9 Matching Results

Search Results

Advanced search parameters have been applied.

Alternative learning algorithms for feedforward neural networks

Description: The efficiency of the back propagation algorithm to train feed forward multilayer neural networks has originated the erroneous belief among many neural networks users, that this is the only possible way to obtain the gradient of the error in this type of networks. The purpose of this paper is to show how alternative algorithms can be obtained within the framework of ordered partial derivatives. Two alternative forward-propagating algorithms are derived in this work which are mathematically equivalent to the BP algorithm. This systematic way of obtaining learning algorithms illustrated with this particular type of neural networks can also be used with other types such as recurrent neural networks.
Date: March 1996
Creator: Vitela, J. E.
Partner: UNT Libraries Government Documents Department

Premature saturation in backpropagation networks: Mechanism and necessary conditions

Description: The mechanism that gives rise to the phenomenon of premature saturation of the output units of feedforward multilayer neural networks during training with the standard backpropagation algorithm is described. The entire process of premature saturation is characterized by three distinct stages and it is concluded that the momentum term plays the leading role in the occurrence of the phenomenon. The necessary conditions for the occurrence of premature saturation are presented and their validity is illustrated through simulation results.
Date: December 31, 1995
Creator: Vitela, J.E. & Reifman, J.
Partner: UNT Libraries Government Documents Department

Recurrent neural networks for NO{sub x} prediction in fossil plants

Description: The authors discuss the application of recurrent (dynamic) neural networks for time-dependent modeling of NO{sub x} emissions in coal-fired fossil plants. They use plant data from one of ComEd`s plants to train and test the network model. Additional tests, parametric studies, and sensitivity analyses are performed to determine if the dynamic behavior of the model matches the expected behavior of the physical system. The results are also compared with feedforward (static) neural network models trained to represent temporal information.
Date: April 1, 1996
Creator: Reifman, J.; Vitela, J.E.; Feldman, E.E. & Wei, T.Y.C.
Partner: UNT Libraries Government Documents Department

PRODIAG: Combined expert system/neural network for process fault diagnosis. Volume 1, Theory

Description: The function of the PRODIAG code is to diagnose on-line the root cause of a thermal-hydraulic (T-H) system transient with trace back to the identification of the malfunctioning component using the T-H instrumentation signals exclusively. The code methodology is based on the Al techniques of automated reasoning/expert systems (ES) and artificial neural networks (ANN). The research and development objective is to develop a generic code methodology which would be plant- and T-H-system-independent. For the ES part the only plant or T-H system specific code requirements would be implemented through input only and at that only through a Piping and Instrumentation Diagram (PID) database. For the ANN part the only plant or T-H system specific code requirements would be through the ANN training data for normal component characteristics and the same PID database information. PRODIAG would, therefore, be generic and portable from T-H system to T-H system and from plant to plant without requiring any code-related modifications except for the PID database and the ANN training with the normal component characteristics. This would give PRODIAG the generic feature which numerical simulation plant codes such as TRAC or RELAP5 have. As the code is applied to different plants and different T-H systems, only the connectivity information, the operating conditions and the normal component characteristics are changed, and the changes are made entirely through input. Verification and validation of PRODIAG would, be T-H system independent and would be performed only ``once``.
Date: September 1, 1995
Creator: Reifman, J.; Wei, T.Y.C. & Vitela, J.E.
Partner: UNT Libraries Government Documents Department

Premature saturation in backpropagation networks: Mechanism and necessary conditions

Description: The mechanism that gives rise to the phenomenon of premature saturation of the output units of feedforward multilayer neural networks during training with the standard backpropagation algorithm is described. The entire process of premature saturation is characterized by three distinct stages and it is concluded that the momentum term plays the leading role in the occurrence of the phenomenon. The necessary conditions for the occurrence of premature saturation are presented and a new method is proposed, based on these conditions, that eliminates the occurrence of the phenomenon. Validity of the conditions and the proposed method are illustrated through simulation results. Three case studies are presented. The first two came from a training session for classification of three component failures in a nuclear power plant. The last case, comes from a training session for classification of welded fuel elements.
Date: August 1, 1997
Creator: Vitela, J.E. & Reifman, J.
Partner: UNT Libraries Government Documents Department

