American Society of Safety Engineers Middle East Chapter 11th Professional Development Conference & Exhibition Kingdom of Bahrain, March 16-20, 2014

Transcription

1 American Society of Safety Engineers Middle East Chapter 11th Professional Development Conference & Exhibition Kingdom of Bahrain, March 16-20, 2014 ASSE-MEC CFD Modeling of Flammable/Hazardous Gas Release and Dispersion for Risk and Safety Assessments Vladimir M. Agranat, PhD, President, Sergei V. Zhubrin, PhD, VP R&D Applied Computational Fluid Dynamics Analysis ( ) Thornhill, Ontario, Canada Andrei V. Tchouvelev, PhD, CEO A.V. Tchouvelev & Associates Inc. ( ) Mississauga, Ontario, Canada Zivorad Radonjic, Senior Environmental Meteorologist ARCADIS SENES Canada Inc. ( ) Richmond Hill, Ontario, Canada

2 Abstract This presentation is an overview of recent customized applications of general-purpose Computational Fluid Dynamics (CFD) software, PHOENICS [1], to the CFD modeling of flammable/hazardous gas release and dispersion (GRAD) for risk and safety assessments. Three application areas are considered: safety analysis of use of flammable gases (hydrogen, methane, etc.), predicting atmospheric pollutant dispersion and modeling two-phase plumes from cooling towers (for air quality assessments). CFD models developed and validated are available as userfriendly templates for environmental and safety engineers. CFD software sales, training, user support and consulting are provided by ACFDA.

3 Overview Introduction: increasing use of CFD in risk and safety assessments CFD modeling approach: governing equations, generalpurpose CFD codes (PHOENICS, ANSYS, CFX, etc.) and customization of CFD models for non-standard tasks Modeling of flammable gas release and dispersion (GRAD) for safety analysis Modeling of atmospheric pollution dispersion for air quality analysis Modeling of two-phase plumes from cooling towers for pollutant deposition prediction Conclusions

4 Introduction: increasing use of CFD in risk and safety assessments CFD modelling of flammable/hazardous GRAD has been used extensively over the last 20 years in safety analyses and for air quality assessments. CFD is applied in risk and safety assessments to compliment the more traditional predictive methods such as AERMOD, CALPUFF and other models (approved by the US Environmental Protection Agency (EPA)) in the areas where these models demonstrate significant deficiencies: near-field dispersion, complex geometries, two-phase plumes, wide buildings at nonperpendicular winds, etc. These limitations are described in [3], [4] and other recent papers. CFD is capable of overcoming the above limitations by solving a comprehensive system of general 3D governing equations applicable for complex geometries and for multi-phase flows under specific atmospheric conditions.

5 Introduction: Increasing use of CFD in risk and safety assessments In this presentation, three different CFD models and corresponding case studies are described: advanced models of flammable GRAD for safety analyses (use of hydrogen, methane, etc.); a single-phase plume model for air quality analyses; a two-phase multi-group model of drift drop plumes from cooling towers. The models have been validated and applied to real-life industrial applications. They are based on customizing the commercial general-purpose CFD software, PHOENICS. PHOENICS is used as a framework and a solver. It has been validated and used for more than 30 years.

6 CFD modeling approach: governing equations and general-purpose CFD codes A commercial general-purpose CFD software package (PHOENICS, FLUENT, CFX, etc.) is equipped with userfriendly pre-processor (problem definition), solver (solution calculation) and post-processor (solution analysis). Governing equations include: conservation/transport equations for mass, momentum and energy for each phase constitutive equations (linkage between phases) turbulence model equations Proper boundary conditions are needed for solving the equations. Software customizations (specific sub-models) are required in non-standard situations. CFD models have to be validated before applying them for risk and safety analyses. Adequate training is required for use of CFD models.

7 GRAD CFD model [2]: prediction of flammable gas clouds Modeling of flammable GRAD scenarios is based on general transient 3D conservation equations (gas convection, diffusion and buoyancy) with proper initial and boundary conditions. The CFD modeling provides: transient behavior of all calculated variables (pressure, gas density, velocity and flammable gas concentration) movement of flammable gas clouds with time safety evaluation by analyzing a flammable gas concentration iso-surface (lower flammability level (LFL)) and total volumes of flammable gas. Three major stages in GRAD modeling: steady-state before-the-release simulations transient during-the-release simulations transient after-the-release simulations. PHOENICS software is used as a framework and solver. It is commonly used and well validated since 1981 has user-friendly interface for incorporating sub-models has many turbulence models (LVEL, MFM, k-ε variants, etc.).

