Journal of Hazardous Materials 189 (2011) 509–518

Contents lists available at ScienceDirect

Journal of Hazardous Materials

journal homepage: www.elsevier.com/locate/jhazmat

Comparison study on qualitative and quantitative risk assessment methods for

urban natural gas pipeline network
Z.Y. Han ∗ , W.G. Weng
Center for Public Safety Research, Department of Engineering Physics, Tsinghua University, Beijing 100084, China

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, a qualitative and a quantitative risk assessment methods for urban natural gas pipeline net-
Received 28 August 2010 work are proposed. The qualitative method is comprised of an index system, which includes a causation
Received in revised form 20 February 2011 index, an inherent risk index, a consequence index and their corresponding weights. The quantitative
Accepted 21 February 2011
method consists of a probability assessment, a consequences analysis and a risk evaluation. The outcome
Available online 26 February 2011
of the qualitative method is a qualitative risk value, and for quantitative method the outcomes are indi-
vidual risk and social risk. In comparison with previous research, the qualitative method proposed in
Keywords:
this paper is particularly suitable for urban natural gas pipeline network, and the quantitative method
Risk assessment
Natural gas pipeline
takes different consequences of accidents into consideration, such as toxic gas diffusion, jet flame, fire
Safety management ball combustion and UVCE. Two sample urban natural gas pipeline networks are used to demonstrate
these two methods. It is indicated that both of the two methods can be applied to practical application,
and the choice of the methods depends on the actual basic data of the gas pipelines and the precision
requirements of risk assessment.
Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.

1. Introduction system, which is based on the basic data of gas pipeline net-
work. The basic data includes pipeline length, flow rate, population
Natural gas is currently one of the most important sources of density, external interference, etc. The outcome of the qualitative
energy. In the European Union, more than 20% of the total energy method is a qualitative risk value. The quantitative method assesses
consumption is from natural gas [1]. However, the accidents caused risk by numerical simulation, including a quantitative calculation of
by gas pipeline rupture are great threats to urban public safety. Due possibilities and consequences of different accidents. The numer-
to the physical and chemical characteristics of natural gas and the ical simulation is based on the physical and chemical models as
complexity of gas pipeline network topology, accidents occurring well as the physiological dose–effect relationship of human. The
in gas pipeline are quite different from other industrial accidents. outcomes of quantitative method are individual risk and social risk
The broken pipelines may cause numerous fatalities and domino [2].
effects, and the derived disasters may cause casualties and prop- Recently, more and more authorities start to be aware of the
erty losses. In 2004, fourteen people were killed and more than security problems in natural gas transmission pipelines. For quali-
two hundred people were injured due to the explosion of a natu- tative assessment, numerous approaches were proposed including
ral gas factory in Belgium. In Paraguay, a conflagration caused by Analytic Hierarchy Process (AHP), Fuzzy logic method (FL), Fault
gas leakage resulted in more than 250 deaths in 2004. In 2009, an Tree Model (FTM), Event Tree Analysis (ETA) and Data Envel-
explosion caused by gas leakage induced the greatest conflagra- opment Analysis (DEA), etc. [3–8]. However, these approach
tion in Moscow ever since the Second World War. Thus, it is very focus only on identifying the causes of the accidents, and fail to
important to assess the risk of natural gas pipeline network. assess the risk. Besides, the Muhlbauer Pipeline Risk Assessment
Risk assessment is defined as a mathematical function of the Method is an approach which using an index system to assess
probability and consequence of an accident. The target of risk the risks of long-distance transmission pipelines outside city. It
assessment is to identify potential accidents, analyse the causa- has been used for more than ten years and works well [9]. But
tions and evaluate the effects of the risk reduction measures [2]. considering the differences of the located environment between
The qualitative and quantitative methods are two aspects of risk long-distance transmission pipeline and urban gas pipeline, this
assessment. The qualitative method assesses risk by using an index existing approach is not suitable for the risk assessment of urban
gas pipeline. For quantitative assessment, many approaches have
been applied to analyse and assess the risk of natural gas pipelines
∗ Corresponding author. Tel.: +86 10 62796323; fax: +86 10 62792863. [10–16]. However, these methods fail in general analysing the
E-mail address: [email protected] (Z.Y. Han). consequences of various accidents, such as the harms of toxicity,

