10,806 research outputs found
COMBINING MULTIPLE NEURAL NETWORKS BY FUZZY INTEGRAL FOR ROBUST CLASSIFICATION
No full textRecently, in the area of artificial neural networks, the concept of combining multiple networks has been proposed as a new direction for the development of highly reliable neural network systems. In this paper we propose a method for multinetwork combination based on the fuzzy integral. This technique nonlinearly combines objective evidence, in the form of a fuzzy membership function, with subjective evaluation of the worth of the individual neural networks with respect to the decision. The experimental results with the recognition problem of on-line handwriting characters confirm the superiority of the presented method to the other voting techniques
RECOGNITION OF LARGE-SET PRINTED HANGUL (KOREAN SCRIPT) BY 2-STAGE BACKPROPAGATION NEURAL CLASSIFIER
No full textA two-stage neural network classifier is described which practically recognizes printed Hangul (Korean script). This classifier is composed of a type classification network and six recognition networks. The former classifies input character images into one of the six types by their overall structure, and then the latter classify them into character code. Furthermore, a training scheme including systematic noises is introduced for improving the generalization capability of the networks. Experiments are conducted with the most frequently used 990 printed Hangul syllables. By the noise included training, the recognition rate amounts to 98.28%, which is better than that of the conventional backpropagation learning. A comparison with a statistical classifier and an analysis of generalization capability confirm the relative superiority of the proposed classification method
Instantaneous reference-free crack detection based on polarization characteristics of piezoelectric materials
No full textA new methodology of guided-wave-based nondestructive testing (NDT) is developed to detect crack damage in a thin metal structure without using prior baseline data or a predetermined decision boundary. In conventional guided-wave-based techniques, damage is often identified by comparing the 'current' data obtained from a potentially damaged condition of a structure with the 'past' baseline data collected at the pristine condition of the structure. However, it has been reported that this type of pattern comparison with the baseline data can lead to increased false alarms due to its susceptibility to varying operational and environmental conditions of the structure. To develop a more robust damage diagnosis technique, a new concept of NDT is conceived so that cracks can be detected even when the system being monitored is subjected to changing operational and environmental conditions. The proposed NDT technique utilizes the polarization characteristics of the piezoelectric wafers attached on both sides of the thin metal structure. Crack formation creates Lamb wave mode conversion due to a sudden change in the thickness of the structure. Then, the proposed technique instantly detects the appearance of the crack by extracting this mode conversion from the measured Lamb waves, and the threshold value from damage classification is also obtained only from the current dataset. Numerical and experimental results are presented to demonstrate the applicability of the proposed technique to instantaneous crack detection.The authors acknowledge the support from the Korea Science
and Engineering Foundation (M20703000015-07N0300-
01510), the Korea Research Foundation (D00462, D0032), and
the US National Science Foundation (CMMI-0529208)
Synthesis and electrode properties of alpha-Fe2O3 from iron phthalocyanine
No full textThe structural evolution of Fe2O3 fine powder synthesized by thermal decomposition of iron phthalocyanine was investigated by X-ray diffraction and Raman spectroscopy. X-ray results indicate that Fe2O3 evolves above 300 degrees C. Raman spectroscopy reveals that it belongs to alpha-Fe2O3. Electrochemical measurements show that only the sample fired at 320 degrees C is of potential interest as an anode in lithium-ion cells, since it displays a recharge capacity of 570 mAh/g in the range of 0.5 to 2.5 V vs. Li. Although the capacity declines during the initial cycles, it becomes stable after ten cycles. (C) 1999 The Electrochemical Society. S1099- 0062( 98) 07- 098- 9. All rights reserved.open1115sciescopu
AN HMM/MLP ARCHITECTURE FOR SEQUENCE RECOGNITION
No full textThis paper presents a hybrid architecture of hidden Markov models (HMMs) and a multilayer perceptron (MLP). This exploits the discriminative capability of a neural network classifier while using HMM formalism to capture the dynamics of input patterns. The main purpose is to improve the discriminative power of the HMM-based recognizer by additionally classifying the likelihood values inside them with an MLP classifier. To appreciate the performance of the presented method, we apply it to the recognition problem of on-line handwritten characters. Simulations show that the proposed architecture leads to a significant improvement in generalization performance over conventional approaches to sequential pattern recognition
RAPID BACKPROPAGATION LEARNING ALGORITHMS
No full textOne of the major drawbacks of the backpropagation algorithm is its slow rate of convergence. Researchers have tried several different approaches to speed up the convergence of backpropagation learning. In this paper, we present those rapid learning methods as three categories, and implement the representative methods of each category: (1) for the numerical method based approach, the Aitken's DELTA2 process, (2) for the heuristics based approach, the dynamic adaptation of learning rate, and (3) for the learning strategy based approach, the selective presentation of learning samples. Based on these implementations, the performance is evaluated with experiments and the merits and demerits are briefly discussed
