Physically motivated models of electromechanical
motion systems enable model-based control theory and facilitate
system interpretation. Unfortunately, the effort of modelling
restricts the usage of model-based methods in many applications.
Some approaches to automatically generate models from
measurements choose the best model based on minimizing the
residual. These model selection attempts are limited due to
ambiguities in reconstructing the internal structure from the
input-output behaviour because usually motion systems have
only one actuator and one sensor. Often, it is unknown if the
resulting model is unique or if other models with different
structure would fit equally well. The set of candidate models
should be designed to contain only distinguishable models but
ambiguities are often unknown to the experimenter. In this
paper distinguishability is investigated systematically for a class
of multiple mass models representing servo positioning systems.
In the analysis a new criterion for indistinguishability is used.
The benefit of additional, structural sensors on distinguishability
of models is demonstrated which suggests to mount them
temporarily for the commissioning phase in order to facilitate
the model selection. It turns out that the best results can be
achieved if synergies among sensor signals are utilized.
© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including
reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
|