All issues
Volume 31, Issue 3
Displaying 1-20 of 20 articles from this issue
- |<
- <
- 1
- >
- >|
-
Mahmoud HASHEMINEJAD, Junichi MURATA, Kotaro HIRASAWA, Setsuo SAGARA1995 Volume 31 Issue 3 Pages 277-283
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSNonlinear systems can be modeled by neural networks. However, choice of suitable network architecture is the most important problem. And “how to find the best activation function” is a persistent aspect of the architecture design. Here we have proposed a sigmoid function with one parameter which provides us not only the reduction of error bound but also the opportunity of obtaining better insight into the systems. The proposed function has the ability of recognizing linear and/or nonlinear parts of the system under study. After automatic training of this parameter along the weights, more information about the system will be available. Using this additional knowledge about structure of the system, one will be well equipped to attack control problems such as controller design using neural network model.View full abstractDownload PDF (2851K) -
Yasuyuki FUNAHASHI, Manabu YAMADA, Shin-ichi FUJIWARA1995 Volume 31 Issue 3 Pages 284-291
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThe discrete time preview control problem is investigated in this paper in the framework of frequency characteristics. Along this line, the most attractive is the Zero Phase Error Tracking Controller, abbr., ZPETC, which can reduce the phase characteristics of the overall system to zero for all frequencies. However the previous methods may produce an excessive gap between the gain of the overall system and unity, which is ideal gain characteristics. This paper proposes a new ZPETC, which is designed so as to minimize the maximum of the gain gap over a specified frequency range under the condition of zero phase error. Our contribution is that the proposed ZPETC can bring the gain of the overall system closer to unity than the previous methods, which is demonstrated by a numerical example.View full abstractDownload PDF (1597K) -
Toshiyuki INAGAKI1995 Volume 31 Issue 3 Pages 292-298
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThis paper investigates responsibility allocation between human and computer in supervisory control of large-complex systems. Strategies for responsibility allocation are analyzed in a probabilistic manner by taking into account human's distrust of a computerized warning system, inappropriate situation awareness, and process dynamics. We show that it is not wise to stick to the principle, “a human locus of control is required, ” even though that is recognized as an essential principle for the human-centered automation. It is proven that responsibility allocation between human and computer should not be fixed but must be changeable dynamically and flexibly depending on the situation, which suggests the need for a new framework on human-centered automation, especially when safety of the process is a factor.View full abstractDownload PDF (2387K) -
Ya Jie TIAN, Nobuo SANNOMIYA1995 Volume 31 Issue 3 Pages 299-301
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSA new mathematical model for the motion of a fish school with many individuals is proposed. The basic assumption for model building is that the behavior of fish school can be decomposed into two components. One is the motion of a fictitious individual called the gravity center of the school. The other is the motion of four individuals which are located at the boundary of the school. The validity of the model is examined from a simulation result.View full abstractDownload PDF (365K) -
Kajiro WATANABE, Takeshi ONISHI1995 Volume 31 Issue 3 Pages 302-307
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThe tone from a small police whistle when blown into weakly, can be listened into in such a way that the frequency is proportional to the amount of blowing. This fact hints to us some whistle element can be a new flowmeter that detects the flow by the frequency. Here we extract a structure of the whistle element that is desirable as the flowmeter. The spectrum of the whistle tone has several peaks and some frequencies with the peaks are linearly proportional to the flow and the other keeps same values. The whistle itself includes several tone generating mechanisms. This research aims at extracting an element and/or mechanism that generates the tone whose frequency is purely proportional to the flow.
