In factory, it is very significant to inspect products automatically. This paper treats automatic inspection of compressor which human does at present. It is hard and difficult to inspect because the difference between good and no-good is very little. We treated the following abnormals: blade error, cylinder's flaw and dust insertion. As the mechanism of abnormal sound production is different, different driving method is required. This paper proposes two methods respectively.
Though the length of a cycle of compressor's sound is stable in the case of normal compressor, the length of blade error is not stable. From this fact, subtraction between nearby cycles is available in inspection of blade error. In the case of cylinder's flaw and dust insertion inspection, variance of wave gives us useful information. When cylinder's flaw of dust insertion occurs, it makes high frequency. To recognize it, variance of wave is effective. The wave has two part: one is stable and the other is not stable. One cycle is divided into 64 blocks and variance of each block is calculated. Block is grouped into two according to being stable or not stable. Two parameters are calculated by adding variance in each group. The combination of these two parameters makes possible to recognize. From experiment, 100% is recognition rate in the case of inspecting about blade error and 93% is the recognition rate in the case of cylinder's flaw and dust insertion.
The automatic factory system can be constructed by automatic acoustic inspection proposed in this paper. The possibility to recognize the error which human can not recognize has a significant meaning in the industry.

抄録全体を表示

A new approach to scanning system for sensor array is proposed. In the multiplex stage of the proposed system, output signals of sensors are modulated by orthogonal sequences, summed to a serial signal, and quantized. The modulating sequences are so selected that the signal power from the sensor is timespread. In the demultiplexing stage, the quantized sequence are multiplied by demodulating sequences and integrated with respect to time.
In this paper, the principle of scanning is described, and the method using Maximal-length binary-sequences (M-sequence) is proposed. Noise effect is evaluated for given length of M-sequence and number of sensors. From the evaluation, we conclude that the proposed system improves the signal to noise ratio compared with conventional time-division-multiplex.
We discuss the application of the method to ultrasound sensor array. Experimental system is constructed using 16 channel ultrasound microphone array and analogue switches. Experimental results demonstrates the ability of the method to data scanning for coherent sound source localization. And computer simulation which assumes the same experimental set up indicates the additive-noise reduction property of the method.

抄録全体を表示

This paper proposes to evaluate the elastic components of force sensors based on three criteria, that is, the strain gage sensitivity, the force sensitivity, and the minimum rigidity. The strain gage sensitivity is defined as the Euclidean norm of a row vector of the sensor compliance matrix. The force sensitivity is the minimum singular value of the sensor compliance matrix. The minimum rigidity means the inverse of the maximum singular value of the compliance matrix at the point to which forces and moments are applied. When the number of strain gages is larger than six, there is variety in inverting strain gage signals into the force information. By using this fact, it is also proposed to design the elastic component so that a Wheatstone Bridge should work as the decoupling circuit of the force information. A six axes force sensor is designed as an example to show the effectiveness of the proposed criteria and signal processing method.

抄録全体を表示

Direct measuring method for soil-water flow rate is developed theoretically and its availability is examined by experiments. In the present paper, temprature response of soil system is analyzed based on a heat equation for mixed material which was developed by the same author. All of soil-water does not move in soil. Some amount of water is trapped on surface of and between soil particles. A ratio of moving-water-content to total-soil-water-content cannot be measured. By the developed method, however, a product of average velocity of soil-water and the ratio can be measured. This product is sectional average flow rate of soil-water.
The method utilizes a phase difference between measured signals at two different sensoring points responding to input sinusoidal temperature signal. Thermal diffusivity of soil should be known in application of the method. In field application, however, the diffusivity changes with such circumstances as velocity of soil-water and soil temperature.
The temperature diffusivity can be eliminated by two phase differences corresponding to two different angular frequencies of input. Square wave input was adopted in the experiments, because the wave includes higher harmonics of primary frequency. Therefore, measured temperature signals are processed by Fourier analysis method to get phase differences corresponding to the primary and the third higher harmonics.
Measured flow rate showed good accordance with actual flow rate in the experiments.

抄録全体を表示

Developments on the ultra-sonic flowmeters, which are used widely in the water loop, have been carried out for the application on the sodium loop in fast breeder reactor. This is because of that ultra-sonic flowmeters have several characteristics needed to the large pipe sodium flowmeter, i.e. good linearity, light-weight, leak tight, low cost etc..
After fundamental improvement about the settling method of transducer to high temperature duct and about measurement circuit were developed, in-water tests were conducted and effectiveness of such improvements were confirmed.
So 12B and 24B ultra-sonic flowmeters were made after additions of several improvements to high temperature problems, and tested in the sodium loop, and it is confirmed that the ultra-sonic flowmeters have superior characteristics as the sodium flowmeters. In the sodium temperature (200∼400°C) and flow rate (0∼6m/s), linearity and repeatability were within ±1%. And effect of upper flow straight pipe length to the ultra-sonic flowmeter output was smaller than in water loop tests.
This paper describes mainly the above results of sodium flow tests which showed that ultra-sonic flowmeters are useful as the sodium flowmeter.

