Professional Documents
Culture Documents
Risk Based Process Plant Design Consider
Risk Based Process Plant Design Consider
Safety Science
journal homepage: www.elsevier.com/locate/ssci
a r t i c l e i n f o a b s t r a c t
Article history: An inherently safer approach is becoming a key parameter of process and plant design. However, a lack of
Received 29 November 2013 established guidelines and methods hinders most industries from utilizing inherent safety concepts to a
Received in revised form 13 May 2014 full extent. This paper presents a risk-based design decision-making tool considering inherent safety. The
Accepted 13 June 2014
tool is called the Risk-based Inherent Safety Index (RISI). The proposed indexing approach is an extension
of the Integrated Inherent Safety Index (I2SI) earlier developed by Khan and Amyotte (2004, 2005). The
RISI incorporates both consequence and probability of accident occurrence reduction through application
Keywords:
of inherently safer design principles throughout the process design life cycle. Unlike other available
Inherent safety
Process life cycle
dimensionless index-based matrices, risk components of the proposed indexing approach are expressed
Inherently safer design principles in terms of SI units. The RISI is applicable at different stages of the process design life cycle. Analytical and
I2SI subjective equations assess the damage potential of major process accidents: fire, explosion and toxic
Risk-based Inherent Safety Index release. The explosion accident scenario is studied separately in terms of vapor/gas explosion and dust
explosion. The decision-making potential based on the quantitative results of the methodology is dem-
onstrated by evaluating alternatives for biodiesel production.
Ó 2014 Elsevier Ltd. All rights reserved.
⇑ Corresponding author. Tel.: +1 7097642354. Process design is a complex activity that is carried out in differ-
E-mail address: samithcr@mun.ca (S. Rathnayaka). ent stages over a period of time. Design at each stage involves
http://dx.doi.org/10.1016/j.ssci.2014.06.004
0925-7535/Ó 2014 Elsevier Ltd. All rights reserved.
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 439
Nomenclature
V volume of the vapor cloud (m3) VU volume occupied by the units in 30 m radius (m3)
DHC heat of combustion (kJ/kg) M mass of the flammable substances (kg)
q density of the flammable material (kg/m3) S burning speed (/s); s = 2.3 Uw
OP operating pressure of the process unit (kPa) Uw wind speed at the elevation of the closed vertical center
c specific heat ratio of mass (m/s)
VP vapor pressure (kPa) CD cloud depth (m)
bC compressibility (kPa1) T0 temperature at the source (°C)
mm molar mass of the chemical (g/mol) P0 pressure at the source (kg/cm2)
DHf enthalpy of the reaction (kJ/mol) A area of the source (m2)
C explosive dust concentration (kg/m3) fv fraction of the liquid that will flash; fv ¼ HC Pv ðT s T b Þ
VC volume of the confinement (m3) CP average heat capacity of the liquid (J/kg °C)
Pmax maximum explosion pressure (kPa) HV heat of vaporization (J/kg)
Patm atmospheric pressure (kPa) Tb normal boiling point (°C)
ca heat capacity ratio of air at maximum explosion tem- Ts operating temperature (°C)
perature qL density of the liquid release (kg/m3)
MIT minimum ignition temperature (°C) Pg pressure inside the vessel (kPa)
D diameter of the pool (m) AP pool area (m2)
u wind velocity at a 10 m height (m/s) MW molecular weight
uc characteristics velocity (m/s) TP characteristics pool temperature (°C)
g gravitational acceleration (m/s2) hL height of the liquid above the release point (m)
m0 burning rate (kg/m2 s) K constant (K = 3.14)
qa density of air (kg/m3)
assessing, analyzing and evaluating design alternatives to enhance primarily on detailed piping and instrumentation design, electrical
safety along with other objectives such as economics, quality, and insulation designs, process control and automation, utilities
productivity, energy conservation and pollution prevention. The and support equipment and safety instrumented systems. During
process design life cycle represents this evolution over time this stage, a design and commissioning team along with engineer-
(Fig. 1). Researchers and regulatory bodies have classified stages ing, procurement and construction management (EPCM) contrac-
for the process design life cycle in different ways as relevant to their tors will carry out construction and plant commissioning.
own studies (CCPS, 2009; Palaniappan et al., 2002a; Mannan, 2005; Finally the operational team will be handed over the plant for
Hurme and Rahman, 2005; Tugnoli et al., 2008). In the present start-up of operations. There is a belief that the applicability of
work, the process design life cycle is divided into five stages and inherent safety strategies is significantly limited during the opera-
they are considered as key design decision-making points. These tion and modification stage. During this stage, which is the longest
are (as shown in Fig. 1): (1) conceptual design, (2) process selection stage of the process design life cycle, many changes in operation,
and design (3) detailed engineering design and commissioning (4) personnel, maintenance and equipment will likely occur. CCPS
operation and modification and (5) decommissioning. (2009) highlighted two main tasks for consideration of inherent
Conceptual design begins with researching an idea for a new
product or process. Research is carried out to determine the tech-
nical, economic and safety feasibility. If the product is practical and
feasible, conceptual design begins. The main purpose of conceptual
design is to study the process chemistry and to evaluate available
chemical synthesis routes. The chemical reactions involved, raw
materials, intermediate and by-products, storage, transportation
and waste treatment associated with each synthesis route are fur-
ther studied.
Once conceptual design efforts lead to configuration of the pro-
cess chemistry and synthesis routes, a process flow-sheet is devel-
oped. This stage is called the process selection and design stage.
Information on desired product rates, product purity, heat transfer
fluids, solvents, catalysts, control and operational methods gath-
ered from conceptual design, laboratory and pilot scale trials and
knowledge of the existing process are used to develop the base
design flow-sheet. During this stage, key decisions on selection of
unit operations, conversion factors, process parameters such as
temperature, flow rate, pressure, and selection of solvents and
catalysts are taken into consideration.
Once the process flow-sheet is developed, further studies are
carried out to improve operating conditions, optimize product
yields and energy usage, improve product quality, and investigate
the need for recycling by using information from process engineer-
ing design principles, computer-aided simulations and expert
knowledge. The detailed engineering design stage focuses Fig. 1. Classification of stages of the proposed process design life cycle.
440 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
safety at this stage: conserving the inherent safety features and entirely on dimensionless index values which may sometimes pro-
practices which were applied during the early stages of the process vide uncertain results and may limit the comparison of design
design life cycle, and seeking opportunities for continued options. I2SI shows less flexibility when applied to different stages
improvement in inherent safety. These can be incorporated and of the process design life cycle. Analysis of other available indices
implemented using inherent safety strategies, where applicable, also indicates similar limitations.
during installing, maintaining and operating the modified In the present work, a risk-based approach is proposed as an
equipment and practices. extension of I2SI development as a key design decision-making
The application of inherent safety strategies during the decom- tool addressing the above mentioned limitations. Key characteris-
missioning stage is equally important. However, this is not tics of the risk-based approach are:
discussed in detail here as the current work is not focused on the
decommissioning stage. Consideration of four most credible accident scenarios sepa-
rately: fire, explosion, and toxic gas and toxic liquid release.
3. Inherent safety tool for decision-making Addressing hazard reduction as well as accident occurrence
probability reduction.
