You are on page 1of 10

Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries


journal homepage: http://www.elsevier.com/locate/jlp

Incident analysis of the “Pajaritos” petrochemical complex


Henry M. Aquino-Gaspar a, Christian O. Díaz-Ovalle b, **, Antioco López-Molina a, *,
Carolina Conde-Mejía a, Luis M. Valenzuela-Gómez a
a
Departamento de Ingeniería Petroquímica, Universidad Juárez Autónoma de Tabasco, Carr. Libre Villahermosa –Comalcalco Km 27, Jalpa de Méndez, Tab., 86205,
Mexico
b
Departamento de Ingenierías, Tecnológico Nacional de México/I. T. Roque, Km 8.0 Carretera Celaya-Juventino Rosas, Gto, 38110, Mexico

A R T I C L E I N F O A B S T R A C T

Keywords: On April 20th, 2016, an explosion occurred in the “Pajaritos” petrochemical plant, in Coatzacoalcos, México. The
Polyvinyl chloride accident caused 32 deaths and almost 100 workers were seriously injured. At present, there is no official in­
Process stream index formation about the cause of the accident and the existent reports are confidential. This paper proposes an
CFD
explanation of what happened at the petrochemical plant facilities by considering the testimony of several
ASPEN PLUS ®
witnesses and the process simulation. The strategy includes three steps: 1) the process simulation via Aspen
Analysis of consequences
Plus® software; 2) The hazardous process streams identification through the process stream index (PSI); and 3)
The consequences description using the TNT method, ALOHA® and ANSYS FLUENT. The results agreed with the
testimony of witnesses and indicated the level of environmental damage and the unavailable prioritization of
process safety over the production and management systems. This indicates that these accidents are recurrent as
a consequence of a low safety culture in developing countries.

1. Introduction • On April 28th of 1973, in Jouetsu, Niigata, Japan, a production plant


of PVC suffered a VCM leak that yielded an explosion with partial
In the last decades, the high demand of petrochemical products has destruction of the chemical plant. The VCM leak was located in the
forced the process facilities to expand (Hassani et al., 2017). Unfortu­ feed stream in the refining tower and came from a storage tank. The
nately, these significant changes in the industry have increased the source of the problem was the sudden opening of a control valve due
process risk and have caused industrial accidents, which have released to maintenance service by corrosion. The released gas reached
many chemicals into the environment affecting the population (He et al., enough temperature to explode and damage 2.2 km around. The
2011). The polymer industry is one example, where the production of incident investigations found a release of around 4 t of VCM (Koseki
polyvinyl c and Tamura, 1973).
hloride (PVC) was 44.4 Mt in 2018 (Business Wire, 2018). This • On April 23rd, 2004, in Illipolis, IL, U.S., a production plant of PVC,
process requires hazardous raw materials and extreme operating con­ owned by Formosa Plastics Corp., exploded and caused the death of
ditions (Gaur and Wunderlich, 1983; Wang et al., 2009). Furthermore, five workers and considerable damages. The U.S. Chemical Safety
the process plants in developing countries disregard the safety measures Board (CSB) investigated the incident. It concluded that: a) causes:
to achieve highly reliable production (de Souza Porto and de Freitas, human error, wrong interpretation of the safety and control systems,
1996). This problem can be observed in the poor safety culture of and b) recommendations: implementing an integrated control sys­
workers, which added to the extreme conditions in the production tem, restrictive access to control system devices, and more infor­
processes generates conditions of high risk catastrophic events. mation dissemination about VCM and its hazards (CSB, 2007).
The PVC process has experimented serious accidents. Two of the
most transcendental events are: In addition, PEMEX suffered an accident in the “Clorados III” plant of
the “Pajaritos” petrochemical complex in Coatzacoalcos, Veracruz,
México (N:18◦ 06′ 47′′ , W:94◦ 23′ 17.5′′ ). This case is analyzed here to

* Corresponding author.
** Corresponding author.
E-mail addresses: christian.do@roque.tecnm.mx (C.O. Díaz-Ovalle), antioco.lopez@ujat.mx (A. López-Molina).

https://doi.org/10.1016/j.jlp.2021.104404
Received 28 October 2020; Received in revised form 19 December 2020; Accepted 18 January 2021
Available online 30 January 2021
0950-4230/© 2021 Elsevier Ltd. All rights reserved.
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