First we investigate what role the each component (a) whistle itself, (b) edge and (c) cylindrical cavity plays in the tone generation. We investigated how the tone frequency changes as flow increases for each element and found the cylindrical cavity played the best role as the flowmeter in the sense of the range ability, stability and linearity.View full abstractDownload PDF (2775K) -
Ippei TORIGOE, Yasushi ISHII1995 Volume 31 Issue 3 Pages 308-314
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSA new method is proposed for measuring fluid density using the equation of motion. Since the pressure gradient along a pipe is proportional to the density and the axial acceleration of fluid in the pipe, the density of fluid can be calculated from the pressure gradient and the acceleration. An exciter is mounted to a sensing pipe and is driven at a constant acceleration amplitude. Two pressure taps are made on the sensing pipe spaced a small distance apart along the pipe axis. The pressure difference between these taps, which is a finite difference approximation for the axial pressure gradient, is picked up by a manometer. A phase-sensitive detector being employed, the output of the manometer is converted to two DC voltages; these outputs are proportional respectively to the fluid acceleration and to the fluid velocity. The fluid density can be known from the output proportional to the acceleration, while the voltage proportional to the velocity can be used to eliminate the error due to the fluid viscosity. A trial device was built which employed a flat diaphragm speaker for the exciter, and several experiments were performed for gases with this device. The experimental results were in good agreement with theoretical predictions.View full abstractDownload PDF (3004K) -
Kazunobu KURIYAMA, Tsutomu MITA1995 Volume 31 Issue 3 Pages 315-323
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn this paper, the authors propose a new method of the spectral factorization. The method is based on the statespace data of a given spectral density matrix and permits one to factor it even if it has zeros on the imaginarity axis including the point at infinity. In order to develop the method, general factorization theorems for a square descriptor system is given. Then, it is shown that the spectral factorization problem is reduced to solving certain quadratic matrix equations which may be termed as generalized Riccati equation. The conditions for the existence of the solution to the equation are also derived.View full abstractDownload PDF (1334K) -
General CaseYoshihiko MIYASATO1995 Volume 31 Issue 3 Pages 324-333
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn most of the studies of model reference adaptive control, it is assumed that an upper bound on the degree of the controlled system is known. It makes the scope of application of model reference adaptive control too restrictive, since the reasonable upper bound on the degree cannot be specified a priori in many practical cases.
In the present paper, we propose a design method of model reference adaptive control systems for nonlinear systems with unknown degrees. The present adaptive controller is composed of high gain feedbacks of hierarchical structures derived from backstepping techniques, and the degree of it is independent of the degree of the controlled system. It is shown that the resulting control system is uniformly bounded, and that the tracking error converges to an arbitrarily small residual region. Finally, several simulation studies also show the effectiveness of the proposed method.View full abstractDownload PDF (1451K) -
Yoshihiro SUNAGA, Shuichi ADACHI1995 Volume 31 Issue 3 Pages 334-340
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSSet-membership identification is one of the most promising method in robust identification which means system identification for robust control systems. Many robust identification methods based on the set-membership method are proposed. Among them, Kosut et al proposed the pioneer work called parameter set estimation method, which evaluated parametric model uncertainty. However, the method is based on a discrete-time identification model, it is not easy to interpret identified uncertain region physically. In this paper, a new robust identification method which evaluates parametric model uncertainty for physical parameters which construct a nominal model. Because the proposed method is based on the approximate discrete-time identification model which consists of parameters of continuous-time system, the method is called continuous-time set-membership identification method. The upper bound for disturbance is necessary in order to apply set-membership method, so an evaluation method for the upper bound is also proposed. Finally, effectiveness of the proposed robust identification method is shown through numerical examples.View full abstractDownload PDF (1389K) -
Koichi KOGANEZAWA1995 Volume 31 Issue 3 Pages 341-346
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSOne method for solving inverse kinematics for redundant manipulators is proposed. A weighted generalized inverse (WGI) is introduced and solved by a recursive algorithm derived in this paper, which enables us to systematically determine the values of weights for preventing singular posture that the manipulator might encounter.
This method is faster than the conventionally used method based on the algorithm of Gauss-Jordan elimination in case of the manipulator having comparatively small number of redundant D. O. F. s. Joint variables are recursively updated at every time when new column of Jacobian is obtained. It implies that we can proceed the calculation parallel to input procedures of each joint variable. The fact that the projection operator is also recurrently obtained with the least burden of computation is another property of the proposed method.
