Automatic modal parameter selection using a statistical model and a Kalman filter

Kurzfassung

The automation of system identification is important for processing large amounts of data without expert user interaction. Automation is also important to maintain consistency in estimates, especially when investigating trends in data which could be masked by variations of mathematical parameters. This research presents a novel idea to obtain automatic modal parameter estimates based on a data driven statistical model and a Kalman filter. A key objective was to make observed data maximally informative. This lead to the development of a sliding predictive model using an optimized linear regression method to use system inputs which are not included in standard system identification. The method was first demonstrated on a numerical data set where it was found to improve system predictions. The method was then tested on full scale data from the German research vessel Polarstern during a voyage to the Arctic. The automatic Kalman estimates showed improved estimates using the combination of statistical model and modal parameters.