Ordinary Neural Network Research Articles

Pruning neural network for architecture optimization applied to near-infrared reflectance spectroscopic measurements. Determination of the nitrogen content in wheat leaves

The pruning neural network, based on the algorithm called optimum brain surgeon, was used for network architecture optimization. This network pruning procedure was applied for estimating the nitrogen contents in wheat leaves, using near-infrared diffuse reflectance spectroscopy. The results obtained with pruning were compared with those obtained by using ordinary procedures with neural networks, partial least squares, polynomial partial least squares and neural networks/partial least squares methodologies. Comparison of the results with those obtained by the conventional Kjeldahl method showed that the results with pruning neural networks were as good as those with ordinary neural networks and with PLS/neural networks, but better than those with the other methodologies. Although the comparison was performed for one data set, the pruning procedure has the advantage of introducing an automatic architecture optimization, which is cumbersome when performed by the other neural network procedures used in this work, generating a simplified model with better generalization abilities.

Analyses of Possible Failure Mechanisms and Root Failure Causes in Power Plant Components Using Neural Networks and Structural Failure Database

This paper describes analyses of case studies on failure of structural components in power plants using hierarchical (multilayer) neural networks. Using selected test data about case studies stored in the structural failure database of a knowledge-based system, the network is trained: either to predict possible failure mechanisms like creep, overheating (OH), or overstressing (OS)-induced failure (network of Type A), or to classify a root failure cause of each case study into either a primary or secondary cause (network of Type B). In the present study, the primary root cause is defined as “manufacturing, material or design-induced causes,” while the secondary one as “not manufacturing, material or design-induced causes, e.g., failures due to operation or mal-operation.” An ordinary three-layer neural network employing the back propagation algorithm with the momentum method is utilized in this study. The results clearly show that the neural network is a powerful tool for analyzing case studies of failure in structural components. For example, the trained network of Type A predicts creep-induced failure in unknown case studies with an accuracy of 86 percent, while the network of Type B classifies root failure causes of unknown case studies with an accuracy of 88 percent. It should be noted that, due to a shortage of available case studies, an appropriate selection of case studies and input parameters to be used for network training was necessary in order to attain high accuracy. A collection of more case studies should, however, resolve this problem, and improve the accuracy of the analyses. An analysis module for case studies using the neural network has also been developed and successfully implemented in a knowledge-based system.

HANDWRITTEN DIGIT RECOGNITION USING TWO-LAYER SELF-ORGANIZING MAPS

In this paper, we present a two-layer self-organizing neural network based method for handwritten digit recognition. The network consists of a base layer self-organizing map and a set of corresponding maps in the second layer. The input patterns are partitioned into subspace in the first layer. Patterns in a subspace are led to the second layer and a corresponding map is built according to the first layer performance. In the classification process, each pattern searches for several closest nodes from the base map and then it is classified into a specified class by determining the nearest model of the corresponding maps in the second layer. The new method yielded higher accuracy and faster performance than the ordinary self-organizing neural network.

Fault location for transmission lines using inference model neural network

The authors propose a new fault location method that uses a neural network to analyze the distribution pattern of the ground wire current along the power line. This method is able to locate the fault section even for secondary power lines with complicated configurations. The method is based on the inference process, i.e., human experts will analyze the distribution pattern of the current amplitude and phase angle. In locating fault sections, higher precision than ordinary three-layer neural network or the expert system of previous development can be obtained. The proposed neural network comprises three sets of three-layer neural networks which follow the back-propagation learning procedure. The 1st and 2nd neural networks calculate the candidate-1 and candidate-2 for the fault section using current amplitude and phase angle distribution patterns, respectively. The 3rd neural network then performs final fault location using these candidates and a current amplitude distribution pattern. The results evaluated with all possible fault causes indicate that the new method is precise to as high as 98.4 percent even when the measured values differ by 30 percent from predicted ones with EMTP.