Showing papers by "Paul M. Frank published in 1992"

Frequency domain approach and threshold selector for robust model-based fault detection and isolation

Abstract: Problems of residual generation and evaluation for uncertain dynamic systems are formulated and investigated with the aid of frequency domain approaches and H∞- optimization techniques. Based on the parametrization of residual generators the design of optimal residual generators reduces to an optimization problem that is solved by H∞-techniques. To improve the robustness of residual evaluation a new evaluation index is introduced, under which thresholds for different types of uncertain systems are derived.

Fault Diagnosis Using Adaptive Observers

Abstract: This paper presents an adaptive observer based fault detection scheme. With the aid of nonlinear adaptive observer theory, an approach of constructing adaptive residual generators is developed. Using such kind of residual generators, we are able not only to detect abrupt faults but also to estimate unknown parameters in the system observed which may represent faults with slow time constants, some kinds of nonlinearities or varying system parameters. This may enhance the robustness of residual generators with respect to model uncertainties and provide a new *ay t o solve fault detection problem for nonlinear systems. The proposed scheme is illustrated in terms of an application example.

Robust component fault detection and isolation in nonlinear dynamic systems using nonlinear unknown input observers

Abstract: In this contribution a well known method for robust Component Fault Detection and Isolation (CFDI) which was mainly developed for linear dynamic systems is extended to a class of nonlinear dynamic systems. The proposed technique is based on a new concept of Nonlinear Unknown Input Observers. The aim is to develop a CFDI-technique which is robust to model-plant mismatches and external disturbances while being sensitive to faults. The component faults as well as the model-plant mismatches are modelled in form of unknown input signals by a Taylor series expansion of the system model. By a nonlinear transformation, the model is decoupled from the unknown inputs. Subsequently, a Robust Component Fault Detection Observer (RCFDO) based on the decoupled model is designed. The RCFDO is used to generate a residual which is not affected by the unknown inputs representing disturbances and model-plant mismatches but still reflects the occurence of faults. Additionally, it is shown how the robustness concept can be utilized for the purpose of Fault Isolation. Based on the nonlinear model of a synchronous machine, the design of a RCFDO is illustrated and tested by simulations.

Robust observer design for dynamical systems under unknown disturbances

Abstract: Problems of designing robust observers in the frequency domain are formulated and solved. Using the factorization approach a characterization of observers and their estimation error dynamics is derived. It is shown that the problem of designing robust observers can be therefore transformed into the well-known model matching problem. An effective approach is furthermore proposed to solve this H∞-optimization problem. It is finally shown that this approach can be interpreted as a general realization of the Luenberger observer scheme and the observer designed can be taken as a modification of an identity observer.

Advanced Optimal Predictive Control

Abstract: In this contribution an extended version of the well known DMC predictive control concept is presented. The definition of an extended DMC cost function leads to a LQR problem with predescribed weighting matrices. Solving the LQR problem the optimal predictive control (OPC) law can be found. Analysing the weigthing matrices it can be seen that the OPC control law is capable of guaranteeing closed-loop stability even for unstable or nonminimum phase plants. In order to achieve disturbance attenuation a disturbance model is used within the OPC method. Also using a reference model the extended OPC method yields a straight forward controller design with a predescribed disturbance attenuation as well as a good tracking behaviour. As one advantage the utilization of a reference model avoids the necessity of known future values of the reference signal.