All issues
Volume 17, Issue 2
Displaying 1-26 of 26 articles from this issue
- |<
- <
- 1
- >
- >|
-
Yoshiharu OHTANI1981 Volume 17 Issue 2 Pages 161-167
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThis paper examines the controllability of homogeneous-in-the-state bilinear systems onRn-{0}dx/dt=Ax+uBx (B) where x∈Rn-{0}, u is a real-valued control function and A, B are n×n matrices specified for some real numbers a, b(≠0) and matrix U by Bt+B=2bI, A=Z+aI, Z=(Bt-bI)U+U(B-bI) under a constraint, tr (Z)=0.
For the system we consider the vector fields associated with systems generated by the projection of trajectories from Rn-{0} onto a sphere of n-1 dimension. It is shown that there exist controls such that the induced vector fields become conservative and that if the system (B) is accessible then it is controllable on the sphere because of the compactness of a spherical surface.
Since we can take a course on the sphere and choose a stable mode or an unstable mode of the system (B), every state is attainable to any point on Rn-{0}. This proves that if the system has the accessibility property it is controllable on Rn-{0}.View full abstractDownload PDF (925K) -
Nobuaki KOBAYASHI, Takayoshi NAKAMIZO1981 Volume 17 Issue 2 Pages 168-175
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThe problem to be considered in this paper is to find a linear state feedback, by which the system having maximum unobservable-subspace will be realized. First a necessary and sufficient condition for an asymptotically stable maximal unobservable system to exist is presented under the condition that the original system is invertible. The appropriate state feedback for achieving the maximal unobservale system is also given. The present results are then applied to derive a sufficient condition for the decoupled system to be stable via a linear state feedback when the number of inputs are more than that of outputs. Furthermore a sufficient condition for the stable system localized from disturbance to exist is presented in terms of system invariant zeroes. It can be readily shown that this sufficient condition is also necessary, provided the system satisfies the decoupling condition.View full abstractDownload PDF (960K) -
Shin-ichi TASHIMA, Kazuaki ANDO, Bunji KONDO1981 Volume 17 Issue 2 Pages 176-181
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThe optimum feedforward control for a system with a measurable disturbance input is discussed.
Suppose that the optimum state feedback is already synthesized using the linear regulator technique for the system with augmented state variables by adding integrators. In this case, the value of the performance index is given by the quadratic form of the initial value of the state. Let the final value of the augmented state subjected to a step disturbance be x∞. By changing the origin of the state into x∞, the response of the system to the step disturbance is transformed to the response of the system with the initial state -x∞. Then the value of the performance index is given by the known quadratic form of x∞. As x∞ changes with the change of the feedforward, the optimum feedforward can be synthesized to minimize the value of the performance index.
As an example, the synthesis procedure is checked to be valid by the simulation for the control of a drum-type boiler for the case in which the step change of the output steam flow takes place.
The feedforward matrix proposed here is proved to become identical with that of Davison when the weights in the performance index to the added components of the state in the augmentation become zero.View full abstractDownload PDF (715K) -
Tadao KAWAGUCHI, Kenshi KATAYAMA, Katsuhiko KASHIBA1981 Volume 17 Issue 2 Pages 182-189
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSWe have established the big advantage of introducing Pontryagin's Maximum Principle into an optimal gauge control system for cold rolling mill.
Automatic gauge control (AGC) for cold rolling mills is said to have been perfected for many years. But recently, we have a new possibility of improving the system with the aid of developing digital computers, high performance control devices and many highly acculate sensors.
In this paper, the optimal control theory is applied to the electro-hydraulic screw down device controller which contributes to the response time of the Sendzimir mill AGC system. The dynamics of the electro-hydraulic screw down system can be sufficiently approximated as the second order system which is constructed by the combination of the servo-valve and the hydraulic screwdown cylinder. By use of the Maximum Principle, a time minimum control of this system can be realized by the Bang-Bang control, which has one switching point at maximum.
