scispace - formally typeset
Search or ask a question

Showing papers on "Hybrid neural network published in 2001"


Journal ArticleDOI
TL;DR: The results of experiments of recognition of different types of beats on the basis of the ECG waveforms have confirmed good efficiency of the proposed solution and show that the method may find practical application in the recognition and classification of different type heart beats.
Abstract: Presents the application of the fuzzy neural network for electrocardiographic (ECG) beat recognition and classification. The new classification algorithm of the ECG beats, applying the fuzzy hybrid neural network and the features drawn from the higher order statistics has been proposed in the paper. The cumulants of the second, third, and fourth orders have been used for the feature selection. The hybrid fuzzy neural network applied in the solution consists of the fuzzy self-organizing subnetwork connected in cascade with the multilayer perceptron, working as the final classifier. The c-means and Gustafson-Kessel algorithms for the self-organization of the neural network have been applied. The results of experiments of recognition of different types of beats on the basis of the ECG waveforms have confirmed good efficiency of the proposed solution. The investigations show that the method may find practical application in the recognition and classification of different type heart beats.

519 citations


Journal ArticleDOI
TL;DR: Ten types of ECG beats obtained from the MIT-BIH database and from a real-time ECG measurement system are classified with a success of 96% by using the hybrid structure.

217 citations


Journal ArticleDOI
Dong-Jin Choi1, Heekyung Park1
TL;DR: The result shows that the hybrid ANN technique can be used to extract information from noisy data and to describe the nonlinearity of complex wastewater treatment processes.

169 citations


Journal ArticleDOI
TL;DR: It was found that the SOM/ARIMA hybrid approach out-performs all individual ARIMA models, whilst the SOM-MLP hybrid approach achieves superior forecasting performance to all models used in this study, including three naïve models.
Abstract: This paper describes an application of hybrid neural network approaches and an assessment of the effects of missing data on highway traffic flow forecasting. Using a self-organizing map (SOM), two hybrid approaches are developed for classifying traffic into different states. In the first hybrid approach, four auto-regressive integrated moving average (ARIMA) models are included. The second approach uses two multi-layer perception (MLP) models. The effects of missing data on neural network performance when forecasting traffic flow are analyzed, and options to replace the missing data are discussed. It is concluded that overall, ARIMA models are more sensitive to the percentage of missing data than neural networks in this context.

153 citations


Proceedings ArticleDOI
01 Jul 2001
TL;DR: The hybrid diagnostic technique takes advantage of the nonlinear estimation capability provided by neural networks while improving the robustness to measurement uncertainty through the application of Genetic Algorithms.
Abstract: In this paper, a model-based diagnostic method, which utilizes Neural Networks and Genetic Algorithms, is investigated. Neural networks are applied to estimate the engine internal health, and Genetic Algorithms are applied for sensor bias detection and estimation. This hybrid approach takes advantage of the nonlinear estimation capability provided by neural networks while improving the robustness to measurement uncertainty through the application of Genetic Algorithms. The hybrid diagnostic technique also has the ability to rank multiple potential solutions for a given set of anomalous sensor measurements in order to reduce false alarms and missed detections. The performance of the hybrid diagnostic technique is evaluated through some case studies derived from a turbofan engine simulation. The results show this approach is promising for reliable diagnostics of aircraft engines.

