Abstracts integral equations. The estimation process is described by a linear integral equation of the second kind. An example of estimation in stochastic systems is considered. Simulation results are presented.
089 Discrete Frequency Formats for Linear Differential System Identification A.E. Pearson, Y. Shen, J.Q. Pan, pp 417-422 The classical Shinbrot method of moment functionals is reexamined for linear differential system identification using explicitly defined Fourier modulating functions. It is shown that a discrete frequency format attends the following least-squares formulations: (i) the parametric identification problem for singleinput single-output systems, with consideration given to both timeinvariant and time-varying models, and (ii) the nonparametric frequency transfer function identification problem for muhivariable systems. The algorithms are demonstrated to be robust in the presence of white measuremeut noise, and a comparison is made via simulation with the prediction error method for parameter identification. 090 Time-Optimal Stochastic Positional Control P. Kulczycki, pp 423-428 This paper deals with the synthesis of time-optimal positional control for objects whose dynamics are described by a different inclusion with discontinuous right-hand side. The probabilistic concept of solving that problem is presented. The existence and characteristic of the time-optimal control and solutions of the random differential inclusion generated by that control, which have been proved in recent work, are commented upon. Finally, remarks concerning the practical consauction of a suboptimal controller based on the above results are formulated.
127
094 Optl~ai Amdllary Input for On-Line Fault Detection and Fault Diagmsis K. Uosaid, N. Takata, T. Hatnnak% p p 441-446 An optimal auxiliary input is introduced to detect system faults quicHy. The auxiliary input is designed to enlarge the distance measured by the Kullback discrimination information, between the system models corresponding to the normal and the fault modes in the single-fanit-mode case. Furthermore, the idea is extended to the multiple-fault-mode case (fanlt-diagnosis problem). Numerical simulation results indicate that the proposed auxiliary input reduces the mean detection time in fault detection using the backward sequential probability ratio test, without having much effect on the original system behaviour, and on false alarm and diagnosis rates.
095 Fault Tolerant Control in the Presence of Noise: A New Algorithm and Some Open Problems A.W. OHbrot, pp 447-4150 A novel approach to noise-insensitive fault-tolerant control is proposed. It is assumed that a system model is known in each failure mode. Adopting the robust simultaneous stabilization as a starting point, a new algorithm with learning capability is developed. All models are assumed to be simultaneously validated on a parallel computer, while their outputs are compared to noisy measurements. A validation algorithm rejects models contradicting the real data. The valid models are used by a system stabilizer, a periodic simultaneous stabilization algorithm which can handle dramatic system changes. A two-dimensional example (an oscillator with unknown frequency) is used to demonstrate the excellent properties of the proposed algorithm. 096 PID Control Revisited P. Persson, K J . Astrrm, p p 451-454
091 Robust Nonlinear Observer-Based Fault Detection for an Overheed Crane R. Seliger, P.M. Frank, pp. 429-432 The concept of nonlinear unknown input observers is utilized here for robust component and instrument fault detection for an industrial overhead crane. The major model uncertainty is modelled as an unknown input signal. Because it is possible to decouple the model completely from this uncertainty by using a nonlinear state transformation, successful supervision of the crane is possible without any information about the actual load mass. Based on the decoupled model, a nonlinear unknown input observer is designed, allowing for the generation of residuals that remain unaffected by load mass variations. These robust residuals are subsequently used for instrument and component fault detection. The concept is tested by laboratory experiments. 092 Fault Diagnosis of Dynamic Systems Using Neural Networks T. Sorsa, J. Suontausta, H.N. Koivo, p p 433436 A method for detecting faults in nonlinear systems is developed using dynamic nonlinear predictor models which are realized with neural networks. Separate predictor models are identified for the process operating normally, and for situations where one of the potential faults has occurred. In monitoring the state of the process, the best fitting predictor of the bank is selected according to the Bayes rule. Radial basis function networks are used in identifying dynamic predictor models and the parameters of the networks are estimated with the orthogonal least-squares algorithm. The performance of the proposed method is demonstrated in simulation studies of a jacketed reactor.
093 Nonlinear Filtering Schemes for Continuous-Time Stochastic Hybrid Systems G. Kalmanovlch, A.H. Haddad, p p 437-440 Hybrid systems are used to represent systems that switch their dynamic behaviour in response to failures, or external and internal variations. This paper considers stochastic hybrid systems in continuous time, which switch among a finite number of known stochastic models based on an underlying finite-state Markov process. The transition probabilities of the process are assumed to depend on the system states. The representation of the optimal filtering equations for the system observed in white Ganssian noise is derived. Approaches to suboptimal implementation of the filters are then considered.
This paper presents a new method for tuning PID controllers. Specifications are given in terms of load disturbance attenuation, set-point following, robustness and noise sensitivity. The method is based on placing a few closed-loop poles. An implementation of the design procedure is presented, and appfications of the method are given. 097 Dynamic Transfer Among Alternative Controllers S.F. Graebe, A. Ahl~n, pp 455-458 Advanced control strategies and modern consulting provide new challenges for the classical problem of bumpless transfer. It can, for example, be necessary to transfer between an only approximately known existing analogue controller and a new digital or adaptive controller without accessing any states. Transfer ought to be bi-directionai and not presuppose steady state, so that an immediate back-transfer is possible if the new controller should drive the plant unstable. This paper presents a scheme that meets these requirements. By casting the problem of bi-directional transfer into an associated control problem, systematic analysis and design procedures from systems theory can be appfied. The paper includes laboratory and industrial applications. 098 GH~tling~ Force Control Using Nonlinear Adaptive Strategy L. Gun, A. Schrne, X. Ding, pp 459-462 This paper presents a nonlinear adaptive control scheme for the grinding force control of grinding processes. First, a nonlinear control model is introduced. Based on this, a nonlinear controller is derived by using a model-following scheme. To increase the robustness of the control system against parameter uncertainty, a nonlinear adaptive observer which estimates the model parameters is used. The complete concept can be considered as a nonlinear adaptive control scheme, which guarantees the stability of the whole closed-loop system. The simulation results show that the controller provides the desired performance. 099 MulUvariable Generaluized Predictive Control with Multiple Reference Model: A Fhxible A n n AppUmflon P. Codron, P. Boncher, p p 463-466 This paper is concerned with multivariable generalized predictive control with a multiple reference model algorithm (MIMO GPCJMRM), and its robust stability analysis via non-structured singular values. The choice of a new reference model ensuring