0304-3894/$ – see front matter. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.jhazmat.2011.02.067
510 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518

failure, ground movement, other and unknown causes including

Nomenclature incorrect operation and maintenance error; secondly, the occurring
probabilities of gas pipeline ruptures depend on the basic charac-
F(t) unreliability function, (103 km year)−1 teristic parameters of the pipelines, for instance, pipeline diameter,
i accident scenario operation pressure, operation flow rate, service life, wall thickness,
L total length of pipelines, m and depth of cover; besides, the consequences of the accidents rely
N time of pipelines in used, year on the environmental conditions including the pipelines location,
n total number of objects to be evaluated population density, economy conditions and so on. So the index
M times of the accidents system constituted in this paper can be divided into three groups,
m total number of observation data i.e. an inherent risk index, a causation index and a consequence
R(t) reliability function, (103 km year)−1 index. The first two indices are used to describe the probabilities
r failure rate function, year−1 of gas pipeline accidents including inducements and frangibility,
t time of the pipelines has been used, year while the last index is used to describe the potential losses. Fig. 1
Yi (k) sequence of data shows the framework of this index system.
ωi weight of accident scenario index i
 ij (k) grey correlation coefficient 2.2. Index classification
 distinguishing coefficient
ij (k) sequence of deviation of the reference sequence The causation index shows how the natural gas pipeline acci-
dents occur. The failure probabilities of the pipelines vary according
to different environmental conditions and inducement mechanism.
So the second level indices of the causation index include an exter-
combustion and explosion. In fact, these accidents have different
nal interference index, a corrosion index, a material defect index,
physical and chemical effects, which cause different harms to peo-
an operation error index and a ground movement index. The third
ple and influence the spatial distributions of individual risk and
level indices include a construction digging index, a ground works
social risk in different ways.
index, a public works index, a agriculture index, a drainage works
In this paper, a qualitative and a quantitative risk assessment
index, a pitting index, a galvanic corrosion index, a stress corrosion
methods are proposed. The qualitative method is comprised of
index, a internal corrosion index, a material failure index, a incor-
an index system, which includes many indices and their corre-
rect operation index, a incorrect operation index, a maintenance
sponding weights. The indices are used to describe the factors
error index, a dike break index, a flood index, a landslide index, a
that influence the probabilities and consequences of gas pipeline
river erosion index and a geologic activity index. All of these indices
accidents, and the weights describe the significance of the corre-
are the factors that may cause pipeline rupture.
sponding indices. The quantitative method consists of a probability
The inherent risk index describes the vulnerability of the
assessment, a consequences analysis and a risk assessment, in
pipeline, so the failure probabilities of the pipelines are determined
which the analysis for the consequences of different accidents is
by the inherent conditions of operation and installation. The sec-
included.
ond level indices of the inherent risk index include an operation
In the second and third sections of this paper, the qualitative
index and an installation index. The operation index represents the
method and the quantitative method are presented, respectively.
operating conditions of the pipeline such as flow rate and pressure,
Section 4 demonstrates the applications of these two methods by
and the installation index represents the construction conditions of
using a small and a large sample urban natural gas pipeline net-
the pipeline such as depth and diameter. So the third level indices
works. Conclusions are given in the last section.
include an operation flow rate index, an operation pressure index, a
wall thickness index, a pipeline diameter index, a service life index
2. Qualitative risk assessment method and a depth of cover index. All of these indices are determined by
the basic engineering designs of the pipelines.
This section proposes a qualitative risk assessment method The consequence index demonstrates the harms and damages of
which is suitable for urban gas pipeline network. The qualitative gas pipeline accidents that include casualties and property losses.
method presented here contains an index system, including three Accidents occurred in different environments may lead to differ-
first level indices such as a causation index, an inherent risk index ent degrees of losses. The second level indices of the consequence
and a consequence index. The selection of indices is based on the index include a leakage hazard index and an effect hazard index.
characteristics of gas pipeline accidents according to the statistical The leakage hazard index represents the fatalness of the substance
analysis of historical accident data. And the Reliability Engineering in the transmission pipelines. The effect hazard index represents
Theory and the Grey Correlation Theory are used to calculate the the potentially devastating losses in the influence area when trans-
weights of the corresponding indices. Combining the evaluation mission pipelines rupture. The third level indices include a medium
values of the indices with the corresponding weights, the criterion hazard index, an environment hazard index, a population density
and standard of risk management can be constituted. index, a property distribution index and an other urban lifeline
distribution index. These indices reveal the consequences of gas
2.1. Index selection pipeline accidents, such as casualties, property losses and domino
effects in the urban lifeline system.
Based on the historical accidents database, especially the statis-
tical analysis on how accidents occurred and how many incidents 2.3. Weight calculation
were induced, the indices can be selected and the weights can be
calculated. In this paper, the database of European Gas pipeline Inci- The weights of indices indicate the differences in the impact
dent data Group (EGIG) is used for the constitution of the qualitative of the inducements to accidents and the potential loss when acci-
risk assessment index system [17]. This database indicates the dents occur. For the first level indices, the weight of each index is
characteristics of gas pipeline accidents: firstly, the causations of usually set to 1, which means the causation index, the inherent risk
gas pipeline accidents can be classified into many categories, such index and the consequence index make equal contribution to the
as external interference, corrosion, construction defect/material total risk. For the second and third level indices, in order to insure
 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518 511

