Академический Документы
Профессиональный Документы
Культура Документы
Individuals working in process facilities containing toxic South Korea, on September 27, 2012 [2], and the anhydrous
gases may face gas releases and poisoning risks. Many studies ammonia release accident in Theodore, Alabama, on August
regarding individual risk (IR) have been carried out on a 23, 2010 [3]. Also, in China, according to the statistical data
worst scenario basis. However, the worst scenario-based from 2002 to 2012, there were 55 poisoning accidents that
approach cannot represent realistic release risks and may occurred in process industry, which caused 81 deaths and
overestimate the IR. In this study, an approach based on 917 injuries [4]. These above accident cases show that people
complete accident scenario set (CASS) and computational who work in the process facilities or live in adjacent areas are
fluid dynamics (CFD) is proposed to quantitatively assess IR of facing risks of toxic gas poisoning. Therefore, it is very impor-
toxic gas release in process facilities. By combining the gas tant to develop strategies to predict accident scenarios and
leakage probability and joint distribution probability of the evaluate the risk of poisoning for providing mitigation mea-
wind direction and speed, a CASS can be built. The CFD sures in process plants.
method is used to predict the concentration field of gas release Many researches concerning both the individual risks
and dispersion. Then, the toxic gas concentration can be con- (IR) and society risks (SR) inside the process plants have been
verted to poisoning fatality probabilities according to the dose– carried out [5–7]. Normally IR was defined as the probability of
response model. A virtual IR contour can finally be defined by death to an individual at any particular location due to all
the accumulative assessment of all release scenarios. A case undesired events [8,9]. The total IR at each point was equal to
study of an IR area classification in a natural gas process and the sum of the IR of all scenarios effects at that point [10],
carbon dioxide recycle terminal processing facility that con- whereas SR expressed the cumulative risk to groups of people
tains an ammonia refrigeration system is also investigated. who might be affected by such accident scenarios [11]. This
With the proposed methodology, the quantitatively classified IR paper will focus on the IR inside the process plants since the
level in process facilities can provide scientific references for individuals may be facing a more direct exposure to toxic
safety decision makers. © 2018 American Institute of Chemical Engi- gases under release conditions.
neers Process Saf Prog 38: 52–60, 2019
The IR evaluation is also in the scope of quantitative risk
Keywords: complete accident scenario set; computational
analysis (QRA) [10]. QRA is the prevailed risk assessment
fluid dynamics; toxic gas; individual risk; dose–response
method and has been widely used in process safety related
model; ammonia poisoning
areas, including design of safety measures [12], safety manage-
ment and decision making [11,13,14], developing risk-based
1. INTRODUCTION AND BACKGROUND maintenance and inspection strategies [15–17], and so forth.
Currently, the safety of toxic chemical process facilities has The implementation of QRA mainly contains two steps: sce-
aroused great concern since the toxic chemicals in these facili- nario envisaging and consequence analysis, in which scenario
ties may severely impact workers, the public and the environ- envisaging serves as the basis for the whole risk quantification
ment. Several accidents occurred in recent years have brought [18]. Some scenario envisaging techniques have been pro-
many casualties and enormous economic losses, such as the posed, such as worst accident scenarios (WAS) [19],
chlorine release accident in Atchison, Kansas, on October maximum-credible accident scenarios (MCAS) [20], reference
21, 2016 [1], hydrogen fluoride leakage accident in Gumi, accident scenarios (RAS) [21], worst maximum credible acci-
dental scenarios (WMCAS) [22], and so forth. As we know, the
WAS approach is the most commonly used method to assess
release consequences [19,23]. However, the WAS may be the
This study was supported by the Natural Science Foundation of
Shandong Province (ZR2016EEM27) and the Fundamental Research exaggerated situation and may overestimate the consequences.
Funds for the Central Universities (18CX05028A). In order to overcome this shortage, the concept of MCAS has
† been proposed and widely used [18,20,24]. Another approach
These authors contributed equally to this work.
