Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Process Safety and Environmental Protection 171 (2023) 353–364

Contents lists available at ScienceDirect

Process Safety and Environmental Protection


journal homepage: www.journals.elsevier.com/process-safety-and-environmental-protection

An inherently safer design approach based on process safety time for batch
chemical reaction processes
Yi Wu, Haotian Ye 1, *, Hong-guang Dong
School of Chemical Engineering, Dalian University of Technology, Dalian 116024, China

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

Keywords: The economic attractiveness of an industrial process depends a large extent on its safe and reliable operation. The
Multi-objective optimization inherently safer design can help reduce the potential hazard on a fundamental level and eliminate escalation at
Inherent safer the early stage of process development. In this work, an inherently safer design method for batch chemical re­
Process safety time
actors is proposed, which maximizes the yield of target product while minimizes Dow’s fire and explosion index.
Batch reactor
Thermal runaway criterion
Process safety time is introduced as criterion to respond to the thermal runaway risk risen by unexpected failures.
Finally, two examples for specific reaction systems with given kinetics are performed to demonstrate the
feasibility and validity of the proposed method. The results indicate that the product yield optimized with the
proposed method is significantly higher than that obtained from literature approach involving temperature limit
or divergence criterion under the same PST. With the same yield, the PST optimized with the proposed method is
longer.

technically complex and costly, but also may lead to catastrophic acci­
dents once the additional safety systems fail. The concept of inherently
1. Introduction safer design (ISD) is originally proposed by Kletz (1985), which is the
most critical layer in the layers of protection. There are four basic
Process safety is always a core issue in the design and operation principles of minimizing, substitute, moderate and simplify in ISD,
stages of a production process (Ahadh et al., 2021). Recent research which are widely used in process and product safety design, manage­
showed that 26% of major chemical plant accidents resulted from ment and accident investigation (Eini et al., 2015). The ISD can help
thermal runaways (Amine et al., 2020; Jiang et al., 2018, 2016). Batch reduce the potential hazard on a fundamental level and eliminate in­
and semi-batch reactors are widely applied in the chemical industry cidents at the early stage of process development (Pasha et al., 2017).
benefit from their simple structure, low manufacturing cost, and ease of Therefore, it is necessary to access abnormal situations and achieve an
operation and maintenance (Zavala et al., 2005; Chen et al., 2021). inherently safer optimal design of process parameters aiming at simul­
Thermal runaway in batch and semi-batch reactors bear high risk taneously maximizing product yield and minimizing process risk within
because of its high capacity of reactants and limited efficiency of heat the entire operation period for batch and semi-batch reactors.
removing (Strozzi and Zaldivar, 1994). Furthermore, the chemical re­ There are a huge number of researches devoted to the inherently
actions running within these kinds of reactors are usually strongly safer design of chemical process (Wang et al., 2021b, 2021b; Ubrich
exothermic. Once there are faults such as stirring system abnormity, et al., 2001). Zhu et al. (2022) proposed a general framework by inte­
cooling circulation pump or pipeline blockage, the heat generated by grating advanced process optimization and inherently safe operation
reaction may be much more than the heat removing by the cooling condition identification criterion for semi-batch reactors, which is
system. Consequently, there will be an important quantity of accumu­ applied to homogeneous solution polymerization. Castillo-Landero et al.
lated heat within the reactor to lead to rising temperature. This may (2022) proposed an inherently safer design method based on modularity
further induce thermal runaway (Jiang et al., 2018; Chen et al., 2020). to strengthen integration among process units while improving both
There are many ways to reduce risk to an acceptable level. The tradi­ safety and energy efficiency to intensify the chemical process. Jano­
tional design methods rely on additional safety systems to improve šovský et al. (2022) utilized process data and five safety indices and
safety such as safety valves and pressure gauges. This method not only is

* Corresponding author.
E-mail addresses: wuyiwuyi@mail.dlut.edu.cn (Y. Wu), htye@dlut.edu.cn (H. Ye), hgdong@dlut.edu.cn (H.-g. Dong).
1
ORCID: 0000-0003-3137-020X

https://doi.org/10.1016/j.psep.2023.01.007
Received 21 November 2022; Received in revised form 30 December 2022; Accepted 3 January 2023
Available online 5 January 2023
0957-5820/© 2023 Institution of Chemical Engineers. Published by Elsevier Ltd. All rights reserved.
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

Nomenclature Vbrf volume of material in the PST constraint process.


temf0 reaction temperature of material in the PST constraint
hcp heat capacity of material in the design process. process.
tem reaction temperature of material in the design process. hcpf0 heat capacity of material in the PST constraint process.
Vbr volume of material in the design process. Vbrf0 volume of material in the PST constraint process.
tem0 starting reaction temperature of material in the design qbrf heat releasing rate in the PST constraint process.
process. cbrf1 heat removing rate at the fault occurrence stage in the PST
hcp0 starting heat capacity of material in the design process. constraint process.
Vbr0 starting volume of material in the design process. cbrf2 heat removing rate at the fault repairing stage in the PST
tbr residence time of reactants. constraint process.
Ω collocation constant of point at the collocation matrix. FP the set of thirty fault points.
qbr heat releasing rate in the design process. FE the set of finite element.
cbr heat removing rate in the design process. xbrf molar composition in the PST constraint process.
tc cooling temperature. Sf effective heat transfer area in the PST constraint process.
dVbr feeding volume flow rate. fbrcf mole in the PST constraint process.
CP set of collocation points. fbrcf0 starting mole in the PST constraint process.
ΔH reaction enthalpy. rbrf reaction rate in the PST constraint process.
k0 pre-exponential factor. t1 total annual operation time.
xbr molar composition in the design process. t2 batch auxiliary time.
α, β reaction orders. R gas constant.
U overall heat transfer coefficient. fbroc mole at the end of the reaction process.
S effective heat transfer area in the design process. fbrc0 mole at the start of the reaction process.
fbrc mole in the design process. MAT maximum allowable temperature.
fbrc0 starting mole in the design process. hcpcc heat capacity of component c.
rbr reaction rate in the design process. Ea activation energy of reaction.
fed feeding mole flow rate. i finite element.
ν stoichiometric coefficient. j collocation point.
MV molar volume. c component.
temx reaction temperature of material in the fault point k fault point.
identification process. G penalty of material.
hcpx heat capacity of material in the fault point identification yield production yield.
process. FEI Dow’s fire and explosion index.
Vbrx volume of material in the fault point identification process. TB total heat.
temx0 reaction temperature of material in the fault point F1,A Penalty on exothermic chemical reactions.
identification process. F1,B Penalty on endothermic processes.
hcpx0 heat capacity of material in the fault point identification F1,C Penalty on material handling and transfer.
process. F1,D Penalty on enclosed or indoor process units.
Vbrx0 volume of material in the fault point identification process. F1,E Penalty on access.
qbrx heat releasing rate in the fault point identification process. F1,F Penalty on drainage and spill control.
cbrx1 heat removing rate at the normal operation period in the F2,A Penalty on toxic material.
fault point identification process. F2,B Penalty on sub-atmospheric pressure.
cbrx2 heat removing rate at the inherent safety period in the fault F2,C Penalty on operation in or near flammable range.
point identification process. F2,D Penalty on dust explosion.
xbrx molar composition in the fault point identification process. F2,E Penalty on relief pressure.
Sx effective heat transfer area in the fault point identification F2,F Penalty on low temperature.
process. F2,G Penalty on quantity of flammable/ unstable material.
fbrcx mole in the fault point identification process. F2,H Penalty on corrosion and erosion.
fbrcx0 starting mole in the fault point identification process. F2,I Penalty on leakage-joints and packing.
rbrx reaction rate in the fault point identification process. F2,J Penalty on use of fired equipment.
temf reaction temperature of material in the PST constraint F2,K Penalty on hot oil heat exchange system.
process. F2,L Penalty on rotating equipment.
hcpf heat capacity of material in the PST constraint process.

