In order to measure the temperature of the silicon wafers in a diffusion furnace (which is usually used for semiconducter heating process such as oxidization, annealing and diffusion), we design a radiation thermometry suitable for this purpose. The feature of this thermometry is measuring the temperature of the wafers to be usually out of sight by an optical guide, which consists of two quartz prisms. The radiation from the objective wafer is measured by a pyrometer with measuring wavelength of 0.9μm. The reason why the wavelength is selected is that a silicon wafer is opaque at this wavelength. The emissivity of a silicon wafer at 0.9μm is measured, and it is confirmed that the emissivity dose not depend on temperature within the range of 600∼1, 000°C. It is also confirmed that the emissivity of a silicon wafer with a thin film of silicon oxide or silicon nitride can be calculated by considering the interference effect of multiple reflection inside the thin film.
The accuracy of the thermometry is examind by comparing the measured value of the pyrometer with that of the thermo-couple. The two measured values agree within the difference of ±2°C in steady state. But when wafers are inserting into or drawing out from the furnace, an error is caused by veiling glare at the optical guide and the wafer. The error is 10∼20°C when the wafer temperature is 250°C lower than that of the inside wall of the furnace, and is about 5°C when the wafer temperature is 150°C lower.

View full abstract

This paper describes the error in the heat flux measurements with an expected transducer. In the heat flux measurements from a wall to environment with the usual heat flux transducer placed on the wall, the original heat flux q₀ and temperature θ₀ simultaneously change by placing the transducer. For the purpose of improvement on the measurement accuracy, the transducer with the function witch compensates the temperature change is expected. In this transducer, it is the essential matter that the measurement error Δq depends on incompleteness of the temperature compensation Δθ.
The relational formula Δq/q₀=-E(Δθ/θ₀) and this coefficient E are derived from the heat conduction equation and a certain integral formula of Bessel function. The analysis of E, as an example, is performed with a wall whose inner structure is composed of surface layer and inner layer. Then E depends on thermal conductivity and thickness of the surface layer λ₁, d, thermal conductivity of the inner layer λ₀, heat transfer coefficient of the wall surface α₀, and size (radius) of the transducer R.
E is represented by three non-dimensional parameters H=λ₀/(α₀R), λ₁/λ₀, d/R in general case, by two parameters (λ₁d)/(λ₀R), H in the case of d<<R and λ₀<<λ₁, and by one parameter λ₁/(α₀R) in the case of d>>R or λ₁=λ₀. E is found numerically in 0.05∼10, 000 on an approximate estimate. The required accuracy of temperature compensation is determined from E and the desired measurement accuracy, and numerical values of this are also described with typical wall materials and others.

View full abstract

The paper presents the development of probe-rotation-type eddy current defect detector for hot rolled round steel bars. On-line defects detection in hot condition was tested by using conventional probe-rotation-type eddy current defect detector with heat shield.
The result of this test was not so good because detection rate of the defects having above 0.3mm depth were only about 48%. Then, these test result were analyzed and main noise factors were extracted. One was the vibration of round steel bars, and the other was their pass line change. Probe rotation speed was increased as the countermeasure for former noise factor, and probe profiling system to round bars was added to probe rotation mechanism for latter one. Consequently, detection rate of the defects having above 0.3mm depth rose to about 100%.

View full abstract

An ultrasonic orientation sensor using the phase difference has been developed. The sensing system measures the orientation angles of a surface normal of an object by phase differences of reflected echoes. The measuring algorithm of the sensor is obtained from the mirror property of reflection from an infinite plate. However, the sound reflection from the actual objects with finite and curved surfaces is influenced by diffracted waves. They cause the significant error on measured orientations.
This paper discusses the errors caused by surface conditions on bases of the theory of sound reflection. The phase difference of two received echoes from the arbitrary shaped surfaces is analyzed using numerical solution of the Fresnel-Kirchhoff's diffraction formula. Using the calculation, the errors of measured orientation are evaluated for small rectangular plates and cylindrically or sinusoidally curved surfaces. The calculated results are compared with the experimental results and a good agreement between both results is obtained.
The spatial properties of the sensing system are also discussed. The orientation measurement along an undulated surface shows the characteristics as low-pass spatial filter. The analysis clarifies the limitation of the orientation measurement of curved surfaces and the ability of reducing the influence of small undulations on a surface.

