Investigation of the Modified Dobson's method and proposal for improvement

This thesis investigates the Modified Dobson’s method. This method for nonlinear system identification works well when the system’s eigenmodes are far apart in frequency. However, when two modes are close together, the method lacks accuracy in the modal parameters that it generates. To better understand this, a computational model of the Modified Dobson’s method was made and first used on linear systems. Modes that are in closer proximity to the mode of interest resulted in increased inaccuracy of the modal parameters. To improve this method, the shape of the resonance of a nearby mode was estimated and subtracted from the total system’s response. This Improved Modified Dobson’s method performed better than the original method. The modal parameters that were generated were more accurate. These improvements are expected to work partly for nonlinear systems, where the Modified Dobson’s method was created for.