explain the main concerns about the catastrophe, which is discussed in Table 1
the next sections. Graphic information sources on the consequences of the incident.
Description Internet route
2. Production of PVC
Exact location of the event, depiction of https://www.youtube.com/watch?v%20
material damages, data about the =%20WE_DDDVgoJM&t%20=%20351s
PVC comes from the systematic transformation of ethylene. This layout and size of the plant.
process starts in the production of 1,2-dichloroethane (EDC) from Dimension of the smoke cloud https://i.ytimg.com/vi/7H8OLg5iczY/
ethylene, C2H4, through two possible paths (Rossberg et al., 2006): a) generated by the blast maxresdefault.jpg
Visible material damage after the https://i2.wp.com/lopezdoriga.com/wp
direct chlorination: injection of pure Cl2, and b) oxy-chlorination: use of explosion and detection of the -content/uploads/2016/04/Explosi%
fresh HCl with recycling stream. Later, EDC suffers an exothermal possible release point. C3%B3n-Coatzacoalcos.jpg?resize
decomposition reaction to yield vinyl chloride monomer (VCM), which =960%2C480&ssl=1
is characterized by: a) high flammability, b) heavy gas behavior, and c) Damage to the reactors and heat https://xalapaenlanoticia.files.wordp
exchangers areas. ress.com/2016/04/adolfoooo_jasso
heat of combustion similar to butane (Guido et al., 2016). The final step
_ajm_04541.jpg?w=547
is the polymerization of VCM in order to obtain PVC. Herein, the anal­
ysis focuses on the steps before polymerization, as Fig. 1 depicts.
inhabitants to avoid outdoor activities. Through a statement, PEMEX
3. Incident description reported the progress of control over the accident. At 6:20 p.m. that day,
the fire was controlled.
The “Clorados III” plant of the “Pajaritos” petrochemical complex
was designed to yield 900 t/day of PVC (Reuters Staff, 2016). The 4. Identification of hazards
company Mexichem S.A.B. de C.V. managed the plant in association
with PEMEX. This chemical plant operated both pathways of EDC pro­ Process streams are high-risk elements in a chemical plant. The
duction. On April 20th, 2016, at 15:45 a massive explosion took place. hazard identification technique determines their level of risk. A suitable
During the blast 32 workers were killed and around one hundred people strategy is the process stream index (PSI), which demands a succinct
were injured. Previously, the plant had several accidents, but the safety description of the streams through the following parameters: a) density,
measures were not modified. At the time of the incident, welding works b) pressure, c) thermal energy, and d) combustibility. The last parameter
were being carried out to install equipment. The catastrophic event was refers to the difference between the lower (LFL) and upper (UFL)
captured by eyewitnesses, who provided photographs via internet; some flammability limits of the mixture. High combustibility values indicate
of them are in Table 1. The chronological description of the event can be high explosiveness in the stream (Heikkilä, 1999). Finally, the PSI re­
summarized as follows: lates all parameters to obtain a hazard score where high values reflect
the severity of a process stream (Shariff et al., 2012). However, to obtain
• 08:00: The leak alarm sounded in the feed stream to R3 and the the stream parameters is a hard task and demands the use of a process
safety personnel applied the evacuation plan to the personnel. simulation software. In this work, this issue applied ASPEN PLUS ®
• 11:00: The leak alarm sounded for the second time in the same area. v10.0 into the PVC production process.
The overall works were stopped, and the personnel was evacuated
until 14:00. 4.1. Simulation process of PVC
• 15:00: The leak alarm and the flame alarm were activated in the
same area. The safety personnel verified the alarms and later on This simulation included the two ethylene reaction options (direct
authorized the overall works. The safety staff said the situation was chlorination and oxy-chlorination). Fig. 2 shows the complete process
under control without explosion risk. described in ASPEN PLUS ® v10.0, in which consecutive numbers
• 15:15: An explosion occurred in the same area and released EDC and indicate the process sections. The main items are reaction and refining
VCM. The blast produced an expansive wave, which destroyed the systems. The operating conditions for each process unit come from the
plant and severely damaged different areas 6 km away. literature. The exact process condition at the time of the disaster is un­
known, except for 900 t of PVC’s daily production. On the other hand,
the CPROB stream presented the leak in the boiler area. It came from the
3.1. Emergency response exothermal decomposition reactor (Noticias, 2016; Sanchez, 2016),
indicated by R3 in Fig. 1 and B19 in Fig. 2.
Internal safety protocols were applied after the explosion, which
included the closing of ducts and valves in the plant and the total 4.1.1. Reaction systems
evacuation of personnel. Additionally, the alarm was extended to The simulation of reaction systems depends on the information of the
neighboring plants and the oil extraction zone and the Coatzacoalcos 1 reaction kinetics. This information has been documented in most cases,
bridge was closed. The visible toxic cloud and corrosive dust forced the but its absence suggests the use of final conversions, Xreactive. For the

Fig. 1. Section of the production process for PVC: R1. Direct chlorination reactor, R2. Oxy-chlorination reactor, S1. Refining of EDC, R3. Exothermal decomposition
reactor and S2. Refining of VCM.

2
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Fig. 2. Flow diagram for the PVC process production in ASPEN PLUS ® in three sections.

simulation of the process, reactions of both types were used. Table 2 by combustion (Ranzi et al., 1992); this reactor consists of two sections,
contains the chemical kinetic data for reactors R1 and R3. Table 3 preheating and radiation. Therefore, this reactor was simulated as two
contains the reaction conversion data for reactors R2 and R3. reactors, PFR, and RStoic, which are in Section 2 of Fig. 2 with the labels
The direct chlorination reactor, R1, was simulated as a plug flow B13 and B19, respectively. Table 4 contains the design features and
reactor (PFR), which contains the reactants, and the conditions of 90 ◦ C operating conditions of all the reactors. Additionally, the reaction sys­
and 1.5 atm. This is the DCRT item in Section 1 of Fig. 2. The oxy- tem requires a fast cooling of the product. Flash tanks with heat ex­
chlorination reactor, R2, was considered to be a fluidized bed (Mor­ changers are included before the refining section. These equipments are
eira and Pires, 2010; Wachi and Morikawa, 1986). However, the items B22, B24, B25, and B28 from Section 3 of Fig. 2.
particular lack of knowledge of the operating conditions suggested a
stoichiometric reactor (RStoic). This corresponds to item B11 in Section 4.1.2. Refining systems
1 of Fig. 2. The exothermal decomposition reactor, R3, requires energy Four distillation towers are involved in this process (Rossberg et al.,

