Identification of the foundation parameters of turbomachinery

Abstract

The dynamics behaviour of the entire rotating machinery is significantly affected by the foundation and an accurate modelling of the foundation can optimise the balancing and operating efficiency. In the complex and large rotating machinery, the removal of the rotor is usually difficult and costly so that the identification of foundation model has to be incorporated into the overall rotor bearing foundation system (RBFS). In this thesis the techniques are developed to identify the equivalent foundation of the rotating machinery in situ, which can substitute for the actual foundation and reproduce the equivalent vibration responses of the RBFS over the operating speed range. The vibration measurement data excited by the existing unbalance in the rotor are used to identify the modal parameters, which construct the equivalent foundation. Modified modal matrix is introduced to the identification equation, which enables the modal parameters of each vibration modes to be solved independently and therefore simplifies the solving procedure. Numerical foundations established by computer modelling were first identified. The numerical experiment shows that the high DOF foundations can be identified by using limited DOF equivalent foundation accurately even with the input data corrupted by noise. It was shown that equivalent foundation constructed by the modal parameters of the vibration modes within or near the operating frequency range is sufficient to estimate the vibration behaviour of the RBFS over the same frequency range. The developed identification techniques were applied to identify two experimental rigs. The equivalent foundation of the first rig, of which the excitation forces are measured by force transducers, is identified accurately as it estimates the vibration behaviour well. The equivalent foundation of the second rig, of which the excitation forces are derived by the rotor displacement and the rotor model, provides fairly good frequency response estimations, though not as good as those in the first rig. It is concluded that the developed identification procedure can be used to identify an equivalent foundation, which predicts the vibration behaviour of the original foundation well.