Fig. 1. The framework of qualitative risk assessment index system.

the accuracy and practicability of the index system, the evaluation where L is the total length of pipelines, N is the time of pipelines
of weights is based on the real data of gas pipeline network such in used, M is the times of the accidents. For the accident scenario i,
as operation information, environment information and statistical the failure rate can be expressed as the following equation:
analysis of historical accident data. And the approaches of weight
ln(1 − Fi (t)) ln(1 − Mi /(N · L))
calculations are the Reliability Engineering Theory and the Grey ri (t) = ri = − =− (3)
t t
Correlation Theory.
so the weight of accident scenario index i can be estimated by the
following equation:
2.3.1. The weights of the causation index
r
The weight of the causation index can be calculated by using ωi = i (4)
the Reliability Engineering Theory. Historical records of accidents ri
can be obtained according to the report of EGIG [17], which gives i
the data on categories and number of accidents during certain time
where ωi is the weight of accident scenario index i.
periods. For the natural gas pipelines being used, the unreliability
function of pipelines can be estimated in the following equation
based on the reliability engineering theory [18,19]: 2.3.2. The weights of the inherent risk and consequence index
For the inherent risk index and consequence index, the infor-
F(t) = 1 − R(t) = 1 − e−rt (1) mation needed for calculating the weights includes the operation
flow rate, operation pressure, depth of cover, wall thickness, as well
where F(t) is the unreliability function (103 km year)−1 , R(t) is the as the population density and economy conditions in the affected
reliability function, r is the failure rate function, t is the time of the area. Usually, it is difficult to obtain the sufficient data [20]. Based on
pipelines has been used. The unreliability of pipelines is also can be this status, the weights of the inherent risk index and consequence
expressed as the following equation: index can be achieved by the Grey Correlation Theory [21–23].
The Grey Correlation Theory aims at finding out the comparabil-
M ity among a mass of dissimilar data. For the risk assessment of
F(t) = (2)
N·L urban gas pipeline network, changes of different kinds of basic data
512 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518

Fig. 2. The framework of quantitative risk assessment method.

influence the evaluation value of risk differently. By using the Grey where ij (k) is the grey correlation coefficient.  is the distinguish-
Correlation Theory, the quantity representing the contribution of ing coefficient,  ∈ (0, 1). Generally  is taken as 0.5 [21]. ij (k)
each kind of basic data to the total risk value can be evaluated. It is the sequence of deviation of the reference  sequence Yi (k) from
is clearly that these quantities of contribution can be used as the the sequence Yj (k) for comparison. ij (k) = Yi (k) − Yj (k), max =
   
weights of the index system in qualitative risk assessment. max Yi (k) − Yj (k), min = min Yi (k) − Yj (k), then the grey corre-
The fundamental principle of grey correlation analysis is to lation grade is an average of the grey correlation coefficients and is
determine whether a relationship exists among a series of data defined as:
based on the degree of similarity among the geometric shapes of
1 
n m
the data series’ curves. Similar curves indicate a stronger correla-
ri = ij (k) (7)
tion between these series of data. The weights are determined from mn
j=1 k=1
the grey correlation grades, which measure the degrees of similar-
ity among sequences [22]. According to this method, the actual data and the weights can be calculated by the followed equation:
depicted as [Xi (k)] for every index of each gas pipeline forms the
foundation for the calculation of weights. Then data preprocess- 
n

ing converts the original sequence to a comparable sequence using ωi = ri / ri (8)

dimensionless transformation [24]. Usually, each series is normal- i=1

ized by dividing the data in the original series by their average as

follows: 2.4. Risk assessment
 
Xi (1) Xi (2) Xi (m) According to the actual operation and environment information,
Yi = , , · · ·, = (Yi (1), Yi (2)· · ·Yi (k))
X1 X2 Xm the value of each index can be evaluated by comparing with the
characteristic parameter of each pipeline and the criterion of risk
1
n
assessment. Then, based on the index system constituted above,
Xk = Xi (k) (5)
n integrating the values of indices and the corresponding weights,
i=1 the total risk value can be expressed as the following equation:
where i = 0, 1, 2, . . ., n and k = 1, 2, . . ., m. n is the total number of

Q
objects to be evaluated, and m is the total number of observation At = atq ωtq
data, respectively. Yi (k) is the vector quantities of grey matrix used
q=1
for the calculation. (9)
 T
After the data have been preprocessed, a grey correlation coeffi- A= At
cient is determined by using the preprocessed sequence as follows:
t=1
min + max where A is the total risk value, At is the risk value of the first level
ij (k) = (6)
ij (k) + max index t, T is the number of the first level index and in this paper
 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518 513

Table 1
Causation index evaluation for pipeline 1 of small urban natural gas pipeline network.

