Stator vanes which are found at different stages in axial compressors are subject to major aerodynamic load fluctuations, due to cyclic excitations induced by the wakes of the rotors located upstream and also due to the pressure waves generated by the blade rows located downstream. In a vibratory fatigue context, it is important to ensure the structural strength of these stator vanes. A new technology of monoblock sectorised stator vanes has been developed to provide an easier manufacturing process and lower costs; in return, cyclic symmetry properties, which make the prediction studies easier, are lost in this case. Besides, these kinds of monoblock sectors are characterized by low damping which increases the criticality of the blade modes localization. The division into sectors of stator vanes creates a high density of eigenmodes in some short frequency ranges; this phenomenon is accompanied by non-negligible modal participations of neighbouring eigenmodes. Moreover, stator vanes sectors have shown a strong sensitivity to mistuning, making their vibratory behaviour prediction more difficult by means of deterministic prediction methods. In this paper, a phenomenological approach is followed to quantify the mistuning. Thus, several sensitivity to mistuning numerical studies were performed. They allowed to have first results about the relative predictable errors that could be done, depending on the mistuning. Some engine tests were also performed with gauges-equipped stator vanes sectors.

Abstract. Stator vanes which are found in axial compressors are subject to vibratory fatigue. Their division into monoblock sectors makes the prediction of their vibratory behaviour difficult by deterministic methods due to the loss of the cyclic symmetry properties and also to a high sensitivity to mistuning. The purpose is to present a robust calculation strategy based on a stochastic modelisation of the structure. The methodology has been developed first on a simplified model and then applied to an industrial case. Polynomial chaos based results are in good agreement with reference Monte Carlo simulations.