One computer simulation of seven D. O. F. s anthropomorphic-type manipulator is examined to compare the computational speed and resultant joint variables between the proposed method and the conventional method.View full abstractDownload PDF (1336K) -
Osamu KATAI, Hiroshi KAWAKAMI, Tetsuo SAWARAGI, Tadataka KONISHI, Sosu ...1995 Volume 31 Issue 3 Pages 347-356
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThis paper proposes a method of deriving meta-planning knowledge from artifacts. The key idea behind this is that the object in question can be regarded to be a result of successive strategically rational decisions and actions which are subject to the Design Axioms, i. e., the principles of “good” design introduced by Suh in his Axiomatic Design Theory. The structural features of designed objects are analyzed by domain specific knowledge, and then the EBG (Explanation-Based Generalization) of this analysis yields a general and systematic explanation of “how” they function and attain their design goals by the use of their functional compositions together with the structures supporting these functions. For deeper understanding of design beyond this “how explanation”, it is necessary to explain the reason “why” these functional and structural compositions of the objects are selected from among other possible alternatives. Our method extracts such a deep explanation from artifacts analysis based on Design Axioms. The deep explanations show strategic or meta-planning knowledge for conceptual design by which the order of attaining design goals and the way of resolving interactions among design goals can be elucidated.View full abstractDownload PDF (3985K) -
Yuichi SAKURABA, Junichi IDE, Takamichi NAKAMOTO, Toyosaka MORIIZUMI1995 Volume 31 Issue 3 Pages 357-363
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn recent years, odor recognition systems have been studied very actively. However, this field is difficult to research since there are a lot of kinds of odors, and they have not been characterized well. Because of these difficulties, odor property can be revealed from inspection such as sensory test performed by human. As the odor inspection by human fundamentally has fluctuation based on physical or climatic condition, an artificial odor recognition system is required in many industries.
The authors proposed a neural network called “Fuzzy Learning Vector Quantization (FLVQ)”, and applied it in the odor recognition system to identifying odor kinds and expressing sensory quantity obtained from the human sensory test.
FLVQ network receives the output pattern from the sensors, and is trained by the result of the human sensory test, such as a triangle test for whiskeys and a semantic differential (SD) method for flavors. The network can output odor quantity after learning. It has been confirmed from the outputs of the estimated sensory quantity that FLVQ has satisfactory ability to express human odor quantity.
Moreover, adsorption membranes for odor sensors were selected to improve the odor recognition system. The authors proposed three statistical methods for membrane selection. It was found that the combination of discrimination analysis and multiple linear regression method was suitable here. Using selected membranes, the estimation accuracy of sensory quantity was raised.View full abstractDownload PDF (1445K) -
Yoshizumi WAKASUGI1995 Volume 31 Issue 3 Pages 364-373
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThis paper describes a new global maximum searching method for one-dimensional multimodal unknown functions. The searching method is composed of three stages and selects sequentially a subdomain based on a criterion function in each stage. In the first stage, the subdomain which length is the largest, is selected. If the lengths of subdomains are equal, the observed values at the two end points are also considered. In the second stage, subdomains in which a local maximum exists are searched and a subdomain is selected considering the values at the two end points and the lengths of subdomains.
Measures of error for local maximum searching are introduced to check if a local maximum has converged to a specified accuracy or cannot be the global maximum. In the third stage, the local maximum, which is determined to be the global maximum at the end of the second stage, is estimated using a second order interpolation polinomial.
It is clearly explained that through the three stages, the method can search maximum points surely and successfully with any required accuracy.
