probability analysis of damage to offshore pipeline …docs.trb.org/prp/13-0923.pdfyutao liu 1 1...

Yutao Liu 1

Probability Analysis of Damage to Offshore Pipeline by Ship Factors 1

2

Yutao Liu 3

Ph.d. Candidate 4

School of Naval Architecture, Ocean and Civil Engineering 5

Shanghai Jiaotong University 6

800 Dongchuan Road, Shanghai, P.R. China 7

TEL: 021- 54748155, FAX: 8621-62933163 8

Email: [email protected] 9

10

11

Hao HU ∗ 12 School of Naval Architecture, Ocean and Civil Engineering 13


800 Dongchuan Road, Shanghai, P. R. China 15

TEL: 8621-62933091, FAX: 8621-62933163 16

E-mail: [email protected] 17

18

19

Di Zhang 20

School of Naval Architecture, Ocean and Civil Engineering 21


800 Dongchuan Road, Shanghai, P. R. China 23

TEL: 86-13472652951, FAX: 8621-62933163 24

E-mail: [email protected] 25

26

Submitted to the Transportation Research Board 92th Annual Meeting 27

for Presentation and Publication 28

Submission date: July 31, 2011 29

30

Word Count: 4247(text) + 3250(13 figures) = 7497 words 31 32

∗ Corresponding Author

TRB 2013 Annual Meeting Paper revised from original submittal.

Yutao Liu 2

ABSTRACT 33

The transport of hydrocarbons by offshore pipeline is threatened by the rapid expansion of pipe 34

networks and the increasing frequency of maritime activities. Risk management is thus necessary to 35

manage and prevent ship-related hazardous events that may damage offshore pipelines. Probability 36

analysis is the key to assessing the risk associated with ship operations on offshore pipelines, and 37

decision making in managing that risk. Bayesian Network (BN) models are proposed in this paper to 38

determine the probability of anchor damage and trawling damage to subsea pipelines. The BN 39

models are developed by integrating directed acyclic graphs, and three computational methods 40

(Boolean operation, standard and historical statistical analysis, and fuzzy set theory) to elicit 41

marginal probability tables and conditional probability tables. A case study illustrates the utilization 42

of two BN-related functions – probability prediction and probability updating – to determine final 43

probabilities of damage to a subsea pipeline. The results of the analysis support risk ranking and risk 44

reducing decisions associated with maritime operations in the area of offshore pipelines. 45

46

KEY WORDS 47

Offshore Pipeline; Ship Factor; Bayesian Network; Damage Analysis; Probability Analysis 48


Yutao Liu 3

INTRODUCTION 49

The transportation of hydrocarbons by subsea pipelines is highly efficient and convenient, while 50

requiring minimal cost. As a result, offshore pipelines are gradually becoming the primary mode of 51

transportation of oil and gas at sea. China currently has 3,000 kilometers of installed offshore 52

pipelines, and is planning to increase that length threefold in the next decade. Some hazards 53

associated with operating offshore pipelines include leakage, rupture and even bursts that result in 54

interruption to transportation and production of hydrocarbons, require clean-up operations, and can 55

cause catastrophic health, environment and safety accidents (1, 2). Consequently, maintaining the 56

integrity of offshore pipeline transportation networks is of vital importance to a nation’s economy 57

and peoples’ lives. Historical data (3) illustrate that a large number of accidents to offshore pipelines 58

were caused by impact, ship anchoring and corrosion, as shown in Fig.1. 59

60

FIGURE 1 Accident breakdown of offshore steel pipeline. 61

Obviously, the majority of anchoring and impact accidents are associated with passing 62

ships, which can be categorized as damage to offshore pipeline by “ship factors”. With the rapid 63

extension of offshore pipeline networks and the increasing frequency of maritime activities, it can 64

be reasonably expected that accidents to offshore pipelines by passing ships will become more 65

frequent. Therefore, risk assessment for ship factor hazards is necessary, and the results from such 66

assessments could support risk ranking and then serve as the main basis to judiciously divide 67

resources for inspection, maintenance and protection among different pipeline networks, pipeline 68

segments, or related assets. 69

In addition to the scoring-type algorithm method (4), many qualitative or 70

semi-quantitative methods, such as the analytic hierarchy process (AHP), fuzzy logic, and neural 71

networks, have been utilized in risk assessment models for onshore and offshore pipelines (5-9). 72