3
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Table 2 4.2. Application of the PSI


Chemical kinetic data.
Reactor Reactions Reaction rate Value of k Reference Quantification of inherently safe design is recommended to prevent
expression the risk of accidents and learn from the occurred events. The PSI de­
R1 C2H4 + Cl2→ r1 = k1[C2H4] 0.132 m3 [9] scribes the level of risk in a process plant by taking into account the flow
C2H4Cl2 [Cl2] mol− 1 s− 1 streams, whereby the interaction of the components and all the flow
C2H4 + 2Cl2→ r2 = k2[C2H4] 0.0239 m6 parameters make a precise quantification (Leong and Shariff, 2009;
C2H3Cl3 + HCl [Cl2]2 mol− 2 Shariff et al., 2012). Through this method, each stream, j, takes part into
R3 C2H4Cl2→ HCl + r7 = k1 0.0143 s− 1 [12]
C2H3Cl [C2H4Cl2]
the analysis by considering its parameters, ϕj, and the average value of
C2H3Cl → HCl + r8 = k2[C2H3Cl] 2.39 × 10− 5 the parameters for all streams, Φ. Normalization provides a represen­
C2H2 s− 1 tative index for each parameter in the stream j, Iϕ,j, such as Iϕ,j = ϕj/Φ.
5
C2H4Cl2→Cl2 + C2H4 r9 = k3 7.62 × 10− Hence, the PSI value for each stream j is computed by the product of all
[C2H4Cl2] s− 1
the stream parameters: PSIj = Iρ,j⋅ I.P.,j⋅ I.E.,j⋅IFL,j. A high PSI value in­
dicates a higher risk stream in the process.

Table 3 5. Incident analysis


Reaction conversion data.
Reactor Rections Conversion Reference The case has a lack of information because there is no official report
R2 r3: C2H4 + 2HCl +0.5O2→ C2H4Cl2+ H2O XHCl = 0.96 [11] of the event. Thus, the only available source of information is the public
r4: C2H4 + 3O2→ 2CO2+ 2H2O XC2H4 = 0.08 description of the event by eyewitnesses. The inhabitants of Coat­
r5: 2HCl + 0.5O2→Cl2+ H2O XHCl = 0.03 zacoalcos City experienced an overpressure wave. The incident pro­
r6: C2H4Cl2 +Cl2 → C2H3Cl3 + HCl XEDC = 0.09 duced an explosion that fractured glass windows in the center of
R3 r10: 2C2H4Cl2→ Cl2 + C4H8Cl2 XEDC = 0.3 [12]
Coatzacoalcos, 4500 m away from the plant. This description is equiv­
r11: C2H3Cl + Cl2→C2H3Cl3 XCl2 = 0.5
r12: C2H4Cl2 + Cl2→C2H3Cl3 + HCl XCl2 = 0.2 alent to an overpressure between 0.7 and 1 kPa (Clancey, 1972).
r13: C2H2 + 2Cl2→C2HCl3 + HCl XCl2 = 0.3 To describe the possible scenarios at the time of the accident in an
r14: C2H3Cl + C2H2→C4H5Cl XC2H2 = 0.5 acceptable way, this analysis was carried out following the chronolog­
ical events, which were treated with various simulation tools such as:

Table 4 1. Leak from the R3 outlet stream: this event comes from the CPROB
Design features and operating conditions of the reactors involved. stream data, as indicated in the previous section.
Parameter DCRT B11 B13 B19 2. Toxic and explosive cloud: the development of gas dispersion re­
quires suitable computational packages. ALOHA ® is a relevant
Temperature (◦ C) 90 305 530 530
Pressure (atm) 1.5 5 18 18 computation software used to describe industrial incidents. During
Volume (m3) 94.24 297.51 141.37 156.79 dispersion, physical obstacles and barriers modify the dynamic
Length (m) 30 – 45 – behavior of the cloud, which is not considered by ALOHA ® despite
its acceptability. Therefore, the CFD (computational fluid dynamics)
tool is recommended to complement the results of ALOHA ®. In this
2006): a) refining of EDC previous to R3 (B5 and B20 from Section 2 in
analysis, ANSYS ® FLUENT 2020 R2 solves the case in a 3D scenario.
Fig. 2) and b) refining of VCM from impurities after R3 (B26 and B33
3. Explosion event: this phenomenon demands proper methods to
from Section 3 in Fig. 2). In the simulation, the Soave-Redlich equation
consider the physical characteristics: turbulence, heat dissipation,
of state provided the phase equilibrium in polar solutions, which is valid
volume expansion, etc. The TNT method represents a well qualified
for these components and operating conditions. Table 5 contains the
short-method to obtain information from an explosion. Also, ALOHA
features of each separator column and all columns processed with a
® contains an appropriate procedure to obtain the effect of this event
pressure drop of 0.5 atm.
based on an explosive cloud scenario. Both cases have been experi­
mentally validated, but they do not consider the physical effect that
4.1.3. Simulation results
produces the obstacles and barriers through the explosion. There­
ASPEN PLUS provided the main parameters of the streams: mass
fore, the CFD tool includes 3D behavior and the time-variability of
density, pressure, thermal energy and ΔFL, difference of UFLmix and
the expansive wave.
LFLmix, which are the upper and lower flammability limit in a mixture
respectively. The last parameters are related to an ideal gas mixture
The next sections thoroughly explain the use of the proposed
through FLmix=(Σi yi/FLi)− 1, where FLmix is either UFLmix or LFLmix, y is
computation softwares for this analysis. For each event, the results ob­
the mole fraction and i is the component index (Crowl and Louvar,
tained from each tool are compared in order to complement the analysis.
2019). Table 6 contains the results for the main streams analysis. Table 7
contains the mass fraction of the CPROB stream, where the EDC (wt%) is
5.1. Use of ALOHA ® software
the highest value and indicates that CPROB is a high-risk stream.

