Parametric modelling of unsteady load for turbine cascade and its application in clocking effect optimization and load-reduction

Abstract

The circumferential phase differences of adjacent stators or rotors will lead to different unsteady aerodynamic loads of turbine (i.e., the clocking effect), which makes the modelling of unsteady load and clocking effect very complex, so an accurate and efficient modelling method for its optimization becomes essential. In this paper, a new modelling method for the unsteady aerodynamic load of turbine cascades is proposed based on the reduced-order model (ROM) and parametric training and applied to optimizing the clocking effect of a 2-stage low-pressure turbine cascade. Firstly, using the sub-block method, snapshot matrices that include the unsteady aerodynamic load on stator/rotor surfaces in different clocking positions are established, and the ROM of aerodynamic load in different clocking positions is presented based on the proper orthogonal decomposition (POD) method. Secondly, the POD coefficient matrices and the basis matrices are reconstructed based on the parametric Gaussian process regression (PGPR). The PGPR surrogate models between clocking positions and aerodynamic load are obtained. Finally, based on the established model, an efficient method is proposed for the clocking effect optimization and the aerodynamic load-reduction of a 2-stage cascade. Results show that the maximum optimization ranges for the variation altitude of unsteady aerodynamic lift and torque are 44.06% and 30.51%, respectively, which verifies the efficiency and feasibility of the proposed theories and methods.