We have made further attempts to eliminate logarithmic and other complicated caluculations and reduce the amount of steps in the program. A suboptimal control has been developed and adopted into the No. 3 ZM of Hikari Works Nippon Steel Corporation. This system extremely simplifies the control's algorithm while keeping the quick response time.
The improvement has been such that the response time is two to five times shorter when it is compared with that in a conventional feed back control systems. It is typical of an open loop system that an accuracy of system modelling directly infuluences the performance of control. So the parameter error must be kept within a few percent.
As result of practical operation, this suboptimal control system was proven to be very useful in improving the response and accuracy of the AGC system.View full abstractDownload PDF (921K) -
Toru OKANUMA1981 Volume 17 Issue 2 Pages 190-196
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThe trajectory estimation of a satellite by Kalman filter is well known and its filter requires various a priori error covariance matrices to be ready before estimation begins. In particular, it is difficult to set up precise a priori error covariance matrix with respect to the state variables.
In this paper, to avoid this difficulty, a nonlinear information filter which uses U-D factorization filter is constructed and operated under a condition that the initial a priori information matrix is zero.
Because of the condition, this U-D factorization information filter becomes sub-optimal filter. Therefore, a relationship between the calculated state error covariance matrix and the actual state error covariance matrix based upon the actual estimation states is derived assuming linearity. The validity of the suboptimal filter is certified from this relationship.
The capability of this filter is examined by a precise simulation program in regard to a trajectory estimation of the lowest-flying earth satellite. As a result of the simulation, it is shown that the estimation error is small enough and that the validity of the sub-optimal filter is confirmed.View full abstractDownload PDF (940K) -
Masashi YAMADA, Yoshimi MONDEN, Masatake HIROOKA, Suguru ARIMOTO1981 Volume 17 Issue 2 Pages 197-201
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn this paper we shall propose a fast recursive algorithm for tracking time-varying autoregressive models by the direct use of observed date sequences. Though this algorithm consisting of a triple of recurrences. (order-ascending, descending and time-updating one), in form resembles Morf's algorithm, it is also varid for tracking time-varying autoregressive models in contrast to Morf's one. It further wins an advantage over the usual adaptive filtering algorithm in that it can track not only fast time-varying model parameters but also their orders based on Akaike's information criterion. Finally a result of computer simulation will be given to illustrate the efficiency of this algorithm.View full abstractDownload PDF (634K) -
Kazuo YAMANAKA, Etsujiro SHIMEMURA1981 Volume 17 Issue 2 Pages 202-205
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSA problem concerning the stochastic stability of linear systems with a random time-delay is discussed. The delay is assumed to be a stationary and ergodic process. It is shown that the system is almost surely a symptotically stable, if 1) there exists a positive constant such that a system in which the delay is fixed to be the constant is exponentially stable and 2) the first absolute moment of the delay around the constant is sufficiently small. An example is presented to explain the practical application of the result.View full abstractDownload PDF (484K) -
Bumpei NAKANO, Yasuo KUWATA, Yasuhiko TAKAHARA1981 Volume 17 Issue 2 Pages 206-209
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSNo one has ever successfully given a satisfactory characterization of a finite automaton, that is, no complete answer has been given to the problem to determine whether a system given by its input-output behavior is realizable as a finite automaton. Gill and Heun studied the problem under the condition that an input and output sequence is finite. This paper treats the problem in a general framework and gives a sufficient condition for a set of input and output sequences to be realized by a finite automaton. Specifically, the paper shows that if a system has finite input and output alphabets and if it is stationary and past-determined, it is realizable as a finite automaton. The paper also clarifies the relation between the result of this paper and that of Gill.View full abstractDownload PDF (628K) -
Yasuo TANAKA, Hidenori KIMURA1981 Volume 17 Issue 2 Pages 210-216
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn this paper we consider the problem of output deadbeat control with internal stability (ODCIS) in discrete-time linear multivariable systems which include uncontrollable external inputs. First, we consider the case in which the state feedback is allowed. It is shown that the problem is solvable if and only if the state space can be decomposed into a controllable subspace of (A, B) and an (A, B)-invariant subspace in Ker D. The minimal number of steps for attaining the deadbeat control is also derived. A procedure for computing the minimal-time state feedback control law for ODCIS is given.
