It is very important to analyze the characteristics of pneumatic pipe line connected to chamber for the purpose of designing pneumatic systems. Linearized mathematical models of the system under small pressure amplitude have been established by many authors. As for the actual pneumatic actuation systems, however driving amplitude is relatively high so that it gives the considerable temperature change, and the established models cannot be applicable to this phenomenon.
In this paper we make clear the unsteady phenomena in the pipe-chamber system with respect to air charging or discharging. Firstly by measuring the pressure and temperature response of the system with respect to air charging or discharging, we observe that the large change in temperature and ensuing heat transfer occur in the pipe. Secondly we propose a mathematical model which describes the phenomena well. The proposed model gives the special importance to the mass fllow according to the motion equation and the energy equation considering heat transfer which is caused by the temperature change. Comparing the numerically calculated response by using the model to the measured response, we show the validity and the applicability of the model. Finally we give some mathematical analysis of the system by the model.

View full abstract

Measurement methods in the space stations must be different from those under the earth gravity in many cases. The level meters for example, do not work under the micro-gravity condition and methods other than the level meters are required to measure the liquid content in the micro-gravity.
This paper treats the problems occurred when the pneumatic method is applied to measure the liquid content in a closed tank under the micro-gravity condition and describes the results obtained experimentally under the micro-gravity condition.
In the pneumatic method, the measurements are carried out by the following procedures. (a) Periodic change of the pressure in the tank subjected to the external periodic small displacement is measured. (b) The volume of the ullage gas in the tank is estimated by the measured pressure by applying the ideal gas low for the isentropic change. (c) The liquid content is calculated by subtracting the ullage gas volume from the tank capacity. This method is different from the level meters and thus can measure the liquid content in the micro-gravity condition. However the following essential problems for the measurement of the pressure change were not investigated when applied under the micro-gravity condition.
(1) What happens when many bubbles exist in the liquid?
(2) What happen when pressure sensor immerges in the liquid?
(3) How does the vaporized liquid influence to the pressure?
(4) How does the thermal time constant of the tank influence to the measurement?
We investigated these problems theoretically and Problem (1) is not essential. Further, we found Problems (2), (3), (4) can be solved by selecting the optimal frequency of the volume change. The maximum error of the measurement under the micro-gravity condition was 3% of the tank capacity of 13.99L.

View full abstract

There have been a number of interesting papers on H_∞ type control problems of sampled-data systems.
In order to desgin a digital controller to minimize the L₂/l₂(or l₂/L₂) induced norm of a closed-loop system, Chen and Francis gave some farmulas for the norms of SP and PH; P is a linear continuoustime system, S is an ideal sampler, and H is a zero-order hold.
Hara and Kabamba investigated the H_∞ type control problems of sampled-data systems in the state spase representations and they poposed the discrete-time systems, the H_∞ control problems of which have the same solusions as the original problems do. These discrete-time systems are here called to be equivalent to sampled-data systems.
This paper shows that the equivalent discrete-time systems can be derived very simply by characterizing the class of minimum energy disturbances.

View full abstract

The deadbeat control, also known as, finite time settling control is supposed to be effective exclusively in sampled data controls. In 1957, O. J. M. Smith proposed Posicast control to obtain the finite time settling response to a command input for continuous control systems. He mentioned in his paper the finite time settling response to a disturbance as well. His method is, however, impractical because inverse transfer functions of controlled objects would be required.
The author has found out a new controller configuration with a delay element for continuous control systems, which has the finite time settling responses both to a step command input and to a step disturbance. This new method is named Continuous deadbeat Control (CdbC) and was published.
This paper describes a design of CdbC. Conditions to realize controllers based on CdbC are shown, and the following is cleared. For a high order controlled object without unstable zeros, only one delay element is required in the controller independent of functions of a command and a disturbance.

