In the study of direct design methods using plant transient responses, model matching methods have been proposed. Model matching is useful for the specication of the closed-loop property from the reference input to the plant output, but the main problem is that information about the plant dynamics is necessary to select the reference model and hence the selection often requires trial and error. We have proposed a loop-shaping method that aims at disturbance rejection with a stability margin. The loop-shaping method can be applied to stable plants with less information about the plant dynamics, but it cannot specify the closed-loop property. In this paper, we propose a method of making an appropriate reference model based on the design result of our loop shaping method. In our proposed method, the reference model is a second order system with time delay, and it is selected so that it may have similar properties with respect to the maximum sensitivity and the low sensitivity at low frequencies of the feedback system designed by the loop-shaping method.
This paper deals with a multiple-input multiple-output dualrate system, where the sampling intervals of the plant outputs are an integer multiple of the hold intervals of the control inputs. In such a multirate system, ripples may occur in intresample behavior even if sampled outputs converge to reference inputs. In order to resolve this problem, a control law is enhanced using free parameters. As a result, ripples are eliminated in the steady state without changing the sampled outputs using an existing designed control law. Furthermore, a design method in state-space representation is proposed. In this paper, an observer is used because all state variables cannot be observed. Finally, numerical examples demonstrate the effectiveness of the proposed method.
Data-oriented controllers are emerging as an alternative to model-based controllers designs, due to the fact that data-oriented controllers enhance the closed-loop behavior of the system by using very simple structures. This paper studies how to evaluate the control performance of a closed-loop data as well as how to adjust the control parameters to improve the control performance. PID control method based on Generalized Minimum Variance Controller (GMVC) is considered. When the control performance is deteriorated due to changes in control parameters, the closedloop data are utilized to re-adjust the direct system parameters. Simulation results are provided to demonstrate the effectiveness of the proposed control scheme.
The idea of control performance assessment is becoming very important in the process control area. However, the low control performance index denotes only the deterioration of control performance. The strategies of the parameter retuning are not suggested. And, the control performance of the new controller can not be estimated, because the calculation of control performance index needs to apply the new controller into the real systems. In the real plant, when the performance index shows the bad value which mean the control performance is deteriorated, the retuning of the controller is done by plant operator based on their experiments. However, if this retuning is failed, the performance of this plant is more deteriorated. This paper proposes the retuning method of PID controller based on the controller performance assessment. The objective of the proposed scheme is to close the target performance of PID controller when controller performance deteriorates. The retuning mechanism is based on the estimated controller performance with the on-line system identication. Finally, numerical evaluations demonstrate the practically and utility of this idea.
The predictive controls are one of the major control strategies in the process control. However, in general, an exact model of the control system is required in order to predict the system's outputs accurately. In this paper, we propose a design method of an adaptive output predictor based on an adaptive output estimator with a simple structure, and an adaptive predictive control system design method using the proposed output predictor will be proposed for uncertain controlled systems.
In this paper, for plants with unknown sensor failures, a new design method for a self-repairing control system with a nonlinear detection filter is developed that can automatically replace the failed sensor with the backup. The nonlinear detection filter can be stabilized by the controller in the control system if there is no failure. However, it has a finite escape time in the faulty situations. Then, from this blowup property, by detecting the exploding signal of the filter, a sensor failure can be exactly detected within an arbitrarily prescribed time earlier than the finite escape time. Furthermore, to improve the control performance, the high-gain feedback controller is presented to cope with nonlinearity of the detection filter.
The direct control parameter tuning methods derive the control parameters from the one-shot experimental input and output data with no need of help from the plant model. The paper proposes a direct control parameter tuning method which employs the variance evaluation for the performance criterion to solve the disturbance attenuation problem. The proposed method tunes the control parameters so that the variance of the plant output approaches the variance of the reference model output. The proposed method can avoid an additive experiment where test signals are fed to the closed loop system for collecting the initial input and output data. The feature facilitates applying the proposed method to the industrial applications. Finally, the efficiency of the proposed method is demonstrated from a numerical example.
In this paper, we consider the simultaneous improvement of a controller and a model for an integral type servo system with a full state observer. Particularly, by employing the fictitious reference iterative tuning (FRIT), we give an effective method of a parameter tuning with only one-shot experimental data for the improvement of a controller that achieves more desired tracking property than the initial controller and a model that reflects the dynamics more accurately than the nominal model, respectively. Finally, we give an illustrative example in order to show the validity of the proposed method.