Next, we generalize these results to the case in which only the output feedback is allowed. The solvability condition, the minimal number of steps and the procedure for synthesizing the minimal order deadbeat observer for attaining the minimal-time ODCIS are shown. An illustrative example is given to show the feasibility of the algorithm.View full abstractDownload PDF (936K) -
Ikuro MASUDA, Hideo MAEJIMA, Tsutomu OHMAE1981 Volume 17 Issue 2 Pages 217-222
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSWhen the computer-based direct digital control (DDC) extends its range of application from the process control to the mechanical control, high-speed operations beyond a computer's ability are often required. In this paper, a firmware technology is applied to increase the operation speed of DDC, and its effect and limitation are discussed.
At first, the special instruction set for DDC is derived from transfer functions and block diagrams, and they are microprogrammed. According to the analytical result, the operations by special instructions (firmware) are 4 to 7 times as fast as those by standard ones (software). One experimental result shows that the execution speed in a motor control system is improved by firmware to become about 5 times faster compared with the speed to be obtained by software.
However, the application will be limited to high-speed and repetitive operations such as DDC or sensing, due to its limit of programming productivity.View full abstractDownload PDF (877K) -
Yajiro MORITA, Yuzo OKA, Yoshiteru OGATA1981 Volume 17 Issue 2 Pages 223-230
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThere is a growing need for a generalized theory of measurement and for a systematic study of measurement technology so diversified nowadays in its objective, scope, forms and methods that virtually every important branch of engineering, physical and social sciences has its own version of measurement technology. In such a broad sense of measurement, man may become an integral part of the measurement system and his evaluation a tool to endow the set of measured objects with an ordered structue.
There may be very often the case in which the transitive property of an order relationship does not hold even if the criterion for evaluation is rational and accurate enough. Therefore the given objects cannot be arranged in a linear order. Such an intransitive ordering is called a loop. This paper deals with how to treat a loop and derives the various properties of the loop.View full abstractDownload PDF (1032K) -
Toshihiko INOUE, Masato AGUI, Goichi MATSUI, Suguru ARIMOTO1981 Volume 17 Issue 2 Pages 231-236
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSTo classify gas-liquid two-phase mixtures according to their“flow patterns”helps qualitative understanding of the flow and leads us to a better prediction of two-phase flow parameters such as pressure drop, quality, and heat-transfer coefficient. Accordingly it is important to develop some technique to discriminate the two-phase flow pattern in a pipe quantitatively.
In this study we propose two kinds of techniques, which are applied to data in the range of 0.0-10.0m/s mixture velocity. In the first technique, measurement data of normalized differential pressures at four locations along the flow direction are used for flow pattern discrimination. In the second technique, the transit length of a bubble, which is measured at the center line of a tube by the use of a miniature optical probe, is chosen. The set of statistical parameters such as average, variance, and skewness factor of the measurement data is adopted as the set of key indexes, in order to establish objective and quantitative standards of flow pattern discrimination among the five dominant flow patterns namely, the bubbly flow, the hemispherical bubble flow, the slug flow, the froth flow, and the annular flow. All key indexes can be obtained by processing directly the measurement data without calculating the PDF (probability density function).