© 2018 American Institute of Chemical Engineers to accident scenarios identification was RAS, which was
Process Safety Progress (Vol.38, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2019 53
Table 1. Typical toxic gas leak accident. occurrence frequency, and the other various speeds can be
described by representative values between the maximum and
minimum speed with a certain interval. Then, the wind field
Date Location Consequences
set can be described by a combination of wind direction and
09/18/2015 Shandong Province, 215 people needed speed, as shown in Eq. 2.
China medical treatment,
5 people injured W ¼ ½wθv ð0 ≤ θ ≤ 16,0 ≤ v ≤ mÞ ð2Þ
seriously
09/07/2014 Ningxia Hui 41 people suffered
Autonomous Region, from acute ammonia Because of the independence of leakage sources and wind
China poisoning field, both variables are combined randomly to construct the
08/31/2013 Shanghai, China 15 deaths, 25 injured scenario set, as shown in Eq. 3.
06/03/2013 Dehui city, Jilin 121 deaths, 76 injured
Province, China
S ¼ fSθveh j0 ≤ θ ≤ 16,0 ≤ v ≤ m,0 < e ≤ q,0 < h ≤ 4g ð3Þ
10/22/2012 Honghu city, Hubei 479 people suffered
Province, China from ammonia
poisoning
08/05/2009 Chifeng city, Inner 246 injured, including 2.2 Assessment of the IR of Poisoning
Mongolia, China 21 ammonia The IR is the risk to an individual near the hazard, which
poisoning cases considers the nature, the likelihood and the time period of a
possible injury to an individual [42]. In this paper, the occur-
rence frequency of each scenario and corresponding individ-
ual exposure FR can be calculated to describe the IR values in
2.1.2 Wind Fields each interested or given location of a process facility. The con-
To set up the wind field set, the local historical meteorolog- centrations during the gas dispersion predicted by CFD can be
ical data including wind speed and direction were considered. used to evaluate the FR by the dose–response model. For
The wind direction can be divided into 16 directions with an instantaneous, time-varying releases, the toxic dose is esti-
interval of 22.5 (Figure 2) [41]. The wind speed, which should mated by integration or summation over several time incre-
be first considered, is the average speed since it has the largest ments [42], as shown with Eq. 4.
54 March 2019 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.38, No.1)
Table 2. Leakage rate of the ammonia refrigeration system (per year).
ð t1 X
k
according to their frequencies and wind speeds. In this case,
V¼ C dt ≈
j
Ci Δti
j
ð4Þ the average wind speed scenarios were chosen since they
t0 i¼1
have an occurrence frequency of 83%. Additionally, special
attention was given to the maximum and second largest wind
Probit equations are derived as lines of best fit to percent- speed because they may enable the released toxic gas to dis-
age fatalities versus concentration and duration using log- perse in a wider range. Finally, eight scenarios were chosen to
probability plots or standard statistical packages, as shown establish the wind fields, which consisted of 92% of all possi-
with Eq. 5. ble situations (Table 3).
Based on the above approach, with a combination of
Pr ¼ A + B lnðV Þ ð5Þ 40 leakages and 8 wind fields, a total of 320 potential leak sce-
narios were set up. Considering the computation cost and the
For single exposure, the probit variable Pr can be related to frequency of possible scenarios, a selection criterion was set,
the FR P by Eq. 6 [37]. and the results are shown in Table 4. The scenario occurrence
probability is defined as the product of leakage probability
ð Pr − 5 2 and wind field occurrence probability. Finally, 82 scenarios
1
e−
x
P ¼ pffiffiffiffiffi 2 dx ð6Þ were chosen, which account for a portion of the 99.2% of all
2π −∞ scenario cumulative occurrence probabilities.
Considering the occurrence frequency of each gas release
scenario, the IR of scenario u in the location (x, y) is calculated
by Eq. 7. 3.3 Simulation of Gas Leak and Dispersion in the
Process Facility
IR u ðx,yÞ ¼ fu Pu ð7Þ Ammonia release and dispersion simulation are one of the
most important steps in this study. A facility geometry model
Finally, the IR of a given point can be obtained by accumu- was created by the CFD preprocessor GAMBIT. The computa-
lating the IRs of all potential scenarios, as shown with Eq. 8: tional domain has dimensions of 800 m in length, 700 m in
X
n
Table 3. Wind field set.