This software is acceptable to analyze industrial events. It provided a


global visualization of the level of affectation by the explosion and
dispersion of the chemical cloud. ALOHA ® required specific data on
Table 5 meteorology and storage conditions to represent the environment and
Features of the separation tower for the PVC process in Fig. 2. release scenarios succinctly. In detail, the CPROB stream was indicated
Feature B5 B20 B26 B33 as VCM (CAS 75-1-4) at ground level with a sudden release of 50 t, and
Number of trays 9 32 41 16 the meteorological data are: a) wind speed 7.72 m/s from the south at
Top pressure (atm) 13 6 10 5 10 m above ground, b) stability class D without thermal inversion, c)
Top temperature (◦ C) 33 152.09 5.05 47.88 partly cloudy with a relative humidity of 75% at 32 ◦ C and d) urban
Bottom temperature (◦ C) 195.56 162.50 128.91 149.8 roughness. The overpressure simulation followed an ignition by
Reflux ratio 3 1.757 3.1 0.6
detonation.

4
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Table 6
Results for the main streams from Fig. 2.
Stream Density (kg/ Pressure Thermal energy (Mcal/ ΔFL (% Stream Density (kg/ Pressure Thermal energy (Mcal/ ΔFL (%
m3) (kPa) s) vol) m3) (kPa) s) vol)

DCERECI 888.5 506.63 16.226 8.331 Q 836.5 1317.23 24.897 9.778


DCERECI2 890.5 1367.88 16.211 8.331 S 863.3 1317.23 41.108 9.237
D 1044.0 1013.25 13.714 11.151 T 874.7 607.95 17.843 8.178
H 844.5 6313.63 39.89 3.792 U 927.0 607.95 24.154 9.976
I 997.4 1013.25 31.921 1.66 V 930.2 1925.18 24.121 9.976
K 1038.5 1013.25 16.039 17.592 AC 873.4 1519.88 27.341 9.525
M 1041.2 1013.25 29.753 9.587 AF 873.4 1519.88 13.67 9.525
N 1043.9 1013.25 28.96 10.886 AH 995.1 1013.25 22.505 14.801
P 981.2 1519.88 27.882 10.886 AI 865.6 1215.9 15.951 12.833
AD 873.4 1519.88 13.67 9.525 CPROB 21.8 1823.85 8.193 9.209

(1.25 m), and c) Zone 3: surrounding area (5 m). The tetrahedral ele­
Table 7 ments formed the mesh and reached the limit on an academic license,
CPROB stream (wt%).
whereby a test of mesh was omitted. The total number of elements was
Component wt% Component wt% 461,802 with an average skewness of 0.289, which is a mesh quality
C2H4Cl2 (EDC) 52.053 C2H3Cl3 6.1E-3 parameter where the value close to zero indicates a good mesh quality.
C2H3Cl (VCM) 16.381 C2HCl3 5.88E-3 Thus, the number of elements and the skewness value per each zone are:
C4H8Cl2 14.316 C2H2 5.92E-4 a) Zone 1: 323,380–0.2638, b) Zone 2: 44,160–0.5273, and c) Zone 3:
HCl 9.227 C2H4 3.48E-4
Cl2 7.992 CO2 2.32E-5
94,262–0.2623.
C4H5Cl 0.017 O2 7.14E-10 In this simulation, the meteorological data from Section 3.1 was
taken into account. The wind speed profile followed the model: v = 7.72
m/s⋅(z/10 m)0.38, which describes a stability class D with z as the ver­
5.2. Application of TNT method tical position (Mannan, 2012). The scenario was under buoyancy effects,
and the gases followed the ideal gas equation. The leak corresponded to
This method calculates overpressure at distances greater than 5 km, CPROB stream and was 10.6 kg/s at the Release boundary. ASPEN PLUS
in which the shockwave is equal in all directions (TNO, 2005). The provided the stream properties, such as: a) density: 21.772 kg/m3, b)
calculation considers the equivalent mass of the explosive component heat capacity: 2.93 kcal/kg⋅K, c) viscosity: 2.6E-5 Pa s, d) thermal
with regard to the TNT mass, which is related as mTNT = η⋅mC⋅HC/ETNT, conductivity: 0.04882 W/m⋅K, e) molecular weight: 78 kg/kg mol y f)
where ETNT is the energy of explosion of TNT (4686 kJ/kg), η is the diffusivity: 0.02 m2/s. On the other hand, the overpressure analysis
empirical explosion efficiency (~10%), H.C. is the energy of explosion of consisted in the description in Section 3.2. Here, the simulation followed
the explosive component, m is the mass, and the subscript C corresponds the strategy proposed by Díaz-Ovalle et al. (2016), where an arbitrary
to the component. The mTNT value relates the desired affectation dis­ spherical zone (radius of 5 m) contained the overpressure energy at
tance to the explosion, r, as: Ze = r/(mTNT)1/3, where Ze is a parameter to 1500 k and 160 atm for an ideal gas. In addition, the model SST κ-ω
calculate the overpressure using equation (1). represented the turbulence phenomenon. This model is suitable for a