compared to each other to acquire the inherently safer process design the dynamics changes of complex hazardous scenarios and update the
alternatives considering fire and explosion hazards as well as the toxicity probability of risk events to provide guidance for the ISD (Li et al.,
level. Martino et al. (2021) provide a structured tool of an inherent 2021). In addition, the quantification of uncertainty contributes to
safety decision making framework, which can classify ISD alternatives improve the accuracy of risk calculation and guarantee safety in the
as purely technical and non-technical to examine possible interactions chemical process. Kalantarnia et al., (2009, 2010) modeled the BP Texas
between proposed alternatives. City refinery accident using the Bayesian failure updating mechanism
Due to the uncertainty and dynamic characteristics of the chemical with consequence assessment and later updated the likelihood of event
reaction process, the traditional static risk assessment cannot reflect the occurrence. Wang et al. (2016) presented a dynamic quantitative risk
ever-changing risks to lead to working by experience in process industry. assessment method to analyze the operational performance of chemical
Dynamic risk assessment can employ a variety of methods to simulate process, which evaluated the probability and consequences considering

354
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

multiple key variables. introduced as thermal runaway criteria to guarantee the safe operation
Chemical reaction thermal runaway is as one of the high-risk acci­ of the batch reactor in case of fluctuations or failures occurred during
dents, which is the important aspect to consider in the inherently safer the operation. The appropriate schemes in a set of optimal solutions was
process optimization design. Runaway criteria can classify operation accessed by using the method of the “Analytic Hierarchy Process +
state as runaway or non-runaway to obtain the safe state-variables and Entropy Weight method” based on Technique for Order of Preference by
parameters of the chemical reaction process (Ni et al., 2020; Kummer Similarity to Ideal Solution (“AHP+EW” TOPSIS). The main contribu­
and Varga, 2018). Kummer and Varga (Kummer and Varga, 2016; tion of this work is that it extends the scope of inherently safer design for
Kummer et al., 2020) used all kinds of criteria (Maxi criterion, Van batch chemical reactors from a novel perspective of PST constraint to
Heerden criterion, Lyapunov stability, etc.) to determine optimal response the thermal runaway risk risen by unexpected failures, which is
feeding trajectory in a pilot plant semi-batch reactor. Kähm and Vassi­ capable of providing reasonable and attractive solutions.
liadis (Kähm and Vassiliadis, 2018a, 2018b) modified the divergence The proposed method includes some assumptions and
criterion and resulted in better controllability and stability in intensified simplifications:
processes when it was embedded in model predictive control schemes.
Guo et al. (2016) can detect the region of thermal runaway on-line and (1) The kinetic parameters employed by chemical reactions, are
determine the real-time accumulation in semi-batch reaction process by considered to be strictly accurate within the composition and
using a new criterion. temperature range.
The inherently safer design of chemical process aims at fundamen­ (2) There is no mixing enthalpy, only reaction enthalpy is considered
tally not only eliminating the accidents conditions but also ensuring in the reaction system.
economic efficiency. Garcia and Toulouse (Garcia et al., 1995; Toulouse (3) The variation of reactant volume in the reactor only depends on
et al., 1996) proposed an optimization scheme that aimed to maximize the feeding operating.
the product yield and ensure the reaction temperature not exceeding (4) There is no phase transition during the chemical reaction process,
MAT, and optimized the batch operating parameters of temperature which is homogeneous in liquid phase.
trajectory, final batch time and feed flow rate. Ubrich et al. (2001) (5) The feeding temperature is equal to the starting reaction tem­
revealed the optimal time-varying feeding trajectory and used a heat perature and cooling temperature, which is a constant within the
balance criterion as safety constraint for an exothermic second-order reactor operating period.
esterification reaction, aiming at maximization of conversion. Thakur (6) The heat producing effect is only caused by the chemical reaction.
et al. (2022) obtained the optimal operating conditions to ensure the The heat removal effect is related with cooling system.
maximization production at a minimal cost while guaranteeing safety (7) The overall heat transfer coefficient is constant in the reaction
for a multi-objective optimization study. process.
When thermal runaway criteria were taken as safety constraints to (8) Heat capacity of the reactants is additive.
avoid the phenomenon of thermal runaway, some of them were too strict
to apply at the expense of economy in the inherently safer optimization The solution of this design method needs give the following
design process. The optimization of the operating and equipment pa­ information:
rameters for batch reactors cannot ensure the safe operation, so that
unexpected equipment or device failure and operation error are also (1) The critical process safety time for the abnormal scenario
necessary to be considered. In the incident of the reactor system failure, considered;
the operator must take steps to ensure the safety of the reactor within a (2) The type of the chemical reactors;
limited period of time, called process safety time (PST). At present, there (3) The flow rates and compositions of all process streams;
are few researches on the study of process safety time. Barnard and Creel (4) The Dow’s fire and explosion index of reactors;
(2016) showed some basic concept, such as, what is PST? How is it (5) The operation and investment parameters of reactors.
determined? When? And by whom? Varga and Abonyi (2010) intro­
duced a numerical analysis-based methodology to estimate PST which The paper is constructed as follows. The present section explored the
was calculated in different scenarios, such as the feeding volume of re­ research background and justified the necessity of the research. The
agent A was decreased. The necessary safety actions were determined to second section is the methodology including process safety time, Dow’s
avoid the undesired states of operation. Li et al. (2019) established a fire and explosion index, the optimization procedure and decision-
steady-state model of coke oven gas (COG) to methanol and calculated making method. The third section is mathematical formulations.
PST under the cases of compressor failure shutdown, different oxygen Finally, the proposed design method is illustrated by two case studies in
flow rate and COG flow rate. Kummer et al. (2020) calculated PST in the fourth section, and the conclusions are provided in the last Section.
different initial reagents concentrations. PST is not a simple specific
value, different safety actions have different PST demand to avoid the 2. Methodology
escalation of the unsafe situation, and the process operators should
know which safety actions should be taken in PST. The inherent safer design approach is to maximize the product yield
However, the current study of PST still remains in its own meaning and minimize the hazard of the reaction process along with the proposed
and calculation so as to determine the potential available time for all thermal runaway criterion based on PST as safety constraints. The multi-
protection layers to response. Therefore, it is necessary to consider PST attribute decision-making method was finally employed to access
as the safety constraint in the batch process optimization to analyze the appropriate schemes from a set of Pareto solutions. The explanation of
abnormal production conditions and dangerous conditions of potential each segment of the proposed method is mentioned in the steps below.
incidents and achieve inherent safety operation.
Hence, in this work, an inherently safer design and optimization 2.1. Process safety time
method for batch chemical reactor is developed to achieve trade-off
between safety and economy. Besides, the abnormal conditions of PST was defined by CCPS Guidelines for Safe and Reliable Instru­
different fault modes related to cooling capacity are considered. From mented Protective Systems, which was the time period between a failure
the perspectives of severity, the Dow’s F&EI is widely used as the safety occurring in the process or its control system and the occurrence of the
metric in the process design (Jensen and Jørgensen, 2007) and is hazardous event (CCPS, 2007). In this paper, the reaction temperature
employed as an optimization objective in this work. From the aspect of was selected as process variable, which was most closely associated with
likelihood, process safety time along with the temperature limit were the occurrence of the hazardous event, to determine PST. PST is the

