Aero acoustic Analysis and Noise Mitigation for Axial Server Fans

In today’s world, a rapid increase of the global data sphere has led to the need to produce high-capacity cloud-based computer storage systems, which in turn gives rise to the need for cooling systems. Typically, most electronics, data centers included, employ fan air cooling. Unfortunately, such fans also generate high noise levels and vibration at higher speeds, which are hazardous in these workspaces. Thus, it is vital to gain an understanding of the acoustic signatures of the noise emanating from such server fans and explore noise mitigation methods to lower the noise levels below those prescribed by existing federal regulations.
This thesis presents the aero-acoustic modeling, and measurements performed on an axial server fan and implementation of appropriate noise mitigation techniques on the same. Initially, a literature survey was performed to explore the various acoustic modelling approaches, acoustic mitigation techniques as well as experimental methodologies. An aero-acoustic model of a commercial axial server with six fans is set up emulating an existing server in a data center lab at Georgia Tech. The modeled fan geometry is tested by obtaining the computed fan curve and assessing the trends against the existing fan curve provided by the manufacturer. Acoustic measurements are taken for the server fans in operation and are subsequently used to validate the Computational Fluid Dynamic (CFD) model.
Based on the acoustic signature plots obtained, the frequencies with peak amplitude are noted and acoustic insulating foam is lined along the inside of the server walls to absorb the noise at those frequencies, and therefore bringing the overall Sound Pressure Level (SPL) down to the levels prescribed by the federal regulatory authorities.
The present aero-acoustic model can hence be used as a benchmark method, helping support a variety of aero-acoustic analyses including parametric studies; and measurements can also be made for validation purposes.