Safety is a key factor in making design decisions, from the Inherent safety risk and base design risk expression in units (SI)
beginning through to the end of the process design life cycle. An rather than a dimensionless index, which provides a better per-
unsafe plant cannot be profitable due to production loss, capital spective for decision making.
loss, insurance cost and public liability. The risk is known as the Improved flexibility, providing applicability at different stages
quantitative representation of safety. The systematic method to of the process design life cycle.
manage process risks is called process risk management (PRM). It
is the application of a wide variety of strategies, techniques, proce- 4. Risk-based Inherent Safety Index (RISI)
dures, policies and systems that can reduce process hazards and
the probability of an accident. Inherent safety is becoming an RISI includes two distinct risk estimations. Figs. 2 and 3 graph-
attractive proposition in process risk management, placed at the ically illustrate the estimation of RISI. The procedure starts with
top of the hierarchy of risk management strategies. the estimation of risk for base design (RiskBD) and subsequently
Inherent safety is an approach that focuses on eliminating or the inherent safety risk for alternative designs (ISRisk). Both risks
reducing hazards by using the properties of a material or charac- are eventually integrated to develop the Risk-based Inherent Safety
teristics of the process. Inherent safety differs from three other Index (RISI). The RISI is defined as the ratio of the inherent safety
strategies (passive, active and procedural safety) because it seeks risk of the selected alternative to the risk of the base design as
to remove the hazard at the source rather than accepting the haz- given by Eq. (1).
ards and implementing add-on systems to control or to mitigate
ISRisk
them. ISD principles are applicable to a process at all stages of its RISI ¼ ; 0 RISI 1 ð1Þ
life cycle. As the process goes through different stages of the design RiskBD
life cycle, the possibility of implementing inherent safety may vary RISI is then used as the key decision-making parameter to select
(Kletz, 1991; Khan and Amyotte, 2004). the optimum design with maximum inherent safety. ISRisk can
A comprehensive review of the development of inherent safety also be used for decision making considering specific design
principles, matrices and models developed during the last decade requirement criteria.
was carried out by Srinivasan and Natarajan (2012). They summa- The value of the RISI varies from 0 to 1. As RISI approaches 0, the
rized many important factors associated with inherent safety alternative is called ‘‘perfect’’ inherently safer design. If it
developments. Their study revealed that the majority (87% of tech- approaches 1, the alternative is called ‘‘inoperative’’ inherently
nical articles) on inherent safety metrics which were developed to safer design. The design improvement of an alternative with
quantify safety associated with the process have focused on respect to the base design can be estimated using Eq. (2).
hazards rather than risk. Key developments in inherent safety
% improvement of inherent safety ¼ 100 ð1 RISIÞ ð2Þ
metrics/indices are the Prototype Index for Inherent Safety (PIIS)
(Edwards and Lawrence, 1993), Inherent Safety Index (ISI) Therefore, the terms ISRisk and RISI can be successfully utilized
(Heikkilä et al., 1996), i-safe (Palaniappan et al., 2002a,b), Inte- for design decision-making during different stage of the process
grated Inherent Safety Index (I2SI) (Khan and Amyotte, 2004, design life cycle.
2005), inherent benign-ness indicator (Srinivasan and Nhan,
2008), Process Route Index (PRI) (Leong and Shariff, 2009), the 4.1. Inherent hazard assessment and accident sequence analysis
inherent occupational health index (Hassim and Hurme, 2010)
and Inherent Safety Key Performance Indicators (IS-KPIs) method- Prior to risk estimation, the design stage should be specified. As
ology (Tugnoli et al., 2012). Inherent safety assessment in terms of the design stage has different activities, attributes and design crite-
the concept of risk rather than focusing on hazard reduction is also ria, the information required to evaluate safety will depend on the
discussed by authors such as Shariff and Leong (2009) and Shariff stage and the nature of the process. This is an iterative process that
and Zaini (2013). However, their methodologies are applicable only involves a combination of synthesis, analysis and evaluation of
for the preliminary design stage. Among these approaches, I2SI is process alternatives. The design team first carries out initial design
one of most referenced method by researchers and industrial to develop the base design.
practitioners (Srinivasan and Natarajan, 2012). Once the initial design is completed, hazard identification is
I2SI is a structured guideword approach that is used to measure carried out to identify all potential inherent hazards associated
the inherent safety of process units. This approach is composed of with the base design, and subsequently alternative designs are
two main sub-indices: hazard index (HI) and inherent safety developed. Inherent hazards may be associated with materials,
potential index (ISPI) which specify hazard potential, inherent equipment and the process itself (Hendershot, 1997; Palaniappan
safety potential and add-on controls. Though it is capable of per- et al., 2002a).
forming the inherent safety evaluation along with an economic Fire, explosion and gaseous and liquid toxic releases are studied
evaluation, the index has certain limitations. I2SI addresses hazard as they are considered the most credible accident scenarios associ-
reduction rather than risk reduction. Decision-making depends ated with the process industries. These accident scenarios are fur-
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 441
Fig. 2. The framework for estimating the base design risk, RiskBD.
ther analyzed to construct an accident sequence. Further, parame- 5. Risk estimation for base design (RiskBD)
ters that affect inherent safety, such as inventory, temperature,
pressure, toxicity and flammability, are also listed and analyzed The risk of the base design (RiskBD) is computed using three
during this step. main factors: damage radius in meters (DR), occurrence probability
A methodology for the identification of major accident hazards (PR) and risk control index (RCI). The framework and sequence of
(MIMAH) developed by Delvosalle et al. (2006) under the ARAMIS steps involved is shown in Fig. 2. Risk for the base case is expressed
project can be used to perform hazard identification. MIMAH used in terms of damage distance and is computed using Eq. (3).
the bow-tie technique to represent the logical cause-consequence
relationship of major accidents; hence accident scenarios can be
DR PR
RiskBD ¼ ð3Þ
generated. To identify the atypical hazards that may arise in a sys- RCI
tem, a methodology called DyPASI (dynamic procedure for atypical The RiskBD denotes the risk level of the base design after taking
scenario identification) (Paltrinieri et al., 2013) can be used along into consideration risk control measures (add-on safety controls)
with MIMAH. applied to the base design. The risk value of the base design can
442 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
Fig. 3. The framework for estimating inherent safety risk (ISRisk) and selection of optimum design.
also be used for decision making. If the base design risk is accept- 5.1. Damage radius (DR) estimation
able, further changes are not necessary. However, it is noted that
inherent safety is not taken into account in this risk calculation. Damage radius addresses the damage or harm caused due to
Therefore, subsequent steps are taken to generate possible fire, explosion, and gaseous and liquid toxic releases. It is measured
alternatives considering inherent safety and to proceed with risk in terms of the area having a 50% probability of complete destruc-
estimation. tion. Subjective and analytical equations have been developed to
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 443
estimate the damage radii for each accident scenario and are dis- as maximum explosion pressure (Pmax) and maximum rate of pres-
cussed in subsequent subsections. sure rise ðdP Þ
dt max
with the parameters that effects to inherent safety
are studied using the information available in the literature
5.1.1. Explosion damage radius estimation (EDR) (Amyotte, 2013; Eckhoff, 2003; Mannan, 2005). The conclusions
The equation to estimate the EDR is derived using a multi- from this study are used to decide the penalties and the guidelines
energy method. The main assumptions considered here is: defla- to estimate penalties are listed in Appendix B.
grative combustion and explosion blast are composed of a number
of sub-blasts corresponding to the number of potential blast 5.1.2. Fire damage radius estimation (FDR)
sources identified within the vapor cloud (Assael and Kakosimos, The FDR is derived using a point-source model. The principal
2010). The coefficient of the strength of the blast is considered as assumptions considered here are:
its maximum, 10, and the maximum overpressure generated by
deflagration is considered as 2 atm (200 kPa). Considering these Heat flux from the center of the fire to the target is at ground
facts, Eq. (4) is developed to estimate the EDR. level and the wind effect is insignificant.
1=3
Transmissivity of the atmosphere sa is approximately equal to
EDR ¼ 0:90 ðhazard potentialÞ ð4Þ one (1).
All target receptors are considered to behave as black bodies.
where the hazard potential is the total energy released during the
Damage due to fire is a direct consequence of the heat flux.
explosion and is expressed in kilo Joules. The hazard potential for
explosion is formulated using two elements: energy factor (F) and
Similar to EDR estimation, the thermal radiation intensity limit
penalty (pne). Four energy factors, F1, F2, F3 and F4 are defined to
is chosen as 37.5 kW m2 considering the design for a worst-case
estimate the hazard potential of a premixed gas/vapor cloud explo-
scenario. This intensity limit could cause property damage and
sion, which take into consideration chemical energy, physical
100% lethality in 1 min (1% lethality in 10 s) (Assael and
energy (isentropic gas expansion and liquid expansion below boil-
Kakosimos, 2010). Eq. (7) estimates the damage radius due to fire.
ing point) and energy released due to chemical reaction, respec-
tively (Khan et al., 2001). FDR ¼ 0:05 ðhazard potentialÞ
1=2
ð7Þ
Process units are divided into five categories: storage units, units
involving physical operations such as absorptions, adsorption, dis- where hazard potential is the total heat release rate during the fire
tillation, evaporation, units involving chemical reactions, transpor- and is expressed in kilo Watts. The damage radius for jet fire is spe-
tation units and other hazardous units such as furnaces, boilers and cifically determined by taking half the value obtained by Eq. (7).