( ( )2 )[( ( )2 )( ( )2 )( ( )2 )]− 0.5


Po Ze Ze Ze Ze
= 1616 1 + 1+ 1+ 1+ (1)
Pa 4.5 0.048 0.32 1.35

scenario with obstacles and different spatial scales (Sklavounos and


where Pa is the ambient pressure and Po is the peak side-on overpressure Rigas, 2004).
at the r distance. In this case, the total released mass was 50t from the The numerical solution was possible with the SIMPLE algorithm. The
CPROB stream, and Hc was obtained by the molar average combustion scenario was transitional with time steps of a) 1 s for dispersion and b)
energy of the mixture 11,754.38 kJ/kg. The TNT equivalent mass was 0.0001 s for overpressure. A numerical test was developed to determine
2548.75 kg. the number of iterations per time step, where the tested values were 10,
20, 40, and 160. The numerical test collected the values of mass fraction
of the mixture after obstructions, which corresponded from 15 m after
5.3. Analysis via CFD the leak in northerly direction. Fig. 4 shows the trial results and indicates
a suggested value of 40 iterations per time step. On the other hand, the
CFD is a valuable tool for process safety. Shen et al. (2020) described scenarios of dispersion and overpressure were independently solved.
a set of published works under CFDs applied to fire, dispersion, explo­ The solution was possible in a computer with 24 GB RAM and an Intel-i7
sion, accident investigation, etc. Herein, the ANSYS FLUENT 2020 R2 ® 3.9 GHz processor, which required 6 h to simulate 120 s for dispersion
software provided the dynamic behavior of the event by considering the and 13 h to simulate 3 s for overpressure.
dispersion and overpressure.
The event scenario was presented using approximated geometry, and 6. Analysis of results
details were avoided to reduce computational efforts. Furthermore, the
computational domain presented different spatial scales, and a sectional The ASPEN PLUS ® software provided the values of the parameters
mesh treatment was adopted. Fig. 3 shows the computational domain in for ρ, P, E and ΔFL for all the process streams: 885.197 kg/m3, 1425.12
three zones. The mesh zones and the average length of the elements per kPa, 22.7 Mcal/s and 9.74 %vol, respectively. These average values
each zone are: a) Zone 1: process area (0.6 m), b) Zone 2: piping systems

5
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Fig. 3. Computational domain for the scenario with boundary conditions.

were obtained by PSI method. Table 8 contains all the PSI values for the testimonials, the ALOHA® results overpredicted overpressure. Fig. 6 b)
main streams, where the H stream presented the highest value in schematizes the CFD results contrasted with the results of the TNT
contrast with the CPROB stream. The PSI method indicates that a stream method and ALOHA® at an overpressure of 100 kPa. It can be observed
with a low value has low risk. Nevertheless, CPROM stream is a highly that the times and distances at which this overpressure value is reached
flammable mixture, but the low PSI value is due to the density value vary with the method used; notice that the time predicted by the TNT
(gas). This suggests that the PSI method considers liquid streams to be method is shorter than that of ALOHA®, 0.8s 1.09s, respectively. CFD
more dangerous, since the gas streams might produce either explosive or provided iso-surfaces of 100 kPa at different times in comparison with
toxic clouds. results by TNT method and ALOHA ®. These iso-surfaces do not describe
The leak from the R3 outlet stream yielded a yellowish cloud (Acp, a perfect sphere, which is a consequence of obstructions in the process
2016; Sanchez, 2016). This corresponds to chlorine emission, which area. Additionally, Fig. 6 c) presents a temperature profile on the iso-
coincides with the mixture obtained in the CPROB stream by ASPEN surface of 100 kPa at 0.1 s, where the central area of the structure
PLUS ®; see Table 7 for mass fraction values. The results of the disper­ received the highest temperature. This is approximate to the combustion
sion determined the greatest risk areas for an explosion. ALOHA ® temperature as shown by the images in the area.
software calculated a wide vulnerable zone, as shown in Fig. 5 a). Fig. 5
b) shows the CFD results in comparison with the section obtained by 7. Health and environment risk
ALOHA ® at 25 ppm. This indicates that CFD under-predicted the con­
centration values but considered the equipment obstruction, which is an VCM is a carcinogenic agent that has been listed by the U.S. Envi­
important factor. In both cases, the cloud covers a considered length in ronmental Protection Agency (EPA) with an ERPG-3 value of 20,000
response to the wind direction. This increased the risk of an explosion ppm (Becker et al., 2001). This indicates that long exposures can cause
due to invading a neighboring plant without announcement. liver cancer and genetic alterations. Short exposures can cause dizziness,
The explosion caused glass fracture in the center of Coatzacoalcos, vomiting, and suffocation. VCM is an unstable compound when inter­
Veracruz, 4500 m away from the plant. At that distance, the TNT acting with the environment due to the degradation of chlorinated
method predicted an overpressure of 0.86 kPa, which is consistent with compounds with volatile organic compounds. On the other hand, EDC is
that reported by witnesses. Fig. 6 a) depicts the obtained overpressure an existing compound in the manufacture of PVC. This agent has a
scenario, where the ALOHA® results located the same pressures ob­ similar effect to VCM under an ERPG-3 value of 40,000 ppm (Hughes
tained by the TNT method at a greater distance. Thus, based on et al., 1994). In the environment, EDC affects the following: a) acid rain
due to chlorine free radicals, b) atmospheric pollution due to its pro­
longed permanence (up to 5 months before degradation), c) immediate
absorption into the soil, and d) filtration to groundwater.
The results obtained by ALOHA ® and CFD coincided in the spatial
location of the dispersion, where a point with 10,000 ppm was 200 m
from the release point. This confirms an imminent risk for VCM. The
exposed area is large and neighboring companies are affected. The
process area contained the highest concentration. Fig. 5 shows the ex­
istence of an open area with environmental risk.

