Title: Parameter Estimation based on FMI
Authors: Rüdiger Kampfmann, Danny Mösch and Nils Menager
Abstract:In order to stay competitive the requirements on machinery in the producing industry have enormously increased. Within the automation industry these demands, like higher throughput or better energy efficiency, result in increasing complexity of the installed plants. Additionally, Industry 4.0 and the Internet of Things continuously increase the amount of software. Using model-based development methods is one approach to deal with this complexity. But model-based methods can also be utilized during the operational phase of a plant in order to generate additional value for the plant operator. Introducing smart services based on the usage of physical models enables new control and diagnosis features, e.g. the utilization of inverse plant models for feedforward control or comparing the output of a model with measurements of the plant in order to prove for correct behavior. For all these services the accuracy of the considered models is crucial. With an inexact model neither the future behavior can be foreseen nor the control quality can be improved. The used models don't have to be built up from scratch, existing models already created for sizing can be reused. However, these models cannot be used directly. First a reparametrization is necessary, because effects like friction or manufacturing tolerances cannot be taken into account correctly during sizing. For this special kind of problem dedicated optimization algorithms are available for parameter estimation, which take randomly distributed measurement errors and the special structure of this problem class into account. In this paper a work flow for parameter estimation based on open source tools is presented, in which the considered models are provided as Functional Mock-up Unit. Afterwards the performance of this work flow is demonstrated on a real industrial problem: A three arm Delta Robot.
Links: Full paper