355
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

T/K T/K

MAT MAT
Temperature
rise threshold Temperature
rise threshold

fault for a period


Normal of time Normal fault for a period
operation range operation range of time
safe running
Tc
Tc operation safe running
operation
fault point repair point fault point repair point
t1 t3 t1 t2 t3
t/s t/s
Process safety time Process safety time

(a) (b)
Fig. 1. Schematic diagram of determining process safety time.

general process hazards including exothermic chemical reactions (F1,A),


Table 1
endothermic processes (F1,B), material handling and transfer (F1,C),
Degree of hazard for Dow’s F&EI.
enclosed or indoor process units (F1,D), access (F1,E) and drainage and
F&EI index range Degree of hazard spill control (F1,F), which are applicable to most process situations. It is
1–60 Light noted that it may not be necessary to take penalties for each of them at
61–96 Moderate the early stage of process design (Ye et al., 2022).
97–127 Intermediate
128–158 Heavy ∑
F
159-up Severe F1 = 1 + PenaltyF1 ,i (2)
i=A

F2 consists of specific process conditions which have been proved to


period of time from the time point of the temperature exceeding the
be major causes of fire and/ or explosion incidents. There are twelve
normal operation range to the time point of the temperature exceeding
items (F2,A-F2,L) for special process hazards, including toxic material (F2,
the maximum allowable temperature, that is, fault-tolerant time of the
A), sub-atmospheric pressure (F2,B), operation in or near flammable
chemical reaction process. When faults occur at any time of a batch
range (F2,C), dust explosion (F2,D), relief pressure (F2,E), low temperature
chemical reaction process, there are two situations: in the first situation,
(F2,F), quantity of flammable/ unstable material (F2,G), corrosion and
the reaction system fails from t = t1 and is repaired at t = t3.The reaction
erosion (F2,H), leakage-joints and packing (F2,I), use of fired equipment
temperature reaches the MAT simultaneously and then turns to a
(F2,J), hot oil heat exchange system (F2,K), and rotating equipment (F2,L).
downward shift shown in Fig. 1(a); in the second situation, the reaction
system fails from t = t1 and is repaired at t = t2. The reaction temper­ ∑
L

ature continues to rise to the MAT at t = t3 and then turns to a downward F2 = 1 + PenaltyF2 ,i (3)
shift shown in Fig. 1(b).
i=A

For the calculation of the F&EI in this paper, the assumptions and
recommendations from the classification guide are followed. Almost
2.2. Dow’s fire and explosion index every penalty factor included in these calculations will remain constant
for each of the potential scheme, except the F2,G item, which considers
Inherent safety index (ISI) is considered to be a most promising tool the additional exposure as quantity of flammable/ unstable material in
in the chemical process design and optimization. The superiority of ISI the process unit is increased. For the estimation of the Penalty F2,G, the
lies in its quantification, robustness and indication of potential incidents quantity of all flammable and unstable material in the process unit is
associated to process characteristics (Gao et al., 2021a,2021b), so that it considered in this work. Moreover, penalties of F2,D, F2,E, and F2,F in the
has been widely used in process industry with hazardous materials. The general process hazards, and F2,H and F2,I in the special process hazards
Dow’s fire and explosion index (F&EI) is a leading index recognized by are not considered in this work, since the information needed for their
the chemical industry in all the evaluation indices, which provides a estimations is usually not available at this stage. The degree of hazard
basis for relative ranking of the individual process units and helps for Dow’s F&EI is as shown in Table 1.
evaluate the overall risk from fire and explosion (AIChE, 1994; Dong
et al., 2022; Pan et al., 2020). In this paper, the F&EI is employed as an
objective for inherent safety considerations from the perspective of 2.3. Decision-making on multi-objective optimization
severity for the optimization design schemes. The index is calculated
based on the material factor (MF) and process unit hazards which is the The multi-objective methods usually produce a set of optimal solu­
product of the general process hazards factor (F1) and the special process tions, called Pareto optimal solutions. But it does not tell the decision
hazards factor (F2). maker which is the best solution. To access the best solution, a tradi­
tional approach is based on single weight assignment which has some
min F&EI = MF⋅F1 ⋅F2 (1)
degree of subjectivity and does not always reflect the designer prefer­
MF is a measure of the potential energy released from combustion or ences or even the specific and usually important particularities of a
chemical reaction by material under evaluation involved in the process given process. Fortunately, a method of comprehensive weighting,
units, which can be obtained according to flammability (NF) and reac­ which combined analytic hierarchy process (AHP) and entropy weight
tivity (NR) based on NFPA 49 and 325 M. (EW) approach has been developed (Zhang et al., 2022). By the use of
F1 is the factor that plays a primary role in determining the magni­ Technique for Order of Preference by Similarity to Ideal Solution
tude of a loss incident. There are six items (F1,A-F1,F) as contributing (TOPSIS), the best solution in multi-objective optimization results can

356
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

objective functions of Dow’s fire and explosion index and the yield of
target product and safety constraints of temperature limitation and
process safety time. A set of optimal solutions are finally acquired,
named as Pareto frontier. At the last step, final optimal solution of
process variables is chosen by using the method of “AHP+EW” TOPSIS,
which are variables including cooling temperature/initial reaction
temperature/feeding temperature, volume of reaction mass, heat
transfer area, feeding rate, reactant feeding ratio and reaction operating
time.

3. Mathematical model

3.1. Basic design equations for semi-batch/ batch reaction process

Orthogonal collocation with finite elements method is employed to


build the mathematical model for the batch and semi-batch reactor
optimization in this paper, which can transform the original differential
equation to an algebraic equation (Biegler, 2010). Hence, the addition of
reactants at the beginning or with a feeding rate during the reaction
process corresponds to the batch and semi-batch reactor, respectively.
The mass balanced and energy balanced equations are introduced to
describe the dynamic behavior of the reaction system, which is
expressed as follow
temi,j ⋅hcpi,j ⋅Vbri,j = tem0i ⋅hcp0i ⋅Vbr0i
∑ ( )
+tbr ⋅ Ωjj,j ⋅ qbri,j ⋅Vbri,j − cbri,j + tc⋅hcpcA ⋅dVbr (4)
jj∈CP
∀i ∈ FE, j ∈ CP