8. Discussion

The witnesses described the continuous occurrence of leaks in the R3


area, where a leak began on the day of the tragedy. Despite this, the
maintenance staff made no immediate changes, but postponed it for
review in scheduled preventive maintenance. This worsened the
Fig. 4. Results of the numerical test for the number of iterations per time step.

6
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Table 8
PSI values for the stream of the VCM process (sorted from the highest to lowest).
Corriente Iρ IP IE IFL PSI Corriente Iρ IP IE IFL PSI

H 0.954 4.33 1.757 0.389 2.826 AH 0.978 0.695 0.703 1.52 0.726
P 1.108 1.042 1.228 1.118 1.586 AD 0.987 1.042 0.602 0.978 0.606
S 0.975 0.903 1.811 0.948 1.513 AF 0.987 1.042 0.602 0.978 0.606
V 1.051 1.32 1.063 1.024 1.51 DCERECI2 1.006 0.938 0.714 0.855 0.576
AI 1.124 0.834 0.991 1.318 1.225 D 1.179 0.695 0.604 1.145 0.567
AC 0.987 1.042 1.204 0.978 1.211 U 1.047 0.417 1.064 1.024 0.476
N 1.179 0.695 1.276 1.118 1.168 T 0.988 0.417 0.786 0.84 0.272
M 1.176 0.695 1.311 0.984 1.054 DCERECI 1.004 0.347 0.715 0.855 0.213
K 1.173 0.695 0.706 1.806 1.04 I 1.127 0.695 1.406 0.17 0.188
Q 0.945 0.903 1.097 1.004 0.94 CPROB 0.025 1.251 0.361 0.946 0.01

Fig. 5. Results of the dispersion: a) ALOHA ® and b) CFD: iso-surface at 25 ppm.

physical characteristics of the leak. Therefore, this area was considered a and the analysis of the accident:
high-risk section within the plant.
With the herein tools proposed, the dispersion and overpressure 1. Process safety must prevail over production in plants with a high risk
analysis coincides with the information from the witnesses. In addition, of accidents.
a sample carried out two days after the incident detected the presence of 2. Identifying the process streams with the highest level of risk is
VCM in the area (Greenpeace, 2016). This was confirmed by the ASPEN essential in a process plant.
PLUS® simulation results, which provided relevant information on the 3. The safety strategy in maintaining equipment (welding, dismantling,
process conditions and steered the PSI index analysis and the simula­ construction, etc.) must be rigorous in plants with a high-risk level.
tions of the consequences for dispersion and overpressure. Despite the 4. Safety culture should be promoted within the personnel to respond
danger of VCM, the PSI index indicated that the released mixture pre­ professionally to the first alarm.
sented a low level of risk, which was due to the low value of the density. 5. Maintenance of process plants under imminent risk of an accident
On the other hand, the consequence analysis acceptably described the should not be delayed.
dispersion and overpressure conditions. This indicated the importance 6. Society loses confidence in the process industry with this type of
of diversifying simulation software (ASPEN PLUS®, ALOHA® and accidents, despite being located in exclusively industrial areas.
FLUENT®) to describe the circumstances of an accident accurately. 7. Toxic gas mitigation systems must exist in plants with chemical
components at risk of harm to the environment and personnel.
9. Lessons learned from the incident 8. Neighboring process plants should be aware of any emergency or
potential accident at a neighboring plant.
This analysis provides important aspects to improve process safety, 9. Preventive maintenance of high-risk streams must be rigorous and
which is useful to prevent accidents, improving inherent safety and continuous.
understanding the situation of process plants similar to the case study.
The following learning points are obtained based on the herein results

7
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Fig. 6. Results of overpressure: a) comparison between ALOHA ® and TNT method (dotted lines), b) CFD: iso-surfaces of 100 kPa at 0.8 s and 1.09 s, and a detailed
view of the temperature profile over an iso-surface of 100 kPa at 0.1 s.

10. Conclusion and administrative processes. Suppose this philosophy is adopted and
the process plants are in an open atmosphere. In that case, these pro­
This work analyzed the accident at the “Pajaritos” petrochemical cesses that handle toxic and corrosive substances can be operated safely.
complex, in Coatzacoalcos, Mexico, which caused U.S. $ 244 million in
property damage. To date government authorities have not indicated the Petrochemical complex
environmental damage. They have reported 32 human losses and more
than 100 injuries. The analysis applied strategies to identify the most All persons who meet authorship criteria are listed as authors, and all
dangerous process flows and the consequences of the accident. Process authors certify that they have participated sufficiently in the work to
simulation and the PSI index defined the most dangerous streams. The take public responsibility for the content, including participation in the
ALOHA ® package, the TNT method, and ANSYS FLUENT software concept, design, analysis, writing, or revision of the manuscript.
described the emission and explosion consequences. The numerical re­ Furthermore, each author certifies that this material or similar material
sults coincided with the testimony of witnesses to the catastrophe. The has not been and will not be submitted to or published in any other
application of these methodologies described in detail the starting point publication before its appearance in the Journal of Loss Prevention in the
and the consequences of the accident. Unfortunately, the responsible Process Industries (ISSN: 0950–4230).
company and the government have not provided a similar analysis.
Therefore, this work provided sufficient elements to understand the Authorship contributions
tragedy and learn from its causes and consequences. It is important to
point out that this type of accidents can be avoided if the directive Conception and design of study: A. López-Molina, C. O. Díaz-Ovalle,
boards consider the safety of the process a priority over the production C. Conde-Mejia, L. M. Valenzuela-Gómez; Acquisition of data: H. M.