Third level index Weights Characteristic information Value

Construction digging 0.272 Frequent 7

Ground works 0.062 Extraordinary frequent 10
Public works 0.062 Not frequent 4
Agriculture 0.062 Scarcely 1
Drainage works 0.056 Frequent 7
Pitting 0.098 Corrosion resistant 5
Galvanic corrosion 0.017 Electric potential: −100 5
Stress corrosion 0.007 Pressure drop: 1.728 kPa 10
Internal corrosion 0.026 Without H2 S 1
Construction defect 0.096 Duration of service: 20 year 10
Material failures 0.064 Pipeline length: 113.44 m 4
Incorrect operation 0.044 Possible 5
Maintenance error 0.064 Possible 5
Dike break 0.001 None 0
Flood 0.015 None 0
Landslide 0.045 None 0
River erosion 0.005 None 0
Geologic activity 0.004 None 0
Risk value 5.662

Table 2
Inherent risk index evaluation for pipeline 1 of small urban natural gas pipeline network.

Third level index Weights Characteristic information Value

Operation flow rate 0.154 10.05 kg/s 0.99

Operation pressure 0.169 9 kPa 10
Wall thickness 0.193 2.2 cm 2.34
Pipeline diameter 0.158 100 cm 2.50
Service life 0.165 20 year 10
Depth of cover 0.161 0.4 m 2.35
Risk value 4.687

Table 3
Consequence index evaluation for pipeline 1 of small urban natural gas pipeline network.

Third level index Weights Characteristic information Value

Substance hazard 0.228 Comparative denseness 7

Environment condition 0.272 Windy 7
Population density 0.167 Extraordinary denseness 10
Property distribution 0.167 Extraordinary denseness 10
Other urban lifeline distribution 0.166 Extraordinary denseness 10
Risk value 8.500

Table 4
Quantitative risk evaluation for pipeline 1 of small urban natural gas pipeline network.

Characteristic information Risk assessment process

−5
Pipeline diameter 100 mm ϕd = 7.5 × 10 Failure rate 3.25 × 10−3
Depth of cover 0.4 m 2.54 Death probability percentage 0.000307
Wall thickness 2.2 mm 1 Fatality probability unit 2.67
Population density Town 18.77 Release rate 1.005 kg/s
Prevention method All methods 0.91 Radius with the individual risk of 10−6 10.6323 m

T = 3, atq is the risk value of the third level index q, ωtq is the weight 3.1. Probability assessment
of index q, Q is the number of the third level index in the first level
index t. Probability assessment focuses on the probabilities of acci-
dents, which depend on the failure assumption caused by different
inducements [11]. The failure probability of a pipeline varies signif-
icantly with design factors, construction conditions, maintenance
3. Quantitative risk assessment method
techniques and environmental conditions, etc. Based on the sta-
tistical analysis of historical accidents database, the failure rate of
This section proposes an integrated quantitative risk assess-
the pipeline for each accident scenario can be estimated by the
ment method, in which the analysis of the consequences of various
modified empirical formula [10]:
accidents such as toxicity diffusion, combustion and explosion are
combined. This method consists of a probability assessment, a 
consequence analysis and a risk evaluation [25]. Fig. 2 shows the ϕ= ϕk Kk (a1 , a2 , ...) (10)
framework of this method. k
514 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518

Table 5
Results of the qualitative risk assessment method of small urban natural gas pipeline.

No. Assessment value No. Assessment value No. Assessment value No. Assessment value

1 225.595 25 235.917 49 279.551 73 211.023

2 329.475 26 249.565 50 276.210 74 220.366
3 314.020 27 237.143 51 272.260 75 225.385
4 278.211 28 242.839 52 278.160 76 226.105
5 218.592 29 227.692 53 235.388 77 221.078
6 265.223 30 253.106 54 253.030 78 260.217
7 262.255 31 232.865 55 257.359 79 325.824
8 225.768 32 236.718 56 227.888 80 314.809
9 213.086 33 251.192 57 231.606 81 346.996
10 255.115 34 229.434 58 265.274 82 263.566
11 220.641 35 236.161 59 227.697 83 244.560
12 186.962 36 286.050 60 262.682 84 251.789
13 220.626 37 242.483 61 204.902 85 266.587
14 204.705 38 256.828 62 245.141 86 244.347
15 217.197 39 281.543 63 205.712 87 269.541
16 233.622 40 244.401 64 248.102 88 357.036
17 235.529 41 222.636 65 266.891 89 400.915
18 268.330 42 259.059 66 254.952 90 406.621
19 244.904 43 226.183 67 318.454 91 353.450
20 273.987 44 252.848 68 322.634 92 359.966
21 287.005 45 220.806 69 213.691 93 402.884
22 270.875 46 227.126 70 227.027 94 242.891
23 286.284 47 252.982 71 210.961 95 270.611
24 278.108 48 254.612 72 219.882

Table 6
Results of the quantitative risk assessment method (individual risk) of small urban natural gas pipeline network.

