An implementation of error localization algorithms for model calibration of dynamical systems

Abstract

The paper discusses the implementation of three Error Localization Algorithms (ELA), test assessment metrics, and an example of model calibration. During model calibration, users routinely perform sensitivity studies to determine critical parameters for model adjustments. Although this makes engineering sense, sensitivity information only reflects parameter importance and not necessarily model deficiencies. Instead, the task of identifying model deficiencies is best suited for ELA in which measured data is used directly to identify potential problem areas in the model. Three ELA methods are applied in this work: (1) Coordinate Modal Assurance Criterion (CoMAC), (2) Robust Model Error, and (3) Analytical Dynamic Model Improvement (ADMI). To establish credibility of the ELA approaches, data from two analytical examples are used with ELA to identify known problem areas. Of the three ELA approaches implemented, only ADMI correctly identified known problem areas in the model. An example application of the ADMI algorithm is presented to identify problem areas of an isogrid-panel model from experimental data. These identified areas are then used to calibrate a model.