other similar units (Khan et al., 2001). Penalties are assigned to Similar to the explosion damage radius estimation, the energy
address the impact of the various process and operating parame- potential is estimated using energy factor E, and penalty pnf for each
ters. The equations for energy factors, hazard potential, and penal- process unit. Considering the above mentioned assumptions, only
ties assigned for different process units are listed in Appendix A. one energy factor is defined to take the radiative heat release rate
There exist similarities of characteristics between a premixed into account. Four distinguishing equations are defined for estimat-
gas/vapor explosion and a dust explosion, except for two basic dif- ing the energy factor of a pool fire, fireball, jet fire and flash fire. The
ferences: physics of the generation and accumulation of a dust most probable fire scenario associated with operating conditions
cloud and flame propagation in the cloud (Abbasi and Abbasi, and process units is considered for design purposes. The penalties
2007; Eckhoff, 2006). Considering these similarities, the equations are also defined to take into consideration the impact of the influ-
developed to estimate the damage radius of a gas/vapor explosion encing factors. Using available literature (Assael and Kakosimos,
can be used to estimate the damage radius of the dust explosion. It 2010; CCPS, 2010; Mannan, 2005; Roberts, 1982), energy factors
is noted that the hazard potential and penalties vary considering and penalties are developed and listed in Appendix C.
the factors influencing a dust explosion. For a dust explosion,
two energy factors FD1 and FD2 are introduced to take chemical 5.1.3. Toxic gas release damage radius estimation (TGDR)
energy and physical energy into account, respectively. Chemical The TGDR is defined as the ground level distance from the
energy liberated due the combustion process (FD1) is computed release source to a receptor at which the downwind toxic concen-
by using the heat of combustion of dust particles (DHc), explosible tration is at the threshold limit value (TLV). To estimate the dam-
dust concentration (C), and the volume of the confinement (VC) as age radius the Pasquill–Gifford plume model is used with the
shown in Eq. (5). FD2 estimates the physical energy release due to following assumptions:
the adiabatic rise in the pressure of the explosion energy. It is com-
puted using Eq. (6). Stability class is chosen as ‘‘slightly stable’’ because it repre-
sents median atmospheric conditions (Khan et al., 2001).
FD1 ¼ DHc C V C ð5Þ
Dispersion coefficients are chosen based on the slightly stable
ðP stability class.
max P atm Þ
FD2 ¼ 2 104 VC ð6Þ Model characteristics considered are: plume, continuous
ca 1 release, steady state, source at ground level in the direction of
where Pmax and Patm denote the maximum explosion and atmo- wind.
spheric pressure, respectively. ca denotes the heat capacity ratio Downwind speed is 2 m/s.
of air at maximum explosion temperature.
Common hazardous operations involved in a dust explosion are The equation developed to estimate the damage radius due to
listed elsewhere (Abbasi and Abbasi, 2007). Using this information, continuous release of toxic gas is:
it is agreed that process units involving chemical reaction and 0:6
TGDR ¼ 5:86 C G ðhazard potentialÞ ð8Þ
physical operation such as absorption, adsorption, distillation, will
not be a candidate for the inherent safety assessment of a dust The hazard potential is defined as the toxic gas (vapor) release
explosion. Three other process units: storage units, transportation rate which is estimated by release factor G and assigned penalties
units and other hazardous units such as furnaces, boilers, have to (png). CG is defined as the maximum allowable concentration
be thoroughly studied. The behavior of dust explosion factors such which is estimated using Eq. (9).
444 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
0:6
1 height of the liquid above the release point, AP is the pool area
CG ¼ ð9Þ and TP is the characteristics pool temperature.
TLV
Similar to toxic gas release, seven penalties are assigned to take
where TLV is the threshold limit value for the released gas. into consideration influencing factors. The guideline to determine
Two main gas release scenarios prevail in process units: (1) flow the penalties and their relevant values is presented in Appendix D.
of gas/vapor through holes and (2) flow of vapor through pipes. It is
assumed that flow through a pipe is controllable and accidental 5.2. Estimation of probability of occurrence (PR) of accident scenarios
release is through a hole. The released gas is assumed to behave
as ideal gas and discharge is classified as isentropic free expansion; Once the hazard assessment is completed, the probability of
the release factor is then estimated using Eq. (10) (Crowl and occurrence of a particular accident scenario is estimated. A generic
Louvar, 2002). bow-tie model to represent the accident scenario is proposed here.
0:5 This will guide users to develop the accident scenario and then to
MW
G ¼ 7:5 103 A P0 ð10Þ estimate the probability of occurrence of a particular accident
T0
scenario. For the case of base design the model is shown in
where A is the area of the source, P0 is the pressure at the source, T0 Fig. 4. It is clear that inherent safety is not the main consideration
is temperature at the source and MW is the molecular weight of the in the base design. The initiating event or the top event is the
substance released. release of gas or liquid under the pre-defined operating conditions.
In this case, penalties are assigned to address the impact of The failure probability for the dispersion prevention barrier is com-
parameters such as operating temperature, pressure, vapor den- prised of the failure of three different barriers mentioned here. By
sity, toxicity of chemical, site characteristics, external environmen- assigning probabilities based on industrial specific data or avail-
tal factors and vulnerability of the area. The expressions to decide able statistics, the probability of failure of the dispersion preven-
the penalties are obtained by referring to the literature: Crowl and tion barrier is estimated by multiplying the failure probabilities
Louvar (2002), Khan et al. (2001), and incorporating authors’ of individual barriers. The event sequence varies according to the
knowledge. Penalties developed here are listed in Appendix D. process condition of release (high pressure and low pressure)
and the physical state of the material (liquid or gas). As well, the
5.1.4. Toxic liquid release damage radius estimation (TLDR) type of ignition decides the type of accident scenario. The probabil-
In the case of toxic liquid release, toxic liquid may reach the ity of occurrence of delayed or immediate ignition depends on the
ground and form a pool that spreads according to the terrain. If nature and availability of the ignition source, and it is also assigned
the vessel is surrounded by a dike, the liquid usually flows to the based on industrial specific data. Three main add-on safety barri-
wall of the dike, and the dimension of the pool is then equal to ers: fire/explosion sprinkler system, fire/explosion wall, and fire-
dimension of the dike. In other cases, the pool is assumed to be cir- fighting are suggested for escalation prevention. The failure
cular. Liquid release may cause damage to soil and water in the probabilities are assigned for each barrier based on industrial spe-
case of an unrestricted flow and lack of barriers. For such scenarios, cific data. The probability of top event occurrence depends on the
it is difficult to estimate the damage radius using a mathematical failure probability of the release prevention barrier. Four main
equation and such an incident must be analyzed independently. immediate causes of escalation prevention barrier failure are
Instead, this work focuses on the damage due to an airborne toxic presented in the bow-tie model. A more comprehensive failure
substance which has evaporated from a liquid pool and/or flashing analysis of the release prevention barrier can be found elsewhere
of the liquid. The damage radius due to liquid release is defined as (Rathnayaka et al., 2011a). Having determined the failure probabil-
the distance from the liquid pool or flashing unit to the threshold ities, simple event tree analysis is performed to estimate the end
limit value (TLV) concentration and is derived using Eq. (11) event probability.
(AIChE technical manual, 1994). It is noted that probability calculation may contain a certain
degree of uncertainty. However this would be obviated due to
0:5
TLDR ¼ 6:51 C L ðhazard potentialÞ ð11Þ the relative analysis. During the conceptual stage, information for
the process units, operating parameters, instrumentation and util-
where
0:5 ities is scarce, whereas information on the chemical reactions and
1
properties is widely available. Therefore, the accident scenario will
CL ¼ ð12Þ
TLV be determined based on the chemical and physical properties and
Hazard potential represents the airborne toxic quantity due to probability is directly obtained through available sources: the
liquid toxic release. As mentioned earlier, the airborne quantity literature and the industrial data.
can be produced by pool evaporation and flashing of liquid, and
depends on the operating temperature of the unit. A momentum 5.3. Estimation of the Risk Control Index (RCI)
balance is used to model the flow of liquid through a hole (Crowl
and Louvar, 2002), and it is assumed that the release of liquid will It is clear that inherent safety is not the sole approach for risk
continue for at least five minutes before the release can be stopped reduction. Hierarchical arrangement of risk management strategies
(AIChE technical manual, 1994). The hazard potential due to flash- highlights that add-on safety is the second layer of protection.
ing (L1) is estimated using Eq. (13), whereas the hazard potential Add-on safety implies the application of both active and passive
due to pool evaporation (L2) is estimated using Eq. (14). safeguards. Therefore, it is important to quantify their effect on risk
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi reduction, and for this purpose, RCI is introduced. The RCI is an
6 1000P g objective analysis that attempts to quantify the process dependence
L1 ¼ 6:0 10 AqL fv þ 9:81hL ð13Þ
qL on risk reduction by means of add-on safety measures. The analysis
is performed similar to the ‘‘Process and Hazard Control Index
MW VP
(PHCI)’’ development earlier proposed by Khan and Amyotte
L2 ¼ 9:0 104 A0:95
P ð14Þ (2004). The framework to estimate the RCI is shown in Fig. 5.
