HAZARDOUS RELEASE SCENARIO ANALYSIS VIA COMPUTATIONAL .
HAZARDOUS RELEASE SCENARIO ANALYSIS VIACOMPUTATIONAL FLUID DYNAMICSbyArunabh BorkatakyRuohan WangA ReportSubmitted to the Faculty of Purdue UniversityIn Partial Fulfillment of the Requirements for the degree ofProfessional Master of Science in Chemical EngineeringDavidson School of Chemical EngineeringWest Lafayette, IndianaAugust 2017
iTABLE OF CONTENTSCHAPTER 1 EXECUTIVE SUMMARY .1CHAPTER 2 SCIENTIFIC ASSEMENT.22.1 FDS and Smokeview .22.2 Source Term Methodology .22.2.1 Source Model .32.2.2 Flash Evaporation .42.2.3 Pool Vaporization .52.2.4 Pool Spreading .62.2.5 Dispersion Model .7CHAPTER 3 RESULT AND DATA INTERPRETATION.73.1 Mathematical Modelling .73.2 FDS Simulation .93.2.1 Defining the Mesh .93.2.2 Specifying the Ambient Conditions and Wind Speed .93.2.3 Defining the Sources of Propane Release .103.2.4 Gas Detectors and Obstructions.113.2.5 Smokeview Representation and Analysis .12CHAPTER 4 DISCUSSION AND CONCLUSIONS.144.1 Validation Studies.144.1.1 Case 1 .144.1.2 Case 2 .154.1.3 Iterative Nature of FDS.184.2 Conclusion .19CHAPTER 5 PROPOSED STEPS FORWARD .20
ii5.1 Improvement of results .205.2 In general.20REFERENCE .21APPENDIX A .22APPENDIX B .23APPENDIX C .29
1CHAPTER 1EXECUTIVE SUMMARYComputational Fluid Dynamics (CFD) has two major roles in risk management:(a) Before building a facility: risk assessment and analysis of potential fires and toxicreleases; to identify optimal placement of detectors, alarms, sprinklers; mitigatewhat might happen.(b) After a release has occurred: root cause analysis of the situation; prediction of therelease path; identify areas of high chemical concentration(s); preventive actionagainst similar future releases.The Chemical Process Industry has started to incorporate CFD into hazardous releasescenario analysis. As part of this integration, it is essential to determine how accurately theexisting methods for CFD modeling can predict an actual release so that it can beeffectively used for risk management in such cases (risk analysis and assessment). The coreof this project was to perform a Root Cause Analysis (RCA), simulate an actual releaseusing CFD, and validate the developed models using first hand data from the release [1].RCA is the very foundation of risk management and is used to determine why an incidentoccurred and to form as basis for actions can be taken to prevent future occurrences ofsimilar situations. This project was suggested by the Purdue Process Safety & AssuranceCenter (PPSAC) steering team. The details of the leak are mentioned below.A far-reaching propane release occurred at a Natural Gas Liquid (NGL) facilitiesβ sales gasmetering skid used for analysis of the quantity and quality of propane transferred thoughthe pipelines. The leak which lasted for about 47 minutes released large quantities of highlyflammable propane into the atmosphere. The propane liquid pool boils vigorously whilealso spreading on the ground due to the massive difference between the ground temperatureand its normal boiling point. Any ignition source near the leak has a potential for a fire orVapor Cloud Explosion (VCE) and hence an accurate CFD model is important for RCAand risk management of such situations.