8
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Aquino-Gaspar, L. M. Valenzuela-Gómez, C. Conde Mejía, A. López- TNT trinitrotoluene


Molina, C. O. Díaz-Ovalle; analysis and/or interpretation of data: H. M. ρ density
Aquino-Gaspar, A. López-Molina, C. O. Díaz-Ovalle, C. Conde-Mejía. ϕ PSI parameters index
Drafting the manuscript: H. M. Aquino-Gaspar, A. López-Molina, C. O.
Díaz-Ovalle, C. Conde-Mejía; revising the manuscript critically for References
important intellectual content: A. López-Molina, C. O. Díaz-Ovalle, C.
Conde-Mejía, L. M. Valenzuela-Gómez. Approval of the version of the Acp, V., 2016. Por sobrecalentamiento, desde la mañana ya temían la explosión; revela
obrero - YouTube. https://www.youtube.com/watch?v=xlQRVyvgskc. (Accessed 20
manuscript to be published (the names of all authors must be listed): September 2018).
Becker, R., Nikolova, T., Wolff, I., Lovell, D., Hüttner, E., Foth, H., 2001. Frequency of
Acronyms HPRT mutants in humans exposed to vinyl chloride via an environmental accident.
Mutat. Res. Genet. Toxicol. Environ. Mutagen 494 (1), 87–96. https://doi.org/
10.1016/S1383-5718(01)00182-6.
CFD Computational Fluid Dynamics Clancey, V.J., 1972. Diagnostic Features of Explosion Damage. Sixth International
EDC 1,2-dichloroethane Meeting of Foresic Sciences, Edinburgh, 1972.
Crowl, D.A., Louvar, J.F., 2019. Chemical Process Safety: Fundamentals with
EPA Environment Protection Agency Applications, fourth ed. Pearson Education.
LFL low flammability limit CSB, 2007. Investigation Report, Vinyl Chloride Monomer Explosion, pp. 1–60.
PSI Process Stream Index de Souza Porto, M.F., de Freitas, C.M., 1996. Major chemical accidents in industrializing
countries: the socio-political amplification of risk. Risk Anal. 16 (1), 19–29. https://
PVC polyvinyl chloride
doi.org/10.1111/j.1539-6924.1996.tb01433.x.
RStoic stoichiometric reactor Díaz-Ovalle, C., López-Molina, A., Vázquez-Román, R., 2016. A CFD-based approach to
UFL upper flammability limit predict explosion overpressure: a comparison to current methods. Chem. Biochem.
VCM vinyl chloride monomer Eng. Q. 30 (4), 419–427. https://doi.org/10.15255/CABEQ.2015.2244.
Gaur, U., Wunderlich, B., 1983. Heat capacity and other thermodynamic properties of
linear macromolecules. VII. Other carbon backbone polymers. J. Phys. Chem.
Declaration of competing interest Ref. Data 12 (1), 29–63. https://doi.org/10.1063/1.555677.
Greenpeace, 2016. Pajaritos, bomba de tóxicos: Greenpeace. Greenpeace México. htt
ps://wayback.archive-it.org/9650/20200422152452/http://p3-raw.greenpeace.
The authors declare that they have no known competing financial org/mexico/es/Prensa1/2016/Junio/Gobierno-Federal-evade-soluciones-verdadera
interests or personal relationships that could have appeared to influence s-a-la-contaminacion-del-aire-en-la-ZMVM-ONG/Pajaritos-bomba-de-toxicos-Green
the work reported in this paper. peace/. (Accessed 20 September 2018).
Guido, M., Sarcognato, S., Pelletti, G., Fassan, M., Murer, B., Snenghi, R., 2016.
Sequential development of hepatocellular carcinoma and liver angiosarcoma in a
Acknowledgments vinyl chloride–exposed worker. Hum. Pathol. 57, 193–196. https://doi.org/
10.1016/j.humpath.2016.07.021.
Hassani, H., Silva, E.S., Al Kaabi, A.M., 2017. The role of innovation and technology in
The authors appreciate the financial support from CONACYT, sustaining the petroleum and petrochemical industry. Technol. Forecast. Soc.
Mexico. The authors also appreciate the help of professor Miguel Frias- Change 119, 1–17. https://doi.org/10.1016/j.techfore.2017.03.003.
Méndez to improve the quality of this paper. He, G., Zhang, L., Lu, Y., Mol, A.P.J., 2011. Managing major chemical accidents in China:
towards effective risk information. J. Hazard Mater. 187 (1), 171–181. https://doi.
All persons who have made substantial contributions to the work
org/10.1016/j.jhazmat.2011.01.017.
reported in the manuscript (e.g., technical help, writing and editing Heikkilä, A.M., 1999. Inherent Safety in Process Plant Design. Ph.D. Thesis, VTT
assistance, general support), but who do not meet the criteria for Publication number 384. Helsinki University of Technology, Espoo, Finland.
Hughes, K., Meek, M.E., Caldwell, I., 1994. 1,2-Dichloroethane: evaluation of risks to
authorship, are named in the Acknowledgements and have given us their
health from environmental exposure in Canada. J. Environ. Sci. Health C. 12 (2),
written permission to be named. If we have not included an Acknowl­ 293–303. https://doi.org/10.1080/10590509409373448.
edgements, then that indicates that we have not received substantial Koseki, H., Tamura, M., 1973. Leakage and explosion of a vinyl chloride monomer
contributions from non-authors. caused due to damage of a valve at distillation column feed piping at a
manufacturing plant of vinyl chloride monomers. Shippai database. http://www.sh
ippai.org/fkd/en/cfen/CC1000041.html.
Nomenclature Leong, C.T., Shariff, A.M., 2009. Process route index (PRI) to assess level of explosiveness
for inherent safety quantification. J. Loss Prev. Process. Ind. 22 (2), 216–221.
https://doi.org/10.1016/j.jlp.2008.12.008.
FLmix flammability limit in a mixture Mannan, S., 2012. Chapter 15 - emission and dispersion. In: Mannan, S. (Ed.), Lees’ Loss
H combustión energy Prevention in the Process Industries, fourth ed. Butterworth-Heinemann, Oxford,
Iϕ índice representativo del parámetro ϕ pp. 752–1074.
Moreira, J.C.S., Pires, C.A.M., 2010. Modelling and simulation of an oxychlorination
m mass of explosive component reactor in a fluidized bed. Can. J. Chem. Eng. 88 (3), 350–358. https://doi.org/
LFLmix low flammability limit in a mixture 10.1002/cjce.20281.
Po peak side-on overpressure at the r distance Noticias, A., 2016. Escenario abrumador tras explosión en torre de clorados. Estados -
YouTube. https://www.youtube.com/watch?v=SO2iHnoGOII. (Accessed 12 August
Pa ambient pressure
2020).
r desired affectation distance for explosion Ranzi, E., Dente, M., Rovaglio, M., Faravelli, T., Karra, S.B., 1992. Pyrolysis and
UFLmix upper flammability limit in a mixture chlorination of small hydrocarbons. Chem. Eng. Commun. 117 (1), 17–39. https://
doi.org/10.1080/00986449208936055.
Xreactive reactive final conversion
Rossberg, M., Lendle, W., Pfleiderer, G., Tögel, A., Dreher, E.-L., Langer, E., Rassaerts, H.,
yi mole fraction of component i Kleinschmidt, P., Strack, H., Cook, R., Beck, U., Lipper, K.-A., Torkelson, T.R.,
Ze parameter to calculate the overpressure Löser, E., Beutel, K.K., Mann, T., 2006. Chlorinated hydrocarbons. Ullmann’s
encyclopedia of industrial chemistry. https://doi.org/10.1002/14356007.a06_233.
pub2.
Griek letters Sanchez, J.E., 2016. Explosión en el complejo clorados 3... se ve la segunda explosión...
η empirical explosion efficiency comparte - YouTube. https://www.youtube.com/watch?v=rRZSPgv_MP8. (Accessed
ϕ PSI parameters 20 September 2018).
Shariff, A.M., Leong, C., Zaini, D., 2012. Using process stream index (PSI) to assess
Φ averaged PSI parameter inherent safety level during preliminary design stage. Saf. Sci. 50 (4), 1098–1103.
https://doi.org/10.1016/j.ssci.2011.11.015.
Subscripts Shen, R., Jiao, Z., Parker, T., Sun, Y., Wang, Q., 2020. Recent application of
Computational Fluid Dynamics (CFD) in process safety and loss prevention: a review.
C explosive component J. Loss Prev. Process. Ind. 104252.
E energy Sklavounos, S., Rigas, F., 2004. Validation of turbulence models in heavy gas dispersion
FL difference between the flammability limits over obstacles. J. of Hazad. Mater. 108 (1), 9–20. https://doi.org/10.1016/j.
jhazmat.2004.01.005.
j stream index
P pressure