TP
Evaluation is conducted considering application of add-on con-
where qL is the density of the liquid release, fv is the fraction of the trol measures to process control and risk control. The requirement
liquid that will flash, Pg is the pressure inside the vessel, hL is the of the process control is first evaluated. Process control measures
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 445
Fig. 4. The accident sequence and probability analysis bow-tie model – base design.
are applied to maintain the process parameters within the desired The relationship between the extent of requirement and the
region. Any deviation may cause the creation of a hazardous condi- index is developed. This is shown graphically using Fig. 6. If the
tion that eventually leads to a major accident. Temperature, pres- system is fully equipped with all control arrangements, it means
sure, flow and level are considered four major process that the system no longer requires further add-on safety measures.
parameters and the rest are considered to be one category. Therefore, the highest value of the index is assigned. The lowest
Risk controls are applied to reduce the subsequent damage or index is assigned if there is no or little control arrangement in
severity due to an abnormal event which is caused by a process the system. In this case, the system requires significant add-on
deviation. Risk controls reduce the probability of a particular acci- safety measures to establish safety of the process. Hence, an
dent occurring. For instance, emergency shutdown (ESD) is improved alternative design will be proposed.
installed to activate on demand to isolate the release; hence the Once indices for each control arrangement are estimated, RCI is
rate of release is controlled or cut off and the probability of a cat- estimated by taking the summation of all indices as per Eq. (15).
astrophic accident will be reduced, preventing events from escalat-
ing into a catastrophic accident. Risk control measures are RCI ¼ ½RCIT þ RCIP þ RCIF þ RCIL þ RCIOTH þ RCIESD þ RCIISO
categorized into five distinct groups to analyze their requirement þRCIECO þ RCIVDL þ RCIAFD ð15Þ
and dependency. They are emergency shutdown system, isolation,
engineering control, venting and dilution, and alarm and fault
detection as shown in Fig. 5. 6. Inherent Safety Risk for Alternative Design (ISRisk)
Process dependence and the requirement of both process and
risk control are measured using a subjective scale. The guidewords Design alternatives are generated when the risk of the existing
to decide the extent of the requirement of the control arrangement design (base design) is not acceptable. Several alternatives can be
and associated scale are presented in Table 1. The guidewords are suggested by adding safety control measures into the base design.
similar to those used in the development of I2SI (Khan and This could be achieved through a systematic approach of risk man-
Amyotte, 2004, 2005). The maximum scale ‘‘10’’ is assigned if the agement strategies: inherent, passive, active and procedural safety.
system is equipped with all necessary control arrangements and As an inherently safer process design is the core protection layer
the guideword is termed ‘‘Not required’’. Scales 1 and 2 are (Kletz and Amyotte, 2010), the design criterion for development
assigned if the system has no or little control arrangements. The of alternatives is risk reduction through the application of inherent
guideword for this scenario is ‘‘Essential’’. The rest of the values safety design principles.
are assigned based on the analyst’s judgment on the requirement Once the alternative design is generated based on ISD
of the control arrangements. principles, each design is evaluated to estimate the contribution
446 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
Table 1
Guideline to decide the extent of the requirement of process and risk control
measures (adopted and revised from Khan and Amyotte (2004)).
existing barriers are modified adopting ISD principles. Thus the reduce the hazards, and consequently the severity of conse-
branch of the event tree shifts backwards creating more branches. quences. The index aL is used to represent the applicability of
On the fault tree side of the bow-tie model, events that lead to fail- inherent safety principles to reduce the probability of a particular
ure of the release prevention barrier are improved by using ISD accident occurrence. It is difficult to analyze applicability in terms
principles. Add-on safety measures are utilized only if required. of an analytical or mathematical model. Therefore, a subjective
In Fig. 7, the revised immediate event (IE-R) may have a lower fail- scaling method is used based on authors’ knowledge and is
ure probability than the base design which eventually results in a expressed in terms of non-dimensional index value.
comparatively low probability of occurrence of gas or liquid release The applicability of each ISD principle to a process system in
(top event). Further, ISD principles can be applied to prevent igni- different design stages varies. It is not always possible to apply
tion sources or change the nature of the ignition sources, which all ISD principles to one system for the given process conditions.
eventually reduces the probability of the occurrence of delayed Most importantly, a change of design or component that reduces
or immediate ignition. The combined effect lowers the probability one hazard or probability may create a new hazardous condition
of a particular accident occurrence. The probability of alternative or increase the magnitude of an existing condition (CCPS, 2007).
design i (PRi) is estimated using the event tree calculation. The evaluation of such conflicts and tradeoff of hazards using
inherent safety concepts has been discussed elsewhere (Rusli
et al., 2013). Therefore, those factors are also to be considered
6.1. Determination of applicability of inherently safer design options
when deciding on the scale.
In the present work key ISD principles considered are: minimi-
zation, substitution, moderation (attenuation and/or limitation of 6.1.1. Estimation of applicability index of inherently safer design
effects) and simplification. The definitions are readily available. principles to reduce the hazard, aH
In the present work, the authors have referred to definitions from Applicability index aH is obtained based on the applicability
Kletz and Amyotte (2010). score. The applicability score for hazard reduction is assigned by
To estimate the applicability of ISD options, two indices are answering two questions: to what extent a particular ISD principle
used: aH is used to present the applicability of ISD principles to can be applied to a system, and how much hazard reduction can be
Fig. 7. The accident sequence and probability analysis bow-tie model – alternative design.
448 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
Table 2 graphs as shown in Fig. 8. The index varies from 1 to 10. Index 1
Guideline to decide the applicability index of inherently safer design options for is the minimum value which represents the case of no process
hazard reduction, aH (adopted and revised from Khan and Amyotte (2004)).
improvement or no hazard reduction in the system due to the
Guideword Score application of ISD options. Indices 2 to 9 represent the gradual
Extent Hazard increment of system inherent safety as hazards are reduced. Index
Completely applicable Eliminated 10
10 is the maximum value. This indicates that the system is com-
Significantly reduced 9 pletely improved (e.g. process is minimized to a large extent) in
Applicable May be eliminated 8
terms of inherent safety and hazards are no longer present in the
Significantly reduced 7 system.
Reduced 6 The overall inherent safety applicability index for the ith alter-
Applicable but process Reduced 5 native is then estimated by combining all applicability indices
dependent May be reduced 4 using Eq. (17).
May be applicable Reduced 3 h i1=2
May be reduced 2 aH;i ¼ ðami Þ2 þ ðasu Þ2 þ ½aat ali þ ðasi Þ2 ð17Þ
Not applicable No hazards reduced or 1
eliminated
6.1.2. Estimation of applicability index of inherently safer design
principles to reduce the probability of accident occurrence, aL
obtained. Five guidewords are assigned to answer the first ques- Probability of accident occurrence depends on the performance
tion, and six guidewords are assigned for the second question. and availability of safety barriers. The logical arrangement of the
Based on these guidewords, the user develops the scenario for safety barriers and possible consequences are shown in Fig. 7. To
the applicability. Each scenario is then assigned a numerical score change the end event probability, either new safety barriers can
as shown in Table 2. be introduced or existing safety barriers can be improved for better
Score 1, Not applicable, means that hazards are neither reduced performance. The present work examines the applicability of ISD
nor eliminated. The guideword May be applicable is used when the principles to carry out both options. The inherent safety applicabil-
user is uncertain whether a particular ISD option is applicable or ity index for probability reduction aL is introduced to measure the
not. Two hazard reduction guidewords associate with it: May be extent of applicability of ISD principles to reduce the occurrence
reduced and Reduced. Score 2 is assigned if the user is uncertain probability of accident scenarios.
about both the applicability of ISD principles and hazard reduction. Based on accident sequences described in Rathnayaka et al.
