Hinweise zum Einsatz der Google Suche
Personensuchezur unisono Personensuche
Veranstaltungssuchezur unisono Veranstaltungssuche
Katalog plus

Dissertation Perdomo

 Dr.Ing. Maria del Carmen Perdomo Arvizu

"Time domain identification of the mechanical system of a drive for the purpose of diagnostics"


Time domain identification of the mechanical system of a drive for the purpose of diagnostics.
A frequent requirement for self-commissioning of electrically powered industrial drives is the permanently updated quantitative information about the mechanical system parameters. An accurate description of the mechanical part and the prediction of the dynamic behaviour of the system are also essential for condition monitoring and diagnosis of the drives. For this, different methods for the identification of two-mass-systems have been experimentally investigated. The general technique used in the research group in Siegen until now is based on the measurement of the frequency response G(jw) of the mechanical system of the drive, when the system is excited by pseudo random binary signals. In fact it is possible to perform system identification in or by using other methods in the time domain.

The time domain methods have to be examined regarding its application in industrial drives with repetitive cycles of production. The scheme for identification in time domain of the mechanical characteristics of electrical drives implies the utilization of transients that occur in every industrial process. The results can be used for the tuning of the parameters of the control and differences of the response compared with measurements carried out during the erection of the system can be used for the purpose of diagnostics.

One way to achieve the parameter identification in time domain, even in presence of time- varying parameters, is the implementation of an Extended Kalman Filter as estimator. The Kalman filter is a recursive algorithm generally used to obtain the optimal estimate of the states of a linear model. Using priory information about the system and measured data, the filter produces an estimate of the system states, statistically minimizing the error between the estimated and actual states of the system. The Kalman filter can be extended for nonlinear systems through a linear Tailor approximation. This is the Extended Kalman Filter (EKF), which has reported to be satisfactory used for system identification. For this purpose, the parameters are considered as additional states of the system and included in the state vector. This way the system turns to be nonlinear, but the desired parameters are estimated within the predicted states.