Search

This paper deals with the design and implementation of Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF) and Neural State Filter (NSF) for the state estimation of a three-phase induction motor. Extensive simulation studies have been carried out to assess the relative performance of the three filters under various machine operating conditions and model uncertainties. Filter performance for similar conditions was verified with experimental data and found to be consistent with simulation results. The simulation and experimental results indicate that for most conditions EKF estimates are better than UKF while error in NSF estimates is large. However NSF performance is relatively better than other two filters for specific condition like large parameter uncertainty.

Catalytic naphtha reforming is an important process carried out in refineries for upgrading low-octane naphtha to high-octane gasoline. A reformer can meet many product demands through its wide range of design and flexibility in operation. This work deals with fuzzy logic-based optimization of a catalytic reformer. This optimization requires an accurate process model that is valid over wide range of operating conditions. In this work, a simple kinetic model has been developed. This model gives the temperature and concentration profiles of three important hydrocarbons (naphthenes, paraffins, and aromatics) across the reactors. A fuzzy logic-based optimal control scheme has been developed to maximize the aromatics yield, subject to constraints in inlet temperature of the reactors. Four fuzzy logic systems, one in forward path and three in feedback path, have been developed to give setpoints for temperature control loops of the three reactors. Finally, the results are compared with the conventional control scheme and the better performance of fuzzy logic-based control is validated. Copyright © 2008 Curtin University of Technology and John Wiley & Sons, Ltd.

An adaptive window length Recursive Weighted Median filter [ARWMF] for removing the impulse noise with better edge and ¯ne detail preservation is presented. Larger window size may blur the images and the lower window size does not remove the noise at high density. To overcome this, the window size of the RWMF is adaptive based on the presence of noise density. Median controlled algorithm is used to calculate the weights for the RWMF. In median controlled algorithm, the ¯lter gives the smallest weight for the impulse. However, for many weight functions, including the exponential one, this weight is non-zero. Thus the impulse has an effect on the output and the magnitude of the impulse is reduced. The computational complexity for the weight calculation is simple and it is very efficient. The window size of the RWMF is adaptive based on the presence of noise density. The proposed algorithm produces better edge and fine details preservations and reduces blurring at the high density impulse noise.The performance of the proposed algorithm is given in terms of mean square error (MSE), mean absolute error (MAE) and peak signal to noise ratio (PSNR) and it is compared with Standard Median ¯lters, Weighted Median filters, Center Weighted Median ¯lters, Recursive Weighted Median filters and Lins Adaptive length Recursive weighted median filters using Median Controlled Algorithm.

Catalytic naphtha reforming is an important process carried out in refineries for upgrading low-octane naphtha to high-octane gasoline. A reformer can meet many product demands through its wide range of design and flexibility of operation. This work deals with optimization of a catalytic reformer using artificial neural networks (ANN). This optimization requires an accurate process model that is valid over wide range of operating conditions. In this work, a simple kinetic model has been developed. This model gives the temperature and concentration profiles of three important hydrocarbons (naphthenes, paraffins, and aromatics) across the reactors. An optimal control scheme using artificial neural networks has been developed to maximize the aromatics yield, subject to constraints in inlet temperature of the reactors. Two neural networks, one in the forward path and the other in the feedback path, are trained to give set points for temperature control loops of the three reactors. Finally, the result...

In this paper a new algorithm is proposed to remove Gaussian noise with edge preservation. The function of the proposed algorithm is to first find the pixel values along the boundary of the filtering window and then calculate its variance. If this variance is less than a threshold specified, then the corrupted pixel is replaced by the mean of the inside pixels from the filtering window after sorting and trimming. Experimental results shows that the proposed algorithm outperforms with significant improvement in image quality than the arithmetic mean, alpha-trimmed mean filter, wiener filter, K-means filter and adaptive window based method. The proposed method removes the Gaussian noise very effectively even at a noise variance as high as 40 with edge preservation.

The field of system identification is an important discipline of engineering which is widely used in control engineering practices and fault diagnosis. This paper describes a new methodology for parameter identification in a three-tank system through neural network and there by detecting a fault. The parameters are obtained by mapping between the process inputs and its states to the predicted states. The weights of the network represent the parameters of the system. The identified parameters are thus used to diagnose the clogging fault using Kohonen neural network.