Score 3 is assigned for the case where the user assumes that there (2011a), four main stages are considered: release, dispersion, igni-
is moderate hazard reduction if the ISD option is applicable. tion and escalation (Fig. 7). The ISD options can be applied to pre-
The guideword Applicable but process dependent indicates that vent, mitigate and control each of these steps. Hence the
the user compromises the application of the ISD option owing to occurrence probability is reduced. These design options also
process and product limitations, tradeoff and hazards conflicts. reduce the severity of the consequences. It is clear that the applica-
Score 4 is assigned to take the user’s uncertainty of hazard reduction bility of the ISD option to a reduction in occurrence probability will
into consideration under this condition. On the other hand, score 5 is be varied at each stage of the accident sequence. Therefore, each
assigned to consider the user’s certainty of hazard reduction. The stage is assessed separately and consequently combined together.
guideword Reduced is used to indicate moderate hazards reduction. Similar to aH calculation, the graphical approach is used here.
Significant applicability of the ISD option to a process system is The graphs are developed for each ISD option. Applicability score
denoted by the guideword Applicable. Upon significant application is assigned answering two questions: to what extent can a partic-
of the ISD option, the user decides on one of these hazard reduction ular ISD principle be applied to the system and how much reduc-
cases: hazards moderately reduced, hazards significantly reduced, tion of occurrence probability can be obtained. To answer the
and hazards may be eliminated. The guidewords, Reduced, Signifi- first question, the guidelines developed for index aH calculation
cantly reduced, and May be eliminated, denote these scenarios, are used. The guideline to answer the second question is modified
respectively. Scores 6, 7, and 8 are assigned to each scenario, as required. Based on these guidewords, the user develops the sce-
respectively. nario for the applicability and decides the numerical score using
When the system can accept many ISD options, the applicability Table 3. Similar definition used in deciding scores during estima-
is denoted by Completely applicable. This will lead to complete tion of applicability index of inherently safer design principles to
elimination or significant reduction of the hazard. The hazard reduce hazards are used deciding scores for estimation of applica-
reduction guideword Eliminated represents the complete elimina- bility index of inherently safer design principles to reduce proba-
tion of hazard presence in the system. Score 9 is assigned where bility of accident occurrence.
the user believes there is a significant reduction of hazards. The Using authors’ knowledge and expertise in this area, four graphs
maximum score 10 is assigned if the hazard is completely elimi- are developed and presented in Fig. 9. Similar to aH calculation, the
nated. This is an ideal case. inherent safety applicability index varies from 1 to 10. Each index
To estimate an inherent safety applicability index associated is interpreted similarly to the estimation of hazards reduction
with each applicability scenario, the graph based approach is used. applicability index. The index value is decided considering how
Hazard reduction using moderation is assessed considering two ISD principles reduce occurrence probability and balance the
aspects: using hazardous materials under less hazardous condi- trade-off and conflicts of hazards. Therefore, even though a higher
tions (attenuation) and changing the design (limitation of effects). score may be assigned for a particular ISD principle, the inherent
Khan and Amyotte (2004) developed graphs for the ISD principles safety index may still be low.
minimization, substitution, attenuation and limitation of effects. During aL calculation, the inherently safer design options limita-
These graphs are used in the present work. An additional graph tion of effects and attenuation are studied together under the ISD
to estimate the applicability index for simplification is introduced option called moderation. Attenuation is the use of hazardous
here. materials under least operating conditions which has been already
Once the applicability score is decided, the relevant inherent incorporated during hazards reduction process. Thus it has insig-
safety applicability index for hazard reduction is located by using nificant applicability for reduction of probability of occurrence,
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 449
Fig. 8. Graphical relationship to estimate the inherent safety availability index for hazard reduction, aH (adopted and revised from Khan and Amyotte (2004)).
Fig. 9. Graphs to estimate the inherent safety availability index for occurrence probability reduction, aL.
applicability to reduce the occurrence probability of fire is derived produce fatty acid, mainly oleic acid and byproduct glycerol. Glyc-
taking the geometric mean of indices estimated for each stage of erol is separated out from oleic acid using a phase separator. Sep-
the accident sequence. (Eq. (19)) arated fatty acid is then mixed with methanol before being fed into
" #1=j the transesterification reactor. The acid in the presence of excess
jP3
Y methanol and catalyst is heated to reaction temperature to pro-
aL;i ¼ aL;j ð19Þ
duce the biodiesel. After the reaction complete, excess methanol
j¼1
is recovered from the distillation column as the overhead product
where j = 1, 2, 3, 4 represent four accident sequence stages: release, and biodiesel is obtained from the bottom product. The methanol
dispersion, ignition and escalation, respectively. is further purified using a distillation process, recycled and used
as a reactant.
7. Application of the methodology
7.1. Base design risk estimation
Numerical feasibility of this methodology is tested by perform-
ing a simple case study. The case study is based on the design of a In the base design, the plug flow reactor (PFR) is used for the
biodiesel production plant using palm oil as the main raw material. transesterification reaction. An atmospheric pressure distillation
The process description, process modeling and simulation results column is used to separate the methanol from the biodiesel.
can be found in detail elsewhere (Gómez, 2013). For overall under- Fig. 10 presents the process flow sheet of the base design.
standing, a brief process description is presented here. The credible accident scenario associated with each unit is
Palm oil and the methanol are the main feedstock materials for determined based on the operating conditions, properties, and
this study. Refined, bleached and deodorized palm oil which con- state of the materials. The reactor is operated under high pressure
tains triglyceride is produced through a palm oil extraction pro- and temperature. The concentration of biodiesel and methanol
cess. It is noted that triolein is used as the main component in released at any time varies with the conversion rate of the reaction.
this work because of the limitation of the process simulator (Aspen Vapor Cloud Explosion (VCE) is determined as the most credible
Plus) used for the simulation in this work. Triolein is hydrolyzed to accident scenario (MCAS) associated with the plug flow reactor
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 451
Fig. 10. Simplified process flow diagram of base design (Gómez, 2013).
Table 4
Summary of results of the base design.
Unit Description Type of the unit Most credible Damage Probability of Risk control RiskBD (m)
accident scenario radius, occurrence, PR index, RCI
(MCAS) DR (m)
R-102 Plug Flow Reactor (PFR) Units involving chemical reaction VCE 265.3 1.3439E-04 69 5.1671E-04
T-101 Distillation column 1 Units involving physical operation VCE 66.6 1.5269E-04 68 1.4943E-04
T-102 Distillation column 2 Units involving physical operation Pool fire 4.7 1.3970E-04 68 9.5431E-06
(R-102) and distillation column 1 (T-101) where biodiesel separa- than the damage radius produced by the PFR. However, the risk
tion primarily takes place. Distillation column 2 (T-102) is occu- of VCE for both units is slightly different as their probability of
pied with a high concentration of liquid methanol and is VCE occurrence is approximately equal. In this stage, decision mak-
operated in a vacuum at a low temperature. Therefore, pool fire ing is performed based on the risk acceptance criteria. However, it
is determined as the most credible accident scenario associated is noted that inherent safety is not taken into consideration.
with distillation column 2 where methanol purification takes
place. Table 4 summarizes the results of damage radii analysis 7.2. Inherent safety risk analysis for alternative design
for the base design. For the damage radius calculation for pool fire,
the pool diameter is taken to be as 5 m. The operating conditions Alternative designs are suggested considering design improve-
required for detailed calculation are primarily obtained through ment primarily in terms of ISD principles. For this particular case
the process simulation. study, two distinct alternatives are suggested focusing mainly on
Each unit is then analyzed using the bow-tie model (Fig. 4) to the PFR, distillation column 1 and distillation column 2.
estimate the probability of occurrence. The top event is chosen
as the release of flammable material. Basic event failure probabil- 7.2.1. Alternative 1
ities are estimated using the Offshore Reliability Data Handbook Alternative 1 uses the same reaction system as the base design.