IRxy ¼ fu Pu ð8Þ
u¼1
No. Wind Direction Wind Speed (m/s) Frequency (%)
3. CASE STUDY 1 W 4.7 21
2 NNE 4.7 18.4
3.1 Description 3 NE 4.7 16.6
There were 16420 enterprises using liquid ammonia in 4 ES 4.7 14
China by the end of 2017 [43]. Several typical ammonia leak- 5 S 4.7 13
age accidents in China are presented in Table 1 [44]. In this 6 SW 34 4.5
article, an IR level classification in a natural gas process and 7 SSW 38 2.8
carbon dioxide recycle terminal (NGPCRT) processing facility 8 NNW 34 2
that contains an ammonia refrigeration system is introduced as
an example. The diagram of the ammonia refrigeration system
is shown in Figure 3.
Table 4. Number and probability of scenarios under different
3.2 Scenario Set selection criteria.
For the ammonia refrigeration system, 10 units were identi-
fied as ammonia containing. Considering the possibility of Frequency Scenario Number/ Cumulative
small, medium, large and catastrophic leakage, a total of Level Portion (%) Probability/Portion
40 leakage scenarios were confirmed. The leakage rate
10−3 46/14.3 0.0459/80.0
acquired from LEAK software is shown in Table 2.
10−4 82/25.6 0.0527/99.2
The wind field set is the combination of wind direction and
10−5 320/100 0.0532/100
speed. To achieve a balance between scenario integrity and
CFD computational costs, prior scenarios were chosen
Process Safety Progress (Vol.38, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2019 55
width, and 200 m in height. The layout and the gambit model
of the factory are shown in Figure 4.
Mesh quality has an important influence on the CFD solu-
tion convergence. An unstructured grid is used for the mesh
generation of the computational domain. The mesh quality is
controlled by using a much more refined mesh near the sur-
face and leakage source [37]. The meshed grid is shown in
Figure 5. To achieve a balance between the simulation accu-
racy and computing costs, an analysis of grid dependency was
carried out. In this study, four types of grids were constructed
by varying the ratio of the “size function” in GAMBIT, as
shown in Table 5. Figure 6 displays the ammonia concentra-
tion comparisons of the 15 points at a time of 120 s predicted
by these four grids applied at the potential dispersion area,
downstream of the release source. The leakage hole diameter
is 0.15 m and the mass flow rate is 3.0 kg/s. As seen from the
diagram, the predictions are grid independent. Then, a 1:1.6
mesh model was chosen for the simulation. A polyhedral mesh
was also used to reduce the grid cell number and enhance the
computation speed and convergence. The final grid number
was reduced from 6,354,181 to 3,514,487.
Another important step before solving the equations is to
set up monitor points. In order to assess the inhalation toxicity
Cell Number
Before Cell Number After
Grid Size Conversion to Conversion to
Type Function Polyhedra Polyhedra
Finer 1:1.4 6354181 3763974
Figure 4. Industry drawing: (a) CAD drawing and (b) Gambit Fine 1:1.6 5223643 3514487
model. [Color figure can be viewed at Medium 1:1.8 4846391 3423357
wileyonlinelibrary.com] Coarse 1:2.0 4690115 3401698
56 March 2019 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.38, No.1)
Figure 7. Location of the monitoring points. [Color figure can
be viewed at wileyonlinelibrary.com]
Figure 8. Examples of ammonia dispersion simulation results: (a) place of the selected monitoring points, (b) mole fraction of
ammonia, (c) ammonia concentration contours at 600 s, and (d) FR of the selected monitoring points. [Color figure can be viewed
at wileyonlinelibrary.com]
Process Safety Progress (Vol.38, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2019 57
IR contour derived by this method can provide scientific refer-
ences for emergency management and safety policy decision
of the NGPCRT facility.
58 March 2019 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.38, No.1)
Δti exposure time duration (2004), 147–59. https://doi.org/10.1016/j.jhazmat.
Pr probit variable 2004.01.011.