It is shown that the set of these statistical parameters is sufficient for characterizing the flow pattern and gives satisfactory results on the flow pattern discrimination in spite of the short sampling time and the simple data-processing.View full abstractDownload PDF (2032K) -
Nobuyuki AKIYAMA, Yoshimasa OSHIMA, Mitsuyoshi KOIZUMI, Masakuni AKIBA ...1981 Volume 17 Issue 2 Pages 237-242
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSMost of the defects of the semi-conductor patterns are caused by the foreign particles on the wafers. We developed an automatic machine which detects the foreign particles (their diameters are larger than 2∼3μm) on a wafer, whether it has patterns or not. The pattern sizes are usually 3∼5μm wide, 1μm high.
We illuminate the small area on a wafer by means of the polarized laser obliquelly, and detect the light, which is scattered by the pattern or the foreign particle on the wafer, from right above the wafer. We can find the foreign particles only, by detecting the scattered light of which the angle of palarization is changed. By the use of this machine, we can know quantitatively the number of foreign particles in the making process of a small pattern like that on LSI.View full abstractDownload PDF (1772K) -
Mitsuyoshi KOIZUMI, Nobuyuki AKIYAMA, Yoshimasa OSHIMA, Asahiro KUNI1981 Volume 17 Issue 2 Pages 243-248
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSA new technique for detecting with high accuracy the position of alignment patterns on both the wafer and the photomask has been developed. The technique which could be applied in an automatic projection mask aligner for LSI fabrication.
Interference fringes, which are caused by monochromatic illumination (g-line) on the surface of the photoresist layer coated over the alignment pattern on the wafer, are projected against a moving slit through the projection lens. The detected signal is generated by the slit movement on the magnified image of the alignment pattern on the wafer.
In order to detect the center of the interference fringes, which is considered to be coincident with the center of the alignment pattern on the wafer, a pattern matching method is applied on the detected signal. The detecting resolution is 0.04μm and the detecting repeatability is 0.074μ (3σ). The accuracy of detecting the relative displacement between the alignment pattern on the wafer and that on the photomask is 0.19μm (3σ).View full abstractDownload PDF (723K) -
Ichiro TOKUHIRO, Katsuhiko SHIRAI1981 Volume 17 Issue 2 Pages 249-255
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIt is one of the difficult problems in the data processing of plant systems, the automatic seismometry systems or in the speech recognition systems to detect the change of the characteristics of a stochastic process.
In this paper, a new method to detect the change and the change time in the stochastic process is presented based on the log-likelihood ratio function. If the quantity of data is very large and a rapid analysis is required as in the automatic seismometry systems, the algorithm should be simple. In this situation, the most suitable mathematical model of a stochastic process may be an autoregressive (AR) model.
In this method, both the amplitude and the spectral information are effectively utilized and a backtracking process is included which simulates the way taken by a human operator. The suitable order of the AR model and the numbers of the data in the calculation of AR model are discussed from the practical point of view.
The experimental results for seismic waves and speech waves show this method is highly reliable and practical.View full abstractDownload PDF (914K) -
Kan TAGUCHI, Yoji UMETANI1981 Volume 17 Issue 2 Pages 256-262
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIt is the most important problem in non-symbolic pattern recognition of general shapes to decide what kinds of features should be extracted. Since the recognition of ‘general shapes’ depends on the purpose the recognizer has in mind, the ‘general shapes’ cannot be dealt with in general, while symbolic patterns such as letters of alphabet connote their categories a priori, the features that declare each symbolic patterns are evident.
On the other hand, the human being can discriminate not only symbolic patterns but also ‘general shapes’, for example, the random shapes dealt with in this paper. The authors consider the discrimination of the human being as the basis of recognizing of ‘general shapes’.
In this paper, the authors propose psychological experiments to find out what feature properties are given to the eyes of human subjects when they discriminate patterns, i. e. 100 convex and concave random shapes. And the standard searching shapes (categories) are set from the response of the subjects.