2The computational work in this project was carried out in Fire Dynamics Simulator (FDS)version 6.5.3 and Smokeview version 6.4.4 was used to visually represent the models. 65simulations were carried out in total, and the 65th model closely resembled the actual leakfor the first 200 seconds in terms of the propane cloud width and dispersion, velocity ofthe release, height of the release, direction of dispersion, etc.CHAPTER 2SCIENTIFIC ASSESSMENT2.1 FDS and SmokeviewFire Dynamics Simulator is an open source Computational Fluid Dynamics (CFD)software developed by the National Institute of Standards and Technology (NIST). It usesa low Mach number approximation appropriate for low speed applications like fire, vapordispersion, etc. to numerically solve the Navier-Stokes equations [2].Smokeview is a software tool which is designed to visualize the results of an FDSsimulation. It is an essential tool which assists FDS users to monitor and visualize asimulationβs progress [3].2.2 Source Term MethodologyDefining the source term is the first step towards modelling a hazardous release scenario(Fig 1). First, an incident such as the rupture of a pipeline, a hole in a tank or a pipe, etc. isdefined and then subsequent source models are employed to describe the release. Thesource model is used to determine the rate of discharge, the total amount discharged andthe state of the discharge (solid, liquid, vapor or a mixed fraction). Next, a dispersion modelis used to predict the downwind movement and concentration(s) of the released material(s)[4].
3Fig 1 Flow of source term methodology for the Project2.2.1Source ModelMechanical energy balance associated with fluids in motion, [5]: π π·πΜ πππ (ππΆπ ) π π«π π πππΎππΜ(1)P Pressure (force/area)Ο Density of the fluid (mass/volume)π’Μ Average instantaneous velocity of the fluid (length/time)ππ Gravitational constant (length mass/force time2)Ξ± 0.5 for laminar flow, Ξ± 1.0 for plug flow, and Ξ± 1.0 for turbulent flow,gAcceleration due to gravity (length/time2 )zHeight above datum (length)FNet frictional loss term (length force/mass)ππ Shaft work (force length/time)πΜ Liquid discharge rate (mass/time)For our model, pure propane discharged from the leak source, which was an orifice in oneof the pipe fittings of the plant. Hence, the energy balance can be simplified to:πΜ π¨πͺπ« ππππ (π·π π·π )πΜ Liquid discharge rate (mass/time)(2)
4A Area of the hole (length2)πΆπ· Discharge coefficient (dimensionless)π1 Upstream pressure (force/area)π2 Downstream pressure (force/area)The coefficient of discharge was determined using the 2-K method [5]:πͺπ« π(3) π π²π πΎπ is the sum of all excess head loss terms, for Reynolds numbers greater than 10,000,πΎπ 0.5 for entrance and 1.0 for exit, and thus πΆπ· 0.632.2.2Flash Evaporation:When superheated propane comes in contact with atmospheric pressure and temperature atthe point of the leak, it breakdowns into small droplets which is called flashing or flashevaporation. The fraction of propane that flashes is calculated assuming that the sensibleheat contained within the superheated liquid is used to vaporize a fraction of the liquid [4].π π πͺπ(π» π»π)(4)ππππΆπ Heat capacity of the liquid, averaged over T to Tb (energy/mass deg)T Initial temperature of the liquid (deg)Tb Atmospheric boiling point of the liquid (deg)hfg Latent heat of vaporization of the liquid (energy/mass)Fv Mass fraction of released liquid vaporizedFDS does not include any sub-models to determine flashing from the release of asuperheated liquid and hence the above model is used to input the resultant terms intoFDS [6].2.2.3 Pool Vaporization:At steady state, the vaporization rate is given as [4]:πΜ π―π³(5)
