The surface pattern of granular materials in a storage vessel has undulations which vary in accordance with the supply and discharge of the materials. Therefore, in order to measure the content amount accurately, pattern recognition of the surface shape is important.
In this paper a newly developed measuring equipment, using an ultrasonic sensor for observation of surface patterns in storage, has been proposed.
The bimorph type transducer, whose resonance frequency and half width were 66kHz and 5 degrees respectively, was tilted to measure the inclination angle of the surface by detecting the maximum reflection intensity of the ultrasound. The time interval between the sound pulse emission and reception of both the surface reflection and the inclination angle were measured over the surface points. The data was then processed, taking into acclount the granular properties to depict the profile. In a symmetrical hopper, cross section patterns for the surfaces with flat, concave, and convex figures were estimated. The results were satisfactory with maximum measurement error of 1.0cm and 2.4cm for sand and nylonchips respectively.
The effect of the particle size and fractional voidage on the measurement was experimentally investigated. We found that the measurement accuracy increased as both the size and the porosity decreased, mainly due to the scattering effect of the sound on the surface. Fundamental characteristics, applicability of the instrument, and relation between measurement error and material properties were also discussed.

View full abstract

Discriminant analysis has many application fields: assigning patients to a disease, loan applicants to a risk category, potential purchasers to a market segment, and the like. The conventional discriminant problem is to determine a discriminant function, which maps a point in a multi-dimensional feature space to a point in a one-dimensional decision space, using a set of labeled points in the feature space.
In many cases, attribute values of each sample are not constant but fluctuating with time. For example, the temperature and the blood sugar level of a patient are not constant. In this paper, we represent these fluctuating attribute values of each sample by a multi-dimensional interval in the feature space, and propose a discriminant method for multi-dimensional interval data from given two groups. This method is based on an interval linear model which maps a multi-dimensional interval to a one-dimensional interval. A mathematical programming problem is formulated to determine the coefficients of the interval linear model and two group intervals which represent given two groups in the decision space. We show a computational algorithm to obtain the optimal solution of this problem using a linear programming problem. In order to discriminate unknown samples using the interval linear model, we propose a set of discriminant rules.
The proposed method is applied to a smell sensing problem. The interval liner model and two group intervals are calculated to discriminate between aromatic compounds and alcohol using response patterns of nine marketing gas sensors. The effectiveness of our method is demonstrated by this application.

View full abstract

A new pump stroke sensor has been developed for a totally implantable artificial heart system. Pump stroke sensor is essential for control and monitor of the artificial heart. The developed stroke sensor has the following characteristics; linear output to pump stroke, small size fitting into the implantable control unit, and long-term reliability. The pump stroke sensor is comprised of three permanent magnets of specific combination and a Hall effect sensor. The magnets consists of a permanent magnet disk and a pair of rectangular parallelepiped magnets placed on the magnet disk. The magnets are fixed on a pusher-plate of the blood pump. This specific combination of magnets provides the Hall sensor with linear output to pump stroke. The optimum design of the magnets was calculated from magnetic circuit analysis. Simulated result gave extremely good agreement with the experimental results. The new pump stroke measuring method has accuracy of better than 5%. In vitro experimental result of the motordriven artificial heart utilizing this pump stroke sensor showed excellent results.
The pump stroke measuring method using the specificaly designed permanent magnets is useful for not only the artificial heart system but also other electro-mechanical systems.

View full abstract

A new 3-D measurement method is proposed. We use the measurement of deformation of the projected grid pattern on an object. The three level grid is coded by a pseudorandom M-sequence to make it easy to identify the position on the object. We also use the gray image obtained by the projection of uniform light onto the object to normalize the gray level of the grid detected by a TV camera. The coordinates of the object points on the grid are calculated by triangulation based on the decoding of the M-sequence. This method has an advantage that only two images obtained by two projections are required. Experimental results show that the decoding is not influenced by the color and/or the reflectance of the object.