View full abstract

The major cause of defects in the initial process for manufacturing high-integration LSI circuits lies in contaminants which arise during the manufacturing process. Therefore, the accurate assessment of the presence of such contaminants and application of measures to deal these contaminants at an early stage is of great importance. The present research was undertaken with the purpose of developing automatic inspection techniques for the detection of minute contaminants on patterned wafer surfaces during the production process by optically distinguishing the contaminants from the background pattern.
This paper describes an optical technique for revealing the presence of contaminants by utilizing differences in the polarization characteristics of the light scattered from the pattern and from the contaminants. The approach to development of this technique consisted in investigating the appropriate conditions with respect to illumination and polarization analysis for emphasizing (a) the scattered light from contaminants and (b) the reflected light from the background pattern, then illuminating the sample in accordance with the conditions (a) and (b), and detecting the contaminants by computation on the basis of comparison between the two detected signals corresponding to (a) and (b) which generate from the same point of the sample.
This study has proved the following two results. (1) As regards the above conditions (a): in order to obtain a proportional relationship between the size of the contaminant particles and the detected signal indicating the scattered light from these particles, the inclination angle w₁ of the illuminating light must not exceed 5 degrees. As regards the above conditions (b): increasing the illumination angle w₂ tends to increase the result of the comparative computation of the detected signal from the contaminant. (2) Varying the illumination angle w₂ does not cause any great variations in the result of the comparative computation of the detected signal from the pattern.
On the basis of the above facts, experiments to distinguish the contaminants from the pattern were conducted using an experimental apparatus with the settings w₁=2° and w₂=30°. The results of these experiments indicated that contaminant particles as small as 1μm in diameter can be detected and confirmed the effectiveness of the present method.
The present method is new in that objects (i.e., contaminants) can be optically distinguished from the background (i.e., the pattern) by comparative computation with respect to two detected signals from the sample.

View full abstract

Polystyrene latex particles are used generally for calibration of particle size analyzers and electron microscopes, and also for an experiment to verify the light scattering theory, because the particles have excellent characteristics in sphericity and monosize distribution. It is of great importance, therefore, to measure precisely the diameter of the polystyrene particles. This paper describes a light scattering method to measure the particle size, because the method permits an absolute measurement and allow to measure it without any contamination and/or destruction of particles. In this method, the size of suspended particles was determined by fitting theoretical values based on Mie scattering to experimental data for an angular distribution of the intensity of light scattered from particles. A laser scattering photometer used in the experiment has the constitutional features that are insensitive to variation of the intensity of light source (He-Ne laser, 633nm) in measurements of scattered light intensity and to nonlinearity in a photoelectrical detection system, since the photometer adopts a new system of measurement on the basis of a principle of zero method. Light scattering measurements were made for suspensions of 0.9μm particles of the SRM of NIST and 2μm particles. The result for the 0.9μm particles agrees with the value of NIST within 0.3% in the number average particle diameter. The coefficient of variation of repeated measurements are 0.5% for the 0.9μm particles and 0.15% for the 2μm particles. The best fits of measured values for an angular distribution of the scattered light intensity to the theoretical values are within 5% at the scattering angle between 4° and 38°. The experimental results prove good accuracy and usefulness in measurements attained for the polystyrene particles by the laser scattering photometer developed in this work.

View full abstract

Fault diagnosis of sensors is important as well as that in system dynamics. As a diagnosis method for sensors, generalized-likelihood-ratio (GLR) approach is proposed, which models the anomaly function appearing in the anomalous sensor by a step function. But, by the method, the discrimination of the anomalous type, such as gain or dead-zone anomalies, is impossible, and also, more important is that the step hypothesis is not always effective, because the anomaly function does not always have a meaningful bias over a round time.
From this point of view, the paper proposes another GLR method which estimates the anomaly function by using the system information about the state. By taking this approach, the number of the parameters to be estimated can be decreased compared to estimating the anomaly function by a linear combination of orthogonal series. This decrease in the number of parameters and the more accurate modelling of the anomaly function can make the fault diagnosis of sensors more easy. Actually, numerical examples show that the proposed method is superior to the conventional ones in detection rate, detection speed, the estimation of the anomaly occurrence time, and the discrimination of anomalous sensor and the type.

View full abstract

The VSS design problem for a linear system has been investigated using a quadratic performance index, the value of which can be more lowered by logically switching admissible gains {K₀, K₁, …, K_p} than by the single constant gain K₀. A unique guaranteed cost control (GCC) method is introduced to derive the gain switching law proposed here.
The problem considered in this paper can be stated as follows: Given a controllable linear system,
x(t)=Ax(t)+Bu(t), x∈Rⁿ, u∈R^m,
with a set of admissible gains {K₀, K₁, …, K_p} and the quadratic performance index,
J=∫∞_t0e^2α(t-t₀)x(t)'Qx(t)dt, Q>0,
find a variable gain state feedback control law which gives a lower cost than the control law with the constant gain K₀.
The main results we have obtained are as follows:
(i) A solution to the problem has the form,
u(t)=K_j(t)x(t), j(t)={j:min_iV_i=V_j}, V_i=2x(t)'PB(K_i-K₀)x(t),
where the matrix P is the solution of the Lyapunov equation,
(A+BK₀)'P+P(A+BK₀)+2αP+Q=0.
(ii) The designed VSS is stable if there hold the conditions that both the matrices A+BK₀ and A+BK₀+αI are stable and the matrix Q+2αP is positive definite.
(iii) Even the case where the system matrices A and B include some bounded modeling errors can be treated successfully by introducing a parameter σ>0 into the switching law as
The parameter σ represents the size of the modeling errors; the trade-off between robustness and control performance can be performed by selecting the value of the parameter σ.
The results of some simulation experiments are also presented to demonstrate the effectiveness of the proposed design technique and to illustrate the corresponding relationship between the parameter σ and the robustness property.

