Incorporating Data Uncertainty in Rotordynamic Model Reconciliation

Abstract

A method called model reconciliation (MR) has previously been reported which modifies analytic rotordynamic models to make them match experimental data. An identified model derived from the experimental data and a nominal model from the engineering analysis are needed; an auxiliary dynamic system is then synthesized to modify the dynamics of the nominal model to match the identified model. The combination of the engineering model and the auxiliary dynamic system is the reconciled model. Due to numerous experimental uncertainties, the identified model may contain significant errors. These errors are then embedded by the reconciliation process so that the reconciled model is incorrect, and sometimes quite poor. A robust control synthesis method is presented here which accounts for model error bounds in generating the auxiliary dynamics to minimize their impact on the final model. To examine the validity of the process, extra data as a reference are used to compare with the data produced by the reconciled model. Both simulated data and experimental data are used to demonstrate the results of the method. The error model permits the synthesis to ignore spurious features of the identified model, thereby producing more reasonable results.