Gas Dispersion With OpenFoam DANSIS Chris Dixon

8/13/2019 Gas Dispersion With OpenFoam DANSIS Chris Dixon

1/37

GAS DISPERSION WITH OPENFOAM

Fluid Mechanics in Offshore Oil & Gas.

th

Copyright of Shell Research Ltd

Chris Dixon

Major Hazards Management Centre of Expertise

1April 2013


2/37


3/37

AGENDA

Introduction/Motivation

OpenFOAM

Solver

Validation

Test Case


Conclusions

April 2013 3


4/37

The need for a lar e number of CFD dis ersion calculations.

INTRODUCTION

1.0


April 2013 4


5/37

DESIGN OVERPRESSURE

Understanding explosion overpressure is important in the design ofonshore and offshore plant; for example blast walls, protection of

Emergency Shut Down Valves etc. The assessment of explosion hazards due to accidental releases of

flammable gas often requires the quantification of the likely gas cloud


conditions.

There is a need to consider a rangeof scenarios because, althoughthe simplest approach in assessing explosion hazards is to calculate a

worst-case explosion, a genuine worst case is an event of extremelylow probability.

April 2013 5


6/37

EXCEEDANCE CURVES

Results normally presented as exceedance curves:

Overpressure exceedance

Copyright of Shell Research Ltd April 2013 6

Gas cloud size exceedance


7/37

LEAK SCENARIOS

An individual release scenario is typically specified by a number ofparameters such as:

The released material and its conditions (pressure and temperature). The location and orientation and the size of the leak.

The wind speed and direction.


These parameters are not chosen arbitrarily: Each leak is associated with some item of equipment, the frequency of

leaks of differing sizes is found by a parts count which supplies the

number of items of various types valves, flanges etc. and historicaldata on failure rates for these items.

The wind speed and direction are chosen from a probabilitydistribution characterised by the wind rose for the location.

April 2013 7


8/37

THE MAIN PROBLEM

Have two desires:

Would like to do enough gas dispersion calculations to properly

characterise the gas build-up exceedance curve. Would like each calculation to be accurate.


These two desires pull in opposite directions: Using a Monte Carlo approach, over a thousand dispersion

calculations may be required.

CFD calculations have, historically, been too slow to allow this.

April 2013 8


9/37

PREVIOUSLY...

The Shell methodology has historically emphasised the

Monte Carlo aspect of the problem.

Model dispersion using particle based random walk

model called DICE (Dispersion In CongestedEnvironments).

Particle velocity =


velocity due to jet release momentum+background flow velocity +

semi-random turbulent component

Calculate trajectories of 1000s of independent particlesand determine gas concentration

Assume fixed background flow field calculated fromCFD.

9

-20 -10 0 10 20 30 40

-20

-10

0

10

20

0

0.05

0.1

0.15

Concentration

April 2013


10/37

A stead -state PDR solver with sub- rid et source.

OPENFOAM CFD SOLVER(S)

2.0


April 2013 10


11/37

OPENFOAM (1)

Open source CFD code.

FOAM = Field Operation and Manipulation.

Originates from Imperial College.

Around since ~1998.

Available to download.


.

Training available from various suppliers.

Really multiphysics, but seems to have moved to CFD.

Use C++ features to allow solvers to be constructed writing equations

as they are written.

11April 2013


12/37

OPENFOAM (2)

Example:

0=+

TDTt

T

Tu


(

fvm::ddt(T)

+ fvm::div(u, T)

- fvm::laplacian(DT, T)

);

12April 2013


13/37

OPENFOAM (3)

Adopted as platform for next-generation explosion CFD code within

Shell Major Hazards Group to replace existing in-house code (EXSIM).

A Porosity/Distributed Resistance solver (PDRFoam) is part of the

standard distribution. Development was funded by Shell.

OpenFOAM chosen because:

Modern architecture.


Inbuilt Parallelism.

Arbitrary meshing.

Flexible & extensible.

Safe against vendor changes.

April 2013 13


14/37

SOLVER OVERVIEW

Basis:

Compressible SIMPLE-based solver.

Uses Porosity/Distributed Resistance method (if needed). Arbitrary mesh.

Additional species equation.


Tabulated thermodynamics. Sub-grid gas release model.

Note:

Transient PDR solver is part of standard OpenFOAM distribution(PDRFoam).

Solver used here is not PDRFoam, but a steady-state version. Aim is to

get reasonable dispersion results with short run-time. April 2013 14


15/37

POROSITY/DISTRIBUTED RESISTANCE (1)

Convert geometry to cell-wise values of:

Volume porosity

Area Porosity

(Tensor) drag

Turbulence generation

Copyright of Shell Research Ltd 15

Drag on log-scale.

