Model-based diagnosis - Methods and experience

This paper is concerned with the diagnosis of the damage and faults of mechanical systems in the diverse fields of engineering. It takes into consideration the various stages of the life cycle of a system from a diagnosis-orientated point of view and from a methodological one. Holistic modelling is the key to the methodology discussed. The holistic model includes the life behaviour (slow-time coordinate) as well as the classical dynamics (expressed with the fast-time coordinate). Diagnosis is based on mathematical models dependent oil discrete life times of the system which perform a system condition assessment in the past and in the present through an adjusted mathematical model (adaptive model), and for the future by the use of prediction with the help of adaptive models. The principle part of the paper is the knowledge base in the form of verified and validated mathematical models which are adjusted to the states of the system at current life times. These models describe the current state of the system, permit a comparison with previous states, and therefore serve the purposes of fault detection, localization, and the cause-finding of faults and their assessment, predictions due to future forcings and trent predictions. The diagnostic decision-making based on the validated models also is mentioned. Finally, some marginal notes are made on the diagnostics of machines, and an example of a bridge illustrates the model-based diagnosis.