抄録全体を表示

In this paper the consistent estimation problem of single-input-single-output continuous time systems is considered. For a given continuous-time system an algebraic equation containing parameters of the given system can be obtained via Block Pulse Functions (BPF). Thus these parameters can be estimated using the well known Least Squares (LS) method if there is no any measurement noise or the noise level is very low. But in the case that system input-output measurements are contaminated by noises, even if white noises, satisfied parameter estimates cannot be got by the LS method.
In this paper the reason of the above problem is analyzed. To solve this problem the idea of Extended Least Squares (ELS) method in discrete-time system estimation problem is applied to a new kind of discrete-time model, which is derived using one of the properties of BPF transfermation, to get a recursive estimation algorithm. Provided some weak conditions are satisfied, fairly good estimation results can be obtained by this method. The asymptotic property of the newly developed method is analyzed, numerical examples are also given to show the efficiency of the proposed method.

抄録全体を表示

The conditions of existence of a unique, real, symmetric and non-negative definite solution of the matrix Reccati equation is derived when the Hamiltonian matrix composed of coefficient matrices of system equations has eigenvalues on the imaginary axis in the complex plane.
Such conditions are precisely determined based on the concept of the detectability and quasi-stabilizability.
The method established in this paper is useful in deriving stationary solutions of matrix Riccati equations when sinusoidal or constant signals are estimated by mean of Kalman filters.

抄録全体を表示

This paper is concerned with robust stability of the predictor-based optimal regulators for discrete-time linear systems with input delays. First, guaranteed stability margins are obtained from the minimum singular value of the return difference matrix at the input of plants. The influence of the input delays on the guaranteed stability margins is discussed and some useful observations are given. Second, the robust stability for perturbations of the system matrix and/or the gain matrix is considered. Admissible bounds for the perturbations are obtained by using Lyapunov function for the extended system. However, it is pointed out that this method provides extremely conservative bounds. An alternative method for obtaining less conservative bounds is proposed under certain restriction on the nominal system. Numerical examples are given to illustrate the influence of the input delays on the stability margins and the admissible bounds.

抄録全体を表示

This paper considers the problem of designing digital sliding mode control systems. Sliding mode control schemes are often implemented in digital fashion because the complex switching logic can be easily carried out. In digital control systems, the ideal sliding mode does not occur since the structure cannot be switched during the sampling interval, however there can be exist a motion which is confined to a region including the sliding surface and proceeds to the origin along the surface. Such a motion is referred to as quasi-sliding mode. Itkis gave a condition of taking place quasi-sliding mode in a linear digital control system, but it is conservative and much complicated. In this paper, a new sufficient condition is presented, in which the upper bound of the sampling period and the upper and lower bounds of the magnitude of feedback gain are shown explicitly. Based on this result, a practically useful design procedure is proposed.

抄録全体を表示

This report is concerned with the forward and inverse kinematics problem relevant to a six-link robot manipulator. In order to derive the kinematic relationships between links, the vector rotation operator was applied instead of the conventional homogeneous transformation.
The exact algorithm for solving the inverse problem was obtained by transforming kinematics equations into a polynomial. As shown in test calculations, the accuracies of numerical solutions obtained by means of the present approach are found to be quite high. The algorithm proposed permits to find out all feasible solutions for the given inverse problem.

抄録全体を表示

Dynamic control is one of the methods which enable quick and accurate motion of robot manipulators. Dynamic control has been difficult to be performed real time by using presently available microcomputers since it requires huge amount of computation. Various efficient recursive computation algorithms have been proposed, especially for inverse dynamics (ID). However, it requires not only ID calculation but also direct kinematics (DK) and inverse kinematics (IK) calculation. We proposed efficient recursive algorithms for DK and IK and reported that the total amount of calculation can be reduced by eliminating duplications included in DK, IK and ID.
In this paper, new computational algorithms for DK, IK and ID are formulated so that they should include as many same physical values as possible. The total amount of computation can be reduced up to 42% in multiplication and 33% in addition by eliminating duplicated computation. These algorithms are useful for the computational part of versatile robot controller or dynamic robot simulator.

