LAUSR

The natural frequencies corresponding to the particular mode shapes of a large turbo-generator must not be in the resonance region before delivery. Different parameters may clearly affect these dynamic characteristics. Different from other studies, this paper introduces a method for obtaining the inherent characteristics of end winding in a short period of time under multiple simultaneously changing parameters. The proposed process is based on the response surface method (RSM). In this method, the main concerned natural frequencies were taken as the critical index to describe the dynamic behaviors of the end winding, with which the mathematical relationship between the dynamic characteristics and design variables was analyzed. First, the stiffness of the rings, stiffness of the radial braces and number of pins bonded to the radial braces were used as samples for orthogonal experimental design. Then, the natural modes and frequencies of 25 different samples were acquired by conducting modal analysis with ABAQUS software. Utilizing these dynamic results, a second-order polynomial response surface model was established to describe the relationships between natural frequencies and these three different parameters. Then, the quality of this model was verified by calculating the valuating indexes for comparison with support vector regression (SVR). With the response surface model, the variation regularities of the natural frequencies and modes due to the above parameters were discussed. The method proposed in this paper can enable natural frequencies in the whole design space to be quickly determined without finite-element analysis, greatly improving the development efficiency and laying a foundation for dynamic response prediction during normal operation with different parameters and additional large turbo-generator stator end winding optimization.