hcpi,j ⋅Vbri,j = hcp0i ⋅Vbr0i + tbr⋅ Ωjj,j ⋅hcpcA ⋅dVbr
jj∈CP (5)
∀i ∈ FE, j ∈ CP
The number of finite elements in the reaction period are demanded
to not exceed ten to balance the precision and solving difficulty. Three
collocation points and Radau roots are selected in this paper. For the
batch reactor, the feeding rate of reactants is required to zero, therefore
dVbr will be forced to be 0. The ode45 in Matlab is applied to validate the
accuracy of the developed mathematical model in this example, more
detail explanations of ode45 are shown in the Supporting Information.
The heat releasing rate in Eq. (4) is expressed as follow
∑ − Ea
Fig. 2. The step of “AHP+EMD” TOPSIS method. qbri,j = ΔH⋅k0 e R⋅temi,j ⋅xbrαi,j,A xbrβi,j,B ⋯
(6)
∀i ∈ FE, j ∈ CP
be found. In this work, the “AHP+EW” TOPSIS method was applied for
the proposed optimization approach. (Borza and Petrescu, 2016). The cooling capacity is a formulation of heat transfer, and the
In the first place, the subjective weights, ѡAj, and objective weights, equation is as follow
ѡBj, are calculated by using AHP and EW method, respectively (Zhang (
cbri,j = U⋅Si,j ⋅ temi,j − tc
)
et al., 2022). Then, based on the combined weight, the Lagrange func­ (7)
∀i ∈ FE, j ∈ CP
tion is introduced to construct the optimal decision model. Through the
Euclidean distance function (D (ѡAj, ѡBj)), the consistency of subjective Because of the characteristic of time-varying in the semi-batch
and objective weights and the preference coefficient (α,β) were main­ reactor, a correlation relationship is proposed by Wu et al. (2022) be­
tained, and index weights and the relationship between primary and tween the effective heat transfer area and the reactor volume to improve
secondary for evaluation index were obtained. Finally, the distance (D-, the model precision. The equation is as follow
D+) between the ideal solution (Z-) and non-ideal solution (Z+) were Si,j = 0.03899⋅Vbri,j 0.6347
measured and the alternatives were ranked according to the ratio of the (8)
∀i ∈ FE, j ∈ CP
non-ideal solution distance to the sum of the ideal and non-ideal solution
distances (Si) (Wang et al., 2021b, 2021a). The solution procedure is as Due to the semi-batch reaction material is added at a certain feeding
follows. (Fig. 2). rate, the volume of material is not a constant which is different from
batch reaction process. The material volume is composed of the start
volume and the added volume in the reactor. Similarly, the mole of the
2.4. The optimization procedure
material are consist of the start material, added material and generated
or consumed during the reaction process. Thus, there will be
Fig. 3 provides the illustrative procedure of the proposed method­

ology, which is utilized in this study. The first step is to determine the Vbri,j = Vbr0i + tbr ⋅ Ωjj,j ⋅dVbr
possible failure of the reactor, such as a breakdown of the cooling cir­ jj∈CP (9)
culation pump or pipeline blockage et al. In the second step, the multi- ∀i ∈ FE, j ∈ CP
objectives optimization process is displayed, which includes the

357
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

Fig. 3. The Procedure of the proposed methodology.

Table 2 Table 3
The kinetic parameters and operating conditions. Optimized results with different optimization schemes.
Parameter Value Unit Variables Optimization schemes I Optimization schemes II
“AHP+EW” TOPSIS “AHP+EW”TOPSIS
k0 7.6402729 × 107 L/(mol•s)
E 82,308.6 J/mol reaction mode batch batch
-ΔH 66,641 J/mol T0 / K 339.6 333.7
ρcp 2432 J/(L•K) Tc / K 339.6 333.7
U 300 W/(K•m2) Td / K 339.6 333.7
MAT 393.15 K Vd / m3 5.687 5.687
T0 317.15 K A / m2 9.422 9.422
Tc 317.15 K nA0:nB0 54,292:11961 36,400:22579
Td 317.15 K t/s 2178 1951
PST/ s 867 867
F&EI 100 100
Yield / t 6402 5570

)
∑ (
fbrci,j,c = fbrc0i,c + tbr ⋅ Ωjj,j ⋅ rbri,j,c ⋅Vbri,j + fedc
jj∈CP
(10)
∀i ∈ FE, j ∈ CP, c ∈ COM
For the batch reactor, the feeding rate of reactants is required to zero,
therefore fedc will be forced to be 0.
The reaction rate in Eq. (10) is expressed as follow
∑ − Ea
rbri,j,c = νc ⋅k0 e Rtemi,j ⋅xbrαi,j,A xbrβi,j,B ⋯
(11)
∀i ∈ FE, j ∈ CP, c ∈ COM
The start overall material volume is made up of the sum of each
component and the feeding material volume consists of the sum of each
component. The mole of component is the product of overall material
volume and component concentration. The equation expression is as
follow

Fig. 4. The production in two optimization schemes. fbrc0c ⋅MVc = Vbr0 (12)
c∈COM


dVbr = fedc ⋅MVc (13)
c∈COM

fbrci,j,c = Vbri,j ⋅xbri,j,c ∀i ∈ FE, j ∈ CP, c ∈ COM (14)

358
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

Fig. 5. The calculation error between the orthogonal collocation on finite elements method (OCFE) and Matlab Ode45.

Fig. 6. The yield of target product in different optimization schemes and fault modes.

The reactor temperature must be operated below the MAT, as follow


Table 4
The kinetic parameters and operating conditions. temi,j ≤ MAT
(15)
∀i ∈ FE, j ∈ CP
Parameter Value Unit

k0 8. 79 × 1010 L/(mol•s) 3.2. Equations for critical fault point identification


E 93,000 J/mol
-ΔH 58,860 J/mol
(ρcp)A 2149 J/(L•K) In this work, both the safe running and abnormal scenarios should be
(ρcp)B 4200 J/(L•K) considered in the batch reactor. Therefore, it is necessary to identify the
U 166 W/(K•m2) critical fault point of chemical reaction process, which will vary with
MAT 373.15 K different fault modes. It is noted that only cooling capacity related fault
T0 317.15 K
Tc 317.15 K
modes were considered. Thus, the reaction operation time is divided into
Td 317.15 K two periods. The first period is the sensitive operation stage, where the
reactor temperature may exceed MAT when the fault occurs. For the
semi-batch and batch (dVbr=0) reactor models, there will be

359
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

where, tbrx1 is the sensitive operation period, tbrx2 is the inherent safety
period.
The heat releasing rate in Eq. (18) is expressed as follow
∑ − Ea
qbrxi,j = ΔH⋅k0 e R⋅temxi,j ⋅xbrxαi,j,A xbrxβi,j,B ⋯
(20)
∀i ∈ FE, j ∈ CP
The cooling capacity is heat removal formulation in the first stage
and it is considered to be zero in the second stage of adiabatic conditions
for fault mode of cooling failure, which are expressed as follow
( )
cbrx1i,j = U⋅Sxi,j ⋅ temxi,j − tc
cbrx2i,j = 0 (21)
∀i ∈ FE, j ∈ CP