(OREDA, 2002), Crowl and Louvar (2002), HSE (2012), Mannan The simplified process flow diagram is shown in Fig. 11. Alternative
(2005), Rathnayaka et al. (2011b). The results are summarized in 1 uses the ISD principles of minimization, moderation and simpli-
Table 5. fication for hazard reduction. The inventory of methanol flow is
Considering the complexity of the process operation and the reduced by approximately half of the initial value, achieving the
operating conditions, the extent of the requirements of the process same throughput of biodiesel. The operating pressures of the
and risk control measures are subjectively analyzed. For example, methanol purification unit (distillation column 2) are reduced by
assuming the plug flow reactor which is operating under high pres- applying the moderation option; as a result, the boiling point of
sure and high temperature does not have enough pressure and tem- the substances maintains a lower value. The process has been sim-
perature control, the extent of requirement is described as plified further by eliminating the unit used for methanol purging
‘‘essential’’ and ‘‘very important’’. The relevant RCIs for pressure, and piping. Reduction of the probability of accident occurrence is
temperature and flow are then formulated using Fig. 6 as 3, 3 and achieved through the application of the ISD principles of simplifi-
3, respectively. It is estimated that the plug flow reactor required cation and substitution. This can be accomplished by using the
a moderate improvement of control measures. For two distillation bow-tie model. Each basic event of the fault tree part is analyzed
columns, RCIs are estimated as 68 which implies that these two for reliability improvement. Both inherent safety and engineered
units have used both risk and process control measures moderately. safety measures are then suggested to improve the reliability of
Table 4 summarizes the base design risk (RiskBD) for each units. the events. It is required to emphasize the inherent safety conflicts
The PFR shows a comparatively higher risk as it has a large in this stage. Further, additional inherent safety barriers are
damage radius, but a low probability of occurrence. The damage applied to prevent, control and mitigate accident propagation in
radius due to VCE of distillation column 1 is considerably lower the event tree side of the bow-tie model. Subsequently, revised
452 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
Table 5
Suggested safety measures to improve the reliability of basic events of fault tree of distillation column 1of alternative 1.
Event Causal factors (or basic event) Inherent safety measure to prevent the accident Other safety measures
propagation or probability reduction
1 Failure of pressure safety valve Perform periodic functional testing of pressure relief
valves; carry out regular calibration, set the safe
operating limit wider than narrow set point
2 Failure of automatic isolation Substitute existing control instruments with high SIL Use redundant system design; conduct regular proof
(SIL 3 or 4) equipment (substitution) testing without interrupting the operation
3 Failure of tower pressure indicator Use the diaphragm seal for pressure gauges Incorporate an international industrial standards and
(moderation); substitute existing pressure gauges with upgrade accordingly
digital pressure gauges (substitution)
4 Failure of overhead product pressure Substitute existing control instruments with high SIL
transmitter (SIL 3 or 4) equipment (substitution)
5 Failure of overhead product pressure Use digital pressure recorders (substitution)
recorder failure
6 Inadequate detector coverage Install adequate sensors; conduct proper area
classification
7 Delayed response Simplify plant by reducing additional pipework and Apply an efficient visual numbering and sign system;
valves so that operators can easily reach isolation valve conduct regular operator training through a simulated
(simplification) environment as well as actual environment; measure
response time versus different alarm rate and operator
characteristics to establish a proper training protocol
8 Lack of accessibility Simplify plant by reducing additional pipework, valves
and equipment (simplification)
9 Failure to identify high pressure alarm or Simplify the existing alarm system to avoid operator Establish proper alarm management procedures;
misinterpretation overloading the alarms (simplification) conduct regular operator training; improve control
dashboard to easily and timely identify the alarms; use
different operator emphasizing systems
10 Loss of cooling water supply to condenser Monitor cooling water feed
11 Failure of condenser due to rupture Change design using high reliable construction
material (moderation)
12 Fouling or blockage of condenser tubes Install cooling water pre-treatment unit to remove
impurities, hardness materials, salts, etc.; install a
device to measure the overall heat transfer resistance
(OHTR) and predict micro and macro-fouling using
these data, introduce proper and timely cleaning
mechanism
13 Failure of reflux flow control valve Substitute the conventional positioners with highly an Conduct preventive maintenance to avoid the external
advance digital positioners (substitution) leakages of valves; apply tough coating to the linkage
so that linkages are less susceptible to vibration; make
sure that no dirty process air supply
14 Uncontrolled feed Apply control system to properly control the feed
15 Failure of level control valve of the bottom Substitute the conventional positioners with highly Conduct preventive maintenance to avoid the external
product advance digital positioners (substitution) leakages of valves; apply tough coating to the linkage;
make sure that no dirty process air supply
16 Failure of tower level control transmitter Substitute existing control instruments with high SIL
(SIL 3 or 4) instruments (substitution)
17 Failure of tower temperature transmitter Substitute existing control instruments with high SIL
(SIL 3 or 4) instruments (substitution)
18 Failure of re-boiler steam control valve Substitute the conventional positioners with highly Conduct preventive maintenance to avoid the external
advance digital positioners (substitution) leakages of valves; apply tough coating to the linkage
19 Failure of temperature indicator controller Advance the controller with better control algorithm
(moderation); place the sensor at proper place
(moderation); choose solid state relay or DC voltage
since they contain no moving parts (moderation)
20 Failure of temperature indicator Substitute existing control instruments with high SIL
(SIL 3 or 4) instruments (substitution)
21 Failure of flanges and connection attached Minimize the number of connections (minimization) Perform regular weld degradation monitoring; use high
to the distillation column reliable sealing material
22 Leaks through pumps Install piping and tubing according to a manufacturer’s
recommendations; avoid pinching, cocking or incorrect
installing
23 Material defects Promote offsite (before purchasing) inspection to make
sure proper material quality
24 Inadequate strength of material Estimate the material properties required to properly
construct
25 Erroneous maintenance Apply fool proof design (moderation) Estimate optimum number of maintenance intervals
through risk based maintenance process
26 Failure of preventive maintenance of Simplify the design by reducing the number of Implement risk-based maintenance management
valves/pumps/fittings maintenance points/intervals (simplification) system (RBM)
27 Inadequate maintenance program Implement RBM with good safety culture
28 Failure of regular corrosion inspection Identify the minimum inspection points; implement
methods to predict the corrosion rate
29 Failure of leak testing protocol Substitute existing detectors with the high sensitive Perform regular area based leak detection; conduct
leak (gas) detectors (substitution) proper area classification
30 Failure of weld degradation monitoring Reduce connection as much a possible (simplification) Determine the optimum number of degradation points;
S. Rathnayaka et al. / Safety Science 70 (2014) 438–464 453
Table 5 (continued)
Event Causal factors (or basic event) Inherent safety measure to prevent the accident Other safety measures
propagation or probability reduction
develop theoretical models to predict the degradation;
utilize the suitable welding method
31 Inadequate inspection program Implement risk-based inspection (RBIM) with good
safety culture
32 Lack of supervision or training Provide regular awareness about new technology and
operating methods; implement safety oriented work
environment
33 Operator negligence or mistakes Design fool proof design (moderation) Implement safety oriented work environment
34 No safe work procedures or industrial best Establish an international recommended industrial
practices best practice; perform periodical review and updating
35 Leaks during start up and shutdown Perform Start-up shutdown based on the
recommended practice
Fig. 11. Simplified process flow diagram for alternative design 1 (Gómez, 2013).
Table 6
Summary of the results of alternatives.
Unit operation Description Category Most credible accident Damage radius, Probability of Risk control aH aL ISRisk1 (m)
scenario (MCAS) DR (m) occurrence (PR) index (RCI)
Alternative 1
R-102 Plug Flow Reactor (PFR) 3 VCE 248.8 9.7737E-05 62 4.4 4.9 1.8151E-05
T-101 Distillation column 1 2 VCE 60.9 1.0823E-04 73 4.3 5.0 4.2287E-06
T-102 Distillation column 2 2 Pool fire 5.3 1.0151E-04 73 4.3 5.0 3.4747E-07
Alternative 2
RT-101 Reactive distillation column 3 VCE 183.5 9.4411E-05 73 8.9 4.9 3.1105E-07
T-102 Distillation column 2 2 Pool fire 4.2 1.0151E-04 83 4.3 5.0 7.7198E-08
failure probabilities are used to perform the quantitative analysis. broader view of inherent safety measures rather than considering
For illustration, the basic events of the fault tree for distillation col- the formal definition of ISD options. Engineered and procedural
umn 1of alternative 1 or causal factor that lead to vapor cloud safety measures required to improve the reliability of the system
explosion in distillation column 1 are listed in Table 5. Inherently are also listed in column three of Table 5.