View full abstract

Piston cylinder is the most widely used pneumatic actuator, but its behavior is very difficult to control in rapid or precise positioning. One cause of the difficulty is the friction between piston and seal. Especially, rubber sealed cylinder has relatively high friction which makes the driving force discontinuously change. Another cause is the low response of control valves.
Recent technology makes these causes out of data, that is, labyrinth sealed cylinder and fast response proportional servo valve are available. They have low friction and rapid response respectively. This paper deals with the control performance of the servo system used by these elements. Hence the rested problems, delay and overshoot caused by the compressibility of air, are now closed up in rapid control.
To improve the performance, generalized predictive control method, which is often used as one part in the self tuning control, is adopted in this system. The method is easy to implement and it gives the control input analytically.
Controlling the system by this method and an usual feed back method respectively, it is shown that the former is effective to compensate phase lag of the system.

View full abstract

This paper describes a new method to control a pneumatic cylinder with less overshoot.
The pneumatic cylinder has been used as an actuator for industrial robots, because it is simple and economical.
However, it is difficult to decide precisely the position of the piston rod, due to the air compression and the friction of the device itself.
There is optimal control in one of the positioning control methods of the pneumatic cylinder. This is the method that decides feedback gain by which the cost function becomes minimum. In this method the overshoot becomes great due to the selection of the cost function or the weighting matrix.
We propose a method that increases the amplifier gain after the control system has been constructed by the optimal control theory. Furthermore, we examine how the overshoot decreases by this method.
According to the experimental results, the overshoot decreases by increasing the amplifier gain and settling time is faster than in the case that the amplifier gain is 1. This method is also effective when the weight of the load changes.

View full abstract

The study has been done on the optimum design of mulling process for the foundry return sand, using a hybrid control system. First, mulling process of return sand is analysed, compared with the characteristics of new sand one presented in the previous paper. Secondly, the hybrid control system is proposed to optimize both control time and precision in the mulling process. Namely, feedforward control by Fuzzy theory is applied for first water addition conducted in the nonlinear region, while adaptive predictive control is used for the feedback control in the linear region. Finally, the usefulness of the proposed approach is demonstrated for the optimization of the return sand mulling process by control experiments.

View full abstract

This paper deals with a method of active control for suppressing vibrations of a flexible cantilever beam which is subject to a distributed random disturbance and a seismic input at the clamped end. First, the mathematical model of the flexible structure is established by a stochastic partial differential equation which describes the Euler-Bernoulli type distributed parameter system with internal viscous damping and subject to the seismic and distributed random inputs. Secondly, the distributed parameter model is reduced to a finite-dimensional one by using the modal expansion, and the resulting model is split into the controlled part and the uncontrolled (residual) one. Regarding the observation spillover due to uncontrolled part as a colored observation noise, an estimator is presented, and then the optimal control system is constructed. Finally, simulation studies are presented by using a real earthquake accelerogram data.

View full abstract

Safety monitoring systems have failed-safe (FS) failures and failed-dangerous (FD) failures that are caused not only by physical noise but also by limits of state cognition of sensors. To diminish those failures, this paper develops reliable safety monitoring system by the optimal combination of two types of monitoring units; safety monitoring unit and danger monitoring unit.
These two kinds of units are composed by a sensor with an indefinite state which is neither judged to be safe not dangerous for the object system. The safety monitoring unit interprets the indefinite state as a danger state, while the danger monitoring unit does it as a safe one. It is analytically shown that the optimal combination which minimizes an expected loss caused by FS-failures or FD-failures is specified by the only three parameters in a system composed by two units. Moreover, the domains of the parameters which optimize each combination are determined and illustrated. Numerical examples are also given.

View full abstract

There is a growing need for the top-down design methodology for Discrete Event Systems (DES). Establishing a firm foundation of such methodology is the central topic of this paper. In a top-down design of a discrete event system, one starts with a high-level model of the system and refines it stepwise until a sufficient low-level model has been reached. Therefore, in this paper, we study the following two things to make the top-down design useful: The first is a theory of formal models of DES which are suited to refinement, and the second is an adequate notion of refinement relation between the models. Firstly, in order to get a new DES's model, we extend the state machines, a well-known DES's model, by introducing the ‘composite event’ into the set of events. This enables us to refine an event in a high-level model by composing several events in a low-level model. Then, to get an adequate refinement relation, we firstly define a ‘canonical model’ called Proper-model which has the simplest state space in a set of models that have the same set of events and the same state transitions. The reason to define such a canonical model is that to refine a model effectively, it's always expected to refine the model with as fewer elements as possible. So, we define a refinement relation only between Proper-models. Finally, we show that the refinement relation defined in this paper has the transitive property which is indispensable for stepwise refinement.