IRu(x, y) fatality ratio in the location (x, y) of accident scenario 16. A. Hameed, F. Khan, and S. Ahmed, A risk-based shut-
u down inspection and maintenance interval estimation con-
IRxy fatality ratio in the location (x, y) sidering human error, Process Saf Environ Prot 100 (2016),
n total number of accident scenarios 9–21. https://doi.org/10.1016/j.psep.2015.11.011.
fu frequency of accident scenario u 17. V. Villa, N. Paltrinieri, F. Khan and V. Cozzani, Towards
dynamic risk analysis: A review of the risk assessment
approach and its limitations in the chemical process industry,
LITERATURE CITED Saf Sci 89 (2016), 77–93. https://doi.org/10.1016/j.ssci.
1. U.S. Chemical Safety and Hazard Investigation 2016.06.002.
Board, MGPI Processing, Inc., Case Study, Report 18. M.A. Baksh, R. Abbassi, V. Garaniya and F. Khan, A net-
No. 2017-01-I-KS, CSB, Washington DC, 2017. work based approach to envisage potential accidents in
2. S. Yang, J. Kyeongwoo, K. Dongju and H. Chonghun, offshore process facilities, Process Saf Prog 36 (2016),
Accident analysis of the Gumi hydrogen fluoride gas leak 178–91. https://doi.org/10.1002/prs.11854.
using CFD and comparison with post-accidental environ- 19. I.L. Hirst and D.A. Carter, A “worst case” methodology for risk
mental impacts, J Loss Prev Process Ind 48 (2017), 207–15. assessment of major accident installations, Process Saf Prog 19
https://doi.org/10.1016/j.jlp.2017.05.001. (2000), 78–85. https://doi.org/10.1002/prs.680190206.
3. U.S. Chemical Safety and Hazard Investigation Board, 20. F.I. Khan and S.A. Abbasi, A criterion for developing
Millard Refrigerated Services, Inc., Case Study, Report credible accident scenarios for risk assessment, J Loss Prev
No. 2010-13-A-AL, CSB, Washington DC, 2015. Proc 15 (2002), 467–75. https://doi.org/10.1016/
4. X.Y. Zhang, J. Li, X.Q. Qin, Y.Z. Li, B. Li, N. Jia, Q. Miao, W. S0950-4230(02)00050-5.
L. Wang, S.L. Chen and Z.X. Wang, Analysis of poisoning 21. C. Delvosalle, C. Fievez, A. Pipart and B. Debray, ARAMIS
accidents and their causes in literature in China from 2002 to project: A comprehensive methodology for the identifica-
2012, J Ind Health Occup Dis 40 (2014), 338–41. https:// tion of reference accident scenarios in process industries, J
doi.org/10.13692/j.cnki.gywsyzyb.2014.05.006. Hazard Mater 130 (2006), 200–19. https://doi.org/10.
5. J.H. Lee, Y.D. Jo, and L.H. Kim, An approach to estimating 1016/j.jhazmat.2005.07.005.
the individual risk for toxic-gas releases using the 22. F. Zhang, G. Zhao, Z. Wang, J. Yuan and Y. Cheng, Worst
load-resistance model, J Loss Prev Process Ind 31 (2014), maximum credible accidental scenarios (WMCAS) - A new
88–96. https://doi.org/10.1016/j.jlp.2014.07.003. methodology to identify accident scenarios for risk assess-
6. V.R. Renjith and G. Madhu, Individual and societal risk ment, J Loss Prev Process Ind 48 (2017), 87–100. https://
analysis and mapping of human vulnerability to chemical doi.org/10.1016/j.jlp.2017.04.007.
accidents in the vicinity of an industrial area, Int J Appl 23. D.A. Carter and I.L. Hirst, ’Worst case’ methodology for
Eng Res (India) 1 (2010), 135–48. the initial assessment of societal risk from proposed major
7. F. Li, J. Bi, H. Lei, L. Huang, C. Qu, J. Yang and Q. Bu, accident installations, J Hazard Mater 71 (2000), 117–28.
Mapping human vulnerability to chemical accidents in the https://doi.org/10.1016/S0304-3894(99)00075-8.
vicinity of chemical industry parks, J Hazard Mater 179 24. F. Khan, Use maximum-credible accident scenarios for
(2010), 500–6. https://doi.org/10.1016/j.jhazmat. realistic and reliable risk assessment, Chem Eng Prog 11
2010.03.031. (2001), 56–64.