These models give relative values of the assessment results, which could support risk ranking but 73

fail in judging whether the risk is acceptable to the local community. Therefore, current research 74

ranges from qualitative schemes to quantitative probabilistic systems. Also, physical models in 75

association with probabilistic methodology are utilized to analyze failures of offshore pipelines 76

caused by accidental external loads in research and incorporated into standards (10-13). These 77

models offer absolute assessment results but require complex analytical procedures. In addition, the 78

parameters used in these models are not frequently updated when new data become available for 79

statistical analysis of the operation period of offshore pipelines. Fault tree (FT) analysis has also 80

been shown to be an effective method in probabilistic failure analysis and has been employed in 81

21%

30%26%

7%

6%5%

1%1% 1% 1% 1%

AnchoringImpactCorrosionStructuralMaterialNatural HazardConstructionMaintenanceHuman errorOperation problemsOther


Yutao Liu 4

pipeline engineering (14, 15); however, FT analysis is best limited to modeling simple static 82

systems, with complex systems being modeled through other procedures. 83

Bayesian network (BN) has a much more flexible structure than FT, and can fit a broad 84

range of accident scenarios. This paper presents a BN model for analyzing the probability of 85

damage to offshore pipelines by ship factors. The remainder of this paper is structured as follows: 86

Section 2 describe offshore pipelines and the hazards of anchor damage and trawling damage. Next, 87

the establishment of BN models is presented in Section 3. Section 4 introduces the case study of the 88

offshore pipeline from Pinghu Oil-Gas Field to Shanghai. Finally, conclusion and comparison of 89

BN and FT are provided in Section 5. 90

91

DESCRIPTION OF OFFSHORE PIPELINE AND ACCIDENT SCENARIO 92

As shown in Fig.2, the section of offshore pipelines considered in this paper is the middle portion of 93

a subsea pipeline, viz., the section of pipeline away from both the platform and the shoreline 94

(e.g., >500 m from the platform and >300m from the coastline). This middle section of pipeline is 95

minimally affected by offshore platform operations and onshore activities. 96

97 FIGURE 2 Illustration of a typical offshore pipeline. 98

Anchor and trawling accidents (i.e., “ship factor”) to offshore pipelines occur frequently 99

and, furthermore, previous studies (3) show that the frequency of containment loss caused by anchor 100

and trawling accidents were 37% (19/52) and 44% (23/52) respectively. Therefore, this paper is 101

limited to modeling anchor and trawling accidents. 102

In order to establish a complete model for probability analysis, the accident scenarios 103

may consider the following factors. 104

Passing ship: engineering ship (supply boat, crane ship, etc.); transport ship (tanker, 105

commercial ship, cruise ship, etc.); fishing vessel 106

Hazardous activities: emergency anchoring (anchor weight, anchor shape); bottom trawling 107

(type of fishing net, depth of casting net, draw force) 108

Offshore pipeline characteristics: type (steel pipeline, flexible or umbilical); water depth; 109

embedment depth; diameter, wall thickness, coating thickness 110

Possible consequence to pipeline: impact damage, hooking damage 111


Yutao Liu 5

For this paper, the energy (in kilojoules) that a “ship factor” accident imposes onto a 112

subsea pipeline is the hazard to be analyzed, and “damage” is the release of hydrocarbon from a 113

distressed pipeline exposed to hazard. As pipeline characteristics (e.g., diameter, thickness, and 114

material properties) vary significantly among pipelines and, therefore, energy was selected to define 115

damage to facilitate application of the method to a wide range of situations. 116

117

ESTABLISHMENT OF BN MODELS 118

Bayesian Network 119

The representation of BN is a directed acyclic graph, in which the nodes represent variables, and 120

arcs signify direct causal relationships between the linked nodes. If a node doesn’t have any parents 121

(i.e. root node), the node contains a marginal probability table which expresses the prior probability. 122