75 citations


Journal ArticleDOI
TL;DR: This work is developing an optimization method in which historical process data is used to train an artificial neural network for correlation of processing inputs and outputs, and the result of this work will be a general fermentation optimization method that can be applied to fermentation processes to improve quality and productivity.
Abstract: Optimization of fermentation processes is a difficult task that relies on an understanding of the complex effects of processing inputs on productivity and quality outputs. Because of the complexity of these biological systems, traditional optimization methods utilizing mathematical models and statistically designed experiments are less effective, especially on a production scale. At the same time, information is being collected on a regular basis during the course of normal manufacturing and process development that is rarely fully utilized. We are developing an optimization method in which historical process data is used to train an artificial neural network for correlation of processing inputs and outputs. Subsequently, an optimization routine is used in conjunction with the trained neural network to find optimal processing conditions given the desired product characteristics and any constraints on inputs. Wine processing is being used as a case study for this work. Using data from wine produced in our pilot winery over the past 3 years, we have demonstrated that trained neural networks can be used successfully to predict the yeast-fermentation kinetics, as well as chemical and sensory properties of the finished wine, based solely on the properties of the grapes and the intended processing. To accomplish this, a hybrid neural network training method, Stop Training with Validation (STV), has been developed to find the most desirable neural network architecture and training level. As industrial historical data will not be evenly spaced over the entire possible search space, we have also investigated the ability of the trained neural networks to interpolate and extrapolate with data not used during training. Because a company will utilize its own existing process data for this method, the result of this work will be a general fermentation optimization method that can be applied to fermentation processes to improve quality and productivity.

63 citations


Journal ArticleDOI
TL;DR: In this article, a hybrid neural network and genetic algorithm approach is described to determine a set of initial process parameters for injection molding, which is based on a skilled operator's know-how and intuitive sense acquired through long-term experience rather than on a theoretical and analytical approach.
Abstract: Determination of the initial process parameters for injection moulding is a highly skilled task and is based on a skilled operator's know-how and intuitive sense acquired through long-term experience rather than on a theoretical and analytical approach. In the face of global competition, the current trial-and-error practice is inadequate. In this paper, a hybrid neural network and genetic algorithm approach is described to determine a set of initial process parameters for injection moulding. A hybrid neural network and genetic algorithm system for the determination of initial process parameter settings for injection moulding based on the proposed approach was developed and validated. The preliminary validation test of the system has indicated that the system can determine a set of initial process parameters for injection moulding quickly, from which good quality moulded parts can be produced without relying on experienced moulding personnel.

53 citations


Journal ArticleDOI
TL;DR: The hybrid model combining a neural network together with an empirical equation provides a better representation of the estimated parameter values and the outputs predicted by the hybrid neural network are compared with experimental data and some correlations previously proposed in the literature for tanks of different sizes.

46 citations


Journal ArticleDOI
TL;DR: The accuracy of the developed network has been tested by predicting the injection pressure and injection time for few engineering components and found that the overall error is 0.93% with a deviation of 3.93%.

39 citations


Journal ArticleDOI
TL;DR: The present study shows that the categorization process, in which neural networks were used, can be reliable and agree well with the manual categorization and may be useful in patient monitoring.
Abstract: Digitized data from CTG (cardiotocography) measurements (fetal heart rate and uterine contractions) have been used for categorization of typical heart rate patterns before and during delivery. Short time series of CTG data, about 7 min duration, have been used in the categorization process. In the first part of the study, selected CTG data corresponding to 10 typical cases were used for purely auto associative unsupervised training of a Self-Organizing Map Neural Network (SOM). The network may then be used for objective categorization of CTG patterns through the map coordinates produced by the network. The SOM coordinates were then compared. In the second part of the study, a hybrid neural network consisting of a SOM network and a Back-Propagation network (BP) was trained with data corresponding to a number of basic heart rate patterns as described by eight manually selected indices. Test data (different than the training data) were then used to check the performance of the network. The present study shows that the categorization process, in which neural networks were used, can be reliable and agree well with the manual categorization. Since the categorization by neural networks is very fast and does not involve human efforts, it may be useful in patient monitoring.

