Showing papers on "Hybrid neural network published in 2002"

Hybrid neural network modeling of a full-scale industrial wastewater treatment process.

Abstract: In recent years, hybrid neural network approaches, which combine mechanistic and neural network models, have received considerable attention. These approaches are potentially very efficient for obtaining more accurate predictions of process dynamics by combining mechanistic and neural network models in such a way that the neural network model properly accounts for unknown and nonlinear parts of the mechanistic model. In this work, a full-scale coke-plant wastewater treatment process was chosen as a model system. Initially, a process data analysis was performed on the actual operational data by using principal component analysis. Next, a simplified mechanistic model and a neural network model were developed based on the specific process knowledge and the operational data of the coke-plant wastewater treatment process, respectively. Finally, the neural network was incorporated into the mechanistic model in both parallel and serial configurations. Simulation results showed that the parallel hybrid modeling approach achieved much more accurate predictions with good extrapolation properties as compared with the other modeling approaches even in the case of process upset caused by, for example, shock loading of toxic compounds. These results indicate that the parallel hybrid neural modeling approach is a useful tool for accurate and cost-effective modeling of biochemical processes, in the absence of other reasonably accurate process models.

An analysis of a hybrid neural network and pattern recognition technique for predicting short-term increases in the NYSE composite index

Abstract: We introduce a method for combining template matching, from pattern recognition, and the feed-forward neural network, from artificial intelligence, to forecast stock market activity. We evaluate the effectiveness of the method for forecasting increases in the New York Stock Exchange Composite Index at a 5 trading day horizon. Results indicate that the technique is capable of returning results that are superior to those attained by random choice.

Learning emotions in virtual environments

Abstract: A modular hybrid neural network architecture, called SHAME, for emotion learning is introduced. The system learns from annotated data how the emotional state is generated and changes due to internal and external stimuli. Part of the modular architecture is domain independent and part must be adapted to the domain under consideration. The generation and learning of emotions is based on the event appraisal model. The architecture is implemented in a prototype consisting of agents trying to survive in a virtual world. An evaluation of this prototype shows that the architecture is capable of generating natural emotions and furthermore that training of the neural network modules in the architecture is computationally feasible. Keywords: hybrid neural systems, emotions, learning, agents.

Comparative evaluation of various control schemes for fed-batch fermentation

Abstract: The crucial problem associated with control of fed-batch fermentation process is its time-varying characteristics. A successful controller should be able to deal with this feature in addition to the inherent nonlinear characteristics of the process. In this work, various schemes for controlling the glucose feed rate of fed-batch baker's yeast fermentation were evaluated. The controllers evaluated are fixed-gain proportional-integral (PI), scheduled-gain PI, adaptive neural network and hybrid neural network PI. The difference between the specific carbon dioxide evolution rate and oxygen uptake rate (Qc–Qo) was used as the controller variable. The evaluation was carried out by observing the performance of the controllers in dealing with setpoint tracking and disturbance rejection. The results confirm the unsatisfactory performance of the conventional controller where significant oscillation and offsets exist. Among the controllers considered, the hybrid neural network PI controller shows good performance.

Hybrid neural network and support vector machine method for optimization

Abstract: System and method for optimization of a design associated with a response function, using a hybrid neural net and support vector machine (NN/SVM) analysis to minimize or maximize an objective function, optionally subject to one or more constraints. As a first example, the NN/SVM analysis is applied iteratively to design of an aerodynamic component, such as an airfoil shape, where the objective function measures deviation from a target pressure distribution on the perimeter of the aerodynamic component. As a second example, the NN/SVM analysis is applied to data classification of a sequence of data points in a multidimensional space. The NN/SVM analysis is also applied to data regression.