Otherwise, the node contains a conditional probability table that specifies how strongly the linked 123

nodes influence each other (16). 124

A BN methodology is used either to predict the probability of unknown variables or to 125

update the probability of known variables. The processes of probability prediction and probability 126

reasoning are all based on Bayes’ theorem. In the predictive analysis, conditional probabilities of the 127

form | are calculated, indicating the occurrence probability of a 128

particular accident given the occurrence or non-occurrence of a certain primary event. According to 129

the conditional independence and the chain rule, BN finally represents the joint probability 130

distribution of variables , … , included in the network as 131

| 1

where are the parents of in the BN, and reflects the properties of the BN. In the 132

updating analysis, those of the form | are evaluated, showing the 133

posterior probability of a particular event given new information, called evidence E. The evidence is 134

usually operational data including occurrence or non-occurrence of the accident or primary events 135

(17): 136

|, ,

∑ , 2

137

Directed Acyclic Graphs for Anchor Damage and Trawling Damage 138

As shown in Fig.3, anchor damage is defined as damage to an offshore pipeline by anchor impact, 139

where the anchor impact occurs when a ship passes above an offshore pipeline, the ship’s engine 140

fails and an anchor is deployed under emergency conditions. Two primary factors affecting impact 141

energy are 1) the water depth which greatly influences the impact probability of the anchor, and 2) a 142

pipeline’s coating which often reduces the severity of the impact (18). Both impact probability and 143

impact energy are integrated to determine anchor damage to a pipeline. 144

As shown in Fig.4, trawling damage to offshore pipelines is mainly caused by fishing 145

nets hooking on to a pipeline. Fishing boats cast nets to perform bottom trawling, and the nets 146

become caught on subsea pipelines. The nets entangled on the pipeline are dragged by the fishing 147

boats resulting on large tensile loads being placed on the pipeline and causing damage to the 148


Yutao Liu 6

pipeline. Entanglement of a fishing net on a pipeline is a function of net-casting depth, water depth, 149

and the pipeline coating (which mitigates the attachment of nets to a pipeline) (19). 150

151 FIGURE 3 Directed acyclic graph of anchor damage. 152

153

FIGURE 4 Directed acyclic graph of trawling damage. 154

155

Elicitation of Marginal Probability Table (MPT) and Conditional Probability Table (CPT) 156

The elicitation of MPT and CPT is complex due to the large amount of judgments required to fully 157

quantify these relationships in a BN model. They are elicited in the following three ways in this 158

study. 159

Table 1 illustrates the conversion of Boolean operations (disjunction and conjunction 160

) into CPTs, which is applied only to the events are considered binary (with two states: 0 and 1). 161

This conversion process has been introduced in some researches about the comparison between 162

fault tree and BN (20, 21). 163

Historical data and standards (12, 19) offer an alternative method to educe MPT and 164

CPT. One example is the marginal probability of engine. According to an investigative report (22), 165

the failure rate of engines is 2E-05 per hour. As engines have rapidly improved and their reliability 166

increased we choose 2E-06 as a conservative estimate of engine failure rate as a ship passes over an 167

offshore pipeline thereby requiring an anchor to deployed under emergency conditions. In addition, 168


Yutao Liu 7

Table 2 shows the example of the CPT of anchor damage, which refers to the proposed damage 169

classification used for bare steel pipes given by DNV-RP-F107 (13). 170

TABLE 1 CPTs of Ship Passing and Anchor 171

Boolean operation

engineering ship passing no transport ship passing no passing no fishing vessel passing no passing no passing no passing no ship passing

(passing = 1, no = 0)1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 1

Boolean operation

ship passing passing no engine fail work fail work anchor

(anchor = 1, no = 0)1 0 0 0 0 1 1 1

TABLE 2 CPT of Anchor Damage 172

impact energy no

(0 kJ) low

(< 2.5 kJ) medium

(2.5 – 10 kJ)high

(10 – 20 kJ) tremendous

(>20 kJ) impact impact no impact no impact no impact no impact no

Dam

age no 1 1 0 1 0 1 0 1 0 1

minor 0 0 1 0 0 0 0 0 0 0 moderate 0 0 0 0 0.5 0 0.25 0 0 0 major 0 0 0 0 0.5 0 0.75 0 1 0

Minor damage: Damage neither requiring repair, nor resulting in any release of hydrocarbons. Moderate damage: Damage requiring repair, but not leading to release of hydrocarbons. Major damage: Damage leading to release of hydrocarbons, water, etc.