No. Radius with individual No. Radius with individual No. Radius with individual No. Radius with individual
risk of 10−6 risk of 10−6 risk of 10−6 risk of 10−6

1 10.63 25 14.41 49 5.49 73 10.40

2 34.86 26 12.15 50 4.40 74 10.40
3 34.82 27 12.17 51 4.40 75 2.37
4 10.54 28 12.15 52 12.04 76 11.97
5 10.58 29 12.15 53 2.97 77 11.97
6 10.58 30 9.42 54 2.97 78 2.04
7 10.58 31 9.42 55 2.96 79 34.62
8 3.06 32 9.42 56 11.02 80 34.56
9 3.06 33 3.06 57 7.24 81 34.56
10 3.06 34 9.43 58 4.40 82 2.35
11 10.59 35 4.48 59 5.49 83 2.35
12 6.53 36 4.40 60 5.49 84 2.93
13 6.53 37 7.89 61 8.08 85 2.93
14 6.53 38 7.89 62 5.58 86 2.93
15 6.53 39 5.49 63 5.58 87 2.93
16 8.08 40 5.49 64 2.98 88 32.64
17 8.08 41 7.24 65 4.37 89 32.64
18 5.58 42 4.40 66 10.55 90 34.39
19 5.58 43 5.49 67 34.82 91 30.88
20 5.58 44 5.49 68 34.82 92 32.64
21 14.39 45 7.24 69 2.67 93 32.64
22 14.35 46 7.24 70 13.36 94 2.97
23 14.35 47 7.24 71 13.36 95 2.95
24 14.39 48 7.24 72 2.07

where ϕ is the expected failure rate per unit pipeline length composed of a gas release rate calculation, a physics effects calcula-
(1/year·km), ϕk is the basic failure rate per unit length of pipeline tion, a fatality probability unit calculation and a casualty percentage
(1/year·km), Kk is the correction function associated with the failure calculation [27].
causes, a1 , a2 , . . . are the variables of the correction function, and Since the damages of gas such as toxicity, heat and pressure
the subscript k denotes the failure causes such as external inter- depend on the amount of leakage, the gas release rate should be cal-
ference, construction defects, corrosion, ground movement and culated firstly. The calculation approach of gas release rate adopted
others. in this paper is the Hole Model, which has been widely used in the
literatures as a general computational method of quantitative risk
3.2. Consequence analysis analysis [28].
After the gas release rate is obtained, the harms of the acci-
Consequence analysis focuses on the physical effects of the dents are analysed. All the physical effects of the aforementioned
accidents that are harmful to human beings, including toxic gas dif- physical processes have quantitative descriptions. If the gas leak-
fusion, jet flame, fire ball combustion and UVCE (unconfined vapour age does not catch fire, the harm of leakage is related to the toxicity
cloud explosion), etc. The consequence analysis in this paper is of gas and the concentration distribution around the region where
 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518 515

Fig. 3. Results of the qualitative risk assessment method for small urban natural gas pipeline network.