safer design options that can be used to reduce the probability of It is determined that the most credible accident scenarios asso-
accident occurrence or to prevent or control the propagation of ciated with the PFR, distillation column 1 and distillation column 2
accident process are also listed in column tow (2) of Table 5. Prob- are the same as those in the base design. The damage radii for the
ability of accident occurrence highly depends on the failure of three units are then estimated using the equations developed,
safety barriers that apply to prevent and control the accident revised process parameters and operating conditions obtained
sequence process. Application of inherent safety measures to through the process simulator. Damage radii of the PFR and distil-
improve the performance of safety barriers lead authors to think lation column 1 have been reduced whereas the damage radius of
454 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
distillation column 2 has slightly increased even after the applica- to fewer transfer operations and less pipework and the frequency
tion of safety measures. The operating temperature of distillation of shutdown and startup operations (simplification). Further, the
column 2 of alternative 1 has increased significantly compared to atmospheric pressure methanol separation distillation column is
the base design. This leads to the increment of the penalty due to replaced by a vacuum distillation column. As a result fewer vapors
temperature; hence, the damage radius is increased. The occur- will be produced during an unexpected release of methanol. The
rence probability (PR) is obtained using the revised fault tree esti- inventory of methanol used for the complete process has been sig-
mation. The results clearly indicate that the application of safety nificantly reduced as a result of the characteristics of the reaction
measures has reduced the probability of occurrence of VCE and system. The process has been significantly simplified by removing
pool fire associated with the units studied. Risk control indices heat exchangers and additional mixer. Three key ISD options (mod-
for each unit are estimated similarly to the base design. It is noted eration, substitution and simplification) are used to reduce the
that RCIs are estimated as 62, 73 and 73 for the PFR, distillation occurrence probability of a particular accident (ISD options to
column 1, and distillation column 2, respectively. The higher RCIs reduce the accident sequence). For instance, existing control instru-
indicate that the units are equipped with process and risk control mentation is replaced by instrumentation with a high safety integ-
measures to a reasonably large extent. rity level (SIL) (substitution); the existing alarm system is simplified
The inherent safety applicability indices for alternative 1 vary to avoid operator overloading with alarms (simplification).
between 4 and 5. These values imply that inherent safety is slightly For alternative 2, there are only two major units considered: the
below the moderate value. Inherent safety risk (ISRisk) is then esti- reactive distillation column (RT-101) where both reaction and sep-
mated and the results are presented in Table 6. The results demon- aration occur and distillation column 2 (T-102) where methanol
strate that the PFR has highest inherent safety risk which is purification occurs. Considering the process parameters and operat-
1.8151 105 and distillation column 2 has the lowest inherent ing conditions, it is determined that the most credible accident sce-
safety risk of 3.4747 107. The inherent safety risk of distillation narios associated with the reactive distillation column and
column 1 lies between the PFR and distillation column 2. distillation column 2 are a vapor cloud explosion (VCE) and pool fire,
respectively. The damage radii for these two units are then esti-
7.2.2. Alternative 2 mated as 183.5 m and 4.2 m, respectively. The pool diameter for this
Alternative 2 uses a different reactor system with the same case is considered as 4 m as the unit runs with less inventory of
reaction synthesis. To develop alternative 2, all four inherent safety methanol (689.65 kg/h). It is clear that the damage radii of both units
design options are used. The process flow diagram for alternative 2 have been considerably reduced when compared to the base
is shown in Fig. 12. In alternative 2, the plug flow reactor and first design and alternative 1. The occurrence probability of a VCE
separation distillation column have been replaced by one unit, the (9.4411 105) has lessened slightly and the occurrence probability
reactive distillation column. Using the reactive distillation column of a pool fire (1.0151 104) remains the same when compared to
reduces the inventory of hazardous material (methanol and biodie- alternative 2. Estimation of the risk control indices for each unit uses
sel) (minimization), the number of process operations which leads the procedure explained in the methodology. RCIs are estimated as
Fig. 12. Simplified process flow sheet for alternative design 2 (Gómez, 2013).
Table 7
RISI results for each process unit of alternatives 1 and 2.
73 and 83 for the reactive distillation column and distillation column ing two alternative processes of biodiesel manufacturing. Three
2, respectively. It is noticed that RCIs for alternative 2 also reach a key hazardous units are selected for evaluation. Based on the risk
higher value, indicating that the units are equipped with process estimation, it is found that both alternative 1 and 2 are better
and risk control measures to a large extent. options than the base design. With the implementation of inherent
The estimation of the applicability index of ISD principles to safety design options and necessary engineered safety measures,
reduce both hazards (aH) and their probability of occurrence (aH) damage radius and probability of accident occurrence of both alter-
follows the procedure described in the methodology. Values for
aH and aL of the reactive distillation column are obtained as 8.92
and 4.87, respectively. It is clear that the designed reactive distilla-
tion column uses the inherently safer design options to a consider-
ably higher extent in terms of hazards reduction, whereas the
inherently safer options are used to a moderate extent in terms of
the probability of occurrence reduction. Values for aH and aL of dis-
tillation column 2 remain similar to alternative 1. Inherent safety
risk (ISRisk) is then estimated and presented in Table 6. The inher-
ent safety risk of reactive distillation column (3.1105 107 m) is
slightly lower than alternative 1. Here, reactive distillation is com-
pared with only PFR of alternative 1. Note that the distillation col-
umn 1 used in alternative 1 has been eliminated in alternative 2.
The inherent safety risk of distillation column 2 has also shown a
slight decrease compared with alternative 1.
To provide a better comparison between alternatives and to Fig. B1. Penalty due to size of the dust particle for storage units.
choose the ‘‘best’’ alternative, the risk-based inherent safety index
is estimated. Percentage improvement with respect to base design
is also estimated using Eq. (2). The results are shown in Table 7.
In alternative 1, the risk-based inherent safety index value of all
three units is very low. These are 0.0351, 0.0238 and 0.0364 for
PFR, distillation column 1, and distillation column 2, respectively.
Results are closer to 0 than 1, indicating that units are inherently
safer than the base design. Based on the percentages of improve-
ment, how much risk reduction has been achieved through appli-
cation of the inherently safer design option compared with base
design can be explained. The PFR has obtained 96.5% improvement
after applying inherent and engineered safety design options. Dis-
tillation column 1 has obtained slightly higher improvement than
distillation column 2. In alternative 2, instead of using both PFR
and distillation, one single unit, reactive distillation, is introduced.
In this study, the reactive distillation column is evaluated com-
pared to the plug flow reactor. The RISI for the reactive distillation
column is 0.0106 and percentage improvement is 99. This result
clearly indicates that reactive distillation is a better option than
using integrated PFR and distillation. The RISI of distillation col- Fig. B2. Penalty due to initial pressure of the containment.
umn 2 is also slightly lower than that of alternative 1 as it operates
with a lower pressure and temperature range than alternative 1.
Comparison of the RISI values demonstrates that alternative 2 is
a better option than alternative 1.
Fig. C2. Penalty for proximity of the unit to other hazardous unit due to the fire ball.
Fig. D1. Penalty due to operating pressure for toxic gas release.
Fig. C3. Penalty for operating pressure of the process unit effected to jet fire.
Fig. D2. Penalty due to operating pressure for toxic liquid release.
457
(continued)
458
Appendix A
pne6 to pne8
Estimate similar to storage unit
pne1 – pne8
Unit involving Estimate similar to units involving physical operation (F1 pne1 + F pne2 + F4 pne9 pne10)
chemical reaction pne3 pne4 pne5 pne6 pne7 pne8
pne9 – Nature of the reaction
Estimate as shown in SWeHI method (Khan et al., 2001)
pne10 – Impact of the side reaction
Estimate as shown in SWeHI method (Khan et al., 2001)
Transportation units pne1 – Transportation temperature (TT) (F1 pne1 + F2 pne2) pne3 pne4 pne5
Estimate similar to storage units pne6 pne7 pne8 pne11
Appendix A (continued)
Process unit Penalty Energy factor Hazard potential
pne6 – pne8
Estimate similar to storage units
Other hazardous units pne1 and pne3 F1 pne1 pne3 pne4 pne5
Estimate similar to units involving physical operations pne6 pne7 pne8 pne11
pne4 – Capacity of the unit
NF
pne4 ¼ 1:0 þ 1000 rate ðtons=hÞ
pne5 – Characteristics (flammability or reactivity)
pne5 = max (1, 0.35 (NR + NF)
459
460 S. Rathnayaka et al. / Safety Science 70 (2014) 438–464
Appendix B. Penalties, energy factors and hazard potential to estimate the dust explosion damage radius (D-EDR)
Other hazardous units pnd1 – Operating temperature (OT) (FD1 pnd1 pnd2 pnd3 pnd4
(screening, If OT 6 0.3 MIT, pnd1 = 1.25 pnd5 pnd6 + FD2 pnd4)
classification, If 0.3 MIT < OT 6 0.7 MIT, pnd1 = 1.45 pne3 pne6 pne7 pne8
separations, size If 0.7 MIT < OT 6 MIT, pnd1 = 1.65
reduction, ancillary If OT > MIT, pnd1 = 1.85
units such as dryers,
heaters, and conveyer,
and packaging units)
Table B1
Guideline to assign the penalty due to type of the process equipment.
