Browsing by Issue Date, starting with "1995-03-20"

Now showing 1 - 20 of 87015

A NEURAL FUZZY CONTROL-SYSTEM WITH STRUCTURE AND PARAMETER LEARNING
(1995-03-20)
LIN, CT
;
交大名義發表
;
電控工程研究所
;
National Chiao Tung University
;
Institute of Electrical and Control Engineering
A general connectionist model, called neural fuzzy control network (NFCN), is proposed for the realization of a fuzzy logic control system. The proposed NFCN is a feedforward multilayered network which integrates the basic elements and functions of a traditional fuzzy logic controller into a connectionist structure which has distributed learning abilities. The NFCN can be constructed from supervised training examples by machine learning techniques, and the connectionist structure can be trained to develop fuzzy logic rules and find membership functions. Associated with the NFCN is a two-phase hybrid learning algorithm which utilizes unsupervised learning schemes for structure learning and the backpropagation learning scheme for parameter learning. By combining both unsupervised and supervised learning schemes, the learning speed converges much faster than the original backpropagation algorithm. The two-phase hybrid learning algorithm requires exact supervised training data for learning. In some real-time applications, exact training data may be expensive or even impossible to obtain. To solve this problem, a reinforcement neural fuzzy control network (RNFCN) is further proposed. The RNFCN is constructed by integrating two NFCNs, one functioning as a fuzzy predictor and the other as a fuzzy controller. By combining a proposed on-line supervised structure-parameter learning technique, the temporal difference prediction method, and the stochastic exploratory algorithm, a reinforcement learning algorithm is proposed, which can construct a RNFCN automatically and dynamically through a reward-penalty signal (i.e., ''good' or ''bad'' signal), Two examples are presented to illustrate the performance and applicability of the proposed models and learning algorithms.
KINETIC-STUDY ON SIMULTANEOUS INTERPENETRATING POLYMER NETWORK FORMATION OF EPOXY-RESIN AND UNSATURATED POLYESTER
(1995-03-21)
LIN, MS
;
CHANG, RJ
;
YANG, T
;
SHIH, YF
;
應用化學系
Simultaneous interpenetrating polymer networks (SIN) based on diglycidyl ether of bisphenol-A (DGEBA) and unsaturated polyester (UP) in weight ratios of 100/0, 50/50, and 0/100 were blended and cured simultaneously by using cumene hydroperoxide (CHP) and m-xylenediamine (MXDA) as curing agents. A kinetic study during SIN formation was carried out at 43, 53, 58, and 64 degrees C. Concentration changes for both epoxide and C=C bond were monitored with Fourier transform infrared (FTIR). Rate expressions for epoxide polymerization were established with model reactions for phenyl glycidyl ether (PGE) and benzylamine in dichloroethane containing benzyl alcohol. Experimental results revealed that a lower cure rate constant for the C=C bond during SIN formation was found, compared with pure UP network formation. A model of network interlock is proposed to account for this phenomenon. During simultaneous cure of DGEBA and UP, the network interlock should provide a sterically hindered environment, which subsequently retards cure rate for UP. On the other hand, epoxide cure during SIN formation indicates higher rate constants for both uncatalyzed and catalyzed reactions, compared with those of pure DGEBA cure. Presumably the catalytic effect of hydroxyl end groups in UP overcomes the effect of network interlock. Kinetic parameters were calculated and are discussed. (C) 1995 John Wiley and Sons, Inc.
的法國梧桐不同
(1995-03-24)
管管
家中養一隻小老虎
(1995-03-24)
管管
家中養一隻小老虎
(1995-03-24)
管管
RECESSED-GATE ALGAAS/INGAAS/GAAS PSEUDORMORPHIC HEMT WITH SI-PLANAR-DOPED ETCH-STOP LAYER
(1995-03-30)
LAN, WH
;
LIN, WJ
;
PENG, CK
;
CHEN, SS
;
TU, SL
An improved slot etch technique based on an Si planar doped laver has been applied to gate recessing in the fabrication of AlGaAs/InGaAs/GaAs pseudomorphic high electron mobility transistors (HEMTs). The devices exhibited comparable g(m) with much better breakdown and leakage behaviour than conventional pseudomorphic HEMT devices.