A multilevel mechanical procedure for discriminating the standard searching shapes (categories) is proposed. In the procedure the feature properties selected through the psychological experiments are used. It is shown that random shapes can be discriminated by this mechanical procedure in close agreement with a human judgement.View full abstractDownload PDF (1555K) -
Hidehiko KONO1981 Volume 17 Issue 2 Pages 263-268
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSThis paper deals with a newly developed methodology and an approach for the pattern discrimination of plane shapes using the group-classification value as a feature parameter. Before the group-classification value is decided, the pulse row obtained by the second differentiation of the projection of the shape on the reference line is considered. As the number of the positive and the negative pulses in the pulse row varies with the postural angle of the shape, the maximum numbers of the positive and the negative pulses are decided. The group-classification value is defined from the ratio of the maximum number of the positive pulses to the sum of the maximum numbers of the positive and the negative pulses.
The author proposes a method to discriminate 22 kinds of shapes by applying the group-classification value, the minimum of the positive pulse, the maximum and minimum of the negative pulse and the pulse type. Furthermore, the proposed method is applied to the actual pattern discrimination and the result shows that 19 kinds of shapes can be discriminated when the pattern image is sampled by the 35×35 matrix pattern with every 5 degrees.View full abstractDownload PDF (866K) -
Mitsuo AI, Masatoshi TSUCHIYA, Tatsuji SAKAMOTO1981 Volume 17 Issue 2 Pages 269-273
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSSemiconductor sensors which detect forces or displacements should preferably be used with an elastic metal having a small coefficient of thermal expansion.
The present investigation is concerned with elastic alloys with low thermal expansion coefficients. The alloys contain iron, nickel and cobalt with proper ratios, and their elastic properties can be adjusted by applying cold working and heat treatment.
The elastic properties of several metals of low thermal expansion coefficients are examined by the stress to plastic strain method, and then the elastic property of the iron-nickelcobalt alloys is evaluated in detail by the internal friction method.
The tensile strength of an iron-nickel-cobalt alloy is increased to 1800MPa when the nickel content is less than 28%. However the most suitable composition for use of the semi conductor sensors is 28.7∼30% of nickel and 17∼18% of cobalt. In this case, the thermal expansion coefficient is less than 50×10-7/K, the tensile strength is about 1100MPa and the elasticity is the same as that of a beryllium-copper alloy.View full abstractDownload PDF (676K) -
Motohisa NISHIHARA, Satoshi SHIMADA, Kazuji YAMADA, Masanori TANABE, Y ...1981 Volume 17 Issue 2 Pages 274-279
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn this paper a method of analyzing the nonlinear characteristics of a piezoresistive semiconductor pressure sensor and the results given by this method are described.
The main causes of nonlinearity in a semiconductor pressure sensor are the nonlinear characteristics of the piezoresistive effect in silicon and of the stress in the silicon diaphragm. Experimental data are used to derive a nonlinear characteristics formula and then the formula is used to estimate the nonlinearity of the piezoresistive effect. Approximate formulas for uniformly loaded circular plates with a large deflection are also used to calculate the nonlinearity of the stress in a silicon diaphragm.
Using these formulas, it is shown that the nonlinearity error of a pressure sensor is related to the positions of radial and tangential resistors on the silicon diaphragm. Optimum resistor positions, in which nonlinearities are less than 0.2% F. S., have also been determined.View full abstractDownload PDF (1300K) -
Osamu OYAMA1981 Volume 17 Issue 2 Pages 280-285
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn this paper, a new design of pneumatic band reject filter based on the characteristics of pneumatic coupling condenser circuits1) is described. The coupling condenser circuit has the character of a high pass filter essentially.
A pneumatic band reject filter is constructed by combining two coupling condenser circuits and the two restrictor-capacitor circuits. Features of this device are as follows.
1) The central frequency of the band reject filter can be chosen by changing resistances in the coupling condenser circuits and the restrictor-capacitor circuits.
2) Low central frequency filters having, for instance a corner frequency lower than 1Hz, can be realized.