Dynamic modeling of retort processing using neural networks

Abstract: Two neural network approaches - a moving-window and hybrid neural network - which combine neural network with polynomial regression models, were used for modeling F(t) and Qv(t) dynamic functions under constant retort temperature processing. The dynamic functions involved six variables: retort temperature (116-132C), thermal diffusivity (1.5-2.3 x 10 -7 m 2 /s ), can radius (40-61 mm), can height (40-61 mm), and quality kinetic parameters z (15-39C) and D (150-250 min). A computer simulation designed for process calculations of food thermal processing systems was used to provide the fundamental data for training and generalization of ANN models. Training data and testing data were constructed by both second order central composite design and orthogonal array, respectively. The optimal configurations of ANN models were obtained by varying the number of hidden layers, number of neurons in hidden layer and learning runs, and a combination of learning rules and transfer function. Results demonstrated that both neural network models well described the F(t) and Qv(t) dynamic functions, but moving-window network had better modeling performance than the hybrid ANN models. By comparison of the configuration parameters, moving-window ANN models required more neurons in the hidden layer and more learning runs for training than the hybrid ANN models. Deux approches par reseaux neuronaux combines avec des modeles de regression polynomiaux sont utilisees pour modeliser les fonctions dynamiques F(t) et Qv(t) dans un procede d'autoclavage a temperature constante. Une simulation par ordinateur destinee aux calculs de procedes des systemes thermiques alimentaires est utilisee pour fournir les donnees fondamentales des modeles. Le reseau par fenetre mobile montre des performances de modelisation superieures a celles du modele hybride.

Fault location in transmission line using hybrid neural network

Abstract: The paper presents the application of self‐organizing neural network for the location of the fault in the transmission line and estimation of the parameter of the faulty element. The location of fault is done on the basis of the measurement of some node voltages of the line and appropriate preprocessing it to enhance the differences between different faults. The hybrid neural network is used to solve the problem. The self‐organizing layer of this network is used as the classifier. The output postprocessing MLP structure realizes the association of the place of the fault and its parameter with the measured set of node voltages. The results of computer experiments are given in the paper and discussed.

Hybrid neural networks: An evolutionary approach with local search

Abstract: Considering computational algorithms available in the literature, associated with supervised learning in feedforward neural networks, a wide range of distinct approaches can be identified. While the adjustment of the connection weights represents an omnipresent stage, the algorithms differ on three basic aspects: the technique chosen to determine the dimension of the multilayer neural network, the procedure adopted to specify the activation functions, and the kind of composition used to produce the output. Advanced learning algorithms should be developed to simultaneously treat all these aspects during learning, and an evolutionary learning algorithm with local search is proposed here. The essence of this approach is a synergy between genetic algorithms and conjugate gradient optimization, operating on a hybrid neural network architecture. As a consequence, the final neural network is automatically generated, and is characterized to be dedicated and computationally parsimonious.

Segmentation of MR and CT Images Using a Hybrid Neural Network Trained by Genetic Algorithms

Abstract: A novel hybrid neural network trained by the genetic algorithms is presented. Genetic algorithms are used to improve the neural net's classification performance while minimizing the number of nodes. Each node of the network forms a closed region in the input space. The closed regions, which are formed by the nodes, intersect each other. The performance of the proposed hybrid neural network is compared with the multilayer perceptron, and the restricted Coulomb energy network for the segmentation of MR and CT head images. Experimental results show that the proposed neural network gives the best classification performance with a small number of nodes in short training times.

Neural network process vision systems for flotation process

Abstract: Froth flotation is a process whereby valuable minerals are separated from waste by exploiting natural differences or by chemically inducing differences in hydrophobicity. Flotation processes are difficult to model because of the stochastic nature of the froth structures and the ill‐defined chemorheology of these systems. In this paper a hierarchical configuration hybrid neural network has been used to interpret froth images in a copper flotation process. This hierarchical neural network uses two Pulse‐Coupled Neural Networks (PCNNs) as preprocessors that ‘convert’ the froth images into corresponding binary barcodes. Our technique demonstrates the effectiveness of the hybrid neural network for process vision, and hence, its potential for use for real time automated interpretation of froth images and for flotation process control in the mining industry. The system is simple, inexpensive and is very reliable.

New Hybrid Neural Network Model for Prediction of Phase Equilibrium in a Two-Phase Extraction System

Abstract: A novel approach to modeling prediction of phase equilibrium is presented. The method, evolutionary polymorphic neural network (EPNN), is developed by the authors on the basis of artificial neural ...

Segmentation of ultrasound images by using quantizer neural network

Abstract: A quantizer neural network (QNN) is proposed for the segmentation of ultrasound images. The elements of the feature vectors are formed by the image intensities within the neighborhood of the pixel of interest. The QNN is a hybrid neural network structure, which is trained by genetic algorithms. The genetic algorithms are used to find optimum values for the weights of the nodes. The hybrid neural network is compared with a multilayer perceptron (MLP) for the segmentation of ultrasound images.