THE EFFECTS OF ND2O3 ADDITIVES AND AL2O3-SIO2-TIO2 SINTERING AIDS ON THE ELECTRICAL-RESISTIVITY OF (BA,SR)TIO3 PTCR CERAMICS
(1995-04-01)
LAI, CH
;
WENG, CT
;
TSENG, TY
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
The influences of Nd donors and Al2O3-SiO2-TiO2 (AST) sintering aids on the electrical properties of (Ba,Sr)TiO3 materials have been investigated. The room-temperature resistivity and the temperature coefficient of the anomalous resistivity rise were used to characterize the performance of the samples. From derivations based on the Heywang-Jonker barrier model, both characterizing parameters are expected to increase as a result of higher acceptor-state density at the grain boundary. The surface acceptor density, whose value was extracted from the slope in the Arrhenius plot of resistivity versus 1/(T epsilon(m)), where epsilon(m) is the measured permittivity and T the absolute temperature, was found to decrease with the Nd content and increase with the AST dopant. A satisfactory interpretation of the observed variations in resistivity-temperature curves caused by additions of various dopants was thus obtained in the light of the resultant acceptor state density.
OPTIMAL ASSIGNMENT OF ROBOT TASKS WITH PRECEDENCE FOR MULTIROBOT COORDINATION BY DISJUNCTIVE GRAPHS AND STATE-SPACE SEARCH
(1995-04-01)
LIN, CF
;
TSAI, WH
;
資訊工程學系
;
Department of Computer Science
An approach to optimal assignment of tasks with precedence relationships to multiple robots is proposed. The robots are assumed to share a common workspace and work cooperatively to accomplish a given process plan consisting of a set of tasks. The optimal task assignment is defined to be the one that results in spending the least amount of time to complete the plan under the criterion that no robot collision will occur when the assigned tasks are performed. The ordering of the tasks in the process plan is described by a topological tree, which is then expanded to form a larger state-space tree without redundant tree paths. Each path in the expanded tree represents a partially developed assignment of the tasks to the robots, and a graph formulation scheme is presented for estimating the cost of the assignment. A collision-free motion schedule for each robot based on each task assignment can be obtained by finding the minimaximal path in a disjunctive graph formulated by the scheme. By using the A* algorithm, a search method for finding the optimal assignment with the minimum cost is presented. Some heuristic rules are also proposed to speed up the search process. Simulation results are illustrated to show the effectiveness of the proposed approach. (C) 1995 John Wiley & Sons, Inc.
A NEW GATE CURRENT SIMULATION TECHNIQUE CONSIDERING SI/SIO2 INTERFACE-TRAP GENERATION
(1995-04-01)
WEN, KS
;
LI, HH
;
WU, CY
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
An efficient and accurate 2D numerical simulation technique is developed to study the hot-electron effects on short-channel n-MOSFETs. The 1D substrate injection probability used by Ning el al, has been further modified by considering the channel hot-electron-enhanced injection probability. Moreover, the hot-electron-injection-induced Si/SiO2 interface-trap generation and its effects on MOSFET drain, substrate and gate currents have also been taken into consideration. It is shown that the generated electron traps at the Si/SiO2 interface enhance both the impact ionization rate and the degradation of MOSFET characteristics but retard the injection probability of hot electrons into the gate oxide. The spatial distribution of the generated Si/SiO2 interface traps calculated by our model has been well verified by the charge pumping measurement. In addition, the simulated substrate current, gate current, and degradation of drain current are in good agreement with the experimental results of a short-channel n-MOSFET with the oxide thickness of 100 Angstrom and the effective channel length of 0.45 mu m for wide ranges of drain and gate biases.