the pipeline ruptures [30–32]. If the gas leakage catches fire at the 3.3. Risk evaluation
leakage source, the fire becomes diffusion flame (i.e., jet flame) and
poses a threat to people near the leakage source. The risk of jet The results of quantitative risk assessment take the forms of
flame can be quantitatively determined by measuring the thermal individual risk and social risk [2,26], which quantitatively describe
radiation flux [33]. If the gas leakage catches fire after it forms a per- the death probability and critical level. The spatial distribution of
sistent vapour cloud but not intensively mixing with air, a fire ball individual risk can be calculated by integrating the failure rate, the
ensues. The thermal radiation flux of fire ball combustion can be probability of each accident scenario and the spatial distribution of
conservatively estimated according to the corresponding fire ball death probability for the pipeline [10–12]. Social risk can be cal-
combustion model [34]. If the gas leakage catches fire after it inten- culated by integrating the spatial distribution of death probability
sively mixes with air and forms a persistent vapour cloud, it will and population density within the hazard area [2,10,37]. Social risk
lead to a significant flash fire or unconfined vapour cloud explosion. can be shown in FN-curve [2,25].
For the convenience of calculation, the feasible approach for the
calculation of explosion overpressure is the modified flash model 4. Applications and comparison
according to TNT equivalent weight method [32]. The recorded data
provided by the API indicated that the probabilities of the accidents To validate the proposed methods, a small sample urban natural
after the pipeline ruptures are 0.8 for toxic gas diffusion, 0.1 for jet gas pipeline network containing 95 pipelines and a large sample
flame combustion, 0.06 for fire ball combustion and 0.04 for UVCE urban natural gas pipeline network containing 5421 pipelines are
[29]. presented for demonstration. In order to assess the risk using the
To quantitatively describe the level of damages, the fatality proposed quantitative risk assessment method, the initial accident
probability unit is defined as a mathematical function based on the hypothesis is assumed that the failure in one pipeline causes an
physiological dose–effect relationship between the dose of harms orifice with one-third of the pipeline diameter. For the regional
such as toxicity, heat or pressure and the effects of the recipient urban gas pipeline network, the experimental conditions can be
such as deaths or injuries. The fatality probability unit of thermal assumed as following: (1) the recommended values of the exposure
radiation can be calculated according to the damage referring to the time for people referred to overpressure and thermal radiation is
third degree burns [35,36]. The fatality probability unit of explosion 30 s [36]. (2) The toxicity of the gas can be omitted due to a regional
overpressure can be calculated according to the damage referring urban gas pipeline network carrying nontoxic gas.
to the lung haemorrhage [37]. Then, using the fatality probability For the qualitative risk assessment method, firstly, the index
unit, the death probability percentage can be obtained by looking system must be constituted based on the operation information,
up the corresponding table [36]. the environment information and the statistical analysis of histori-
516 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518

Fig. 4. Results of the quantitative risk assessment method (individual risk) for small urban natural gas pipeline network.

cal accident data. Then the value of each index can be evaluated evaluation process are shown in Table 4. According to the character-
according to the actual operation, the environment information istic information, the failure rate can be calculated as 3.25 × 10−3 .
and the criterion. The result of qualitative risk assessment method So the death probability percentage of the location where the
can be given by evaluating each index, respectively and integrating individual risk is 10−6 should be 10−6 /(3.25 × 10−3 ) = 0.000307. By
the corresponding weight according to Eq. (9). For the quantitative looking up the corresponding table, the fatality probability unit can
method, the result is given by evaluating the probability and con- be obtained, which is 2.67. The release rate can be calculated as
sequence as discussed in Section 3. Detailed process of calculation 1.005 kg/s. Then, based on the physical models of jet flame, fire ball
can be found in Ref. [25]. combustion and UVCE, the radius with the individual risk of 10−6
In this paper, the results of qualitative risk assessment method can be calculated as 10.6323 m.
are compared to those of the quantitative method to validate the The results of risk assessment are shown in Tables 5 and 6
feasibility and practicability of these two methods. Since the indi- and displayed by geographic information system as shown in
vidual risk of 10−6 has been set as a guideline for the boundary Figs. 3 and 4. Table 5 shows the results of the qualitative risk assess-
between the broadly acceptable and the tolerable regions for the ment method. It is indicated that the average risk value for all of
public, the radius with individual risk of 10−6 is used as the pre- the 95 pipelines is 258.718, and there are 37 pipelines with risk
sented result of the quantitative assessment method in this paper. value higher than the average value. Table 6 gives the radius with
the individual risk of 10−6 as the result of the quantitative assess-
4.1. Application in small sample urban gas pipeline network ment method. It is evident that longer radius means higher risk. The
average value for all of the 95 pipelines is 10.82 m, and there are 28
A small sample urban natural gas pipeline network used here pipelines with risk value higher than the average value. Figs. 3 and 4
is a part of a whole network in a city, which includes 95 pipelines. illustrate the results of qualitative and quantitative risk assessment
For the qualitative risk assessment method, the index system is methods for the small urban natural gas pipeline network, respec-
constructed as shown in Fig. 1. tively. In these illustrations, the risk values are equally divided into
Based on the methods proposed above, the risk value can be four levels that are depicted by the color of red, orange, yellow and
assessed. Taking pipeline 1 as an example. For the qualitative blue. It can be seen that the pipelines of each level of risk in Fig. 3 are
method, the characteristic information and the evaluation value of almost the same as those in Fig. 4, especially for the levels colored
each index is shown in Tables 1–3, including the causation index, in red and blue.
the inherent risk index and the consequence index, respectively. The analysis result indicates that both the qualitative and quan-
For the quantitative method, the characteristic information and the titative risk assessment methods can be applied to risk assessment
 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518 517

Fig. 5. Results of the qualitative risk assessment method for large urban natural gas Fig. 6. Results of the quantitative risk assessment method (individual risk) for large
pipeline network. urban natural gas pipeline network.

of real small urban natural gas pipeline network, and the evaluation the risk value. The evaluation results of these two methods for the
results of the qualitative and quantitative risk assessment meth- applications in two sample urban gas pipeline networks are simi-
ods are similar. For risk management, the pipelines with risk value lar. It is indicated that these two methods are feasible and scientific,
higher than the average value should be stressed to carry out risk and can be used in practical application. It is clear that the choice
reduction measures. of the methods depends on actual basic data and precision require-
ments of the risk assessment according to the methods described
4.2. Application in large sample urban gas pipeline network above.

