Identification of variable mechanical parameters using extended Kalman Filters

The automatic operation of processes requires accurate and up-to-date information about the current state of the system parameters, which frequently cannot be measured during operation. Furthermore, this parameters can change in time due to several factors such as the own dynamics of the system. For electrically powered systems a correct description of the mechanical part and its dynamics is a requirement for a good control performance. The present work describes a Kalman Filter approach to the identification of mechanical parameters. The online identification of time variable mechanical parameters is a task of prime importance for the tuning of self-adaptive controls. In this paper a method for the identification of constant and variable mechanical parameters in industrial drives, introduced in the past by other authors, is analyzed and experimentally tested.