8. K. Han, Y.H. Kim, N. Jang and E.S. Yoon, Risk index 25. B. Zhang, L. Wang, and Z. Wang, Area risk level classifica-
approach for the optimal layout of chemical processes tion for hazardous gas release in petroleum refining instal-
minimizing risk to humans, Ind Eng Chem Res 52 (2013), lations, J China Uni Petro 39 (2015), 144–9. https://doi.
7274–81. https://doi.org/10.1021/ie3025104. org/10.3969/j.issn.1673-5005.2015.05.020.
9. W. Frank and D. Jones, Choosing appropriate quantitative 26. B. Zhang, C. Mu, Z.G. Wang and Y.F. Wang, An approach
safety risk criteria: Applications from the new CCPS guide- for dangerous gas detectors placement optimization consid-
lines, Process Saf Prog 29 (2010), 293–8. https://doi. ering multi-factors, Chem Ind Eng Prog 36 (2017), 2809–15.
org/10.1002/prs.10404. https://doi.org/10.16085/j.issn.1000-6613.2016-2279.
10. A. Dormohammadi, E. Zarei, M.B. Delkhosh and 27. K. Cen, T. Yao, Q. Wang and S. Xiong, A risk-based methodol-
A. Gholami, Risk analysis by means of a QRA approach on ogy for the optimal placement of hazardous gas detectors, Chin
a LPG cylinder filling installation, Process Saf Prog 33 J Chem Eng 48 (2017), 87–100. https://doi.org/10.1016/j.
(2014), 77–84. https://doi.org/10.1002/prs.11623. cjche.2017.10.031.
11. J.N. Shah and D.M. Shaffer, Risk-based approach for evalu- 28. S.W. Legg, C. Wang, A.J. Benavides-Serrano and C.D. Laird,
ating safety events in large plants, Proc Saf Prog 31 (2012), Optimal gas detector placement under uncertainty consider-
287–90. https://doi.org/10.1002/prs.11511. ing conditional-value-at-risk, J Loss Prev Process Ind 26
12. F. Khan, T. Husain, and S.A. Abbasi, Design and evalua- (2013), 410–17. https://doi.org/10.1016/j.jlp.2012.
tion of safety measures using a newly proposed methodol- 06.006.
ogy “SCAP”, J Loss Prev Process Ind 15 (2002a), 129–46. 29. DNV, PHAST Software Introduction [Online]. https://
https://doi.org/10.1016/S0950-4230(01)00026-2. www.dnvgl.com/services/hazard-analysis-phast-1675.
13. F. Khan, A. Iqbal, N. Ramesh, and S.A. Abbasi, SCAP: A 30. EPA, ALOHA Software [Online]. https://www.epa.gov/
new methodology for safety management based on feed- cameo/aloha-software (accessed May 12, 2016).
back from credible accident-probabilistic fault tree analysis 31. F.I. Khan and S.A. Abbasi, HAZDIG: A new software pack-
system, J Hazard Mater 87 (2001), 23–56. https://doi. age for assessing the risks of accidental release of toxic
org/10.1016/S0304-3894(01)00276-X. chemicals, J Loss Prev Process Ind 12 (1999), 167–81.
14. F. Khan, R. Sadiq, and T. Husain, Risk-based process safety https://doi.org/10.1016/S0950-4230(97)00043-0.
assessment and control measures design for offshore pro- 32. N. Pandya, N. Gabas, and E. Marsden, Sensitivity analysis
cess facilities, J Hazard Mater 94 (2002b), 1–36. https:// of Phast’s atmospheric dispersion model for three toxic
doi.org/10.1016/S0304389402000043. materials (nitric oxide, ammonia, chlorine), J Loss Prev
15. F.I. Khan and M.R. Haddara, Risk-based maintenance of Process Ind 25 (2012), 20–32. https://doi.org/10.1016/
ethylene oxide production facilities, J Hazard Mater 108 j.jlp.2011.06.015.