8 Flammable GRAD CFD model: governing equations [2] 3D momentum equations ( r u ) P i + div ( r Uu - rn i eff grad u i ) = - t x Continuity equation r + div ( r U ) = 0 t + r f, Flammable gas mass conservation equation ( r C ) + r - r t div ( UC D eff grad C ) = i=1,2,3 Gas mixture density based on flammable gas mass concentration, C, or flammable gas volumetric concentration, α P CR gas r = a = [ CR + ( 1 - C ) R T gas air ] C " Effective viscosity and diffusivity n eff = n l + n t, D eff = n l / Pr l + n t / Pr t l = ar gas n gas + - a r air n r i CR gas i + ( 1 - C ) R air n [ ( 1 ) air ]/

9 Advanced flammable GRAD CFD model features [2] Dynamic Boundary Conditions at Release Orifice: Transient choked mass flow rate m & ( t ) d r = - V = r ( t ) u ( t ) A m & 0 e dt Initial choked mass flow rate 2 m & = 0 C r 0 P g d A 0 ( ) g + 1 Real Gas Law Properties: Turbulence Model Settings: C d A - t V 2 g ( ) g + 1 Abel-Noble Equation of State for hydrogen P r H 2-1 z = = ( 1 - H ) z 2 H = r R T d H 2 H 2 NIST data for methane, propane H g + 1 g LVEL model, k-ε model, k-ε RNG model, k-ε MMK model and MFM Local Adaptive Grid Refinement (LAGR) Iterative technique, accurate capture of flammable cloud behaviors near the release location and large gradient regions g + 1 g - 1 d H RT 2, P R H 2 T

10 Flammable GRAD CFD model validation matrix Validation, calibration and enhancement of GRAD CFD module capabilities for simulation of hydrogen releases and dispersion have been provided using available experimental databases (see table below). Case No. Validation Case Name Conditions Domain Leak Type Process Available Data 1 Helium jet Open Vertical Steady Velocity, concentration, turbulence intensity 2 H 2 jet Transient Concentration Horizontal 3 INERIS jet Steady Concentration 4 Hallway end Transient Concentration 5 Hallway middle Semienclosed Transient Concentration Vertical 6 Garage Transient Concentration 7 H 2 vessel Enclosed Transient Concentration

11 Flammable GRAD CFD model validation case: HYDROGEN SUBSONIC RELASE IN A HALLWAY Concentrations at four sensors for 20 min. duration Domain: 2.9 m 0.74 m 1.22 m Grid size: H 2 leak rate: 2 SCFM (0.944 m 3 /s) Duration: 20 min Concentration: 3 % iso-surface Hydrogen inlet Door vent Roof vent Sensor 1 Sensor 2 Sensor 4 Sensor 3 Published results Our results

12 Flammable GRAD CFD model validation case: HYDROGEN & HELIUM SUBSONIC RELEASE IN A GARAGE WITH A CAR Garage size: 6.4 m x 3.7 m x 2.8 m Leak size: 0.1 m x 0.2 m Two vents: porous material Car size: 4.88 m x 1.63 m x 1.35 m Leak rate: 7200 L/hour Leak direction: downwards Leak location: bottom of the car Helium release simulated by using LAGR (local adaptive grid refinement) Simulations Sensor 1 Sensor 2 Sensor 3 Sensor 4 Swain s CFD results 0.5% 2.55% 2.55% 1.0% Initial coarse grid, % 2.53% 2.52% 1.94% Adaptive refined, % 2.66% 2.62% 1.08% Adaptive refined, % 2.70% 2.67% 1.01%

13 Flammable GRAD CFD model application case: RELEASE IN A HYDROGEN GENERATOR ROOM Before-the-Release Simulation Existence of louver and exhaust fan in the Generator Room creates a steady-state airflow with 3D fluid flow pattern Ventilation velocities before release : During-the-Release Simulation End of 10-min release from the vent line 50% LFL 100% LFL

14 GRAD CFD model: summary GRAD CFD model has been developed, validated and applied for various industrial real-life indoor and outdoor releases of flammable gases (hydrogen, methane, propane, etc.). Advanced modeling features: Real-life scenarios with complex geometries Dynamic release boundary conditions Calibrated outlet boundary conditions Advanced turbulence models Real gas law properties applied at high-pressure releases Special output features (separation distances, flammable volume) Adaptive computational grid refinement tools. Dynamic behaviors of clouds of flammable gas or pollutant could be accurately predicted. GRAD CFD model is recommended for safety and environmental protection analyses.

15 CFD modeling of pollutant dispersion: AIR QUALITY ASSESSMENTS Over the past two decades, CFD modeling of single-phase plumes has been widely used for air quality analyses in addition to use of AERMOD, CALPUFF and other traditional analytical models as the latter models are insufficient and inaccurate in many cases (wide buildings, near-field regions, two-phase conditions, complex geometries, etc.). Studies on comparing these models with CFD predictions are provided in recent papers, e.g. [3] - [5]. A typical cross-validation case study [5] is presented here to illustrate the benefits of combined use of CALPUFF and CFD. In particular, the PHOENICS CFD software was applied in addition to CALPUFF in order to perform the near-field modeling of pollutant dispersion around a backup diesel generating station in Dawson City, Yukon, Canada.