Improvement of electrochemical properties of LiNi0.5Mn1.5O4 spinel
No full textLiNi0.5Mn1.5O4 material with a spinel structure is prepared by a sol-gel method. The material is initially fired at 850 T and then subjected to a post-reaction annealing at 600 T in order to minimize the nickel deficiency. The elevated firing temperature produces materials with a small surface-area which is beneficial for good capacity retention. Indeed, the spinel LiNi0.5Mn1.5O4 not only shows a good cycle performance, but exhibits an excellent discharge capacity, i.e. 114 mAh g(-1) at 4.66 V plateau and 127 mAh g(-1) in total. Cyclic voltammetry and ac impedance spectroscopy are employed to characterize the reactions of lithium insertion and extraction in the LiNi0.5Mn1.5O4 electrode. Excellent electrochemical performance and low material cost make this compound an attractive cathode for advanced lithium batteries. (C) 2002 Elsevier Science B.V. All rights reserved.X1181sciescopu
Production of 3,6-Anhydro-D-galactose from kappa-Carrageenan Using Acid Catalysts
No full textThis study describes acid-catalyzed production of 3,6-anhydro-D-galactose (D-AnG) from kappa-carrageenan, a sulfated polysaccharide with an alternating backbone consisting of D-AnG and D-galactose (D-Gal). We analyzed four hydrolysis products (D-AnG, 5-hydroxymethylfurfural (HMF), levulinic acid (LA), and D-Gal) and reducing sugar contents during acid hydrolysis. Acid screening was carried out using seven acid catalysts which have different acidity. The catalysts showing high D-AnG production and high selectivity were chosen for subsequent experiments. We selected four acid catalysts (HCOOH, CH3COOH, HNO3, and HCl), and studied the effects of catalyst acidity, hydrolysis temperature T, and reaction time t on the production of D-AnG and other hydrolysis products. The optimal condition for maximum production of D-AnG by kappa-carrageenan hydrolysis was T = 100A degrees C and t = 30 min using 0.2 M HCl. Under this condition, 2.81 g/L D-AnG (33.5% of theoretical maximum) could be obtained from 2% (w/v) kappa-carrageenan. In general, the maximum values of D-AnG, D-Gal, and the sum of two by-products (HMF and LA) increased with the acidity of catalysts. However, HNO3 was an exception in that the maximum production levels of HMF and LA were unusually low compared with other acid catalysts. D-AnG was successfully purified from acid hydrolysates using silica gel chromatography and the product was nearly 100% pure. This effective D-AnG production could facilitate future studies on the conversion of D-AnG to biofuels and biochemicals.11Nsciescopuskc
MULTIPLE NETWORK FUSION USING FUZZY-LOGIC
No full textMultiplayer feedforward networks trained by minimizing the mean squared error and by using a one of c teaching function yield network outputs that estimate posterior class probabilities. This provides a sound basis for combining the results from multiple networks to get more accurate classification. This paper presents a method for combining multiple networks based on fuzzy logic, especially the fuzzy integral. This method non-linearly combines objective evidence, in the form of a network output, with subjective evaluation of the importance of the individual neural networks. The experimental results with the recognition problem of on-line handwriting characters show that the performance of individual networks could be improved significantly
Continuous fatigue crack monitoring without baseline data
No full textIn order to overcome the susceptibility of conventional non-destructive testing (NDT) techniques to operational and environmental variations, a new damage detection technique that does not require direct comparison with baseline data was previously developed by the authors for detecting a crack in a plate structure. This reference-free technique employs two pairs of collocated lead zirconate titanate transducers (PZTs) placed on both sides of the plate to generate and measure Lamb waves. Then, the existence of mode conversion due to the crack is identified from the Lamb wave signals instantly measured by PZTs. In this study, the effectiveness of the proposed technique is tested using a steel girder specimen. A cyclic loading is applied to the girder resulting in fatigue cracks, and the proposed technique detects the appearance of fatigue damage solely based on the measured Lamb waves at the present stage. Experimental results are presented to demonstrate the applicability of the proposed technique to fatigue crack monitoring, and issues related to PZT installation are discussed.This research is supported by the Radiation Technology
Program under Korea Science and Engineering Foundation
(KOSEF) and the Ministry of Science and Technology
(M20703000015-07N0300-01510) andKorea Research
Foundation Grant funded by the Korean Government
(MOEHRD, Basic Research Promotion Fund)
(KRF-2007-331-D00462). Any opinions, findings and
conclusions or recommendations expressed in this material
are those of the author(s) and do not necessarily reflect
the views of the funding agencies. The authors also
would like to thank Professors Kent A. Harris and Piervincenzo
Rizzo at University of Pittsburgh for designing
the steel girder specimen and conducting fatigue loading
tests
- …