Finally, expert judgment is another important way of eliciting MPT and CPT. However, 173

experts often prefer to linguistic judgment than probabilistic description, like “safe” and “unsafe”. 174

The integration of fuzzy set theory can help domain experts to elicit MPT and CPT in an efficient 175

manner. For example, in this paper we define the buried depth of the offshore pipeline using three 176

fuzzy numbers, , defined over universe of discourse where each subset represents an depth grade; 177

, , , as illustrated in Fig.6. According to experts’ opinion, 178

the triangular membership function for each fuzzy number is represented by the 179

following set of equations, 180 0, 0 ,

,, ,

, , ,

,, ,

, , ,

0, ,

3

In this example, it can be seen that for a pipeline with buried depth 0.75 meters the 181 membership values are 0.75, and 0.25 and zero for . The fuzzy set 182

representing the buried depth can be written as a MPT, as shown in Fig. 5. 183


Yutao Liu 8

184

FIGURE 5 MPT of buried depth educing by fuzzy set theory. 185

Water depth can similarly be modeled as a fuzzy set, and buried depth and water depth 186

are modeled as fuzzy sets to complete the evaluation of anchor (Fig. 3) and trawling (Fig. 4) 187

damage. 188

The elicitation of CPT could also be carried out from expertise by integrating fuzzy set 189

theory with the AHP method. This method has been described extensively by (21), and is beyond 190

the scope of this paper. 191

192

CASE STUDY 193

Offshore Pipeline from Pinghu Oil-Gas Field to Shanghai 194

An offshore pipeline with a length of 386.2 km transports natural gas from Pinghu Oil-Gas Field to 195

Shanghai. The pipeline is divided into 6 segments in order to get evaluation results with significant 196

difference. The essential data to create the BN model are listed in the Table 3. As shown in Fig.6, the 197

1st and 2nd segments of the pipeline are located in the Zhoushan fishing ground, where fishing 198

vessels appear frequently. A national coastal shipping line with busy transport ships runs through the 199

3rd segment, and the 5th and 6th segments are mainly threatened by engineering ships servicing the 200

Pinghu Oil-Gas field. In addition, the offshore pipeline is set on the continental shelf of the East 201

China Sea, and thus water depth gradually increases from west to east. Finally, among all the 202

pipeline sections, only the 6th one enjoys a covering of submarine soil with a depth of about one 203

meter. 204

TABLE 3 Basic Data of the Offshore Pipeline 205

Pipeline segment

length (km)

water depth (m)

buried depth (m)

frequency of passing ship (per year) engineering transport fishing

1 25.2 0-10 0 50 50 1500 2 75.2 10-20 0 50 150 1000 3 70 30-50 0 50 2000 500 4 105.6 50-80 0 100 300 300 5 48.5 80-100 0 700 150 100 6 61.7 100-110 1 1000 100 50


Yutao Liu 9

206 FIGURE 6 Offshore pipeline from Pinghu Oil-Gas Field to Shanghai. 207

208

Probability Prediction for the Pipeline Segments 209

The procedure for estimating the probability of anchor damage, , is described by the equation: 210

· 4

In the Eq. (4), (1, …, n) means the type of ship. is the frequency of the type of ships 211

passing by per year. And calculated by the BN model reflects the probability of anchor damage 212

once an -type ship passing through. In this paper, we only consider engineering ship, transport 213

ship and fishing vessel. The probability of trawling damage, , is described by the equation: 214 · 5

where is the frequency of fishing vessels running through the offshore pipeline per year, and 215

expresses the probability of trawling damage by one fishing vessel. 216

In this paper, the BN model is analyzed using HUGIN 7.6 (23). By Eqs. (4) and (5), we 217

could obtain probability prediction results of anchor damage and trawling damage, as shown in 218