16 The diesel engines are housed in a low building (blue building on the left), vented through short, curved exhaust stacks such that the exhaust gas is released horizontally rather than vertically. The task of CFD modeling is to predict the pollutant cloud behavior under various meteorological conditions. CFD modeling of pollutant dispersion: backup diesel generation station in Dawson City Geometry of the building housing the generation station and the stacks located on it is shown below.

17 CFD modeling of pollutant dispersion: ground-level pollutant concentrations near power plant Figure below shows a contour of relative (with respect to a value at the source) mass fraction of pollutant, C1, predicted by CFD in the worst case scenario. The power plant is shown in gray and the area of maximum ground-level pollutant concentrations is shown in red.

18 CFD modeling of pollutant dispersion: comparing CFD and CALPUFF predictions CFD modeling results are consistent with those derived from the CALPUFF model, with CFD predicted maximum concentrations being less than 10% higher than those estimated using the CALPUFF model. However, whereas the CALPUFF model predicted the maximum point of impingement to occur at the facility property line beside the power house, the CFD model predicted the maximum concentrations to occur meters from the property line. The analysis provides justification for using the CALPUFF model to represent near-field contaminant concentrations in similar regulatory applications. Additional analysis is required to verify whether the two models would continue to provide similar results for higher building heights, higher stacks or more complex building shapes.

19 CFD modeling of two-phase plumes: multi-group model of drift drop plume Drift drop plumes from industrial cooling towers can impose health concerns due to deposition of water drops containing a pollutant. A new homogeneous two-phase multi-group CFD model was developed for analyses of water deposition from these plumes. PHOENICS was customized to account for the deposition of drops. The model was validated based on the high-quality Chalk Point Dye Tracer Experiment (CPDTE) data described in [3, 6, 7]. A good agreement with experimental data on water deposition rates at selected locations from the cooling tower has been observed. It is better than that reported in [3, 6, 7]. The details of CFD model and case study are provided below.

20 CFD modeling of two-phase plumes: predicting water deposition in CPDTE case study Figure below shows the computational domain of Chalk Point Dye Tracer Experiment (CPDTE), the cooling tower, the ground, the inlet and the outlet. The probe (a red pencil with yellow end) shows the location of the first measuring station (at a distance of 500 m from the cooling tower centerline). The experimental data are described in detail in [3, 6, 7]. The water/sodium deposition fluxes were measured at the sensors located on the 35 arcs at 500 and 1000 m from the cooling tower centerline. The CFD modeling task was to accurately predict the deposition fluxes.

21 CFD modeling of two-phase plumes: features of multi-group model of drift drop plume Conservation equations for mass fractions of water drops having different sizes are solved in addition to standard conservation equations for mixture mass, momentum, energy, water vapor mass fraction, turbulent kinetic energy and its dissipation rate. Phase change effects (condensation, evaporation, solidification and melting) are assumed to be negligible due to specific conditions of the experiment (relative humidity of 93%). Extra terms are provided to the conservation equations for mass fractions of liquid water to account for the drift of water drops due to their gravitational settling. Various formulations for drift velocity and terminal velocity have been tested and compared. The droplet-size distribution [6] available at the cooling tower exit and containing the 25 groups of drops is simplified to 11 groups. Also, the 3-group and 1-group options are considered for comparison. The total deposition flux is calculated as a sum of products of individual group mass concentrations of water drops and corresponding terminal velocities.

22 CFD modeling of two-phase plumes: CPDTE case study modeling results Figures below show the 1% iso-surfaces of relative mass fractions of 110- and 900-micron drops predicted with the 11-group distribution model. 110-micron drops travel above the ground and leave the domain without touching the ground. 900-micron drops fall on the ground at a certain distance from the cooling tower (less than 500 m).

23 CFD modeling of two-phase plumes: comparison of CFD results with CPDTE data CFD predictions agree well with the experimental data on the water vapor plume rise and the total drift deposition fluxes. In particular, the plume rise predictions agree well with experimental values (the errors are from 4% to 33% at different distances from the tower centerline). Table below shows the detailed comparison between the CFD predictions of plume rise, the experimental data and the Briggs s formula available in [6]. Distance from centerline (m) Experimental plume rise (m) Briggs s formula plume rise (m) (17%) 40 (33%) (10%) 48 (-4%) (-13%) 72 (-28%) 500 N/A N/A Plume rise predicted by CFD (778,050 cells, k-ε model) (m)