CATALYTIC INTERFACE EROSION
(1995-04-01)
MENG, HF
;
COHEN, EGD
;
物理研究所
;
Institute of Physics
SOME NEW DESIGNS OF 2-D ARRAY FOR MATRIX MULTIPLICATION AND TRANSITIVE CLOSURE
(1995-04-01)
TSAY, JC
;
CHANG, PY
;
交大名義發表
;
工學院
;
National Chiao Tung University
In this paper, we present some new regular iterative algorithms for matrix multiplication and transitive closure. With these algorithms, by spacetime mapping the 2-D arrays with 2N-1 and [(3N-1)/2] execution times for matrix multiplication can be obtained, Meanwhile, we can derive a 2-D array with 4N-2 execution time for transitive closure based on the sequential Warshall-Floyd algorithm. All these new 2-D arrays for matrix multiplication and transitive closure have the advantages of faster and more regular than other previous designs.
DESIGN AND IMPLEMENTATION OF MOONY - A FAULT-TOLERANT DISTRIBUTED SHARED-MEMORY SYSTEM
(1995-04-01)
LIN, YW
;
YUAN, SM
;
LIANG, DR
;
資訊工程學系
;
Department of Computer Science
Shared memory is a natural high level programming paradigm in particular in distributed computing environments. In this paper, we describe the design and implementation of a fault-tolerant distributed shared memory system (FTDSM) called Moony,which provides shaved memory abstraction and guarantees continuous services even in the presence of site crashes. Moony is implemented at application level with a set of daemon sewers and provides services to users via Moony library function calls. The first prototype is portable to all Unix machines that support BSD socket and System V IPC. The preliminary results demonstrate that FTDSM is viable and, that Moony is portable over a wide variety of distributed environments.
ELECTROOPTIC CHARACTERISTICS OF AMINO-GEL-GLASS-DISPERSED LIQUID-CRYSTAL AND ITS MATRIX FORMATION
(1995-04-01)
CHANG, WP
;
WHANG, WT
;
WONG, JC
;
材料科學與工程學系
;
Department of Materials Science and Engineering
The gel-glass-dispersed liquid crystals (GDLCs) from amino-silicon alkoxides have successfully been prepared in a short time by sol-gel process. Those GDLCs had the same operation mechanism and performances as polymer-dispersed liquid crystal(PDLC). The electrooptic characteristic of the GDLC strongly depended on the composition of raw materials and the concentrations of liquid crystal. Titanium ethoxide was used successfully to adjust the refractive index of the gel-glass matrix, but it caused the gel-glass to be less flexible and the GDLC operated at a higher voltage. GDLC-13 had a reasonable operation voltage of 13 V, a good on-state transmittance of 100%, and a good contrast of 16.2. The introducing of a two-functional amino silane, (3-[diethoxymethylsilyl] propylamine), in the composition of matrix raw materials offered the gel-glass matrix better flexibility, higher conductivity, and greater capacity for trapping liquid crystal. It resulted in a significant improvement of the electrooptic characteristic of the GDLC
A MULTIVALUED BOLTZMANN MACHINE
(1995-04-01)
LIN, CT
;
LEE, CSG
;
電控工程研究所
;
Institute of Electrical and Control Engineering
The idea of Hopfield network is based on the Ising spin glass model in which each spin has only two possible states: up and down. By introducing stochastic factors into this network and performing a simulated annealing process on it, it becomes a Boltzmann machine which can escape from local minimum states to achieve the global minimum. This paper generalizes the above ideas to multi-value case based on the XY spin glass-model in which each spin can be in any direction in a plane, Simply using the gradient descent method and the analog Hopfield network; two different analog connectionist structures and their corresponding evolving rules are first designed to transform the XY spin glass model to distributed computational models. These two analog computational models are single-layered connectionist structures and multi-layered Hopfield analog networks. The latter network eases the node (neuron) computational requirement of the former at the expense of more neurons and connections. With the proposed evolving rules, the proposed models evolve according to a predefined Hamiltonian (energy function) which will decrease until it reaches a (perhaps local) minimum. Since these two structures can easily get stuck in local minima, a multi-valued Boltzmann machine is proposed which adopts the discrete planar spin glass model for the local minimum problem. Each neuron in the multi-valued Boltzmann machine can only take n discrete directions (states). The stochastic simulated annealing method is introduced to the evolving rules of the multi-valued Boltzmann machine to solve the local minimum problem. The multi-valued Boltzmann machine can be applied to the mobile robot navigation problem;problem by defining proper artificial magnetic field on the traverse terrain. This new artificial magnetic field approach for the mobile robot navigation problem has shown to have several advantages over existing graph search and potential field techniques.
