Machine-learning and CFD based optimization and comprehensive experimental study on diagonal flow fan for energy conservation and efficiency enhancement

Abstract

The diagonal flow fan, known for its innovative airflow guiding technique, presents a dual advantage of heightened efficiency and reduced noise levels. Despite garnering attention in energy production systems, research on optimizing outlet guide vanes for diagonal flow fan performance remains limited. This study introduces a novel machine learning approach, utilizing Opt LHD sampling instead of traditional simulations and experiments to accumulate ample sample data. The fusion of LS-SVM and improved GA-PSO is employed for multi-objective optimization of outlet guide vanes, aiming to simultaneously enhance fan performance and reduce energy consumption. Experimental assessments reveal the optimized fan's significant improvements, including a 106 Pa increase in total pressure, a 3.6 dB reduction in noise levels, and a remarkable 16.3% enhancement in total pressure efficiency. These results highlight the robustness of the machine learning approach in optimizing diagonal flow fan exit guide vanes, effectively addressing the research gap in this area. Additionally, numerical analysis of internal flow dynamics and acoustic properties pre and post-optimization uncovers the intrinsic mechanisms influencing the overall performance of the diagonal flow fan.