View full abstract

This paper is concerned with finite spectrum assignment of time-delay systems. One of authors proposed an algorithm for it based on transformation from rational matrices to finite Laplace transform matrices. In the present paper, to improve the algorithm a new method for the transformation is given. The feedback generated by the new algorithm includes a smaller number of finite Laplace transform terms than that by the existing one.

View full abstract

This paper presents a finite spectrum assignment procedure based upon the input-output relation for scalar systems with commensurate time-delays. This procedure is effective under the spectral controllability and the spectral observability of systems.
Existing procedures based upon state space realizations need matrix calculation, but this procedure needs only scalar calculation. It consists of the following two steps.
(1) To solve the polynomial Bezout equation consisting of denominator and numerator of the transfer function.
(2) To design control systems using the solution given in the spep (1).
In the step (1), the algorithm is based upon properties of polynomial ring in two variables. The solution can be obtained over polynomials with coefficients in a special ring of the finite Laplace transform. In step (2), it is possible to design output-feedback-type and observer-type control systems with a finite set of spectrum assigned.

View full abstract

In our preceding studies, we showed a design method for constructing adaptive control systems for Attasi-type 2-dimensional systems and then constructed a kind of learning control system using the notion of 2-dimensional systems. But in those approaches, the convergence properties of adaptive processes on 2-dimensional plane are restricted to one direction only.
In the present paper, we propose a design method for constructing adaptive control systems which have convergence properties of two different directions on 2-dimensional plane. To realize those systems, we introduce two kinds of adaptive laws, that is, interlace-type adaptive laws and hybrid-type adaptive laws, into one adaptive control system, and make it possible to improve one adaptive process with interlace-type adaptive laws, adaptively by utlizing the other adaptive process with hybrid-type adaptive laws. In the following, we apply this technique (named “2-dimensional adaptive control”) to Attasi-type 2-dimensional systems and construct a more skillful learning control system where adaptive processes themselves can be improved adaptively. Finally, some simulation studies show the effectiveness of the proposed method.

View full abstract

It was necessary to improve the manual mode in order to increase the availability of a man-robot system, a part of the human interface technique. Therefore, the Error Recovery Expert System/Cartesian Coordinate System (ERES/CCS), combining the operation by the control program representing autonomy of a robot and the operation by a human representing versatility of a human operator effectively in the view point of human factors engineering, was desigened. The ERES/CCS applying the cartesian coordinate system uses the geometric inverse kinematics for the calculation of the robot joint values. Singularity operation errors and parameter operation errors often occur in this procedure.
These two operation errors increase the movement time of the robot and the coordinate reading time during the error recovery task in the ERES/CCS. An Error Recovery Expert System/World Coordinate System (ERES/WCS) was also developed so as to improve the disadvantages of the ERES/CCS. The ERES/WCS aimed to improve the operation by the control program by following studies. First, the WCS, a new coordinate system, in which the work space is regularly divided into equal motion lines by means of functional analysis of each robot joint about the relationship between the robot and the task object is constructed. Second, the linguistic variables for azimuth and posture are inputted by a human operator in order to reduce the burden of the human operator, and the fuzzy set theory is used to reduce the difficulty of the posture control. Third, heuristic rules are developed to calculate the robot joint values, which enable two operation errors not to occur and to make calculation time faster.
The ERES/WCS improves the operation by the control program and reduces disadvantages of the ERES/CCS by using these studies.

View full abstract

This paper describes a method to control position of a manipulator which is composed of active joints and passive joints. The passive joints have holding brakes instead of actuators. While the brakes are released, the passive joints are indirectly controlled by the motion of the active joints using coupled characteristics of manipulator dynamics. While the brakes are held, the passive joints are fixed and the active joints are controlled. The condition that ensures the controllability of the passive joints is obtained. An algorithm for PTP control of the manipulator is also presented. The feasibility of the method is demonstrated by simulation experiments for a two degrees of freedom manipulator.

View full abstract

The purpose of this paper is to construct a new hierarchical structure stochastic automata for applying to the problem in which the number of outputs and the contents of them are not known in advance.
The variable hierarchical structure stochastic automata (algorithm) is proposed, and the learning behaviors of it are shown, i.e., it is ε-optimal in any structure and that the probability of being selected the best output by it can be approached 1 as much as possible. In numerical simulation, it is known that this algorithm can be successfully used to make the mobile robot find the best path in the unknown structure maze.

View full abstract

Closed form solutions to the full order deadbeat observer and the deadbeat state regulator are derived for the multivariable descriptor system. With the deadbeat obsever-controller implemented for state regulation, the deadbeat action of the overall system is attained.

View full abstract