In this paper, a new multiloop self-tuning PID controller design scheme is proposed. The proposed control scheme has a static matrix pre-compensator in order to reduce the interaction terms of the process transfer function matrix. The static matrix pre-compensator is adjusted by an on-line estimator. The p×p pre-compensated system is then controlled via 'p' self-tuning PID controllers, whose parameters are adjusted by a second identifier placed around the pre-compensated plant. The PID parameters are calculated on-line based on the relationship between the PID control and generalized minimum variance control laws. The proposed scheme is experimentally evaluated on a 2×2 temperature plus level control system. Experimental results illustrate the effectiveness of the new scheme.
Problems of finding linear systems which are consistent with given input-output characteristics are important in system approximation, model reduction and system identification. A conventional problem among them is the partial realization problem. In this paper, we consider a problem of finding MIMO linear discrete-time systems interpolating given input-output characteristics. The input-output characteristics are given by coefficients of the Taylor series of the transfer function at some complex points in the unit circle. Thus our problem is a generalization of the partial realization problem. Here, we provide a parametrization of all the interpolating systems with a norm constraint and a parametrization of all the stable interpolating systems.
In this study, the extended H∞, control, proposed in Ref. 6), is applied to the frequency control of a multi area power generation plant. More specifically, we consider a two area electric power generation system, with two generators, two loads and a tie-line connecting them. The purpose is to reject the frequency deviations due to step load disturbance. This is an interesting control problem because of the existence of a zero at s=0 in the multivariable plant that imposes a constraint on the steady state output y (∞). The paper discusses how to construct a suitable generalized plant for such a problem in order to design an H∞ optimized control system that is able to reject step load disturbances asymptotically. Using the H∞ approach, the controller is derived to achieve the disturbance rejection and to be robust to parameter uncertainty of the plant. The performances of the extended H∞ controller and the LQG controller are compared. It is shown that the H∞ controller is superior to the LQG controller in both nominal case and perturbed case.
In this paper, the separability condition that the separation hyperplanes pass through rectangles surrounding the input sets is formulated by normal vectors. Then distributions of the normal vectors satisfying the condition are depicted for the two-dimensional case. These distributions elucidate that the condition for the first hidden layer varies significantly even when the input patterns are simply translated, and that the conditions (in a wider sense) for the other layers are different depending on whether the unit is unipolar or bipolar, i.e. whether it activates from 0 to 1 or from-0.5 to 0.5. The initial distributions of the normal vectors with the weights initialized ordinarily by random numbers with zero mean are also depicted. Comparison of the initial distributions to the separability conditions leads to the conclusion that the bipolar nets are superior to the unipolar nets in convergence of the back propagation learning initialized in such an ordinary manner. The bipolar nets exhibit better convergence than the unipolar nets even if their input space divisions in every layer by the separation hyperplanes are geometrically the same at the outset of the learning.
A GMDH type neural network algorithm which can identify a nonlinear system whose structure is very complex is proposed. In this algorithm, the heuristic self-organization method of the GMDH algorithm is used and so a neural network structure which has a good prediction accuracy can be organized automatically. Furthermore, instead of the partial polynomials which are used in the GMDH algorithm, partial-neural networks which have four layered hierarchical network structures are used to construct the GMDH type neural network. The proposed algorithm is applied to a nonlinear system identification problem and its advantage over the GMDH algorithm and the conventional networks is shown.
In this paper, we propose a new algorithm for on-line adaptive identification of dynamics of general motor servo systems including winding machines. This algorithm is then incorporated into a feedforward controller to realize the high speed tracking control for a prototype of winding machines. The parameter convergence as well as the stability of the closed-loop tracking system is proved theoretically, and simulations and experiments are conducted to study the feasibility of the adaptive indentification algorithm and the control performance of the feedback loop. It is found that the algorithm not only guarantees the parameter convergence, but also yields satisfactory transient responses of the identified parameters. This enables us to use the simple feedforward controller to realize high speed tracking control of servo systems.