Modelling of turbine blade vibrations via computational intelligence methods

Abstract

A method for modelling turbomachine blade vibration events is proposed, based on computational intelligence algorithms. The method utilises steady thermodynamic data and blade tip-timing data to identify high amplitude vibration events and to draw underlying relationships between steady-thermodynamic input channels and resultant blade motion characteristics. Several computational studies probe specific process aspects in order to improve model prediction accuracy and several methods of data-feature reduction are established to further enhance vibration predictions. Overall, the study shows promise of what prediction capabilities can be achieved with seemingly limited instrumentation. Drawbacks in matters of tip-timing interpretation, quality/quantity of data and process limitations are discussed. Consequential future objectives are outlined to envisage onward predictive accuracy.