Table 4. Both types of damage are classified by the damage degrees of minor damage, moderate 219

damage and major damage, which have been briefly introduced in the Table 2. 220

TABLE 4 Probability Prediction Results of Anchor Damage and Trawling Damage 221

Pipeline segment

anchor damage trawling damage minor moderate major minor moderate major

1 0.00E+00 3.58E-04 1.59E-03 1.50E-01 4.37E-02 1.79E-03 2 0.00E+00 2.01E-04 9.03E-04 4.38E-02 1.11E-02 4.52E-04 3 0.00E+00 2.80E-04 1.36E-03 5.00E-02 1.46E-02 5.95E-04 4 0.00E+00 6.80E-05 3.05E-04 1.07E-03 2.66E-04 1.08E-05 5 0.00E+00 7.44E-05 2.77E-04 4.71E-05 1.17E-05 4.76E-07 6 1.83E-04 8.25E-05 1.09E-04 9.80E-08 2.43E-08 9.90E-10

222

Analyses of Main Causes 223

The main causes of accident scenarios could be analyzed by the probability updating process, as 224


Yutao Liu 10

introduced in the section 2. Here the example of anchor damage to the 1th pipeline segment is 225

utilized to illustrate the analysis process. After inputting the MPTs elicited from the Table 3 (prior 226

probabilities of primary events), then given major anchor damage occurs (top event occurs), the 227

updating of the marginal probabilities of primary events ( | ) could be 228

calculated by HUGIN 7.6. As shown in Table 5, it can be concluded that “engine fail”, “buried depth 229

is shallow”, “fishing vessel passes” and “water depth is shallow” generally have the greatest 230

contribution to the occurrence of the top event, i.e., they are the main causes of the “major anchor 231

damage” accident scenario. 232

TABLE 5 Prior and Posterior Probabilities of Major Anchor Damage to the 1th Pipeline 233

Segment 234

Primary event (per hour) Prior Posterior Primary event Prior Posteriorengineering ship passes 0.0057 0.0249 buried depth is shallow 1 1 transport ship passes 0.0057 0.0268 buried depth is medium 0 0 fishing vessel passes 0.1712 0.9485 water depth is shallow 0.7 0.8033 engine fails 2.0E-05 1 water depth is medium 0.3 0.1967 235

Risk Ranking and Risk Reducing Measures 236

In order to compare the damage probability and the risk of the relevant hazards, an individual 237

ranking from 1 (very low probability) to 5 (very high probability) is proposed (24), as shown in 238

Table 6. Note, however, that the limits given in the Table 6 may be adjusted to comply with case 239

specific requirements. Then the probability analysis results of the offshore pipeline (Table 4) could 240

be ranked. For example, Table 4 shows the probability of major anchor damage to the 1st segment 241

is 1.59E-03, which intervenes between 1E-03 and 1E-02. As a result, we consider it is high 242

according to the ranking in Table 6. Then the probability ranking for major damage and 243

minor/moderate damage are provided, with the analysis results of the main causes for several risky 244

pipeline segments (see Table 7). 245

Balance between safety and cost requires using different strategies for different damage: 246

the occurrence of major damage should be eliminated, while the minor and moderate damage 247

should be reduced as low as reasonably practicable (ALARP). Therefore, the following risk 248

mitigation measures are proposed to address the results presented in Table 7. 249

Risk reducing measures (reduce probability or reduce damage degree) must be taken for the 1st 250

and 3rd segments, so as to practically eliminate major damage to these segments during the 251

pipeline’s lifetime. 252

Risk reducing measures (introduce safe distance or safety areas, introduce extra chaser tug or 253

anchor chain buoys, etc.) are better suited for the 1st segment. As a result, the high probability 254

of major damage and the very high probability of minor/moderate trawling damage could be 255

both decreased by these measures. 256

Since it is impractical to control the frequency of passing ships or the emergency anchoring of 257

transport ships, measures to reduce the damage intensity, such as increasing a pipeline’s 258

concrete coating are most appropriate for the 3rd segment. 259

Furthermore, to evaluate the economic effects of any risk reducing measures, a cost-benefit 260

calculation should be performed. It will be a combination of engineering principles and sound 261

business practices based on economic theory (25), and the combination of BN model with 262

cost-benefit value (CBV) is the major purport of our ongoing research. 263


Yutao Liu 11

TABLE 6 Damage Probability Ranking for One Pipeline 264

Ranking Description Probability (per year) 1 (very low) So low probability that event considered negligible. <1E-05 2 (low) Event rarely expected to occur. 1E-04>1E-05 3 (medium) Event individually not expected to happen, but when

summarized over a large number of pipelines have the credibility to happen once a year.