A CHINESE-CHARACTER THINNING ALGORITHM-BASED ON GLOBAL FEATURES AND CONTOUR INFORMATION
(1995-04-01)
LIN, JY
;
CHEN, Z
;
資訊科學與工程研究所
;
Institute of Computer Science and Engineering
This paper proposes a method of using run-length coding to perform thinning. First, we construct graphs from characters. The attributes (vertical lines, horizontal lines or points) of each node in the graph are determined according to the node's relationship to the nodes above and below it (we will refer to these relationships as global features) and the black runs within the node. Intersections between two adjacent segments are determined on the basis of the graph constructed and contour information. The thinning algorithm thus employs global features and contour information to produce a more accurate skeleton.
HIGH-PERFORMANCE NITRIDED OXIDES FABRICATED BY VERY-LOW-PRESSURE NITRIDATION TECHNIQUE
(1995-04-01)
SU, HP
;
CHENG, HC
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
A never and simple method is proposed to produce high-performance nitrided oxides. Thin oxides nitrided only in NH3 at 900 degrees C for 1 h at 0.1 Torr have excellent interface stability, few electron traps, and excellent reliability. Although the nitrogen concentration [N] at the dielectric/Si substrate interface is kept at a low value, the hydrogen concentration [H] has been shown to be comparable to that of pure oxide at the beginning of the nitridation. Consequently, a high ratio of [N]/[K], which reflects the stable interface, can be attained at the initial stage of the very-low-pressure nitridation (VLPN). This is very different from the case of conventional low-pressure nitridation. The retarded formations of the interfacial Si3N4 and the surface nitrogen-rich oxynitrides are conjectured to be the main cause of reduction in the hydrogen content in the dielectrics during VLPN. Due to the simplicity and wide process window, VLPN is promising for the future ultra-large-scale-integration (ULSI) technology.
INFLUENCE OF PROCESSING PARAMETERS ON THE MICROSTRUCTURE AND ELECTRICAL-PROPERTIES OF MULTILAYER-CHIP ZNO VARISTORS
(1995-04-01)
LEE, YS
;
TSENG, TY
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
Multilayer-chip varistors based on ZnO with lead zinc borosilicate glass instead of Bi2O3 were prepared by tape casting and green-sheet lamination processes using a non-aqueous slurry system. The influences of slurry composition and the degassing process on the microstructure and non-ohmic properties of multilayer-chip varistors were studied. The electrical properties of chip varistors can be influenced substantially by the pore defects resulting from an unsuitable slurry formulation for the tape-casting process. The sintering temperature of the chip varistors was lowered to 1100 degrees C and silver-palladium alloys were employed as internal electrodes. The non-linear coefficients of 27-32 and breakdown voltages of 7.6-24.5 V were achieved by controlling the green-sheet thickness and sintering temperatures in the present study. The Delta V-br/V-br values for the chip varistors lie within +/-10%, indicating excellent surge-withstanding capability.