The gain characteristics of pneumatic band reject filters were tested in the range of stop frequencies 0.7∼5Hz. As a result, it was shown that the band reject filters reduce the signal levels of stop frequencies by more than 15dB in comparison with the signal levels of distant frequencies.
Theoretical values of the gain characteristics of band reject filters were derived from the time constants and amplitudes of the coupling condenser circuits, and the time constants of the restrictor-capacitor circuits. These calculated values and the experimental data show good agreement.
The effects of these time constants and amplitudes on the characteristics of a pneumatic band reject filter are discussed theoretically.View full abstractDownload PDF (1157K) -
Fumio HARASHIMA, Seiji KONDO, Masato KOYAMA1981 Volume 17 Issue 2 Pages 286-293
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSPhase-locked loop technique permits the precise control of motor speed. In the PLL speed control system, the speed accuracy of 0.002% may be achieved. This represents almost a hundredfold improvement over a conventional method of speed regulation.
This paper presents a microprocessor-based digital PLL speed control system and describes several methods to analyse the system performance. In this system, digital techniques provide all the control functions such as detection of speed and phase error of the rotor, signal processing and the determination of the control signal of the power converter. They can be achieved with a simple circuit architecture by using a microprocessor. The completely digital scheme is free from drift and offset error which are inherent in an analog circuit.
Sampling, time delay and quantization of phase error detection affect respectively the stability, dynamics and speed error of the system. For the analysis of system performance, the digital simulation based on the state-space method is used. The stability, dynamics and speed error of the system are additionally analysed with the simplified models.
An experimental apparatus including a 2.2kW DC motor is built and the system is tested. The experimental results cofirm the validity of the analytical methods mentioned above. The speed accuracy observed is 0.028% or less and the possibility to reduce the figure to 0.0018% is shown.View full abstractDownload PDF (2105K) -
Satoru MYOJIN, Takeshi KASAI, Saburo TSUJI1981 Volume 17 Issue 2 Pages 294-299
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSIn a primate oculomotor system, a part of the central nervous system (e. g., superior colliculus and reticular formation) generates velocity command and position command separately in each of the horizontal and vertical directions regardless of the initial eye position.
While in the peripheral muscle and eye ball system, strong position dependent mechanical interactions are expected to exist among the six extraocular muscle forces.
A special controller which virtually decouples the interactions and computes an appropriate motor signal for each muscle would be necessary between the central nervous system (CNS) and the final motor system, otherwise the eye could be rotated in an unintended orientation.
This paper discusses the functional structure of such kind of decoupling controller on a simulated and simplified oculomotor system, in which an eye ball is rotated in three directions (horizontal, vertical and tortional) by a pair of thin wire and a DC motor for each direction.
A decoupling control scheme has been proposed which enables the CNS to control the rotation of the eye ball in any direction by monitoring the temporary eye position and velocity, as if the mechanical system comprised three separate linear systems.
Feasibility of the control scheme has been demonstrated via computer simulations of the system and the controller, and also via a real time control experiment on a hardware oculomotor system model with rotations about two axes (horizontal and vertical) by two DC motors.
Finally, an evaluation has been given on the applicability of the control scheme to the human oculomotor control system.View full abstractDownload PDF (1814K) -
Ryoji ITO, Yoshiyuki SAKAWA1981 Volume 17 Issue 2 Pages 300-302
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSDownload PDF (275K) -
Hayao MIYAGI, Tsuneo TANIGUCHI1981 Volume 17 Issue 2 Pages 303-305
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSDownload PDF (298K) -
Masataro NISHIMURA, Kimihiro MATSUNAGA1981 Volume 17 Issue 2 Pages 306-308
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSDownload PDF (407K) -
Shin-ichi NAKAO1981 Volume 17 Issue 2 Pages 309-311
Published: April 30, 1981
Released on J-STAGE: March 27, 2009
JOURNAL FREE ACCESSDownload PDF (323K)
- |<
- <
- 1
- >
- >|