5πΜ Vaporization rate (mass/time)HTotal heat flux to the pool (energy/time)LHeat of vaporization of the pool (energy/mass)For a spill of liquid with normal boiling point below ambient temperature (231K forpropane), the initial stage of vaporization is assumed to be due to the heat transfer from theground. Hence, the heat flux from the ground to the pool can be expressed in terms of asimple one-dimensional heat conduction equation [4]:π (π» π»)π πππ (π πΆπ)π/π(6)πππ Heat flux from the ground (energy/area)ππ Thermal conductivity of the soil (energy/length deg)ππ Temperature of the ground (deg)T Temperature of liquid pool (deg)πΌπ Thermal diffusivity of the ground (area/time)t Time after spill2.2.4 Pool Spreading:The radius of the pool is an important parameter for the evaporation model as it is directlyproportional to total rate of vapor released into the atmosphere and subsequent dispersionby wind. The pool spread was calculated using Wu and Schroyβs (1979) model whichassumes that the pool growth is radial and uniform from the point of spill, unconstrainedand on a flat surface. [4].πππ« [πͺππ π /ππ ππΈππ¨πππ/π ππ¨π¬ π· π¬π’π§ π·]rPool radius (length)tTime after the spill (time)C Constant developed from experimental data, see below (dimensionless)gAcceleration due to gravity (length/time2)ΟDensity of the liquid (mass/volume)ππ΄πΉ Volumetric spill rate after flashing (volume/time)(7)
6Β΅Viscosity of the liquid (mass/length time)Ξ²Angle between the pool surface and the vertical axis perpendicular to the ground,see below (degrees)The Reynolds number for the pool spread is given by [4]π΅πΉπ ππΈπ¨π π(8)π ππC 2 for ππ π 25, and C 5 for ππ π 25π.ππ πππ§ π [(π. ππ π©)π.π π. π]π ππ.πππππ π(9)(10)πΈπ¨π πThe pool radius is iteratively determined using the above equations. As FDS does notinclude any sub-models for the pool spread, pool dimensions and correspondingvaporization rate(s) are used as inputs for FDS modelling [6].2.2.5 Dispersion ModelAs the density of propane vapor is greater than that of ambient air, a dense gas dispersionmodel should be used for our model. However, dispersion effects can be directly modelledin FDS by specifying certain parameters including wind speed, atmospheric stability,surface roughness etc.CHAPTER 3RESULTS AND DATA INTERPRETATION3.1 Mathematical Modelling:Pure propane was released through a 0.5-inch hole, at a line pressure of 450 psig andtemperature ranging from 60-100F. The details of the release are mentioned in Table 1.Additional details such as the physical properties of propane and sand have been includedin Appendix A.
7Table 1 Parameters defining the leakParameterValueMass flow rate4.78 kg/sTotal mass of propane released13500 kgFraction of propane flashed, ππ0.4569Mass flow of propane flashed2.187 kg/sMass flow of propane into pool2.6 kg/sUsing the mass flow rate of propane going into the pool i.e. 2.6 kg/s and the equationsdescribed in 2.2.3 and 2.2.4, the pool spread and evaporation models were calculated.Based on the location of the release, dry medium sand was assumed as the ground(Appendix A). The pool evaporation rate is directly proportional to the surface area (radius)of the pool as is evident in Fig 2 and Fig 3. As time progresses, the change in evaporationrate gradually decreases as the surface cools. When the leak is contained after 2820s, thepool reaches a radius of 37.5 m.Fig 2. Increase in pool radius with timeFig 3. Increase in pool evaporation rate with timeAccording to the data collected [1], an average wind speed of 3.6 m/s was reported blowingfrom the North South direction (Appendix C). The atmospheric stability was assumed tobe class A due to the low wind speed, strong-moderate insolation and as the release wasduring the day. Wind Class is shown in Fig 4. The credibility of the wind data was initiallyquestioned as the weathering station was located far from the area of the leak. After running
8the FDS simulations we concluded that the wind data was inaccurate, further details ofwhich have been included in Chapter 4.Fig 4. Wind Class3.2 FDS Simulation3.2.1 Defining the MeshAll FDS computations are executed within a region made up of rectilinear volumes calledmeshes. Each mesh is divided into several rectangular cells which depend on the desiredresolution of the flow dynamics [7]. The area of concern was determined to be 150m by104m from the plot overview [1]. For our simulations, we assumed a mesh size of200m*200m*10m with cell dimensions of 1m*1m. The following line is used to initializethe mesh.By default, FDS assumes that the exterior boundaries of the computational domain are solidwalls [7]. For outdoor simulations, each boundary must be explicitly defined OPEN asshown below.