28 citations


Patent
06 Apr 2001
TL;DR: In this article, a computer-implemented method and system for building a neural network is disclosed, where the neural network predicts at least one target based upon predictor variables defined in a state space.
Abstract: A computer-implemented method and system for building a neural network is disclosed. The neural network predicts at least one target based upon predictor variables defined in a state space. First, an input data set is retrieved that includes the predictor variables and at least one target associated with the predictor variables for each observation. In the state space, a number of points is inserted in the state space based upon the values of the predictor variables. The number of points is less than the number of observations. A statistical measure is determined that describes a relationship between the observations and the inserted points. Weights and activation functions of the neural network are determined using the statistical measure.

Journal ArticleDOI
01 Oct 2001
TL;DR: The accuracy of the developed network has been tested by predicting the injection pressure and injection time for few engineering components and showed that the Levenberg algorithm converged rapidly with fewer training cycles when compared with the error back-propagation algorithm.
Abstract: In the present work attempts have been made to determine the process parameters that could affect an injection moulding process based on governing equations of the mould-filling process. Focus is then directed to parameters that require the use of trial and error methods or other complex software to determine the process parameters. The two parameters that are predicted from the developed network are injection time and injection pressure. In this work, the training data are generated by simulation using C-MOLD flow simulation software. A total of 114 data points under different process conditions were collected out of which 94 data points were used to train the network using MATLAB and the remaining information was used for testing the network. Two algorithms are used during the training phase, namely the error back-propagation algorithm and the Levenberg-Marquardt approximation algorithm. Results showed that the latter algorithm is more suitable for this application since the Levenberg algorithm ...

Book ChapterDOI
21 Aug 2001
TL;DR: In this article, a hybrid neural network is proposed for chaotic time series prediction, which combines a traditional feed-forward network and a local model, which is implemented as a time delay embedding.
Abstract: We propose an efficient hybrid neural network for chaotic time series prediction. The hybrid neural network is constructed by a traditional feed-forward network, which is learned by using the backpropagation and a local model, which is implemented as a time delay embedding. The feed-forward network performs as the global approximation and the local model works as the local approximation. Experimental results using Mackey-Glass data and K.U. Leuven competition data show that the proposed method can predict the more long term than each of predictors.

Journal ArticleDOI
TL;DR: In this article, a hybrid neural network forward model is proposed for retrieving near-surface winds from scatterometer observations over the ocean surface, which retains the physical understanding embodied in CMOD4, but incorporates greater flexibility, allowing a better fit to the observations.
Abstract: Current methods for retrieving near-surface winds from scatterometer observations over the ocean surface require a forward sensor model which maps the wind vector to the measured backscatter. This paper develops a hybrid neural network forward model, which retains the physical understanding embodied in CMOD4, but incorporates greater flexibility, allowing a better fit to the observations. By introducing a separate model for the midbeam and using a common model for the fore and aft beams, we show a significant improvement in local wind vector retrieval. The hybrid model also fits the scatterometer observations more closely. The model is trained in a Bayesian framework, accounting for the noise on the wind vector inputs. We show that adding more high wind speed observations in the training set improves wind vector retrieval at high wind speeds without compromising performance at medium or low wind speeds. Copyright 2001 by the American Geophysical Union.

Journal ArticleDOI
TL;DR: A model based on laboratory data for the synthesis of recombinant β-galactosidase was corrupted by adding imperfect mixing and noise in the feed stream to generate data mimicking a real nonideal operation, and it was seen that a hybrid network provides more accurate estimates of both extra-cellular and intra- cellular variables.
Abstract: Fermentations employing genetically modified microbes under industrial conditions are difficult to monitor on line or to describe by simple, good mathematical models. So, a practically convenient approach is to combine mathematical models of some aspects with artificial neural networks of those aspects which are difficult to measure or model. Such hybrid models have been applied earlier to laboratory-scale bioreactors. In the present work, a model based on laboratory data for the synthesis of recombinant β-galactosidase was corrupted by adding imperfect mixing and noise in the feed stream to generate data mimicking a real nonideal operation. These data were used to train a recurrent Elman neural network and a hybrid neural network, and it was seen that a hybrid network provides more accurate estimates of both extra-cellular and intra-cellular variables. The benefit is enhanced by the hybrid network's superiority being more pronounced for the intra-cellular recombinant protein, β-galactosidase, which is the main product of interest.

