Inlet box structure optimization of a large axial-flow fan using response surface methodology

Large axial-flow fans are widely used in many fields. The inlet box is an integral part of large axial-flow fans, and a welldesigned inlet box could effectively improve fan efficiency. However, the inlet box structure is complicated, and the existing inlet box design method severely depends on the design experience. In this study, we propose a structure optimization design system based on a surrogate model technique for researching the critical structure parameters of the inlet box and accomplishing aerodynamic performance optimization. As for this expensive optimization problem, the design system contains twice optimization procedures by using the Response Surface Methodology (RSM) with the orthogonal design method. The optimization object is an existing large axial-flow fan. The optimization objective is the total pressure efficiency of the fan, and the total pressure rise is the restriction condition. We generate eighteen different inlet boxes connect with the same impeller and outlet pipe by the orthogonal design method and calculated fan aerodynamic performance by CFX software. After the first optimization, we find the key structural parameters by the sensitivity analysis and the reselect variables total of 25 cases are adopted in a further RSM optimal process. The ultimate surrogate model estimates the fan with the optimal inlet box has a better aerodynamic characteristic and a 6.7% total pressure efficiency rise. Finally, we compare the aerodynamic characteristics of the ultimate design fan and the initial fan by CFD simulation. The numerical results show that: the total pressure efficiency is 6.5% higher than that of the initial impeller, and the pressure rise is 3% higher than that of the initial impeller. The result demonstrates that some most critical parameters of the inlet box structure decide the aerodynamic performance, and the inlet box optimization effectively increases the fan efficiency in the meanwhile.