A large sample urban natural gas pipeline network used here

5. Conclusions
is a main part of a whole network in a city, which includes 5421
pipelines. The results of risk assessment are displayed by geo-
In this paper, a qualitative and a quantitative risk assessment
graphic information system as shown in Figs. 5 and 6. Fig. 5 shows
methods for urban natural gas pipeline network are proposed. For
the result of the qualitative risk assessment method. It is indicated
the qualitative method, the selection of the indices is based on
that the average risk value for all of the 5421 pipelines is 271.965,
the statistical analysis of accidents database, and the calculation of
and there are 2065 pipelines with risk value higher than the aver-
corresponding weights is according to the Reliability Engineering
age value. Fig. 6 is the result of the quantitative risk assessment
Theory and the Grey Correlation Theory. For the quantitative risk
method, i.e. the radius with individual risk of 10−6 . It is indicated
assessment method, the possibilities and consequences of different
that the average value for all of the 5421 pipelines is 33.01 m, and
accidents are analysed and integrated. Two sample urban natural
there are 1526 pipelines with risk value higher than the average
gas pipeline networks are chosen to validate the presented meth-
value. Like Figs. 3 and 4, it also can be seen that the pipelines of
ods. The risk assessment results of these two methods are similar. It
each level of risk in Fig. 5 are almost the same as those in Fig. 6. The
is indicated that these two methods can be used in practical appli-
analysis result indicated that both qualitative and quantitative risk
cation for risk assessment of urban natural gas pipeline network,
assessment methods can be applied to risk assessment of real large
and the choice of the methods depends on actual basic data and the
urban natural gas pipeline network.
precision requirements of the risk assessment.
In general, the qualitative risk assessment method is more effec-
tive and intuitive than the quantitative method. Comparing with
the quantitative method, the qualitative method takes more fail- Acknowledgement
ure causations into consideration, but it also needs much more
basic data. For the quantitative method, it integrates more details of This paper was supported by National Natural Science Founda-
induced accident consequences. So the analysis result of the quanti- tion of China (Grant No. 70871069). The authors deeply appreciate
tative method is more accurate, but it takes more time in calculating the support.
518 Z.Y. Han, W.G. Weng / Journal of Hazardous Materials 189 (2011) 509–518