93.2.2 Specifying the Ambient Conditions and Wind SpeedThe release occurred on the 6th of December 2013 in the Middle East, and hence an ambientcondition of 20 C and 1 bar were assumed for the simulations. The MISC namelist is usedto specify these conditions in FDS.There are a couple of methods available in FDS to specify a wind field, the one used in oursimulations is the one recommended in the FDS user guide [7]. For this method, eachcomponent of the wind velocity vector is explicitly defined using U0, V0 and W0. Eachvelocity component can be individually varied with time and height. The following code,points a wind of 3.6 m/s in the negative x direction, i.e. from the right to the left of themesh.3.2.3 Defining the Sources of Propane ReleaseThere are two sources for the propane release into the atmosphere: propane flashing fromthe leak point and propane evaporating from the liquid pool on the ground. In FDS we haveto specify separate models for both these sources. Defining a source of release in FDS is athree steps process:1. Defining the leak in terms of the species released with SPEC namelist; the massflux of the release etc. using the SURF namelist.2. Include rectangular obstructions into the domain using OBST namelist.3. Use VENT namelist to inject propane into the computational domain.The physical properties of propane gas are tabulated within FDS and need not be explicitlyspecified. An example of defining the source(s) is include below.
103.2.4 Gas Detectors and ObstructionsThere were 15 propane gas sensors located around the area of the release, Fig 5. Thedetectors were calibrated to alarm if the concentration of propane (vol %) increased above25% of its Lower Flammable Limit (LFL) i.e. 2.1% (vol /vol). DEVC namelist is used todefine the gas sensors in FDS. Each gas sensor has a time dependent output associated withit, which is stored in a comma-delimited ASCII file (.csv). At the end of the simulation, the.csv file was analyzed to note the time at which each sensor first reported a concentrationgreater than or equal to 0.00525 (Appendix C).Fig 5 Location of the gas detectors (not to scale)Any notable obstructions which might affect the flow path of propane vapors and the poolwere included in the simulations using the OBST namelist.
113.2.5 Smokeview Representation and AnalysisFor the base model of our project, we decided to exclude any wind as we didnβt haveaccurate data. Further models with different wind profiles were developed to validate themodel(s) with the actual alarm summary included in Chapter 4. All the data included inthis Chapter are for the no wind condition.The dispersion path of the propane vapors can be visually represented by using the ISOFnamelist. The concentrations (mol/mol) to be displayed in Smokeview must be explicitlyspecified. For our models, we selected concentrations of 2.1 % propane in the atmosphere,which is its LFL and 0.525% which is the concentration above which the detectors go off.Figure 6 shows the direction of the release from 14 secs to 452 secs.Fig 6 Spread of Propane VaporAs propane vapor is heavier than air, the height to which it is dispersed is limited. Todetermine the maximum height of the dispersed gas, we studied the propane concentrationsat various planes and concluded that it is essentially zero above 9 m.Further analysis of the leak was done by examining the variation of propane volumefraction at different planes within the domain, using the SLCF namelist. Fig 7 shows thedistribution of propane mole fraction, 10 minutes into the release, at the plane Z 1, whichis the ground (sand) for our simulation.
12Fig 7 Smokeview rendering of the propane fraction at Z 1In the absence of any wind, the sequence in which the detectors alarm (i.e. reach aconcentration above 25% LFL) is different from the detector alarm sequence collectedfrom the release location as shown in Fig 8. Also, the total time required for all the detectorsto sound is much greater as there is no external driving force to speed up the dispersion ofpropane vapors.Fig 8 Detector alarm summary
13Table 2 Detector alarm summaries (No Wind)Actual SequenceTime between eachFDS SequencedetectorTime between 26134226038260701734 A11261071746 A1261224.51734 B11734 A213.5260741734 B501746 B31746 A2726061201746 B5733.2.6 3.6m/s from NE to SWFor this case, we assumed a wind blowing at an average speed of 3.6 m/s from the NNEdirection for the entire duration of the leak in accordance to the data provided (AppendixC). The North in the plot overview was in the negative x direction and hence the winddirection input into the FDS simulation was adjusted accordingly. Fig 9 depicts how themole fraction of propane on the ground varied throughout the leak.