抄録全体を表示

In order to improve the quality of the program control for a batch process it is necessary to use some adaptation technique which can adapt the controller to the changes in process dynamics. Recently, it has been proposed to adopt the MRAC for this purpose. Since, however, a rough prediction of the changes appearing in a batch process is usually an easy task, it seems more suitable to adopt the programmed adaptation in place of MRAC. From such a point of view it is worked in this paper to apply the programmed adaptation to the problem of concern.
A design technique of the digital programmed adaptive PI-controller containing some nonlinear action is developed. The linear PI-action is designed on the basis of the dominant root specifying technique in which the damping ratio and the undamped natural angular frequency of the dominant roots are specified. The specification of the latter value is changed according to change in the mean delay time of the process. As the control error exceeds a threshold value the control is switched to an on-off control and in addition the integration is interrupted to prevent the process from the reset windup.
Three numerical examples about the reversible reaction, the consecutive reaction and the stylene polymerization reaction taking place in the stirred-tank batch reactor are worked. The results indicated always good control quality and robustness.

抄録全体を表示

This paper is concerned with a new scheme of independent control of two induction motors which are connected in parallel to a single inverter. This method is developed as an AC motor alternative of a currently-used DC motor propulsion system for an unmanned ground carrier vehicle, which is equipped with a pair of identical motor-wheel driving mechanisms at right and left sides of the body. This type of the ground vehicle, named as a Powered Wheel Steering (PWS) Vehicle, because it can be steered by appropriately controlling the speed difference of each motor together with attaining the propulsion force by adjusting the average speed of the two motors, is playing an important role in the recent factory automation as a flexible carrier vehicle.
The paper firstly investigates the required basic running functions of a PWS vehicle and the proposed hardware to realize these functions is described. Followed with the above discussions, the speed control strategy to achieve the independent control of the two motors, under the limitation of the proposed single inverter utilization, is presented. The closed loop speed control program of motors is developed based on this control strategy. The experimental run of the test vehicle, with the specification which is almost the same as a currently-adopted 100kg load-weight carrier vehicle, is executed to examine the effectiveness of the proposed propulsion method.
The results, obtained for three running functions for a straight course, a curved course with a fixed radius of rotation and a spinturn which is defined as a turn of the vehicle arround the tread center, indicate that the presented method of induction motor propulsion assures a satisfactory operation for the vehicle control.
Furthermore, a self-sustained steering control, where the vehicle itself estimates the present position and generates the steering command so that the vehicle follows the pre-determined course on the floor, is tried for the straight course running. The result demonstrates the possibility that the vehicle will be well controlled under the presence of the guidance commands.

抄録全体を表示

A three-dimensional heat conduction model for laser material processings was developed using the finite difference method. The model was represented in a moving slab which was divided into several unequal volume elements, where stationary continuous wave laser beam interacts with a surface of the slab. The model could evaluate a key-hole effect in a laser welding process. Typical four kinds of converged laser beam profile were considered in the model because laser beam profile has critical influence of laser processing characteristics. In order to verify the validity of the model, outputs were compared with the analytical solution of Jaeger. To show more practical applicability, laser surface hardening and laser welding experiments were conducted using 15kW continuous wave CO₂ laser. This experimental results were compared with the model outputs, and good agreement was obtained.

抄録全体を表示

The authors propose a new fuzzy decision making method for multi-objective problems. The main feature of this method is the usage of fuzzy connectives to represent the decisionmaker's preference structure. The fuzzy connectives which we adopt are Zimmermann's γ operation and its extensions. These extended operations have much larger descriptivity of the decision-maker's preference structure than the γ operation.
Our method is summarized in the following 5 steps.
Step 1. The identification of a decision making problem with a fuzzy environment and the construction of an objective hierarchy.
Step. 2. A questionnaire procedure: Questionnaire is used to identify the decision-maker's preference structure in each cluster of the objective hierarchy.
Step 3. A parameter identification procedure: The parameters of the three fuzzy connectives: γ operation, extension 1, and extension 2 are identified by Quasi Newton Projection Method using the questionnaire results. Then one fuzzy connective is selected among the three as the preference structure model on the basis of AIC's value in each cluster.
Step 4. Evaluation of alternatives: The decision-maker assesses the evaluation value of alternatives with respect to the lowest-level objectives. If he cannot assess the unique value, then he uses a fuzzy probability representation.
Step 5. Rating and Ranking: The evaluation values with respect to the lowest-level objectives are aggregated using the preference structure model from the bottom level to the upper level in the objective hierarchy. And then the alternatives are ranked.
Finally, an application of this method is presented in a company-choice problem for students.

抄録全体を表示

Human has two eyes and see objects binocularly. The three dimmensional sensing capacity of human depends on both the eyes and the neuron system behind. In this paper, we made model of human visual system including the eye movement and tried to realize the capacity in a artificial sensor. For this, visual system is separated in two parts-the Eye System and the Neuron System. The former is realized by the sensor mechanisms which is similar to human eyes and the latter by the Parallel Processors System which is consisted of 64 micro-processors. This system can be used as passive three dimmensional sensor in unarranged surroundings such as outdoor.

抄録全体を表示