Sxi,j = 0.03899⋅Vbrxi,j 0.6347


(22)
∀i ∈ FE, j ∈ CP
The volume of the material and the amount of the material in the
reactor are expressed as follow
Fig. 7. The yield of target product in two optimization schemes. ∑
Vbrxi,j = Vbrx0i + tbrx1 ⋅ Ωjj,j ⋅dVbr
jj∈CP (23)
∀i ∈ FE, j ∈ CP
Table 5
Optimized results with different optimization schemes. )
∑ (
Variables Optimization schemes I Optimization schemes II fbrcxi,j,c = fbrcx0i,c + tbrx1 ⋅ Ωjj,j ⋅ rbrxi,j,c ⋅Vbrxi,j + fedc
(24)
“AHP+EW” TOPSIS “AHP+EW”TOPSIS jj∈CP
∀i ∈ FE, j ∈ CP, c ∈ COM
reaction mode semi-batch semi-batch
T0 / K 303.15 314.5 ∑
Tc / K 303.15 314.5 Vbrxi,j = Vbrx0i + tbrx2 ⋅ Ωjj,j ⋅dVbr
Td / K 303.15 314.5 jj∈CP (25)
Vd / m 3 5.687 5.687 ∀i ∈ FE, j ∈ CP
A / m2 9.422 9.422
u / mol⋅s− 1 2.07 1.69 )
∑ (
nA0:nB0 51,187:0 51,231:0 fbrcxi,j,c = fbrcx0i,c + tbrx2 ⋅ Ωjj,j ⋅ rbrxi,j,c ⋅Vbrxi,j + fedc
t/s 22,845 27,820 jj∈CP
(26)
F&EI 100 100
∀i ∈ FE, j ∈ CP, c ∈ COM
Yield / t 6145 5154
The reaction rate and the amount of component are expressed as
follow
temxi,j ⋅hcpxi,j ⋅Vbrxi,j = temx0i ⋅hcpx0i ⋅Vbrx0i
) ∑ − Ea
∑ ( rbrxi,j,c = νc ⋅k0 e Rtemxi,j ⋅xbrxαi,j,A xbrxβi,j,B ⋯
+tbrx1 ⋅ Ωjj,j ⋅ qbrxi,j ⋅Vbrxi,j − cbrx1i,j + tc⋅hcpcA ⋅dVbr (16) (27)
jj∈CP ∀i ∈ FE, j ∈ CP, c ∈ COM
∀i ∈ FE, j ∈ CP
fbrcxi,j,c = Vbrxi,j ⋅xbrxi,j,c
∑ (28)
hcpxi,j ⋅Vbrxi,j = hcpx0i ⋅Vbrx0i + tbrx1 ⋅ Ωjj,j ⋅hcpcA ⋅dVbr ∀i ∈ FE, j ∈ CP, c ∈ COM
jj∈CP (17)
∀i ∈ FE, j ∈ CP At the end of the reaction process, the reaction temperature should
be equal to the MAT
The second period is inherent safety stage, reactor temperature will
not exceed MAT even if the faults occur. The semi-batch reactor model is temxi,j = MAT
(29)
as follow ∀i ∈ max(FE), j ∈ max(CP)

temxi,j ⋅hcpxi,j ⋅Vbrxi,j = temx0i ⋅hcpx0i ⋅Vbrx0i


)
∑ ( 3.3. Constraint equations for process safety time critical equations
+tbrx2 ⋅ Ωjj,j ⋅ qbrxi,j ⋅Vbrxi,j − cbrx2i,j + tc⋅hcpcA ⋅dVbr (18)
jj∈CP
∀i ∈ FE, j ∈ CP In Section 3.2, it is indicated that the reactor temperature may
exceed MAT when a fault occur at the sensitive operation stage. In this
∑ work, thirty discrete fault points based on orthogonal collocation with
hcpxi,j ⋅Vbrxi,j = hcpx0i ⋅Vbrx0i + tbrx2 ⋅ Ωjj,j ⋅hcpcA ⋅dVbr
jj∈CP (19) finite elements are introduced to replace the continuous operation
∀i ∈ FE, j ∈ CP period to improve the solution efficiency. The fault influencing time is
divided into two periods, where the first period is the fault occurrence

Table 6
Comparison optimization results with the same yield and F&EI.
Optimization F&EI Yield PST Tc/ Td/ T0 A V u t
1
schemes /t⋅yr− /s /K /m2 /m3 /mol⋅s− 1
/s

I 100 5154 0 314.5 9.422 5.687 1.69 27,820


II 100 5154 1321 303.9 9.422 5.687 1.7 28,375

360
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

stage. For the batch and semi-batch reactors, there will be reactor should be no higher than the MAT, which should also be the peak
temperature within the period and are expressed as follow
temfk,i,j ⋅hcpfk,i,j ⋅Vbrfk,i,j = temf 0k,i ⋅hcpf 0k,i ⋅Vbrf 0k,i
)
∑ temfk,i,j ≤ MAT
+tbrf 1 ⋅
(
Ωjj,j ⋅ qbrfk,i,j ⋅Vbrfk,i,j − cbrf 1k,i,j + tc⋅hcpcA ⋅dVbr (30) (43)
∀i ∈ max(FE), j ∈ max(CP), k ∈ FP
jj∈CP
∀i ∈ FE, j ∈ CP, k ∈ FP ( )
qbrfk,i,j ⋅Vbrfk,i,j + tc − temfk,i,j ⋅hcpcA ⋅dVbr − cbrf 2k,i,j = 0
(44)
∑ ∀i ∈ max(FE), j ∈ max(CP), k ∈ FP
hcpfk,i,j ⋅Vbrfk,i,j = hcpf 0k,i ⋅Vbrf 0k,i + tbrf 1 ⋅ Ωjj,j ⋅hcpcA ⋅dVbr
jj∈CP (31) The safety constraint corresponds to the proposed thermal runaway
∀i ∈ FE, j ∈ CP, k ∈ FP criterion based on PST is expressed as follow

The second period is the fault maintenance stage. The reactor model is expressed as follow tbrf 1k ≥ PST
(45)
∀k ∈ FP
temfk,i,j ⋅hcpfk,i,j ⋅Vbrfk,i,j = temf 0k,i ⋅hcpf 0k,i ⋅Vbrf 0k,i
)
∑ (
+tbrf 2 ⋅ Ωjj,j ⋅ qbrfk,i,j ⋅Vbrfk,i,j − cbrf 2k,i,j + tc⋅hcpcA ⋅dVbr (32)
jj∈CP 3.4. Objective function
∀i ∈ FE, j ∈ CP, k ∈ FP
In this paper, the ε-constraint method was applied to solve the pro­

hcpfk,i,j ⋅Vbrfk,i,j = hcpf 0k,i ⋅Vbrf 0k,i + tbrf 2 ⋅ Ωjj,j ⋅hcpcA ⋅dVbr posed multi-objective optimization model, which can transform multi-
jj∈CP (33) objective problem into a suite of single-objective problems and other
∀i ∈ FE, j ∈ CP, k ∈ FP objectives converts as the constraints. This method is applicable to very
few objectives and can prevent attractive solutions from being over­
where, tbrf1 is the fault occurrence stage, tbrf2 is the fault repairing looked, which is suitable for the research scope of this paper. The
stage. objective functions are to maximize the product yield and minimize the
The heat releasing rate in Eq. (32) is as follow value of F&EI, and the mathematical expression equation is as follow
∑ − Ea {
qbrfk,i,j = ΔH⋅k0 e R⋅temfk,i,j ⋅xbrf αk,i,j,A xbrf βk,i,j,B ⋯ obj1 = − yield
(34) min (46)
obj2 = FEI
∀i ∈ FE, j ∈ CP, k ∈ FP
The cooling capacity is zero at the fault occurrence stage for the 3.4.1. Net product yield
adiabatic condition corresponding to fault mode of cooling failure and The first objective function is to maximum product yield in Eq. (47).
normal heat removal formulation at the fault repairing stage, which are ( )
t1
expressed as follow obj1 = − yield = − × fbrocc − fbrc01,c (47)
tbr + t2
cbrf 1k,i,j = 0 ( )
cbrf 2k,i,j = U⋅Sfk,i,j ⋅ temfk,i,j − tc (35) 3.4.2. The calculation of F&EI
∀i ∈ FE, j ∈ CP, k ∈ FP The second objective function is minimizing the F&EI, which can be
calculated by the inventory of flammable material as the only variable.
Sfk,i,j = 0.03899⋅Vbrfk,i,j 0.6347 The inventory is calculated in Eq. (48) for batch reactors.
(36)
∀i ∈ FE, j ∈ CP, k ∈ FP ( )
Vol = max Vbri,j (48)
The mass balance of the whole process is expressed as follow
∑ The penalty of flammable/unstable vapor and liquid in the batch
Vbrfk,i,j = Vbrf 0k,i + tbrf 1 ⋅ Ωjj,j ⋅dVbr
reactors can be calculated as follow
jj∈CP (37)
∀i ∈ FE, j ∈ CP, k ∈ FP ( )
log(G) ≥ 0.17179 + 0.42988 log TB − 0.37244 log (TB)2
(49)
) +0.17712 log (TB) − 0.029984 log (TB)4
3
∑ (
fbrcfk,i,j,c = fbrcf 0k,i,c + tbrf 1 ⋅ Ωjj,j ⋅ rbrfk,i,j,c ⋅Vbrfk,i,j + fedc
jj∈CP
(38) TB = f (Vol, ρ, HC ) (50)
∀i ∈ FE, j ∈ CP, c ∈ COM, k ∈ FP
According to the penalty of the batch reactor, the F&EI of it can be
∑ calculated as follow
Vbrfk,i,j = Vbrf 0k,i + tbrf 2 ⋅ Ωjj,j ⋅dVbr
jj∈CP (39) FEI = MF⋅F1 ⋅F2 (51)
∀i ∈ FE, j ∈ CP, k ∈ FP
) 4. Case study
∑ (
fbrcfk,i,j,c = fbrcf 0k,i,c + tbrf 2 ⋅ Ωjj,j ⋅ rbrfk,i,j,c ⋅Vbrfk,i,j + fedc
jj∈CP
(40) 4.1. The esterification reaction in batch reactor
∀i ∈ FE, j ∈ CP, c ∈ COM, k ∈ FP
The first example is an uncatalyzed esterification reaction, which is
The reaction rate and the composition of component in Eqs. (37)-
studied by Bosch et al. (2004). The 2-propanol is esterified with propi­
(40) are expressed as follow
onic anhydride to produce propionic acid and isopropyl propionate. The