14Fig 9 Variation of propane mole fraction on the ground (Case 1)A comparison of the detector alarm summaries in Fig 10, shows the inaccuracy of the winddata. Only five detectors marked in red reached a concentration above 25% LFL propanewhich is not consistent with the actual detector alarm summary. The unalarmed detectorswere located opposite to the direction of the wind.Fig 10 Detector alarm summaries (Case 1)
15Table 3 Detector alarm summaries (Case 1)Actual SequenceTime between eachFDS SequencedetectorTime between 6043.526102260610.5261372610Did not alarm261112613Did not alarm2612122611Did not alarm260382612Did not alarm1734 A112603Did not alarm1746 A11734 ADid not alarm1734 B11746 ADid not alarm260741734 BDid not alarm1746 B32607Did not alarm26061201746 BDid not alarmCHAPTER 4DISCUSSION AND
Fire Dynamics Simulator is an open source Computational Fluid Dynamics (CFD) software developed by the National Institute of Standards and Technology (NIST). It uses a low Mach number approximation appropriate for low speed applications like fire, vapor dispersion, etc. to numerically solve the Navier-Stokes equations [2].
Gap analysis CAV Executive Report Planning Process Planning Process. Scenario Planning Scenario Planning. Key Traffic/ITS Assumptions Scenario 1 Scenario 2 Scenario 3 Scenario 4 CV 15% vehicles 75% of vehicles 50% of vehicles 75% of vehicles . PowerPoint Template Keyword
3.3 Scenario 1: Subcontracting Based on a Use Case Model 3.4 Scenario 2: Subcontracting Based on an Analysis Model 3.5 Scenario 3: Subcontracting Based on a Design Model 3.6 Scenario 4: Programming Subcontracting 3.7 Scenario 5: Test Subcontracting 3.8 Scenario 6: Subcontracting Programming and Tests 3.9 Hybrid, Trivial, and Canonic Scenarios 15 4.
Hazardous Materials Module: NFIRS-7 Objectives After completing the Hazardous Materials Module the students will be able to: 1. Describe when the Hazardous Materials Module is to be used. 2. Demonstrate how to complete the Hazardous Materials Module and identify appropriate other modules, given the scenario of a hypotheti-cal incident.
the hazardous Materials table in 49 CFr 172.101. note: dot hazardous materials include all resource Conservation and recovery act (rCra) hazardous waste and Comprehensive environmental response, Compensation, and liability act (CerCla) hazardous substances. Hazardous Waste as regulated by epa in 40 CFr 261 In general, a waste is hazardous .
Conduct detailed inspection of hazardous areas installations dust atmospheres Conduct detailed inspection of hazardous areas installations pressurisation Manage compliance of hazardous areas Report on the integrity of explosion-protected equipment in a hazardous area UEENEEM078A UEENEEM080A UEE42611 Certificate IV Hazardous Areas - Electrical
Laboratory Hazardous Waste Disposal Procedures 8 Disposal of Biohazardous Waste 9 Disposal of Sharps (and Syringes) 12 . For a list of chemicals that are considered non-hazardous visit the Non-Hazardous Chemical Disposal Guide. Regularly Scheduled Hazardous Waste Pick-up . ESF picks up hazardous waste throughout UBC on a scheduled basis. Pick .
The Sun News View Release San Jose Mercury News View Release The Miami Herald View Release. Star Tribune View Release CEO World News View Release AZ Central . Poteau Daily News View Release The Evening Leader View Release Wapakoneta Daily News View Release Observer News Enterprise. View Release Valley City Times-Record
Scenario 2 is a variant of Scenario 1 where the only difference is the replacement of the activities of Year 2 in Scenario 1 with a two-year phase-in. Under this scenario, more time is available for the development of the storage and transportation infrastructure, and some P application is still allowed in the first of the two years of phase-in.