MECHANISMS OF INTERFACE TRAP-INDUCED DRAIN LEAKAGE CURRENT IN OFF-STATE N-MOSFETS
(1995-04-01)
CHANG, TE
;
HUANG, CM
;
WANG, TH
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
An interface trap-assisted tunneling and thermionic emission model has been developed to study an increased drain leakage current in off-state n-MOSFET's after hot carrier stress. In the model a complete band-trap-band leakage path is formed at the Si/SiO2 interface by hole emission from interface traps to a valence band and electron emission from interface traps to a conduction band. Both hole and electron emissions are carried out via quantum tunneling or thermal excitation. In this experiment, a 0.5 mu m n-MOSFET was subjected to a de voltage stress to generate interface traps. The drain leakage current was characterized to compare with the model. Our study reveals that the interface trap-assisted two-step tunneling, hole tunneling followed by electron tunneling, holds responsible for the leakage current at a large drain-to-gate bias (V-dg) The lateral field plays a major role in the two-step tunneling process. The additional drain leakage current due to band-trap-band tunneling is adequately described by an analytical expression Delta I-d = A exp (B-it/F). The value of Bit about 13 mV/cm was obtained in a stressed MOSFET, which is significantly lower than in the GIDL current attributed to direct band-to-band tunneling. As V-dg decreases, a thermionic-field emission mechanism, hole thermionic emission and electron tunneling, becomes a primary leakage path. At a sufficiently low V-dg, our model reduces to the Shockley-Read-Hall theory and thermal generation of electron-hole pairs through traps is dominant.
ENHANCEMENT OF CRITICAL-CURRENT AND MAGNETIC-FIELD IN A DEFECTED JOSEPHSON TUNNELING JUNCTION
(1995-04-01)
CHOW, TC
;
CHOU, H
;
LAI, HG
;
LIU, CC
;
GOU, YS
A Josephson junction with a length of 10 times the Josephson penetration depth (lambda(J)) and a defect size of 0.5 lambda(J) to 2 lambda(J) in different positions has been studied by a mechanical simulation. It was found that the defect modulated the current distribution tremendously when it was near the edge of the junction. A surprising enhancement of the critical current under the field was observed when the defect was located at 0.5 lambda(J) and had a size from 0.5 lambda(J) to 2 lambda(J), which was much larger than the conventional pinning size: the coherent length xi. This effect could be attributed to a self-field, which was either generated by the current itself or by the applied magnetic field, penetrating into the defect smoothly and being pinned at the defect. A repulsive interaction between the self-field and the external field kept any further flux from abruptly penetrating into the junction. The increment in the critical current under the applied field was up to 125% of the original defect-free one. By fitting two defects of the same size of 2 lambda(J) and positioning them at 0.5 lambda(J) to both edges of the junction, the zero field critical current and the critical magnetic field were enhanced up to 150% of a defect-free junction.
HIGH-PERFORMANCE SUPERTHIN OXIDE/NITRIDE/OXIDE STACKED DIELECTRICS FORMED BY LOW-PRESSURE OXIDATION OF ULTRATHIN NITRIDE
(1995-04-01)
LIU, HW
;
SU, HP
;
CHENG, HC
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
High-performance superthin oxide/nitride/oxide (O/N/O) stacked dielectrics have been successfully achieved by oxidizing ultrathin nitride films in low-pressure dry oxygen at 850 degrees C for 30 min. The low leakage current and the high reliability of this O/N/O structure are attributed to the growth of the top oxide on the nitride, which is confirmed by Auger electron spectroscopy (AES) depth profiles and step-by-step diluted-HF etching results. The commonly used Net oxidation of the nitrides at atmospheric pressure cannot attain an effective oxide thickness of as low as 46 Angstrom. Furthermore, the dry oxidation of nitrides at atmospheric pressure can only yield a nitride/oxide (N/O) structure, which results in high leakage current. Therefore: our novel technique is promising for future dynamic-random-access-memory (DRAM) technology.