Performance of the proposed method is evaluated and compared with the previous method by Kubota for a few function models with different parameters of peaks. The simulated results show that the global maximum points are obtained with remarkably high efficiency and the method saves largely computation for local searching even for functions with extremely steep peaks.View full abstractDownload PDF (2043K) -
Eiji WATANABE, Hikaru SHIMIZU1995 Volume 31 Issue 3 Pages 374-381
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSBack propagation learning rule (BP rule) is used as an effective learning algorithm for multi layered neural network (NN). However, BP rule has such a basic problem that the learning speed becomes very slow. This paper proposes two learning time reduction algorithms of NN for pattern classification problems. We consider the reasons why the learning speed becomes very slow and the back propagation errors for the each learning pattern and output unit don't change at the same speed. An error function is introduced to improve the unbalance of the learning speed among patterns. By using the error function, a step correction algorithm of convergence condition is proposed, in which the convergence condition is switched from the loose condition to the strict one to stop the unbalance of learning speed among the output units and the learning patterns. An efficient learning algorithm is also proposed, in which inefficient correction procedures of weights are omitted in order to save the learning time of NN. From the simulation results for pattern classification problems, it is confirmed that the proposed algorithms are superior to BP rule and other learning algorithms with respect to the reduction of learning time.View full abstractDownload PDF (1645K) -
Manabu KOTANI, Yasuo UEDA, Haruya MATSUMOTO, Toshihide KANAGAWA1995 Volume 31 Issue 3 Pages 382-390
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThe application of hybrid neural networks consisted of an acoustic feature extraction network and a fault discrimination network to an acoustic diagnosis for compressor is described. The acoustic feature extraction network uses a five-layered feed-forward neural network which is able to extract the nonlinear features from the input information. The fault discrimination network uses a Gaussian potential network which is able to adjust the number of hidden units based on the learning algorithm. The experiment is performed under the various experimental conditions in order to examine the performance of the hybrid neural network for a practical use in the plant. The obtained result shows that the network has the superior performance in the discrimination accuracy, the learning speed and the adjustment of the number of hidden units.View full abstractDownload PDF (1724K) -
Eimei OYAMA, Susumu TACHI1995 Volume 31 Issue 3 Pages 391-398
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn order to solve the inverse problems for the system with unknown characteristics, many researchers have used a method that uses an inverse model of the target system, which is acquired by learning. The direct inverse modeling, the forward and inverse modeling and the goal directed model inversion were proposed for the acquisition. However, precise inverse models for some kind of systems cannot be obtained by these method. Furthermore, a limited scale neural networks system has only limited precision. Errors still remains in the output of the inverse model using the neural networks system. The use of the inverse model is not always a good method for solving inverse problems.
Another way to solve the inverse problem is by an iterative method. We proposed a generalized inverse model with output feedback as a model of a human nervous system solving inverse kinematics problem of a human arm. The system acquires the inverse model of the linearized model of the human arm by using neural networks and uses the inverse model as an output feedback system. The system approximates the iterative improvement of Newton's method. We call the system Output Feedback Inverse Model. However, the precision of solutions provided by the system is low.
In order to make the precision high, we improve the system and propose a new configuration of the iterative system using neural networks. We call the improved system Modified Output Feedback Inverse Model. By using a random search technique for the initial value, the proposed system provides more precise solutions than the conventional methods.
The performance of the proposed method are shown by numerical simulations.View full abstractDownload PDF (1961K) -
Masanari OH, Youich TANAKA, Shin-ichi FUJINO1995 Volume 31 Issue 3 Pages 399-401
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSA system of viscosity measurement has been developed. This system is based on the falling ball principle and utilizes image processing. The measurement accuracy has been calibrated by using standard liquid. From measuring the temperature of the liquid the viscosity can be exactly calculated. This system has been applied to fluid experiment equipment to make the experiment more efficient.View full abstractDownload PDF (363K) -
Tomoyuki OSAKI, Byeongdeok YEA, Ryosuke KONISHI, Kazunori SUGAHARA1995 Volume 31 Issue 3 Pages 402-404
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSNeural networks were examined to discriminate the vapor of alcohol and spirits from the city gas. As the result, very good discrimination rates, that is 100% for the spirits and city gas and 95% for the alcohol and city gas, were achieved.
We expect that the incorrect operations of gas detectors will be disappered provided that above method was utilized with microprocessors.View full abstractDownload PDF (400K) -
Shuichi SASA1995 Volume 31 Issue 3 Pages 405-407
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn this paper, a robust state estimator against system uncertainties is proposed by H2 norm optimization. Frequency domain and state space solution is presented including a simple numerical example.View full abstractDownload PDF (291K) -
Tielong SHEN, Katsutoshi TAMURA1995 Volume 31 Issue 3 Pages 408-410
Published: March 31, 1995
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThis paper investigates robust H∞ control problem for affine nonlinear systems with uncertainty. A sufficient condition is given for that the nonlinear system is robust exponential stable and the L2 gain is less than one under the presence of gain bounded perturbation. Using this condition such a state feedback controller is obtained that the closed loop system is robust H∞ sub-optimal. The results given in this paper include the result obtained in linear system.View full abstractDownload PDF (377K)
- |<
- <
- 1
- >
- >|