rbrfk,i,j,c =
− Ea
νc ⋅k0 e Rtemfk,i,j ⋅xbrf αk,i,j,A xbrf βk,i,j,B ⋯ stoichiometric equation is described in Eq. (52). The reaction process
(41) has moderate exothermic heat and no risk of side reaction or decom­
∀i ∈ FE, j ∈ CP, c ∈ COM, k ∈ FP position reaction. The accurate kinetic model has been established in
previous studies. The physical, chemical as well as kinetic parameters
fbrcfk,i,j,c = Vbrfk,i,j ⋅xbrfk,i,j,c
(42) and operating conditions are provided in Table 2.
∀i ∈ FE, j ∈ CP, c ∈ COM, k ∈ FP
(CH3)2CHOH+(CH3CH2CO)2O→CH3CH2COOH+CH3CH2COOCH
At the end of the fault influencing period, the temperature of the
(CH3)2 (52)

361
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

optimization result of proposed method is compared with that of the


literature optimization method for semi-batch reaction process optimi­
4.1.1. Cooling system failure zation considering thermal runaway. Furthermore, divergence criterion
In this example, the proposed inherently safer design approach which is widely applied was taken as the safety constraint in the liter­
developed in this work was applied, which considered the worst sce­ ature approach. The main objective is to maximize the target product
nario of adiabatic conditions. To demonstrate the effectiveness of the production meanwhile to increase the process safety (Wu et al., 2023;
proposed method, the optimization result is compared with a literature Kummer and Varga, 2016). The optimization results from the two
optimization scheme of batch reaction process considering thermal optimization schemes are compared under the same PST condition,
runaway. Therefore, there are two optimization schemes. For the first where the acetic acid production increases from 1987 t/yr to 14,
scheme, optimization is carried with the proposed method. For the 213 t/yr with the proposed method while that increases from 1599 t/yr
second scheme, the aim is to maximize the production and minimize the to 12,447 t/yr with the literature method, respectively, as shown in
F&EI. The temperature limit is widely applied, which is taken as the Fig. 7. It can be seen that the yields of the target product optimized with
thermal runaway criterion in the second scheme. The Pareto solutions of the proposed method are increased by about 15% compared with those
the two optimization methods are shown in Fig. 4. The isopropyl pro­ from the literature optimization approach. Again, the accuracy of the
pionate production increases from 1346 t/yr to 13,927 t/yr for the optimization results was verified and the errors are all within 5%.
literature schemes (Wu et al., 2023; Garcia et al., 1995) and the shortest The optimal parameters are selected by “AHP+EW” TOPSIS, as
PST is calculated with the optimized parameters, which increases from shown in Table 5. The production of the first optimization scheme is
439 s to 1567 s corresponding to an increase of the F&EI value from 91 significantly higher than the second optimization scheme under the
to 120. In the proposed schemes, the isopropyl propionate production same F&EI and PST.
increases from 1512 t/yr to 15,901 t/yr under the same couple of PST Furthermore, from another perspective, the PST is 1321 s in the first
and F&EI as the literature schemes. The production in the proposed optimization scheme and 0 s in the second optimization scheme under
method is increased by about 15% compared with the second optimi­ the same F&EI and the yield, which is as shown in Table 6. The optimal
zation method. It is noted that with the F&EI value increases, both the scheme from proposed method has significantly longer time to respond
yield of isopropyl propionate and the shortest PST increase. Thus, the to various failures compared to the literature optimal scheme.
economy and safety are conflicted while the fault tolerance time is According to the analysis of the above two case studies, it is clear that
longer, which means that more sufficient time is left to take measures. the product yield optimized with the proposed method is significantly
The optimal parameters are obtained based on the method of higher than the yield obtained from literature approach involving
“AHP+EW” TOPSIS from the Pareto solutions of two optimization temperature limit or divergence criterion as thermal runaway criterion
schemes, respectively, which are presented in Table 3. With the same under the same PST. The PST calculated in the proposed method is
F&EI and PST conditions, different optimal process variables of the two longer compared to the literature optimal scheme under the same yield.
optimization schemes lead to different yields of the target product and The inherently safer design approach is applicable to different failure
the yield of the proposed optimization scheme is significantly improved modes of cooling failure, cooling rate slowdown and stirring rate drop,
compared to the literature optimization scheme. etc. Furthermore, the proposed method is also expected to be applied in
Since orthogonal collocation on finite elements (OCFE) approach is continuous production processes to obtain better applicability in the
applied in the proposed method for model description and the accuracy future work. This is also a very challenging job for researchers.
verification of the results is shown in Fig. 5. It is observed that the error
is almost within 5% for the location of the critical cooling failure point 5. Conclusion
and the shortest PST.
In this work, an inherently safer optimal design method based on PST
4.1.2. Stirring system failure for batch/semi-batch chemical reactor was proposed to maximize the
In the above section, the fault mode of cooling failure is considered. product yield and minimize degree of the F&EI. The process safety time
Moreover, the proposed method can also be applied to different modes, was introduced as thermal runaway criterion to optimize the process
such as failure of refrigeration circulating pump, partial blockage of variables. The appropriate scheme was obtained by using the method of
cooling system or stirring out of control, which result in the overall heat “AHP+EW” TOPSIS. To illustrate the validity of the proposed approach,
transfer coefficient decrease. In Fig. 6, for the abnormal scenarios of two case studies of strong exothermic reaction systems under specific
cooling rate slowdown and stirring rate decline, the yields of isopropyl kinetics conditions are carried out. In the first example, considering the
propionate optimized with the proposed method are significantly cooling failure scenario, the product yield optimized with the proposed
increased by about 34% and 30% compared to that obtained from the method is increased by about 15% compared with the scheme obtained
literature optimization method, respectively. from literature approach involving temperature limit as thermal
runaway criterion. Furthermore, considering the abnormal scenarios of
4.2. Acetic anhydride hydrolysis reaction in semi-batch reactor cooling rate slowdown and stirring rate decline, the yield of isopropyl
propionate can be significantly increased by about 34% and 30% by the
In this work, acetic anhydride hydrolysis reaction is considered in developed optimization strategy, respectively. In the second example,
the second case study. The reaction is described in the stoichiometric with the condition of cooling failure, the yield of the target product
equation shown in Eq. (53). The physical, chemical as well as kinetic obtained from the proposed method can also be increased by about 15%
parameters and operating conditions are given in Table 4. A series of compared with that from the literature approach for semi-batch reaction
experimental investigations has been carried out to access the reactor process. Moreover, under the same yield, the optimal scheme of this
behavior, including kinetic parameters and operating characteristic, etc work can extend the PST from 0 s to 1321 s compared with the literature
(Westerterp et al., 2014; Hirota et al., 2010). The feeding mode is scheme, which significantly increases the inherent safety of the process.
considered to be water dosed into acetic anhydride and the conversion The inherent safer design method offers an optimal process scheme to
of the substrate is assumed to be no less than 80%. Besides, there is only achieve trade-off between economy and safety, which also responds to
feeding stage with in the chemical reaction process. the thermal runaway risk risen by unexpected failures based on PST.