24 CFD modeling of two-phase plumes: comparison of CFD results with CPDTE data Table below shows the predicted and measured maximum water deposition fluxes at 500 and 1000 m from tower centerline. CFD results obtained with 3-group and 11-group models are provided for comparison. The error factors are shown in red. Grid Size (number of cells) Location (m) Predicted flux (3 groups) (kg/s/m 2 ) Predicted flux (11 groups) (kg/s/m 2 ) Experimental flux (kg/s/m 2 ) 212, E-7 (0.90) 1.37E-7 (1.01) 1.36E-7 212, e-8 (1.27) 7.71e-8 (1.71) 4.52E-8 778, e-8 (0.59) 1.12e-7 (0.82) 1.36E-7 778, e-8 (1.13) 6.81e-8 (1.51) 4.52E-8 The CFD predicted maximum water deposition fluxes are in a good agreement with the experimental values within a factor of 1.7, which is well within the industry acceptable error limits (a factor of 3) [3].

25 CFD modeling of two-phase plumes: effect of number of groups on model accuracy Figure below shows the comparison of water deposition fluxes predicted with use of 11-, 3- and 1-group options. The 3- and 11-group results are close to each other and to the experimental data available at 500 m and 1000 m from the cooling tower (see table above). The 1-group results are close to the 3- and 11-group results only at large distances (more than 700 m) and are very different from the above results at short distances (smaller than 500 m).

26 CFD modeling of two-phase plumes: multi-group model summary New two-phase multi-group CFD model was developed and validated using the experimental data on water deposition fluxes available at specific distances from a typical cooling tower (CPDTE data). The model is recommended for analyzing the drift drop plumes under the conditions similar to the CPDTE conditions of small Stokes numbers. It is easier to use and not less accurate than the multiphase Eulerian- Lagrangian CFD models used recently in [3, 7] for modeling the CPDTE plume. The multi-group model has a potential to supplant or complement the latter in the computational analysis of gravitational phenomena in engineering equipment and its environment.

27 Conclusions Three CFD models have been developed, validated and applied to real-life industrial applications: 1. flammable gas release and dispersion model for safety analyses (use of hydrogen, methane, etc.); 2. single-phase pollutant dispersion model for air quality analyses; 3. two-phase multi-group model for analyses of drift drop plumes from cooling towers. The models are based on customizing PHOENICS CFD software, which is used as a framework and a solver. User-friendly CFD templates are available for applications by environmental and safety engineers for risk and safety assessments. More information on PHOENICS licensing, training, user support and CFD consulting services is available from info@acfda.org.

28 Acknowledgements The authors gratefully acknowledge the financial support of Natural Resources Canada (NRCan) for development and validation of some single-phase GRAD CFD models. Also, the authors would like to acknowledge helpful communications and information provided by Dr. Steven Hanna (Hanna Consultants), Prof. Robert Meroney (Colorado State University) and Jeffrey Weil (Research Applications Laboratory at the National Center for Atmospheric Research) on the experimental data and modeling results for Chalk Point Dye Tracer Experiment [3,6].

29 References 1. PHOENICS On-line Information System: 2. Agranat, V.M., Tchouvelev, A.V., Cheng, Z. and Zhubrin, S.V. (2007). CFD Modeling of Gas Release and Dispersion: Prediction of Flammable Gas Clouds. Advanced Combustion and Aerothermal Technologies, Edited by Syred, N. and Khalatov, A., Dordrecht, Netherlands: Springer, Meroney, R.N. (2006). CFD Prediction of Cooling Tower Drift. Journal of Wind Engineering and Industrial Aerodynamics, 94 (6), Schulman, L.L. and DesAutels, C.G. (2013). Comparing AERMOD and CFD Downwash Predictions for GEP Stacks. Proceedings of the Air & Waste Management Association s Conference Guideline on Air Quality Models: The Path Forward, Raleigh, North Carolina, USA, March 19-21, 2013, Control # Radonjic, Z., Agranat, V., Telenta, B., Hrebenyk, B., Chambers, D. and Ritchie, T. (2013). Comparison of Near-field CFD and CALPUFF Modelling Results around a Backup Diesel Generating Station. Proceedings of the Air & Waste Management Association s Conference Guideline on Air Quality Models: The Path Forward, Raleigh, North Carolina, USA, March 19-21, 2013, Control # Hanna, S.R. (1978), A Simple Drift Deposition Model Applied to the Chalk Point Dye Tracer Experiment. Environmental Effects of Cooling Tower Plumes, Symposium on, May 2-4, 1978, U. of Maryland, III-105-III Lucas, M., Martinez, P.J., Ruiz, J., Sanchez-Kaiser, A. and Viedma, A. (2010). On the Influence of Psychrometric Ambient Conditions on Cooling Tower Drift Deposition. International Journal of Heat and Mass Transfer 53 (4),

30 Thank you!