A comparison of face recognition algorithms neural network based & line based approaches

Abstract: One of the most successful applications of image analysis and understanding, face recognition has received significant attention. There are at least two reasons for the trend: the first is the wide range of commercial and law enforcement applications and the second is the availability of feasible technologies. In general, few methods of face recognition are in practice: feature based face recognition methods, eigen face based, line based, elastic bunch graph method and neural network based methods. All have their possibilities and features. In the neural network approach automatic detection of eyes and mouth is followed by a spatial normalization of the images. The classification of the normalized images is carried out by hybrid neural network that combines unsupervised and supervised methods for finding structures and reducing classification errors respectively. The line-based is a type of image-based approach. It does not use any detailed biometric knowledge of the human face. These techniques use either the pixel-based bi-dimensional array representation of the entire face image or a set of transformed images or template sub-images of facial features as the image representation. An image-based metric such as correlation is then used to match the resulting image with the set of model images. In the context of image-based techniques, two approaches are there namely template-based and neural networks. In the template-based approach, the face is represented as a set of templates of the major facial features, which are then matched with the prototypical model face templates.

Improving driver pose estimation

Abstract: In this paper, we consider the problem of estimating the pose of a driver from video data. We propose extensions to our previous eigenface and Fisherface-based methods to improve classification performance. In particular, a hybrid neural network/nearest neighbor algorithm is formulated for classification of frames. Experimental results show that the hybrid neural network outperforms the nearest neighbor classifier.

A hybrid neural network-decision tree-based method for transient stability assessment

Abstract: Phasor Measuring Units (PMUs) using synchronization signals from Global Positioning System (GPS) satellite system have evolved into mature tools for power system operation and control. For power system transient stability assessment, a computationally efficient way of processing real-time measurements to determine whether an evolving swing will ultimately be stable or unstable is required. A hybrid pattern classifier that combines neural networks and decision trees has been used to assess a 12-generator power system based on phasor measurements with classification rates of over 99% for the training set and over 94% for the test set.

Hybrid neural network/conventional control of a benchmark process control problem

Abstract: We combine neural network and conventional controllers to form a hybrid that achieves very accurate control in both stable an unstable operating regions of a simulated bioreactor. The neural network handles nonlinearity and generalizes to cover both regions. The conventional controller eliminates the offset error incurred by generalization.

Modeling of supercritical ethane extraction by artificial neural networks

Abstract: In this paper, an artificial neural network that considers the system as a black box is designed for the mass transfer modeling of supercritical ethane extraction In addition, a hybrid model using both neural network and Peng-Robinson state equation is developed for supercritical ethane extraction, where the neural network is used to generate the nonlinear binary interaction parameter of the Peng-Robinson state equation The predictions of the proposed neural network models are compared to a conventional model with a Peng-Robinson equation of state in literature Generally, the results using the proposed models are better than those using the conventional model

Hybrid Neural-Network: Genetic Algorithm Technique for Aircraft Engine Performance Diagnostics Developed and Demonstrated

Abstract: As part of the NASA Aviation Safety Program, a unique model-based diagnostics method that employs neural networks and genetic algorithms for aircraft engine performance diagnostics has been developed and demonstrated at the NASA Glenn Research Center against a nonlinear gas turbine engine model. Neural networks are applied to estimate the internal health condition of the engine, and genetic algorithms are used for sensor fault detection, isolation, and quantification. This hybrid architecture combines the excellent nonlinear estimation capabilities of neural networks with the capability to rank the likelihood of various faults given a specific sensor suite signature. The method requires a significantly smaller data training set than a neural network approach alone does, and it performs the combined engine health monitoring objectives of performance diagnostics and sensor fault detection and isolation in the presence of nominal and degraded engine health conditions.

Evolving artificial neural network models for voltage contingency ranking

Abstract: This paper presents four different models of artificial neural network (ANN), namely Multi-layer feed-forward (BP) model, Modular neural network (MNN), Hybrid neural network (HNN), and Cascade neural network (CNN) for fast voltage contingency screening and ranking. The effectiveness of the proposed method is demonstrated for contingency screening and ranking at different loading conditions on two sample power systems. One of the proposed neural networks, CNN provides fast and accurate screening and ranking for unknown patterns and would be suitable for on-line applications at energy management center.

Application of a novel neural network for identification of nonlinear carbon/carbon gaseous-state deposit process

Abstract: A novel neural network model and algorithm for highly nonlinear C/C gaseous-state deposit process is presented. The network model consists of fuzzy classifier and some wavelet sub-networks called a hybrid neural network. The input samples are trained by a homologous wavelet network after classifying. The results of the identification of the C/C gaseous-state deposit process are desirable.