Proceedings ArticleDOI
06 May 2001
TL;DR: A novel optimization technique for the parameter identification of microwave monolithic integrated circuits is presented, based on a hybrid neural network whose learning process convergence allows the validation of the circuit approximated lumped model.
Abstract: A novel optimization technique for the parameter identification of microwave monolithic integrated circuits is presented. It is based on a hybrid neural network whose learning process convergence allows the validation of the circuit approximated lumped model. The main feature of such a learning process is that no external desired signal is required and the neural network can be considered of the unsupervised type. Furthermore, the neural network output represents the lumped circuit parameter estimation.

Journal ArticleDOI
TL;DR: The objective of this paper is to review some of the most widely used approaches to neural-network-based modelling, including plain black box as well as hybrid neural network — first principles modelling.
Abstract: In recent years, neural networks have attracted much attention for their potential to address a number of difficult problems in modelling and controlling nonlinear dynamic systems, especially in (bio) chemical engineering. The objective of this paper is to review some of the most widely used approaches to neural-network-based modelling, including plain black box as well as hybrid neural network — first principles modelling. Two specific application examples are used for illustration purposes: a simple tank level-control system is studied in simulation while a challenging bioprocess application is investigated based on experimental data. These applications allow some original concepts and techniques to be introduced.

Proceedings ArticleDOI
29 Oct 2001
TL;DR: This work focuses on utterance verification of short keywords, which contain only one syllable, and integrates neural network and hidden Markov models (HMMs) in an attempt to utilize the strength of both.
Abstract: We focus on utterance verification of short keywords, which contain only one syllable. The relative less information makes the task rather tough. We integrate neural network and hidden Markov models (HMMs) in an attempt to utilize the strength of both. A classifier was built for each keyword. Our model differs from other hybrid models in that we feed the neural networks with posterior likelihood generated by a set of HMMs. Therefore, statistical information can be reserved and the size of neural networks is not too large. Another feature is our utilization of word duration. Our experiments were carried out in Mandarin. Two baseline models for comparison were built, one employed the likelihood ratio measure and the other utilized a neural network presented in preceding studies. Two models depending on our approaches were constructed. The results showed that one proposed approach improved by 8.7% and 12.5%, and another by 15.2% and 19.2%.

Proceedings ArticleDOI
19 Jun 2001
TL;DR: In the paper, an algorithm is presented for the construction of representations of 18 object classes, which can be later recognized by a hybrid neural network, and the map of features for the Kohonen layer has been analyzed.
Abstract: In the paper, an algorithm is presented for the construction of representations of 18 object classes, which can be later recognized by a hybrid neural network The preprocessing took place in log-polar space and it included: object centering, binarization, edge detection, normalization of angular position and scaling After the normalization and log-Hough transformation, the maxima have been projected onto respective axes In the paper, results have been discussed of neural network learning using such constructed representations, and the map of features for the Kohonen layer has been analyzed

Book ChapterDOI
TL;DR: A hybrid mathematical model to describe a three-phase reactor behavior that combines neural network architecture––as a predictor block for the liquid solid mass-transfer coefficient––and phenomenological equations describing the mass-conservation principle is presented.
Abstract: Publisher Summary This chapter presents a hybrid mathematical model to describe a three-phase reactor behavior that combines neural network architecture––as a predictor block for the liquid solid mass-transfer coefficient––and phenomenological equations describing the mass-conservation principle. The optimization procedure used in the network training was based on the Fletcher–Powell algorithm. Results of the network training and validation showed the predictive capacity of the proposed model and its great potential to be used as a support for process modeling and control. To explore the potentialities of the artificial neural network (ANN), two different approaches were tested: (i) standard ANN modeling (also called “black-box”), where ANN was used to represent the whole process behavior by mapping its input to output process data and (ii) hybrid ANN modeling, where ANN was used to predict the liquid–solid mass transfer coefficient that is a parameter for the determinist model. The hybrid neural network model is composed of two blocks. The ANN block estimates a process parameter––the liquid–solid mass transfer coefficient––which is used as input to the second block, represented by the deterministic equations of the process––mass and energy balance equations.