References [19] G.C. Bello, S. Avogadri, Offshore pipeline system—how to decide the optimal
redundancies using reliability engineering, Reliab. Eng. 3 (1982) 89–99.
[1] H. Montiel, J.A. Vilchez, J. Amaldos, Historical analysis of accidents in the trans- [20] A.M. Bartenev, B.E. Gelfand, G.M. Makhviladze, J.P. Roberts, Statistical analysis
portation of natural gas, J. Hazard. Mater. 51 (1996) 77–92. of accidents on the Middle Asia-Centre gas pipelines, J. Hazard. Mater. 46 (1996)
[2] S.N. Jonkman, P.H. van Gelder, J.K. Vrijling, An overview of quantitative risk 57–69.
measures for loss of life and economic damage, J. Hazard. Mater. 99 (2003) [21] J.L. Deng, Introduction to the grey system theory, J. Grey Syst. 1 (1989) 1–24.
1–30. [22] Deng. Julong, Control problems of grey systems, Syst. Control Lett. 1 (5) (1982)
[3] E. Cagno, F. Caron, M. Mancini, F. Ruggeri, Using AHP in determining the prior 288–294.
distributions on gas pipeline failures in a robust Bayesian approach, Reliab. Eng. [23] H.S. Gau, C.Y. Hsieh, C.W. Liu, Application of grey correlation method to evaluate
Syst. Saf. 67 (2000) 275–284. potential groundwater recharge sites, Stoch Environ. Res. Risk Assess. 20 (2006)
[4] A.S. Markowski, M.S. Mannan, Fuzzy logic for piping risk assessment, J. Loss 407–421.
Prev. Process Ind. 22 (2009) 921–927. [24] C.P. Fung, Manufacturing process optimization for wear property of fiber rein-
[5] S. Bonvicini, P. Leonelli, G. Spadoni, Risk analysis of hazardous materials trans- forced polybutlene terephthalate composite with grey relation analysis, Wear
portation: evaluating uncertainty by means of fuzzy logic, J. Hazard. Mater. 62 254 (2003) 298–306.
(1998) 59–74. [25] Z.Y. Han, W.G. Weng, An integrated quantitative risk analysis method for nat-
[6] Y. Dong, D. Yu, Estimation of failure probability of oil and gas transmission ural gas pipeline network, J. Loss Prev. Process Ind. 23 (2010) 428–436.
pipelines by fuzzy fault tree analysis, J. Loss Prev. Process Ind. 18 (2005) [26] TNO Purple Book, Guideline for Quantitative Risk Assessment, Committee for
83–88. the Prevention of Disasters, The Netherlands, 1999, Chapter 6.
[7] David Hawdon, Efficiency, performance and regulation of the international gas [27] Z.Y. Han, W.G. Weng, An overview of quantitative risk analysis methods for
industry—a bootstrap DEA approach, Energy Policy 31 (2003) 1167–1178. natural gas pipelines, J. China Saf. Sci. 19 (2009) 1–13, in Chinese.
[8] P.K Dey, An integrated assessment model for cross-country pipelines, Environ. [28] H. Montiel, J.A. Vılchez, J. Casal, Mathematical modeling of accidental gas
Impact Assess. Rev. 22 (2002) 703–721. releases, J. Hazard. Mater. 59 (1998) 211–233.
[9] W. Kent, Muhlbauer, Pipeline Risk Management Manual, Gulf Publishing Com- [29] A PI RP 581, Recommended Practice 581, Risk Based Resource Document, API,
pany, 1992. Washington, DC, 2000, November.
[10] Y.D. Jo, B.J. Ahn, A method of quantitative risk assessment for transmission [30] A. Mazzoldi, T.H.J.J. Colls, CFD and Gaussian atmospheric dispersion models: a
pipeline carrying natural gas, J. Hazard. Mater. A123 (2005) 1–12. comparison for leak from carbon dioxide transportation and storage facilities,
[11] Y.D. Jo, D.A. Crowl, Individual risk analysis of high-pressure natural gas J. Atmos. Environ. 42 (2008) 8046–8054.
pipelines, J. Loss Prev. Process Ind. 21 (2008) 589–595. [31] Z. Yang, X.H. Li, J.B. Lai, Analysis on the diffusion hazards of dynamic leakage of
[12] S. Sklavounos, F. Rigas, Estimation of safety distances in the vicinity of fuel gas gas pipeline, J. Reliab. Eng. Syst. Saf. 92 (2007) 47–53.
pipelines, J. Loss Prev. Process Ind. 19 (2006) 24–31. [32] Centre for Chemical Process Safety of the American Institute of Chemical
[13] J. Arnaldos, J. Casal, H. Montiel, Design of a computer tool for the evaluation Engineering, Guidelines for Evaluation the Characteristics of Vapour Cloud
of the consequences of accidental natural gas releases in distribution pipes, J. Explosion, Flash Fires, and BLEVE, New York, 1994.
Loss Prev. Process Ind. 11 (1998) 135–148. [33] A PI RP 521, Recommended Practice 521, third ed., API, Washington, DC, 1990,
[14] P.L. Metropolo, A.E.P. Brown, Natural gas pipeline accident consequence anal- November.
ysis, in: 3rd International Conference on Computer Simulation in Risk Analysis [34] J.H. Luo, M. Zheng, X.W. Zhao, C.Y. Huo, L. Yang, Simplified expression for esti-
& Hazard Mitigation, Sintra, Portugal, 2004, pp. 307–310. mating release rate of hazardous gas from a hole on high-pressure pipelines, J.
[15] Joel Krueger, Duncan Smith, A practical approach to fire hazard analysis for Loss Prev. Process Ind. 19 (2006) 362–366.
offshore structures, J. Hazard. Mater. 104 (2003) 107–122. [35] CPQRA, Guideline for Chemical Process Quantitative Risk Analysis, The Center
[16] A. Jaffee, A. Suardina, Jeff McPhate Jr., Fire and explosion assessment on oil and for Chemical Process Safety of American Institute of Chemical Engineers, 1988,
gas floating production storage offloading (FPSO): an effective screening and ISBN 0-8169-0402-2.
comparison tool, Process Saf. Environ. Prot. 87 (2009) 147–160. [36] A.H. Rausch, N.A. Eisenberg, C.J. Lynch, Continuing Development of the Vulner-
[17] European Gas Pipeline Incident Data Group, Gas Pipeline Incidents, 7th Report ability Model (VM2), Department of Transportation, United States Coast Guard,
(1970–2007). Washington, DC, 1977, Report No. CG-53-77, February.
[18] M. Mesarovic, Reliability theory of large systems as applied to electrical engi- [37] IGE, Steel pipeline for high pressure gas transmission, IGE Code TD/1, fourth
neering, Elektrotehnika 24 (1981) 3–10. ed., Communication, vol. 1670, 2001.