Gas turbines enclosures entail a high number of auxiliary systems which must be preserved from heat, ensuring therefore the long term operation of the internal instrumentation and of the data acquisition system. A dedicated ventilation system is designed to keep the enclosure environment sufficiently cool and dilute any gas coming from potential internal leakage to limiting explosion risks. These systems are equipped with axial fans, usually fed with air coming from the filter house which provides air to the gas turbine combustion system, through dedicated filters. The axial fans are embedded in a ducting system which discharges fresh air inside the enclosure where the gas turbine is housed. As the operations of the gas turbine need to be guaranteed in the event of fan failure, a backup redundant system is located in a duct parallel to the main one. One of the main requirements of a ventilation fan is the reliability over the years as the gas turbine can be installed in remote areas or unmanned offshore platforms with limited accessibility for unplanned maintenance.
For such reasons, the robustness of the ventilation system and a proper understanding of coupling phenomena with the axial fan is a key aspect to be addressed when designing a gas-turbine system. Here a numerical study of a ventilation system carried out with RANS and LES based methodologies will be presented where the presence of the fan is synthetized by means of static pressure discontinuity. Different operations of the fans are investigated by means of RANS in order to compare the different operating points, corresponding to 1) clean and 2) dirty filters operations, 3) minimum and 4) maximum pressure at the discharge section. Large Eddy Simulations of the same duct were carried out in the maximum loading condition for the fan to investigate the unsteady response of the system and validate its correct arrangement.
All the simulations were carried out using OpenFOAM, a finite volume open source code for CFD analysis, treating the filters as a porous medium and the fan as a static pressure discontinuity according to the manufacturer’s characteristic curve. RANS modelling was based on the cubic k-ε model of Lien et al. while sub-grid scale modelling in LES was based on the 1 equation model of Davidson.
Computations highlighted that the ventilation system was able to work in similarity for flow rates between 15 m3/s and 23.2 m3/s and that the flow conditions onto the fan suggest that the aerodynamic stress on the device could be reduced introducing in the duct flow straighteners or inlet guided vanes.