View full abstract

Generalized-Likelihood-Ratio (GLR) method is well-known for its most rapid failure detection speed and its easier implementation for the systems whose mathematical models are known. But, it has a drawback such that we must assume several hypotheses a priori which are able to model actual anomalies and failures well. The step hypothesis, which models unexpected anomaly vector function appearing in the system by a step function, has often been used for its convenience. But, we cannot always expect to have such a successful situation where the step hypothesis can be well applied. Thus, we previously proposed a GLR method based on the construction of a tracking function space so that any anomaly function due to dynamics or sensor faults is easily tracked and estimated.
The present paper proposes another convenient detection method also in the framework of the GLR technique. It is shown that the proposed method can detect any anomaly easily by recognizing the pattern of the curve of the maximum GLR which is calculated by the conventional step-hypothesized GLR method. This means that we introduce a concept of robustness into the step-hypothesized detection scheme. Because in actual systems the anomaly vector function cannot always be well modeled by a single step function, despite its appearance in the system equations constantly after the anomaly.

View full abstract

A design method of digital type 1 servo-systems is proposed for n-th order single-input single-output linear systems under perfect state observation and L-step delays. The structure of the proposed controller resembles to that of Mita's digital optimal type 1 servo-system but differs in the point that it contains a stable zero. By using pole assignment technique, the controller parameters are determined so that the poles of the closed loop system consist of L zeros and n arbitrary positions with cancellation of the zero in the controller. The computation of the controller parameters is easily performed since it requires matrix inversions only for upper triangular matrices. It is shown that the zero in the controller can be used to improve feedback properties without changing the transfer function from the reference to the output. In addition, effects of the delays on feedback properties are discussed. A simple numerical example is presented to illustrate the effectiveness of the proposed method.

View full abstract

This paper presents a new LTR method which recovers a robustness property at plant inputs to a specified degree. In LQG/LTR procedure a design problem of controllers for which the sequence of the open loop transfer functions of an LQG regulator converges to the open loop transfer function of the corresponding LQ regulator is examined, but for non-minimum phase plant this asymptotic property does not hold and the effectiveness of the procedure is not clear. In this paper, first, it is pointed out that the controller is assumed to be stable implicitly for LTR method so LTR method can be only applied to strongly stabilizable plants. Second, in stead of utilizing the asymptotic property, a difference of those transfer functions is estimated quantitatively by H^∞ norm and a design problem of an observer which makes the norm less than a specified value is proposed. The meaning of this setting is clear also for non-minimum phase plants. A condition for the existence of a solution is given as the existence of a positive definite solution of a Riccati matrix equation. The relation between this method and LQG/LTR method is clarified.

View full abstract

Differently from the flexible one-link arm case, the flexible multi-link manipulator is a nonlinear distributed parameter system, whose dynamics is described by a bundle of nonlinear coupled partial differential equations, with nonhomogeneous natural boundary conditions and geometric boundary conditions.
Even if the model is linearized, it is not easy to obtain the eigenvectors due to it's coupled terms, and properties of the eigenvalues generally can not be analyzed.
In this paper, under the assumption that each joint is rigid, a nonlinear lumped-parameter model including centrifugal and Coriolis forces is derived by appling Hamilton's principle and Galerkin's model expansion method. A control scheme called PDS (PD-Strain) control for flexible multi-link manipulators is presented.
The PDS control law is composed of a conventional PD control and a feedback of strain detected at the root of each link. Stability conditions of the closed-loop system are obtained by using the moment equilibrium equation for the root of each link and the direct method of Liapunov. The effects of stabilization of this scheme are also tested for a linearized model. Experimental results are provided for illustration of the effects of PDS control.

