Analysis of complex flows in safety valves using CFD

The aim of this thesis was to utilise Computational Fluid Dynamics (CFD) tools available to industry - primarily ANSYS FLUENT and ANSYS Workbench - to develop a modelling methodology which could allow CFD to be relied upon and be used as an analysis/design tool within the Pressure Relief Valve (PRV) industry. This work will also investigate the capability and resilience of modern day CFD solvers to accurately capture the array of complex flow dynamics which exist during the operation of various types of PRV’s. Furthermore, to achieve confidence in the capability of the CFD models developed; it was required for experimental validation work to be performed. Initially a 3D single phase (air only) steady state CFD model was developed for a right-angled type 3511E PRV manufactured by Broady Flow Control and validated against experimental results generated at an industrial testing facility. In addition, by developing upon previous 2D CFD work at the University of Strathclyde for a smaller through flow type PRV geometry; the lessons learned from the Broady PRV research allowed a 3D CFD model to be developed for a 5231BX through flow type PRV manufactured by Henry Group Industries. Experimental validation for the 5231BX PRV was performed at a testing facility developed within the flow laboratory at Strathclyde. For both cases, good correlation for air mass flow rate and disc/piston aerodynamic force with experimental results was achieved. Furthermore, from the validation analysis performed for the Henry PRV it was highlighted that a significant improvement in CFD modelling accuracy could be achieved by adopting a 3D CFD model when compared with 2D. This was due to a significant variation in complex flow features found at the piston surface and presence of symmetry breaking flow phenomena. Due to the success of the 3D steady state CFD models to accurately capture flow features for both the Broady and Henry PRV’s, a transient single phase CFD model was developed to enable the use of a dynamic moving mesh to capture each PRV’s operational characteristics. The experimental facilities developed for both PRV’s allowed a validation of the CFD model’s prediction for overpressure, blowdown and overall dynamic characteristics (disc lift vs time). For both cases it was found that overpressure and blowdown could be predicted closely as well as a general consensus being achieved for lift vs time behaviour. Therefore, it was concluded that for both types of PRV geometry, a moving mesh was capable to accurately capture the dynamic PRV response. For the Henry case however there was vibration in the CFD results which was attributed due to numerical errors induced by the RANS CFD numerical model. Following the dynamic mesh analysis for the Henry PRV; a study of the validity of the commonly used quasi steady design approach to PRV’s was examined by comparing results from steady state to dynamic operating conditions. The conclusion from this study highlighted a potentially significant issue with current typical PRV design practices as different magnitudes of disc force could be attributed with disc velocity when compared with steady state results. This would therefore cause a difference in expected performance from quasi steady based simple dynamic models often used in initial valve design. In addition, two-phase (air-water) 3D CFD analysis was performed for the Henry 5231BX PRV to extend previous 2D CFD research at Strathclyde. Both steady state and transient experimental testing was performed for a range of water injection rates; however only two phase steady state CFD validation was undertaken due to computational restrictions. The two-phase steady state validation results highlighted that a 3D CFD model, which utilised the homogenous mixture model, was capable of achieving a reasonably accurate correlation to experimental data across the full lift range when subjected to a water injection rate of 0.96 L/min and 2.1 L/min. Degradation in accuracy could be found for two-phase CFD modelling when compared to single phase modelling; however correlation was generally within acceptable limits. Furthermore, from two phase transient experimental testing; it was possible to establish a direct trend of water injection rate/water mass fraction to the dynamic characteristics of the Henry PRV as well as blowdown pressure. The effect of two phase operation on dynamic stability was also established.The aim of this thesis was to utilise Computational Fluid Dynamics (CFD) tools available to industry - primarily ANSYS FLUENT and ANSYS Workbench - to develop a modelling methodology which could allow CFD to be relied upon and be used as an analysis/design tool within the Pressure Relief Valve (PRV) industry. This work will also investigate the capability and resilience of modern day CFD solvers to accurately capture the array of complex flow dynamics which exist during the operation of various types of PRV’s. Furthermore, to achieve confidence in the capability of the CFD models developed; it was required for experimental validation work to be performed. Initially a 3D single phase (air only) steady state CFD model was developed for a right-angled type 3511E PRV manufactured by Broady Flow Control and validated against experimental results generated at an industrial testing facility. In addition, by developing upon previous 2D CFD work at the University of Strathclyde for a smaller through flow type PRV geometry; the lessons learned from the Broady PRV research allowed a 3D CFD model to be developed for a 5231BX through flow type PRV manufactured by Henry Group Industries. Experimental validation for the 5231BX PRV was performed at a testing facility developed within the flow laboratory at Strathclyde. For both cases, good correlation for air mass flow rate and disc/piston aerodynamic force with experimental results was achieved. Furthermore, from the validation analysis performed for the Henry PRV it was highlighted that a significant improvement in CFD modelling accuracy could be achieved by adopting a 3D CFD model when compared with 2D. This was due to a significant variation in complex flow features found at the piston surface and presence of symmetry breaking flow phenomena. Due to the success of the 3D steady state CFD models to accurately capture flow features for both the Broady and Henry PRV’s, a transient single phase CFD model was developed to enable the use of a dynamic moving mesh to capture each PRV’s operational characteristics. The experimental facilities developed for both PRV’s allowed a validation of the CFD model’s prediction for overpressure, blowdown and overall dynamic characteristics (disc lift vs time). For both cases it was found that overpressure and blowdown could be predicted closely as well as a general consensus being achieved for lift vs time behaviour. Therefore, it was concluded that for both types of PRV geometry, a moving mesh was capable to accurately capture the dynamic PRV response. For the Henry case however there was vibration in the CFD results which was attributed due to numerical errors induced by the RANS CFD numerical model. Following the dynamic mesh analysis for the Henry PRV; a study of the validity of the commonly used quasi steady design approach to PRV’s was examined by comparing results from steady state to dynamic operating conditions. The conclusion from this study highlighted a potentially significant issue with current typical PRV design practices as different magnitudes of disc force could be attributed with disc velocity when compared with steady state results. This would therefore cause a difference in expected performance from quasi steady based simple dynamic models often used in initial valve design. In addition, two-phase (air-water) 3D CFD analysis was performed for the Henry 5231BX PRV to extend previous 2D CFD research at Strathclyde. Both steady state and transient experimental testing was performed for a range of water injection rates; however only two phase steady state CFD validation was undertaken due to computational restrictions. The two-phase steady state validation results highlighted that a 3D CFD model, which utilised the homogenous mixture model, was capable of achieving a reasonably accurate correlation to experimental data across the full lift range when subjected to a water injection rate of 0.96 L/min and 2.1 L/min. Degradation in accuracy could be found for two-phase CFD modelling when compared to single phase modelling; however correlation was generally within acceptable limits. Furthermore, from two phase transient experimental testing; it was possible to establish a direct trend of water injection rate/water mass fraction to the dynamic characteristics of the Henry PRV as well as blowdown pressure. The effect of two phase operation on dynamic stability was also established