Proceedings ArticleDOI
25 Jun 2001
TL;DR: On-line reactive impurity estimation is combined with batch reactor optimal control to form a novel re-optimisation control strategy and this approach is illustrated on the optimisation control of a simulated batch MMA polymerisation process.
Abstract: A hybrid recurrent neural network model based on-line re-optimisation control strategy is developed for batch polymerisation reactors. The hybrid model contains a simplified mechanistic model covering material balance and simplified reaction kinetics only and recurrent neural networks. Based on this hybrid neural network model, optimal control policy can be calculated. A difficulty in the optimal control of batch polymerisation reactors is that optimisation effort can be seriously hampered by unknown disturbances such as reactive impurities and reactor fouling. A technique for on-line estimation of reactive impurity and reactor fouling has been developed by Zhang et al. (1999). In this contribution, on-line reactive impurity estimation is combined with batch reactor optimal control to form a novel re-optimisation control strategy. When there exists an unknown amount of reactive impurities, the off-line calculated optimal control profile will be no longer optimal. On-line impurity estimation is applied to estimate the amount of reactive impurities during the early stage of the batch. Based on the estimated amount of reactive impurities, on-line re-optimisation is applied to calculate the optimal reactor temperature profile for the remaining time period of the batch reactor operation. This approach is illustrated on the optimisation control of a simulated batch MMA polymerisation process.

Proceedings ArticleDOI
15 Jul 2001
TL;DR: A neural network model that deals with both geometric and functional variables, which have been shown to play an important role in the comprehension of spatial prepositions, is described for the study of spatial language.
Abstract: Describes a neural network model for the study of spatial language. It deals with both geometric and functional variables, which have been shown to play an important role in the comprehension of spatial prepositions. The network is integrated with a virtual reality interface for the direct manipulation of geometric and functional factors. The training uses experimental stimuli and data. Results show that the networks reach low training and generalization errors. Cluster analyses of hidden activation show that stimuli primarily group according to extra-geometrical variables.

Journal ArticleDOI
01 Nov 2001
TL;DR: A hybrid neural network with supervised learning is used to extract the contact event from the measured parameters, which presents a very important step towards the full control of grinding operations in a computer numerically controlled (CNC) environment.
Abstract: Eliminating the gap between the grinding wheel and workpiece is a major time-wasting step during grinding operations. Reducing the time of this step has attracted many investigations. Several researchers have investigated the variation in some process parameters during the wheel-workpiece contact. These parameters include grinding force, grinding power and acoustic emission. During the approach of the grinding wheel to the workpiece, there are three primary stages which have an effect on these parameters: hydrodynamic stage, grit contact stage and wheel contact stage. A few researchers introduced a method to identify the start of the wheel contact stage, which is the practical contact stage. Most of these methods depend on predefined threshold values for some measured parameters. This paper introduces a new methodology to identify the wheel-workpiece contact event in surface grinding operations. A hybrid neural network with supervised learning is used to extract the contact event from the measured parameters. It consists of two neural nets. The first is a self-organizing map neural network with unsupervised learning and the second is a feedforward neural network with supervised learning. Using this hybrid network produces first self-organized clusters for the input data at the first network and then the second network recognizes these clusters. This results in the detection and classification of the contact events automatically from the measured data. This presents a very important step towards the full control of grinding operations in a computer numerically controlled (CNC) environment.