(CH3 CO)2 O + H2 O→2CH3 COOH


(53) Declaration of Competing Interest

Similarly, under the abnormal condition of cooling failure, the The authors declare that they have no known competing financial

362
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

interests or personal relationships that could have appeared to influence Kähm, W., Vassiliadis, V.S., 2018. Stability criterion for the intensification of batch
processes with model predictive control. Chem. Eng. Res. Des. 138, 292–313.
the work reported in this paper.
https://doi.org/10.1016/j.cherd.2018.08.017.
Kähm, W., Vassiliadis, V.S., 2018. Thermal stability criterion integrated in model
Acknowledgments predictive control for batch reactors. Chem. Eng. Sci. 188, 192–207. https://doi.org/
10.1016/j.ces.2018.05.032.
Kalantarnia, M., Khan, F., Hawboldt, K., 2009. Dynamic risk assessment using failure
This work is supported by the National Natural Science Foundation assessment and Bayesian theory. J. Loss Prev. Process Ind. 22 (5), 600–606. https://
of China (NO. 52204208). doi.org/10.1016/j.jlp.2009.04.006.
Kalantarnia, M., Khan, F., Hawboldt, K., 2010. Modelling of BP Texas City refinery
accident using dynamic risk assessment approach. Process Saf. Environ. Prot. 88 (3),
Appendix A. Supporting information 191–199. https://doi.org/10.1016/j.psep.2010.01.004.
Kletz, T.A., 1985. Inherently safer plants. Plant/Oper. Prog. 4, 164–167. https://doi.org/
Supplementary data associated with this article can be found in the 10.1002/prsb.720040311.
Kummer, A., Varga, T., 2016. Feeding trajectory optimization in fed-batch reactor with
online version at doi:10.1016/j.psep.2023.01.007. highly exothermic reactions. Comput. Chem. Eng. 98, 1–11. https://doi.org/
10.1016/j.compchemeng.2016.12.008.
References Kummer, A., Varga, T., 2018. Completion of thermal runaway criteria: two new criteria
to define runaway limits. Chem. Eng. Sci. 196, 277–290. https://doi.org/10.1016/j.
ces.2018.11.008.
Ahadh, A., Binish, G.V., Srinivasan, R., 2021. Text mining of accident reports using semi-
Kummer, A., Varga, T., Nagy, L., 2020. Semi-batch reactor control with NMPC avoiding
supervised keyword extraction and topic modeling. Process Saf. Environ. Prot. 155,
thermal runaway. Comput. Chem. Eng. 134, 106694 https://doi.org/10.1016/j.
455–465. https://doi.org/10.1016/j.psep.2021.09.022.
compchemeng.2019.106694.
Amine, D., Lamiae, V.H., Dimitri, L., Lionel, E., 2020. Early detection and diagnosis of
Li, X., Qi, M., Jiu, J., Zhao, D.F., Wang, X.T., 2019. Research on safety requirements of
thermal runaway reactions using model-based approaches in batch reactors.
critical safety instrumented function in coke oven gas to methanol process using
Comput. Chem. Eng. 140, 2. https://doi.org/10.1016/j.
aspen simulation. J. Loss Prev. Process Ind. 60, 253–263. https://doi.org/10.1016/j.
compchemeng.2020.106908.
jlp.2019.05.011.
AIChE, 1994. Dow’s Fire and Explosion Index Hazard Classification Guide, 7th ed.
Li, Y., Wang, H., Bai, K., Chen, S.M., 2021. Dynamic intelligent risk assessment of
AIChE, New York.
hazardous chemical warehouse fire based on electrostatic discharge method and
Barnard, G., Creel, W., 2016. Impacts of process safety time on layer of protection
improved support vector machine. Process Saf. Environ. Prot. 145, 425–434. https://
analysis. Process Saf. Prog. 34 (4), 383–388. https://doi.org/10.1002/prs.11759.
doi.org/10.1016/j.psep.2020.11.012.
Biegler, L.T., 2010. Nonlinear Programming Concepts, Algorithms, and Applications to
Martino, D.Y., Duque, E.S., Reniers, G., Cozzani, V., 2021. Making the chemical and
Chemical Processes. SIAM, Philadelphia.
process industries more sustainable: innovative decision-making framework to
Borza, S., Petrescu, V., 2016. The Olt River pollution monitoring, using spatial analysis,
incorporate technological and non-technological inherently safer design (ISD)
analityc hierarchy process and technique for order preference by similarity methods.
opportunities. J. Clean. Prod. 296, 126421 https://doi.org/10.1016/j.
Process Saf. Environ. Prot. 101, 9–18. https://doi.org/10.1016/j.psep.2016.01.002.
jclepro.2021.126421.
Bosch, J., Kerr, D.C., Snee, T.J., Strozzi, F., Zaldivar, J.M., 2004. Runaway detection in a
Ni, L., Cui, J., Jiang, J.C., Pan, Y., Wu, H., Shu, C.M., Wang, Z.R., Mou, S.J., Shi, N., 2020.
pilot-plant facility. Ind. Eng. Chem. Res. 43 (22), 7019–7024. https://doi.org/
Runaway inhibition of styrene polymerization: a simulation study by chaos
10.1021/ie049540l.
divergence theory. Process Saf. Environ. Prot. 135, 294–300. https://doi.org/
Castillo-Landero, A., Aburto, J., Sadhukhan, J., Martinez-Hernandez, E., 2022. A process
10.1016/j.psep.2020.01.015.
modularity approach for chemical process intensification and inherently safer
Pasha, M., Zaini, D., Shariff, M.A., 2017. Inherently safer design for heat exchanger
design. Process Saf. Environ. Prot. 168, 54–66. https://doi.org/10.1016/j.
network. J. Loss Prev. Process Ind. 48, 55–70. https://doi.org/10.1016/j.
psep.2022.09.054.
jlp.2017.04.002.
CCPS, 2007. Guidelines for Safe and Reliable Instrumented Protective Systems, Center
Strozzi, F., Zaldivar, J.M., 1994. A general-method for assessing the thermal-stability of
for Chemical Process Safety. American Institute of Chemical Engineers, New York,
batch chemical reactors by sensitivity Calculation based on lyapunov exponents.
NY.
Chem. Eng. Sci. 49, 2681–2688. https://doi.org/10.1016/0009-2509(94)E0067-Z.
Chen, Q., Ni, L., Jiang, J.C., Parker, T., Chen, Z.Q., Chen, Z., Jiang, W., Wang, Q.S., 2020.
Thakur, A.K., Gupta, S.K., Kumar, R., Banerjee, N., Chaudhari, P., 2022. Multi-objective
Inhibition of exothermic runaway of batch reactors for the homogeneous
