RICERCA ITALIANA     

Mechanical system modelling for diagnostics purposes

Politecnico di Milano

Abstract

Model based diagnostics of structural defects or faults in mechanical systems for different applications (structures,rotors,mechanisms and plant and machine components), by means of measurements of some characteristic quantities of its dynamical behaviour, generally of its vibrations measured in some points of the system, and of a comparison with results from simulations is showing increasing success in different applications but requires accurate modelling of the mechanical system and of the fault or defect.

Principal Investigator

Nicolo' BACHSCHMID Politecnico di MILANO

Research Objectives

The goal of this research is identify structural defects or faults in mechanical systems, from observation of its dynamical behaviour, by means of measurements of the vibrations in some points of the system.
The identification is performed basically by comparing the measurement results with calculated results obtained by simulating the dynamical behaviour by means of a mathematical model of the system and of a suitable model of the fault.
The research is related to diagnostics in general and to model based diagnostics in particular, which actually does not exploit completely the knowledge of reliable and accurate models which can be developed in the different fields of mechanics.
The fault identification methods are different and specific to the different types of mechanical systems: rotor systems, mechanisms, electric drives, components of moving mechanical systems. The excitation of its dynamical behaviour is obtained naturally through the motion of the system.
The comparison of experimental results with calculated results can be performed automatically, e.g. in a least square fitting approach in case of linear or quasi linear systems which are analysed in the frequency domain, such as rotor systems in industrial plants (research group of MI), in which the vibrations in correspondence of the bearings are continuously monitored, or can be effected by an operator by means of an optimized model of the non linear system in the time domain, as it >>>

First Results

Tuned models and related software.
Fitting out of prototypes and/or test-rigs.
Sensitivity analysis results.Validated models and related software.
Validated identification techniques and related software.
Indications about validity and limits of models and identification techniques.

Timescale

24 months

National and international background

The necessary background for the identification of faults in mechanical systems considered in the present project which is directed towards model based identification, is obviously the model of the analysed mechanical system, together with the model of the defect or fault which should be identified. This technique can be considered included in the model based diagnostic approach, which is a rapidly increasing branch of Diagnostics in general, which has developed enormously in the last years through the contribution of researchers of different sectors (see e.g. the proceedings of the yearly Congress COMADEM - Condition Monitoring and Diagnostic Engineering Management or of the International Conference on Acoustical and Vibratory Surveillance Methods and Diagnostic Techniques, CETIM, Senlis).
The models of the considered systems are generally well known both in the international scientific community and in the national community, because they have also been used in previous works by the different research groups (which are reported in their bibliographic references). This belongs both to the linear or quasi-linear models, and to the strongly non linear models. The models which have been developed generally do not require further refinements.
The modelling of faults or defects or of the system or component affected by a fault is instead a goal of each research group, which can be reached only through experimental validation performed by the different research >>>