An artificial neural network controller for intelligent transportation systems applications

Description: An Autonomous Intelligent Cruise Control (AICC) has been designed using a feedforward artificial neural network, as an example for utilizing artificial neural networks for nonlinear control problems arising in intelligent transportation systems applications. The AICC is based on a simple nonlinear model of the vehicle dynamics. A Neural Network Controller (NNC) code developed at Argonne National Laboratory to control discrete dynamical systems was used for this purpose. In order to test the NNC, an AICC-simulator containing graphical displays was developed for a system of two vehicles driving in a single lane. Two simulation cases are shown, one involving a lead vehicle with constant velocity and the other a lead vehicle with varying acceleration. More realistic vehicle dynamic models will be considered in future work.
Date: April 1, 1996
Creator: Vitela, J.E.; Hanebutte, U.R. & Reifman, J.
Partner: UNT Libraries Government Documents Department

Automatic inspection for remotely manufactured fuel elements

Description: Two classification techniques, standard control charts and artificial neural networks, are studied as a means for automating the visual inspection of the welding of end plugs onto the top of remotely manufactured reprocessed nuclear fuel element jackets. Classificatory data are obtained through measurements performed on pre- and post-weld images captured with a remote camera and processed by an off-the-shelf vision system. The two classification methods are applied in the classification of 167 dummy stainless steel (HT9) fuel jackets yielding comparable results.
Date: June 1, 1995
Creator: Reifman, J.; Vitela, J.E.; Gibbs, K.S. & Benedict, R.W.
Partner: UNT Libraries Government Documents Department

A parallel neural network training algorithm for control of discrete dynamical systems.

Description: In this work we present a parallel neural network controller training code, that uses MPI, a portable message passing environment. A comprehensive performance analysis is reported which compares results of a performance model with actual measurements. The analysis is made for three different load assignment schemes: block distribution, strip mining and a sliding average bin packing (best-fit) algorithm. Such analysis is crucial since optimal load balance can not be achieved because the work load information is not available a priori. The speedup results obtained with the above schemes are compared with those corresponding to the bin packing load balance scheme with perfect load prediction based on a priori knowledge of the computing effort. Two multiprocessor platforms: a SGI/Cray Origin 2000 and a IBM SP have been utilized for this study. It is shown that for the best load balance scheme a parallel efficiency of over 50% for the entire computation is achieved by 17 processors of either parallel computers.
Date: January 20, 1998
Creator: Gordillo, J. L.; Hanebutte, U. R. & Vitela, J. E.
Partner: UNT Libraries Government Documents Department

Prodiag--a hybrid artificial intelligence based reactor diagnostic system for process faults

Description: Commonwealth Research Corporation (CRC) and Argonne National Laboratory (ANL) are collaborating on a DOE-sponsored Cooperative Research and Development Agreement (CRADA), project to perform feasibility studies on a novel approach to Artificial Intelligence (Al) based diagnostics for component faults in nuclear power plants. Investigations are being performed in the construction of a first-principles physics-based plant level process diagnostic expert system (ES) and the identification of component-level fault patterns through operating component characteristics using artificial neural networks (ANNs). The purpose of the proof-of-concept project is to develop a computer-based system using this Al approach to assist process plant operators during off-normal plant conditions. The proposed computer-based system will use thermal hydraulic (T-H) signals complemented by other non-T-H signals available in the data stream to provide the process operator with the component which most likely caused the observed process disturbance.To demonstrate the scale-up feasibility of the proposed diagnostic system it is being developed for use with the Chemical Volume Control System (CVCS) of a nuclear power plant. A full-scope operator training simulator representing the Commonwealth Edison Braidwood nuclear power plant is being used both as the source of development data and as the means to evaluate the advantages of the proposed diagnostic system. This is an ongoing multi-year project and this paper presents the results to date of the CRADA phase.
Date: March 1, 1996
Creator: Reifman, J.; Wei, T.Y.C.; Vitela, J.E.; Applequist, C. A. & Chasensky, T.M.
Partner: UNT Libraries Government Documents Department