optimization of an industrial slurry phase ethylene polymerization reactor. Int. J.
esterification using nano-encapsulated phase change materials. Appl. Therm. Eng.
Chem. React. Eng. https://doi.org/10.1515/ijcre-2021-0196.
178, 115531 https://doi.org/10.1016/j.applthermaleng.2020.115531.
Toulouse, C., Cezerac, M., Cabassud, M., Lann, M.V.Le, Casamatta, G., 1996.
Chen, Q., Ni, L., Jiang, J.C., Wang, Q.S., 2021. Modeling of runaway inhibition in batch
Optimisation and scale-up of batch chemical reactors: Impact of safety constraints.
reactors using encapsulated phase change materials. Renew. Energy 170, 387–399.
Chem. Eng. Sci. 51, 2243–2252. https://doi.org/10.1016/0009-2509(96)00081-4.
https://doi.org/10.1016/j.renene.2021.01.132.
Ubrich, O., Srinivasan, B., Lerena, P., Bonvin, D., Stoessel, F., 2001. The use of
Eini, S., Abdolhamidzadeh, B., Reniers, G., Rashtchian, D., 2015. Optimization procedure
calorimetry for on-line optimisation of isothermal semi-batch reactors. Chem. Eng.
to select an inherently safer design scheme. Process Saf. Environ. Prot. 93, 89–98.
Sci. 56, 5147–5156. https://doi.org/10.1016/S0009-2509(01)00183-X.
https://doi.org/10.1016/j.psep.2014.05.002.
Varga, T., Abonyi, J., 2010. Novel method for the determination of process safety Time.
Gao, X., Raman, A.A.A., Hizaddin, H.F., Bello, M.M., Buthiyappan, A., 2021. Review on
Chem. Biochem. Eng. Q. 34 (3), 283–293. 〈https://hrcak.srce.hr/59065〉.
the inherently safer design for chemical processes: past, present and future. J. Clean.
Wang, H.Z., Khan, F., Ahmed, S., Imtiaz, S., 2016. Dynamic quantitative operational risk
Prod. 305, 127154 https://doi.org/10.1016/j.jclepro.2021.127154.
assessment of chemical processes. Chem. Eng. Sci. 142, 62–78. https://doi.org/
Gao, Z.J., Miao, J.C., Zhao, J.F., Mesri, M., 2021. Comprehensive economic analysis and
10.1016/j.ces.2015.11.034.
multi-objective optimization of an integrated gasification power generation cycle.
Wang, L.P., Yan, F., Wang, F., Li, Z.J., 2021. FMEA-CM based quantitative risk
Process Saf. Environ. Prot. 155, 61–79. https://doi.org/10.1016/j.
assessment for process industries—⋅A case study of coal-to-methanol plant in China.
psep.2021.09.007.
Process Saf. Environ. Protect. 149, 299–311. https://doi.org/10.1016/j.
Garcia, V., Cabassud, M., Lann, M.V.Le, Pibouleau, L., Casamatta, G., 1995. Constrained
psep.2020.10.052.
optimization for fine chemical productions in batch reactors. Chem. Eng. J. 59,
Wang, R., Wang, Y.F., Gundersen, T., Wu, Y., Feng, X., Liu, M.X., 2021. A multi-objective
229–241. https://doi.org/10.1016/0923-0467(94)02949-0.
optimization method for industrial park layout design: the trade-off between
Guo, Z.C., Bai, W.S., Chen, Y.J., Wang, R., Lin, H., Wei, H.Y., 2016. An adiabatic criterion
economy and safety. Chem. Eng. Sci. 235, 116471 https://doi.org/10.1016/j.
for runaway detection in semibatch reactors. Chem. Eng. J. 288, 50–58. https://doi.
ces.2021.116471.
org/10.1016/j.cej.2015.11.065.
Westerterp, K.R., Lewak, M., Molga, E.J., 2014. Boundary diagrams safety criterion for
Hirota, H.W., Rodrigues, R.B., Sayer, C., Giudici, R., 2010. Hydrolysis of acetic
liquid phase homogeneous semibatch reactors. Ind. Eng. Chem. Res. 53, 5778–5791.
anhydride: non-adiabatic calorimetric determination of kinetics and heat exchange.
https://doi.org/10.1021/ie500028u.
Chem. Eng. Sci. 65, 3849–3858. https://doi.org/10.1016/j.ces.2010.03.028.
Wu, Y., Ye, H.T., Dong, H.G., 2023. A multi-objective optimization for batch chemical
Janošovský, J., Rosa, I., Vincent, G., Šulgan, B., Variny, M., Labovská, Z., Labovský, J.,
reaction processes: the trade-off between economy and safety. Chem. Eng. Sci. 265,
Jelemenský, Ľ., 2022. Methodology for selection of inherently safer process design
118231 https://doi.org/10.1016/j.ces.2022.118231.
alternatives based on safety indices. Process Saf. Environ. Prot. 160, 513–526.
Ye, H.T., Gao, W., Dong, H.G., Bi, M.S., 2022. An inherently safer development approach
https://doi.org/10.1016/j.psep.2022.02.043.
for thermally coupled distillation sequences: Application in hazardous chemical
Jensen, N., Jørgensen, S.B., 2007. Taking credit for loss control measures in the plant
separation. Process Saf. Environ. Prot. 160, 786–802. https://doi.org/10.1016/j.
with the likely loss fire and explosion index (LL-F&EI). Process Saf. Environ. Prot. 85,
psep.2022.02.044.
51–58. https://doi.org/10.1205/psep06001.
Zavala, V.M., Flores-Tlacuahuac, A., Vivaldo-Lima, E., 2005. Dynamic optimization of a
Jiang, J.C., Wu, H., Ni, L., Zou, M.Y., 2018. CFD simulation to study batch reactor
semi-batch reactor for polyurethane production. Chem. Eng. Sci. 60, 3061–3079.
thermal runaway behavior based on esterification reaction. Process Saf. Environ.
https://doi.org/10.1016/j.ces.2005.01.020.
Prot. 120, 87–96. https://doi.org/10.1016/j.psep.2018.08.029.
Zhang, L.L., Wu, J.S., Zhang, J., Su, F., Bian, H.F., Li, L., 2022. A dynamic and integrated
Jiang, J.J., Yang, J.Z., Jiang, J.C., Pan, Y., Yu, Y., Zhou, D., 2016. Numerical simulation
approach of safety investment decision-making for power grid enterprises. Process
of thermal runaway and inhibition process on the thermal polymerization of styrene.
Saf. Environ. Prot. 162, 301–312. https://doi.org/10.1016/j.psep.2022.04.007.
J. Loss Prev. Process Ind. 44, 465–473. https://doi.org/10.1016/j.jlp.2016.10.017.
Zhu, J.X., Lin, H., Bai, W.S., Zhu, Z.X., Zhang, B., Wei, H.Y., 2022. A design framework
for optimized economic and inherently safe operation conditions for isoperibolic

363
Y. Wu et al. Process Safety and Environmental Protection 171 (2023) 353–364

semi-batch reactors. Process Saf. Environ. Prot. 168, 166–179. https://doi.org/


10.1016/j.psep.2022.09.048.

364

You might also like