parameters. The performance of two classes of GFM has been tested on two systems and it is shown that class II GFM is the best out of all the fuzzy models tested.

Local recurrent sigmoidal-wavelet neurons in feed-forward neural network for forecasting of dynamic systems: Theory

In this paper different structure of the neurons in the hidden layer of a feed-forward network, for forecasting of the dynamic systems, are proposed. Each neuron in the network is a combination of the sigmoidal activation function (SAF) and wavelet activation function (WAF). The output of the hidden neuron is the product of the output from these two activation functions. A delay element is used to feedback the output of the sigmoidal and the wavelet activation function to each other. This arrangement leads to proposed five possible configurations of recurrent neurons. Besides proposing these neuron models, the presented paper tries to compare the performance of wavelet function with sigmoid function. To guarantee the stability and the convergence of the learning process, upper bound for the learning rates has been investigated using the Lyapunov stability theorem. A two-phase adaptive learning rate ensures this upper bound. Universal approximation property of the feed-forward network with the proposed neurons has also been investigated. Finally, the applicability and comparison of the proposed recurrent networks has been weathered on two benchmark problem catering different types of dynamical systems.

Fuzzy modeling of fluidized catalytic cracking unit

The paper deals with the fuzzy system identification of reactor-regenerator-stripper-fractionators (RRSF) section of a fluidized catalytic cracking unit (FCCU). The fuzzy system identification based on the data collected from an operating refinery of FCCU of capacity, 1.2MMPTA, with a sample time of 10min. A generalized fuzzy model (GFM) and identification of structure and model parameter for multi-input/single output is presented. The GFM has the capability of representing both the CRI model and TS model under certain conditions. The structure identification and the parameter estimation are carried out using hybrid learning approach comprising modified mountain clustering and gradient descent learning with least square estimation (LSE) for the identification of a fuzzy model. The modified mountain clustering considers every data point as a potential cluster center in xxy hyperspace. The optimum number of clusters, which leads to an optimum number of rules, is determined with the help of validity function that guides the search. The obtained result from the modified mountain clustering initializes the GFM. Further hybrid of the gradient descent technique and LSE is aimed at learning of the GFM parameters in two phases. In the first phase of an epoch of learning gradient descent tunes the premise parameter and index of fuzziness of each rule. In second phase, LSE utilizes the results of first phase for evaluating the coefficient of local linear model of corresponding rules.

Artificial wavelet neuro-fuzzy model based on parallel wavelet network and neural network

From the well-known advantages and valuable features of wavelets when used in neural network, two type of networks (i.e., SWNN and MWNN) have been proposed. These networks are single hidden layer network. Each neuron in the hidden layer is comprised of wavelet and sigmoidal activation functions. First model is derived from adding the outputs of wavelet and sigmoidal activation functions, while in the second model outputs of wavelet and sigmoidal activation function are multiplied together. Using these proposed networks in consequent part of the neuro-fuzzy model, which result summation wavelet neuro-fuzzy and multiplication wavelet neuro-fuzzy models, are also proposed. Different types of wavelet function are tested with proposed networks and fuzzy models on four different types of examples. Convergence of the learning process is also guaranteed by performing stability analysis using Lyapunov function.

Modified Queen-Bee Algorithm-Based Fuzzy Logic Control for Real-Time Robust Load Matching for a Solar PV System

The maximum power point (MPP) of a solar photovoltaic (PV) module fluctuates over the whole day with the variation of weather conditions. In order to track it precisely and harvest it efficiently, an appropriate value of load has to be matched. A prerequisite to successful load matching is the knowledge of available real-time maximum power. For real-time maximum power transfer and robust load matching, a control strategy that combines the use of a dc-dc boost converter, fuzzy logic control, and a modified queen bee genetic algorithm has been proposed. Simulation results show the promise of the approach. It is believed that this research will lead to improvement in the efficiency of PV modules considering real-time weather conditions and intelligent control for robust load matching.

Generalized Wavelet Neuro-Fuzzy Model and its Application in Time Series Forecasting

The advantages of wavelets when used in neural networks and fuzzy are well known. The new notion is to combine wavelet networks and neuro-fuzzy models. In this paper two models namely summation wavelet neural network (SWNN) and multiplication wavelet neural network (MWNN) are proposed. These two generalized wavelet neural network (WNN) models are used in neuro-fuzzy model are tested by using time series prediction