Appendix C. Penalties, energy factors and hazard potential to estimate the fire damage radius (FDR)
Process unit Pool fire Fire ball Jet fire Flash fire
Storage units pnf1, pnf5, pnf8, pnf9 pnf1, pnf5, pnf9, pnf9 pnf1, pnf5, pnf8, pnf9 pnf1, pnf5, pnf8, pnf9
Estimate similar to EDR Estimate similar to EDR Estimated similar Estimate similar to EDR
EDR
Units involving reactionx If u/uc > 1, pnf3 = u/uc pnf 2 ¼ 1:1 þ ðOPVP
OP Þ 0:2
pnf4 pnf4
Case 2: VP > AP Estimate using Negligible effect
Fig. C1
If VP > OP P AP,
pnf 2 ¼ 1:1 þ ðOPVP
OP Þ 0:4
If OP > VP,
Transportation units u
¼ ðgm
0D 1=3
pnf 2 ¼ 1:1 þ ðOPVP
OP Þ 0:6
uc q Þ
a
Energy Factor Epf = 0.785 D2 m0 DHC Efb = 2.2 DHC M0.67 Ejf = m0 DHC S
Eff ¼ M DHC ðCD Þ
Hazard potential Epf pnf1 pnf3 pnf4 Epf pnf1 pnf3 pnf4 Ejf pnf1 pnf2 Eff pnf1 pnf3 pnf5
pnf5 pnf6 pnf7 pnf5 pnf6 pnf7 pnf4 pnf5 pnf6 pnf6 pnf8 pnf9
pnf8 pnf9 pnf8 pnf9 pnf7 pnf8 pnf9
pnf1 – operating temperature, pnf2 – operating pressure, pnf3 – metrological condition (wind velocity), pnf4 – distance to nearest hazardous unit, pnf5 – quantity of chemical,
pnf6 – possibility of domino effect, pnf7 – physical state of the material, pnf8 – external environmental effect, pnf9 – vulnerability of surrounding
Table C1
Penalties due to the physical status of the chemicals to estimate the damage radius due to fire.
Appendix D. Penalties, energy factors and hazard potential to estimate the toxic release damage radius (TDR)
Process unit Toxic gas release (TGDR) Toxic liquid release (TLDR)
Storage units png1 – Operating temperature pnl1 – Operating temperature
Units involving If(OT > 4 AT), png1 = 1.55 Case 1: OT 6 BP, then fv = 0 (Pool evaporation only)
physical If(OT > 2 AT), png1 = 1.35 L = L2
operation Else, png1 = 1.11 pnl1 ¼ 1:1 þ ðBPOT
BP Þ
Other L = L1
hazardous png4 – Toxicity of the released chemical If fv > 0.2, then there is no any pool is formed. It is logical
units png4 = 1.0 + 0.6 NH that the amount of airborne should be less than or equal to
liquid release rate. Therefore, in a case of fv > 1.0, the value is
set up to one.
png5 – Characteristics of the proximity of plant
pnl2 – Operating pressure
Urban area with high density population = 1.55 Estimate using Fig. D2
Suburb area with moderate density population = 1.35
Rural area with low density population = 1.11 pnl3 – pnl7
Estimate similar to toxic gas release
Hazard G png1 png2 png3 png4 png5 png6 png7 Hazard potential due to the pure liquid pool evaporation:
potential L2 pnl1 pnl2 pnl3 pnl4 pnl5 pnl6 pnl7
Hazard potential due to the pure flashing
L1 pnl1 pnl3 pnl4 pnl5 pnl6 pnl7
Hazard potential due to both liquid pool evaporation and
flashing
(L1 + L2 pnl2) pnl1 pnl3 pnl4 pnl5 pnl6 pnl7
Khan, F.I., Husain, T., Abbasi, S.A., 2001. Safety weighted hazard index (SWeHI): a Rahman, M., Heikkilä, A.M., Hurme, M., 2005. Comparison of inherent safety indices
new, user-friendly tool for swift yet comprehensive hazard identification and in process concept evaluation. J. Loss Prev. Process Ind. 18 (4–6), 327–334.
safety evaluation in chemical process industries. Process Saf. Environ. Prot. 79 Rathnayaka, S., Khan, F., Amyotte, P., 2011a. SHIPP methodology: predictive
(2), 65–80. accident modeling approach. Part I: Methodology and model description.
Khan, F.I., Sadiq, R., Amyotte, P.R., 2003. Evaluation of available indices for Proc. Safety Environ. Protect. 89 (3), 151–164.
inherently safer design options. Process Saf. Prog. 22 (2), 83–97. Rathnayaka, S., Khan, F., Amyotte, P., 2011b. SHIPP methodology: predictive
Kletz, T.A., 1978. What you don’t have can’t leak. Chem. Ind. 6, 287–292. accident modeling approach. Part II: Validation with case study. Process Saf.
Kletz, T.A., 1984. Cheaper, Safer Plants, or Wealth and Safety at Work. IChemE, Environ. Prot. 89 (2), 75–88.
Rugby. Roberts, A.F., 1982. The effect of conditions prior to loss of containment on fireball
Kletz, T.A., 1991. Plant Design for Safety: A User-friendly Approach. Hemisphere behaviour. IChemE Symp. Series 71, 397–429.
Publishing Corporation, New York, USA. Rusli, R., Shariff, A.M., Khan, F.I., 2013. Evaluating hazard conflicts using inherently
Kletz, T., Amyotte, P., 2010. Process Plant: A Handbook for Inherently Safer Design, safer design concept. Saf. Sci. 53, 61–72.
second ed. CRC Press, Taylor & Francis group, FL, USA. Shariff, A.M., Leong, C.T., 2009. Inherent risk assessment – a new concept to
Leong, C.T., Shariff, A.M., 2009. Inherent safety index module (ISIM) to assess evaluate risk in preliminary design stage. Process Saf. Environ. Prot. 87 (6), 371–
inherent safety level during preliminary design stage. Process Saf. Environ. Prot. 376.
86 (2), 113–119. Shariff, A.M., Zaini, D., 2013. Inherent risk assessment methodology in preliminary
Mannan, S., 2005. Lee’s Loss Prevention in the Process Industries, vol. 3. Elsevier Inc. design stage: a case study for toxic release. J. Loss Prev. Process Ind. 26 (4), 605–
OREDA, 2002. SINTEF Industrial Management: Det Norske Veritas. 613.
Ouattara, A., Pibouleau, L., Azzaro-Pantel, C., Domenech, S., Baudet, P., Yao, B., 2012. Srinivasan, R., Natarajan, S., 2012. Developments in inherent safety: a review of the
Economic and environmental strategies for process design. Comput. Chem. Eng. progress during 2001–2011 and opportunities ahead. Process Saf. Environ. Prot.
36 (10), 174–188. 90 (5), 389–403.
Palaniappan, C., Srinivasan, R., Tan, R., 2002a. Expert system for the design of Srinivasan, R., Nhan, N.T., 2008. A statistical approach for evaluating inherent
inherently safer processes. 1. Route selection stage. Ind. Eng. Chem. Res. 41 (26), benign-ness of chemical process routes in early design stages. Process Saf.
6698–6710. Environ. Prot. 86 (3), 163–174.
Palaniappan, C., Srinivasan, R., Tan, R., 2002b. Expert system for the design of Tugnoli, A., Khan, F., Amyotte, P., 2008. Inherent safety implementation throughout
inherently safer processes. 2. Flowsheet development stage. Ind. Eng. Chem. the process design lifecycle. In: International Conference on Probabilistic Safety
Res. 41 (26), 6711–6722. Assessment and Management (PSAM), 18–23 May, Hong Kong, China.
Paltrinieri, N., Tugnoli, A., Buston, J., Wardman, M., Cozzani, V., 2013. Dynamic Tugnoli, A., Landucci, G., Salzano, E., Cozzani, V., 2012. Supporting the selection of
procedure for atypical scenarios identification (DyPASI): a new systematic process and plant design options by inherent safety KPIs. J. Loss Prev. Process
HAZID tool. J. Loss Prev. Process Ind. 26 (4), 683–695. Ind. 25 (5), 830–842.