An Intelligent Optimization Method for Vortex-Induced Vibration Reducing and Performance Improving in a Large Francis Turbine

Abstract

In this paper, a new methodology is proposed to reduce the vortex-induced vibration (VIV) and improve the performance of the stay vane in a 200-MW Francis turbine. The process can be divided into two parts. Firstly, a diagnosis method for stay vane vibration based on field experiments and a finite element method (FEM) is presented. It is found that the resonance between the Kármán vortex and the stay vane is the main cause for the undesired vibration. Then, we focus on establishing an intelligent optimization model of the stay vane’s trailing edge profile. To this end, an approach combining factorial experiments, extreme learning machine (ELM) and particle swarm optimization (PSO) is implemented. Three kinds of improved profiles of the stay vane are proposed and compared. Finally, the profile with a Donaldson trailing edge is adopted as the best solution for the stay vane, and verifications such as computational fluid dynamics (CFD) simulations, structural analysis and fatigue analysis are performed to validate the optimized geometry.

Keywords

• intelligent optimization method
• stay vane
• vortex-induced vibration (VIV)
• extreme learning machine (ELM)
• computational fluid dynamics (CFD)