9
H.M. Aquino-Gaspar et al. Journal of Loss Prevention in the Process Industries 70 (2021) 104404

Staff, Reuters, 2016. Pemex Raises Death Toll at Petrochemical Plant Explosion to 28 - Wang, H., Chen, B., He, X., Tong, Q., Zhao, J., 2009. SDG-based HAZOP analysis of
Reuters. https://www.reuters.com/article/us-mexico-pemex/pemex-raises-death-t operating mistakes for PVC process. Process Saf. Environ. 87 (1), 40–46. https://doi.
oll-at-petrochemical-plant-explosion-to-28-idUSKCN0XK02O (accessed September org/10.1016/j.psep.2008.06.004.
20th 2018). Wire, Bussines, 2018. Global Polyvinyl Chloride (PVC) Market Size, Demand Forecasts,
TNO, 2005. Methods for the Calculation of Physical Effects -Yellow Book. CPR 14E. TNO. Industry Trends and Updates (2018-2025) - ResearchAndMarkets.Com. Business
Wachi, S., Morikawa, H., 1986. Liquid-phase chlorination of ethylene and 1, 2- Wire. https://www.businesswire.com/news/home/20181024005331/en/Global-Po
dichloroethane. J. Chem. Eng. Jpn. 19 (5), 437–443. https://doi.org/10.1252/ lyvinyl-Chloride-PVC-Market-Size-Demand (accessed August 20th 2020).
jcej.19.437.

10

You might also like