Proceedings ArticleDOI
25 Jul 2001
TL;DR: This work proposed a hybrid system that uses an initial computed kinematics move followed by a visual servoing correction, thereby providing a compromise between speed and accuracy.
Abstract: There are two primary methods for mapping an input image to robot motion: computed kinematics and visual servoing. Computed kinematics uses a kinematic transform between the image plane and the world frame. Computed kinematics algorithms require only a single iteration, but are sensitive to calibration errors. Visual servoing uses a control law to regulate the image to a desired state. Visual servoing is more robust, but requires more computation to reach a solution. To balance these opposing factors, we proposed a hybrid system that uses an initial computed kinematics move followed by a visual servoing correction, thereby providing a compromise between speed and accuracy. A linear approximation model and a neural network were used to approximate the kinematic transform between the image and world frames. A PD control system is used to regulate the image to its final state.

Proceedings ArticleDOI
20 Sep 2001
TL;DR: A hybrid neural network system for the recognition of handwritten character using SOFM,BP and Fuzzy network is presented and the recognition rate is improved visibly.
Abstract: A hybrid neural network system for the recognition of handwritten character using SOFM,BP and Fuzzy network is presented. The horizontal and vertical project of preprocessed character and 4_directional edge project are used as feature vectors. In order to improve the recognition effect, the GAT algorithm is applied. Through the hybrid neural network system, the recognition rate is improved visibly.

Proceedings ArticleDOI
25 Oct 2001
TL;DR: A hybrid neural network used to perform visual search classification to classify the various human visual search patterns into predetermined classes, which signify the different search strategies used by individuals to scan the same target pattern.
Abstract: Visual search describes the process of how the eyes move in a visual field in order to acquire a target. Visual search needs to be quantified to improve future search strategies. This paper describes a hybrid neural network used to perform visual search classification. The neural network consists of a Learning vector quantization network (LVQ) and a single layer perceptron. The objective of this neural network is to classify the various human visual search patterns into predetermined classes. The classes signify the different search strategies used by individuals to scan the same target pattern. The input search patterns are quantified with respect to an ideal search pattern, determined by the user. A supervised learning rule, Learning vector quantization1 (lvq1) is used to train the network.

Proceedings ArticleDOI
20 Aug 2001
TL;DR: The paper focuses on the aspects of successful use of learning methods and human expert interactivity in analyzing unstructured data coming from industrial application.
Abstract: Performing the diagnosis of technical plants results in most of the cases in analyzing huge amounts of unstructured sensor data. If additionally the gathered sensor measurements are noisy or partial faulty and the knowledge about the underlying system or plant is incomplete, than adaptive, learning methods are required in order to interpret the measurements automatically. This paper gives an overview about our diagnosis tool. Two classification kernels, the one based on hybrid, neural network and the other on support vector machines are compared. The paper focuses on the aspects of successful use of learning methods and human expert interactivity in analyzing unstructured data coming from industrial application.

Proceedings ArticleDOI
14 Aug 2001
TL;DR: The hybrid neural network controller was applied to a nonlinear process involving controlling the position of a bouncing ball over a rough moving surface, a typical example of an uncertain model subjected to various types of noises.
Abstract: Presents a new, simple, and versatile neural network controller paradigm, which applies a hybrid learning approach. The major advantage of this controller is that the network learning process is faster. The controller first applies a nonlinear neuron computation scheme to make functional association of input signals to an internal representation in frequency domain. It then maps the internal representation to desired patterns as the output of the controller. Unsupervised learning is conducted for tuning the synaptic weights from the input layer to the internal layer. Supervised learning is employed to tune the synaptic weights for output pattern matching. The hybrid neural controller is especially capable of handling highly noise-corrupted signals in many real-world control applications, such as real-time robot motion planning and control. The hybrid neural network controller was applied to a nonlinear process involving controlling the position of a bouncing ball over a rough moving surface. This system is a typical example of an uncertain model subjected to various types of noises. The simulation was done in a MATLAB environment.