(NB Small obstacles omitted in picture)

April 2013


16/37

POROSITY/DISTRIBUTED RESISTANCE (2)

Copyright of Shell Research Ltd 16April 2013


17/37

THEMODYMAMICS

Solver accepts tables, but these produced externally (by any method).

In this case using Shell in-house thermodynamics code.

Modified Redlich-Kwong. Accounts for condensation/evaporation of ambient water.

Single species equation (no requirement for water vapour equation etc.)

Copyright of Shell Research Ltd 17 April 2013


18/37

Testin the solver a ainst ex erimental data.

VALIDATION

3.0


18April 2013


19/37

FULL-SCALE: JIP RIG

Physical rig


CAD representation

April 2013


20/37

RESULTS: VOLUME FLOW RATE IN JIP RIG

Experimental volume flow rate found from anemometers within rig at10 locations.


Series A has no blockage at rig ends, series B ~80% blockage.

Significant uncertainty in wind speed.

April 2013


21/37

MESH SENSITIVITY: VOLUME FLOW RATE IN JIP RIG

Pre-release volume flow rate in series A approximately linear withwind speed.


Volume flow rate shown for two mesh sizes.

Topology changes with different mesh resolution as blockage resolved.

April 2013


22/37

SUB-GRID JET MODEL


Free choked methane jet

22

Full scale methane jet in cross-flow

April 2013


23/37

SUB-GRID JET MODEL FLASHING CASE (CO2)

Concentration vs. distance

Horizontal liquid CO2 release at 150 bar.


Temperature vs. distance

April 2013


24/37

EXPERIMENTAL ARRANGEMENT

Series A & B


Series C

April 2013


25/37

RESULTS: FLAMMABLE GAS VOLUME

Experimental gas cloud volume found from number of measurementlocations with concentration above lower flammable limit.


CFD results are improvement on previous methodology less outliers.

April 2013


26/37

RESULTS: CASES WHERE DICE AND CFD DIFFER

Some cases where DICE gives poor results:

Case C05: obstacle omitted.


Case A28: flow reversed.

April 2013


27/37

CASE C05 CFD RESULTS

Concentration on 4 planes at 1, 3, 5 and 7 m above ground.


CFD results shown by contours, experiment by spheres on same scale.

Spheres placed at measurement locations. April 2013


28/37

RESULTS: CASES WHERE CFD IS POOR

Some cases where CFD gives poor results are: For A2, A22, A24 CFD under-predicts.

For A8 CFD over-predicts.



29/37

CASE A02 CFD RESULTS (OPENFOAM UNDER-PREDICTS)

Concentration on vertical plane though release.


CFD results shown by contours, experiment by spheres on same scale. Spheres placed at measurement locations.

Black point shows release location.

Results actually quite reasonable. April 2013


30/37

CASE A17 CFD RESULTS (OPENFOAM OVER-PREDICTS)

Locations with concentration above 2.5%.


CFD results shown by isosurface (right). Spheres placed at measurement locations (left).

Cloud shape is reasonable.

Experiment has large volume a little below LFL, CFD a little above.April 2013


31/37

MESH SENSITIVITY: CLOUD VOLUME

Cells with mass fraction above 5% shown for two meshesCase on left has 5,250 cells in enclosure volume.

Case on right has 32,256 cells in enclosure volume.


Plume shapes qualitatively very similar.

Flammable cloud volume differs by 7%.

April 2013


32/37

Usin the solver to roduce an exceedance curve.

APPLICATION

4.0


32April 2013


33/37

SAMPLE EXCEEDANCE CURVE

Use JIP rig geometry. Use wind rose from a previous project.

Assume some plant items in 10 distinct 2 m cubes within the rig.

Assume that each item has 25 and 50 mm releases. Assume some leak frequencies.

Discretize wind into 8 directions N NE etc . For each leak scenario


do at least two releases for each wind direction one in lower half ofwind speed probability distribution one in lower half.

Leak direction is random. Leak location random within cube.

Number of repeats for each leak scenario depends on leak rate andfrequency.

33April 2013


34/37


A typical resulting exceedance curve is shown below.



35/37


A typical resulting exceedance curve is shown below.



36/37

SUMMARY

OpenFOAM can provide a reasonable platform for gas dispersion CFD:

PDR and resolved regions can be mixed and matched.

PDR regions can be locally refined. Flexible, extensible...

Generally gas cloud exceedance calculations accentuate the need for


either accurate individual dispersion calculations or a statisticallymeaningful number of calculations. The word accurate should be

taken in context.

Until recently it has not been possible to undertake a sufficiently large

number of CFD calculations to solve the Monte-Carlo problem.

It is now becoming possible using a steady-state PDR approach to solvethe Monte-Carlo problem directly using CFD calculations.

36April 2013


37/37

Date post:	04-Jun-2018
Category:	Documents
Upload:	userscribd2011
View:	222 times
Download:	1 times

Gas Dispersion With OpenFoam DANSIS Chris Dixon

Documents