A new algorithm is proposed to estimate parameters of autoregressive (AR) processes with output data corrupted by white noise. This algorithm is based on the bias compensated least-squares (BCLS) principle which is introduced by Sagara and Wada for estimation of pulse transfer function models. The BCLS method requires the estimate of the unknown variance of the noise to compensate the biases of the LS estimates. For this end, a variance estimation algorithm is developed to obtain the consistent estimate. This algorithm is joined with the BCLS algorithm to provide consistent estimates of the AR parameters. The algorithm proposed here is a modified version of Sakai and Arase's algorithm which often gives negative values of the variance estimate. Several simulation results indicate that the proposed algorithm is more effective than Sakai and Arase's from statistical and computational points of view.
Suspention of a rotor in totally active DC-type magnetic bearings, which is unstable without any control, can be stabilized by feedback control. Existing controllers for the magnetic bearings are composed of analog circuits, because the stabilization requires high-speed-control. But it is difficult for analog controllers to give the magnetic bearings various functions such as force-sensing, rotor-positioning and variable suspention-stiffness as well as contactless suspention, because of the lack of flexibility and adaptability of analog controller. The authors have developed a digital controller for a totally active magnetic bearing system. The controller has two CPU : one is a digital signal processor (DSP) to process the feedback control at high speed for stabilization, and the other is a personal computer to process the flexible and adaptable control. The authors have demonstrated through experiments that the controller enables the magnetic bearing system both to be stable and to have various functions.
Recently the possibilistic linear models whose coefficients are fuzzy parameters have been proposed by Tanaka et al. The features of this model are to be able to deal with fuzzy data and to represent a fuzzy relation between input and output variables. In this paper, we extend the possibilistic linear models to nonlinear interval models in order to analyze the complex system using interval data. This extension is achieved by proposing the interval GMDH. Two interval regression models are obtained by this method. One model is obtained so as to include interval data. Thus interval data are included in the intervals estimated by this model. Another model is obtained so as to be included in interval data. Thus interval data include the intervals estimated by this model. Interval data from grinding experiments of fine ceramics by diamond grinding wheels are analyzed to illustrate the proposed method.
In the design of digital control systems, the time delay due to computation time of a processor cannot be often neglected. In this paper, we consider the problem of synthesizing a type-1 optimal digital servosystem taking account of controller delay of one sampling period. It is first shown that the optimal digital servo problem in which a quadratic criterion function is defined for a given continuous-time plant can be reasonably reformulated to a discrete optimal regulator problem. To assure the solvability of this problem, some conditions for the preservation of stabilizability and detectability of the plant under sampling are given. In the discrete optimal regulator problem considered here, not only control input but also the initial state of the servocompensator are referred as free design parameters. The obtained optimal digital control law contains a feedforward term of reference signal in addition to a feedback term. It is also shown that the optimal control law has a function of estimating the state variables at the following sampling period in order to recover the controller delay.
This paper will introduce a new sensor coupling device for milling machines, drilling machines and for the machine center. Within the industry there has always been the need for a practical application of the sensor coupling device to detect the cutting process related signals from the “tool side”. A mounting device which enables the sensor to be fixed close to the rotating multipoint cutting tool has been developed to meet this need. Thus a more practical method for detecting the tool fracture of twist drills and/or tool chipping of an end mill cutter becomes possible. The machining process associated signals are high frequency Acoustic Emission (AE) signals detected from the spindle top of the machine being used. In this study two kinds of AE-signals will be compared. The first are those detected from the spindle housing. The second are signals detected from the spindle top using the newly developed sensor coupling device. Furthermore, the structure of an AE-monitoring system for supervising multipoint cutting tools will be described. In addition, results of experiments which monitor the tool health coefficient, derived from the power spectrum of the machining signals, will be discussed.