This paper discusses a new roll eccentricity control method which has the following features ; the roll eccentricity is modeled by a sine wave equation for each roll and the eccentricity parameters, such as amplitude and initial phase angle are estimated using the recursive estimation technique. In this method, a square root filter is used to prevent the parameters from diverging. In addition, Kalman filter and exponentially weighted least square estimation are applied to decrease the estimation error due to the variance of observation noise and the gradual variation in force offset respectively. Furthermore, the convergence of parameters is automatically checked observing the trace of error convariance matrix. This control method is aplied to the first stand of tandem cold mill as a roll eccentricity control system. The convergence of recursive computation is confirmed and both the force and thickness deviation due to roll eccentricity have been reduced to less than half.
A linear cursor key whose cursor speed on a CRT display is proportional to the force applied on the key top is presented in this paper. Three results were obtained as follows. Firstly, the positioning time by the linear cursor key can be made shorter than the time by the standard cursor key in the difficult positioning task with smaller size of the targets and longer distance to the targets. Secondly, positioning behavior consists of the approach behavior and the settlement behavior. In the approach behavior, subjects move the cursor to the edge of the targets as fast as possible. In the settlement behavior, subjects tap the cursor key repeatedly until the cursor moves in the targets. Each keystroke is performed without feedback control, because the duration of its behavior is shorter than the reaction time of the subjects. Finally, faster cursor speed is suitable for the approach behavior. On the other hand, slower cursor speed is suitable for the settlement behavior.
In this paper, a low sensitive position control system driven by a dc servo motor using a learning identifier is proposed. The low sensitive position control system is designed for the change of moment of inertia and the learning identifier is used to make up the inverse system of the low sensitive system. The responses of the system are investigated experimentally. Moreover, these responses are compared with the responses of a proportional position control system using the learning identifier. As a result, it is confirmed that the responses of the proposed system are significantly improved by the learning identifier and have low sensitivity for the large change of moment of inertia about fourteen times as large as that of the motor itself. Therefore, the proposed system has a good responsibility for the change of moment of inertia without re-learning.
Stability of a model reference adaptive positioning servo system which have adaptive velocity loop and nonadaptive position loop is discussed. Such structure of model reference adaptive controller has been applied to real systems and some successful results have been reported. Although asymptotic stability of the system with gradient type adaptation law can be proved under the ideal conditions, the system become unstable when a disturbance exists inside the position loop. In this paper, stability of the system is analyzed by singular perturbation method, supposing that the speed of the adaptation is small enough. Further, the system with two different adaptation laws proposed to improve the stability of the system is analyzed similarly.