Optimization of Low Pressure Axial Fans and Effect of Subsequent Geometrical Modifications

A2 - Optimization & Design Methods

A low pressure axial fan is optimized by an evolutionary optimization algorithm. The target function is maximization of total-to-static efficiency at one design point while fulfilling a complete targeted performance curve. Additional constraints are limitation of axial depth, number of blades, and a fixed blade sweep which is selected for low sound generation. The target function is evaluated by artificial neural networks of the multi-layer perceptron (MLP) type. The MLPs were trained in earlier studies by the authors and are based on around 13,000 performance curves simulated by the Reynolds-averaged Navier-Stokes (RANS) method. While most of the common geometrical parameters of axial fans were varied for the simulations, the tip clearance was hold constant at a rather small value (0.1% of the fan diameter) and a cylindrical hub extending over the complete axial length of the blade was assumed. Hence, the MLPs are only valid for designs with these two features meaning that they are also assumed for the optimized fan of this paper.
After the optimization, the MLP-predicted performance curves are compared with RANS simulations. Additionally, a prototype is built to determine experimental performance curves. The results of the three methodologies (MLP, RANS, Experiment) show very good agreement with respect to total-to-static pressure rise and efficiency proving the adequacy of the optimization method used. However, the aforementioned properties regarding tip clearance and hub design are often violated in practice. The consequent aerodynamic penalty is determined experimentally. Firstly, the tip clearance is increased to 0.3% and 0.5% resulting in a reduction of peak efficiency by 3.5 and 7 percentage points, respectively. In addition, the operating range is decreased significantly. Secondly, the cylindrical hub is replaced by a thin disk which is only constructed to connect the blades with the shaft, but no longer separates the blade's suction and pressure side. The consequence is significantly increased secondary flow in the hub region which leads to a reduction of peak efficiency by around 7 percentage points as compared to fans with equal tip clearance but cylindrical hub.
It is concluded that we have developed an extremely quick and reliable optimization method, which was successfully validated. However, the direct practical application requires a small tip clearance and a hub which separates the blade's suction and pressure side. Ongoing work is dealing with measures to avoid the dramatic drop of efficiency when violating these features.