View full abstract

In this paper, we consider the extension of the β-type learning automata to achieve conditionally optimal performance in a stationary random environment with some response characteristics of P- and Q-models.
The learning scheme of β-type automata is based on the finite models of the probability distributions which characterize the environment, as in the Bayesian learning. In this learning scheme, we have proposed a new class of finite models, by extending the probability distributions to the density functions which belong to the exponential family, so that the automata will act in S-model environment. Moreover, a new output function is defined to determine the rule for selecting an action of the automata. We have shown the conditional optimality of the new β-type automata, by using the martingale convergence theorem. Also, some useful properties of the β-type automata are discussed with simulation tests. It is shown that the β-type automata have a good performance on the learning curves which indicate the nature of evolution of the action probability that corresponds to the optimal action.

View full abstract

A task teaching for robots on a predictive display is an effective way to overcome the time delay problem in space teleoperation. Several graphic systems for this purpose have been introduced in teleoperation systems. In almost systems, however, operators manipulate the displayed environments unilaterally and cannot feel any constraint force during the teaching tasks. This reduces the operationability of the teleoperation system. To solve the problem, this paper proposes a method of virtually generating the constraint forces in the model world. It is assumed that the environment model consistes of several polyhedral objects without friction on the surface and a slave arm with a point at the tip. Also, the objects are assumed to have springs distributed on the surface which can produce forces normal to the surface. In such environments, the method first calculates the interacting location between the arm and an object in the model world. Secondly, at the location, how much the spring is pressed by the arm is calculated. Using the spring displacement and its stiffness parameter specified, the constraint force vectors applied to the tip of arm are estimated, and are transmitted to the operator through a master joystick with four joints driven by direct drive motors. In experiments using a predictive display in which the method was implemented, the operator was requested to manipulate the displayed arm and to trace the surface of a wall and a pillar models with the large spring stiffness constructed in the display, and it was found that the method could produce enough sensation to make the operator recognize the existence of a hard wall with reality. Finally, the predictive display was applied to a master slave system with the 0.5s time delay, and it was confirmed through a simple task that the display worked well to improve the operationability.

View full abstract

Fault diagnosis of printed circuit boards has become more and more difficult as the number of parts on one board grows larger considerably. The relationships between various test results and the truely failed part has become so complicated that only experts can locate the failed part using both the test results and their own experiences. Shortage of such experts and the low diagnosis accuracy for a newly designed circuit board are serious problems in this field.
In order to aid human in this fault diagnosis task, a learning method using a decision tree is proposed in this paper. Our inductive type learning method generates a decision tree, which represents a certain decision sequence for inferring the class, from the relational data of test results (attributes) and failed parts (classes). When this type learning is applied to such field as fault diagnosis where the attributes take discrete values, efficient decision rules generally cannot be obtained because there is no measure for the distance relationships of the attribute values. We develop a new algorithm for deriving such distance relationships only from the training data set. The conventional matching algorithm for class inferrence is also modified so that several ordered candidates of class can be indicated from a single set of test results.
The method is applied to the real fault diagnosis task, using the data obtained from the actual plant of manufacturing and testing certain printed circuit boards. The experiment results well confirm the effectiveness of the proposed algorithm. The fault diagnosis system based on the proposed method is already active in the plant, and it is reported that acceptable performance, though not comparable with human experts, is achieved.

View full abstract

The improvement of a control performance can be expected by actuator selections. However, for the combinatorial property of the optimal control problem with actuator selections, it is difficult to solve the problem. We study a sufficient condition to dissolve the combinatorial property.

View full abstract

In this paper, it is shown that the addition of derivative term to SAC algolithm significantly improves the tracking performance of MIMO SAC systems. Especially, this method is effective to SAC system with parallel feedforward compensator.

View full abstract

When the rotor of a step-motor returns after going forward by a microstep technique, the hysteresis error is appeared, that is, each path of the rotor between the going and returning is different.
In the present work, the precise micro-step technique reducing the hysteresis error is investigated using a four phases VR step-motor. Using the method that all coils of four phases are simultaneously excited, the micro-step operation within the error of 2% to one fundamental step is realized.

View full abstract