In this paper a new efficient algorithm for the removal of Gaussian noise in gray scale and color images using adaptive window is presented. The function of the algorithm is to replace each corrupted pixel by a mean value of the pixels inside an adaptive window. The adaptive window is formed using a threshold calculated form noise variance. The proposed algorithm is simple and it works very effectively in removing Gaussian noise compare with the other techniques. The proposed algorithm is tested for both gray scale and color images corrupted with Gaussian noise. The visual and quantitative results show that the proposed algorithm performs well in removing Gaussian noise and preserve edge details.

This paper presents the hardware realization of fast and flexible feed forward neural network which is capable of dealing with fixed point arithmetic operations using VHDL with minimum number of CLB slices and good speed of performance. The hardware architecture of neural network with two input, one output and three hidden neurons occupies only 44% of CLB slices. An efficient and fast carry look-ahead adder and Booth multiplier are the essential building blocks of the processing elements to perform parallel computation in the neural network. The activation function has been carried out based on piecewise linear approximation only with combinational logic circuits

In this paper, we propose a hybrid soft computing technique called adaptive neuro-fuzzy inference system (ANFIS) to estimate the interference and to separate the electroencephalogram (EEG) signal from its electrooculogram (EOG), electrocardiogram (ECG) and electromyogram (EMG) artifacts. This paper shows that the proposed method successfully removes the artifacts and extracts the desired EEG signal

In this paper, artificial intelligence like hybrid neuro fuzzy logic technique is proposed to cancel the major non-linear interference called maternal electrocardiogram for the extraction of fetal electrocardiogram (FECG). Conventional filtering techniques are not suitable due to an overlap in spectral content of the fetal and the interference. The performance evaluation of the proposed technique is done on the extracted fetal signal in terms of signal to noise ratio, mean square error, and number of membership functions, learning rates and processing time. Comparison is made between the proposed technique and the neural network. It shows that neuro fuzzy logic successfully cancels the interference in fetal electrocardiogram.

This paper deals with a method of estimation of robust dead-time for first order plus time delay (FOPTD) process which does not violate the closed loop stability when Proportional-Integral (PI) parameters and Proportional-Integral-Derivative (PID) parameters as suggested by popular tuning techniques, have been used. Using the Hermite-Biehler Theorem applicable to quasi polynomials, we estimate the lower bound value of (T/L) and then the range of L for the two controller modes. This investigation will help to infer the amount of system parameters that can be tolerated before control system becomes unstable for the two controller modes.

Fetal ECG (IECG) monitoring enables accurate measurement of fetal cardiac performance including transient or permanent abnormalities of rhythm. Thie fetal signal obtained from the maternail abdomnen is mixed with much interference. The major source of this interference is maternal ECG (mECG). This paper describes the cancellation of mECG in fECG using Adaptive Netiro fuzzy logic technique (ANFIS).

The hardware implementation of fuzzy logic controller(FLC) on FPGA is very important because of the increasing number of fuzzy applications requiring highly parallel and high speed fuzzy processing. This paper describes the hardware implementation of two input, one output fuzzy logic controller using VHDL. The architectural design is tested using Spartan FPGA chip xc2s300e-pq208. The maximum frequency of the clock running the FLC and the total number of gates required for the hardware implementation of fuzzy logic controller are compared for the proportionally increasing number of rules.

A fault can be defined as an unexpected change or malfunction in a system and a supervision system should be used to detect and identify the fault and its severity. A fault diagnosis system for online application must provide guaranteed response with its severity, so that a quick decision can be taken either for periodic maintenance if the severity of the fault is less or an immediate shut down if the severity is more. In this paper, a model-based fault diagnosis in a three-tank system has been done using artificial intelligence (AI). Two clogging faults are introduced at different locations with different magnitudes of severity. The A and B parameters of the state space model are estimated using single layer neural network and four AI techniques are used to detect the clogging fault along with severity. The results of different techniques of fault diagnosis are also presented and compared.