Process Safety Progress (Vol.38, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2019 59
33. N.J. Duijim, B. Carissimo, A. Mercer, C. Bartholomè, and Hazard Mater 197 (2011), 285–93. https://doi.org/10.
H. Giesbrecht, Development and test of an evaluation proto- 1016/j.jhazmat.2011.09.086.
col for heavy gas dispersion models, J Hazard Mater 56 41. QX/T 51, Meteorological Industry Standard of the People’s
(1997), 273–85. https://doi.org/10.1016/S0304-3894 Republic of China, Specifications for Surface Meteorologi-
(97)00069-1. cal Observation. Part 7: Measurement of Wind Direction
34. F.I. Khan and S.A. Abbasi, Modelling and control of the and Wind Speed, 2007.
dispersion of hazardous heavy gases, J Loss Prev Process 42. CCPS, Guidelines for Chemical Process Quantitative Risk
Ind 12 (1999), 235–44. https://doi.org/10.1016/ Analysis, 2nd ed., Wiley-AIChE, New York (1999).
S0950-4230(98)00009-6. 43. SAWS, State Administration of Work Safety Report. http://
35. B. Sun and K. Guo, LNG accident dynamic simulation: www.chinasafety.gov.cn/awhsy/awhdt/201803/
Application for hazardous consequence reduction, J Loss t20180306_213102.shtml.
Prev Process Ind 26 (2013), 1246–56. https://doi.org/10. 44. CCSA, China Chemical Safety Association. http://www.
1016/j.jlp.2013.06.005. chemicalsafety.org.cn/.
36. F. Khan, S. Rathnayaka, and S. Ahmed, Methods and 45. The Standard of People’s Republic of China,
models in process safety and risk management: Past, pre- GB10000-1988, Chinese Adults Body Size, China Institute
sent and future, Process Saf Environ Prot 98 (2015), of Standardization and Information Coding, Beijing (1988).
116–47. https://doi.org/10.1016/j.psep.2015.07.005. 46. The Standard of People’s Republic of China,
37. B. Zhang and G.M. Chen, Quantitative risk analysis of toxic gas SY/T6610-2005, Recommended Practice for Oil and Gas
release caused poisoning—A CFD and dose–response model Well Servicing and Workover Operations Involving Hydro-
combined approach, Process Saf Environ Prot 88 (2010), gen Sulfide, National Development and Reform Commis-
253–62. https://doi.org/10.1016/j.psep.2010.03.003. sion, Beijing (2005).
38. L. Dong, H. Zuo, L. Hu, B. Yang, L. Li, and L. Wu, 47. HSE, Reducing Risks, Protecting People, UK Health and
Simulation of heavy gas dispersion in a large indoor space Safety Executive, HSE Books, Sudbury, UK (2001).
using CFD model, J Loss Prev Process Ind 46 (2017), 1–12. 48. SAWS, Interim Provisions on Supervision and Administra-
https://doi.org/10.1016/j.jlp.2017.01.012. tion of Major Hazard Sources of Hazardous Chemicals.
39. S.W. Legg, A.J. Benavides-Serrano, J.D. Siirola, J.P. Watson, http://old.chinasafety.gov.cn/fgzc/gz/201508/t2015
S.G. Davis, A. Bratteteig and C.D. Laird, A stochastic 0827_174230.shtml.
programming approach for gas detector placement using 49. HSE, The Tolerability of Risk from Nuclear Power Stations,
CFD-based dispersion simulations, Comput Chem Eng 47 UK Health and Safety Executive, HMSO, London, UK
(2012), 194–201. https://doi.org/10.1016/j.compchem (1988) Revised 1992.
eng.2012.05.010. 50. P. Xin, F. Khan, and S. Ahmed, Dynamic hazard identifica-
40. F. Scargiali, F. Grisafi, A. Busciglio, and A. Brucato, tion and scenario mapping using bayesian network, Pro-
Modeling and simulation of dense cloud dispersion in cess Saf Environ Prot 105 (2017), 143–55. https://doi.
urban areas by means of computational fluid dynamics, J org/10.1016/j.psep.2016.11.003.
60 March 2019 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.38, No.1)