View full abstract

This paper proposes a feedback controller design method for linear time-invariant plants that are subject to perturbations of a physical parameter vector.
First, the paper investigates the relationship between stabilizing state feedback gains and positive definite solutions of a Lyapunov equation which play an important role in stability analysls.
Next, the paper describes a design procedure in which the positive definite solution of a Lyapunov equation is used as a design parameter.
The advantages of this method are as follows: stability of nominal closed-loop systems is automatically ensured, the nonlinear optimization to solve this problem is feasible and the lower bound of the damping rate of perturbed closed-loop systems is assignable.
Finally, it is shown that the stability of perturbed cloosed-loop systems can be recoverd by a LTR method when state variables can not be measured.

View full abstract

In this paper, we have proposed the simple compensator which can be easily adjusted in order to stabilize the system. The proposed method is fundamentally based on the inverse system of multivariable system. The proposed compensators are used by the feedback compensation, or the feed-forward compensation with feedback compensation. The feedback compensator is composed of the diagonal matrix. The element of diagonal matrix is the phase lead compensation with adjustable gain, and the structure of compensator is simple. Finally, the simulation results for design of compensators based on the proposed method are presented.

View full abstract

This paper deals with determination of a firing sequence of transitions in reachability problem of a Petri net. In general, the reachability problem can be solved by checking whether there exists a non-negative integer solution such that the matrix equation of Petri net holds true. Even if the solution exists, however, there exists a problem for finding a firing sequence which makes it possible to reach the given final marking on the places set of Petri net. The determination problem of a firing sequence is very important from the point of practical view, especially in the control of event-driven systems, in the failure diagnosis of sequence control systems modelled by means of Petri net.
In the paper, the method of determining a firing sequence of transitions is proposed. First, the determination problem discussed in this paper is defined. Second, the two kind of notions, that is, the reverse graph of primal Petri net and the transition matrix representing the paths of firing transitions, are newly introduced in order to determine the firing order of transitions. Third, an algorithm to determine the firing sequence by tracing reversely from the final marking to the initial one on the places set is presented. Finally, the method proposed in this paper is applied to the failure diagnosis of a simple type of sequence control system modelled by Petri net.

View full abstract

The issues addressed here are systematic intransitivity of uncertain preferential judgments and their descriptive modeling by fuzzy utilities. Throughout this paper the “uncertainty” of preferential judgments is associated with the degree of support for the judgments.
First the degree of support for pairwise comparative judgments was understood as a membership grade in a fuzzy preference relation. Then three levels of fuzzy transitivities are introduced, and we hypothetically supposed the conditions under which empirically well-known typical stochastic intransitivities occur are same for the violation of such fuzzy transitivities. In this context some experiments were designed as paired comparisons between multiattribute alternatives. Their analyzed results suggested the occurrence of several types of intransitivities of fuzzy preferential judgments.
Next, to comprehend the confirmed intransitivities, some properties on human judgmental processing were induced. Then, for representation of such properties the concepts of interactivity and separability between fuzzy utilities were employed. This makes fuzzified additive utility difference structures effective to describe the intransitive preferential judgments.

View full abstract

Job-shop type, time optimal scheduling problems for repetitive processes are modelled by the Petri net with timed places (PL-TPN). Assuming the stationary periodic solution, the optimal or the sub-optimal solution is obtained by LP and the branch and bound method.

View full abstract

The adjust rule with a dead zone is effective in the robustness of an adaptive observer of an unknown plant under uncertain input disturbance and output measurement noise. But it is difficult to decide the size of the dead zone. In this paper we propose that the size of the dead-zone is simply equal to a predetermined rate of the maximum output error between the plant and the adaptive observer. The stability of the system is considered and a computer simulation is presented.

View full abstract

This paper presents a variable adaptive law, which is robust to bounded noise. Instead of the dead zone which is proposed by Peterson, It uses σ-modified adaptive law in the small estimation error region for estimation of unknown parameters.

View full abstract

In digital image processing, calculations of arctangent are done many times in a limited time for extracting straight lines from a contour line by using a line-fitting method. We proposed a method of successive calculations of Arctangent (y/x) whose y and x have successive discrete values.

View full abstract