1E-03>1E-04

4 (high) Event individually may be expected to occur during the lifetime of the pipeline.

1E-02>1E-03

5 (very high) Event individually may be expected to occur more than over during lifetime.

>1E-02

TABLE 7 Probability Ranking and Main Causes analysis for the Offshore Pipeline 265

Segment major damage minor/moderate damage main causes of risky segments anchor trawling anchor trawling

1 high (1) high (2) medium very high (1) “buried depth is shallow”, “fishing

vessel passes”, “water depth is shallow”

(2) “fishing vessel passes”, “buried depth is

shallow”, “water depth is shallow”

(3) “buried depth is shallow”, “transport

ship passes”, “water depth is shallow”

2 medium medium medium very high 3 high (3) medium medium very high 4 medium low low medium 5 medium very low low low 6 medium very low low very low 266

CONCLUSION 267

This paper analyzed the probability of damage to offshore pipelines by passing ships using Bayesian 268

Network. Firstly, anchor damage and trawling damage are identified as the main accident types to be 269

described and analyzed. This is then followed by the creation of BN models. Three methods 270

(Boolean operation, standard and historical statistic, fuzzy set theory) were used to elicit marginal 271

probability table and conditional probability table. Subsequently, the Pinghu-Shanghai offshore 272

pipeline is evaluated by BN models. Two important functions of BN – probability prediction and 273

probability updating - were used to analyze damage probability and find the main cause of damage, 274

respectively. Finally, the evaluation results were combined to support risk ranking and risk 275

reduction measures. 276

The risk assessment framework presented in this paper is of use to oil and gas operators 277

so that the risk of anchor and trawling damage can be effectively managed. The quantification of 278

risk allows an engineered design of pipelines and adjustment of maritime operations so that risk is 279

controlled. The largest concern of operators, related to offshore pipelines, is the disruption of 280

hydrocarbon delivery to the evacuation point. Anchor and trawling damage has previously caused 281

interruptions in hydrocarbon deliveries, and operators could use the BN model to quantify damage 282

frequency, consequences, mitigation measures, and cost of mitigation to achieve a specified risk of 283

trawling and anchor damage. By doing so, the expected loss of hydrocarbon and expected costs of 284

construction (depending on acceptable risk level to the operator) can be determined for establishing 285

budgets for design, construction and installation, and also for operations and maintenances. 286

287

288


Yutao Liu 12

REFERENCES 289

1. Liu, Y. T., H. Hu, and Y. B. Song. Pipeline Integrity Management System Based on Dynamic 290

Risk Assessment. Journal of Shanghai Jiaotong University, Vol.45, No.5, 2011, pp. 687-690. 291

2. Liu Y T and H. Hu. Functional Modeling for an Integrated Pipeline Integrity Management 292

System Using IDEF0 Approach. Proceedings of the 1st International Conference on 293

Transportation Information and Safety, 2011. 294

3. The Update of the Loss of Containment Data for Offshore Pipeline. Publication PARLOC 295

2001. Mott MacDonald Ltd., Offshore Operators Association, U.K. and the Institute of 296

Petroleum, U.K., 2003. 297

4. Muhlbauer, W. K. Pipeline Risk Assessment Manual (Third Edition).Gulf Publishing Co., 298

Houston, TX, U.S., 2003. 299

5. Dey, P. K., S. O. Ogunlana, and S. Naksuksakul. Risk-based maintenance model for offshore 300

oil and gas pipelines: a case study. Journal of Quality in Maintenance Engineering, Vol. 10, 301

No. 3, 2004, pp.169-183. 302

6. Rajani, B.B., Y. Kleiner, R. Sadiq. Translation of pipe inspection results into condition rating 303

using fuzzy synthetic valuation technique. Journal of Water Supply Research and Technology: 304

Aqua, Vol. 55, No.1, pp.11-24. 305

7. Markowski, A. S., M. S. Mannan. Fuzzy Logic for Piping Risk Assessment (pfLOPA). 306

Journal of Loss Prevention in the Process Industries, Vol. 22, 2009, pp.912-927. 307

8. Singh, M., T. Markset. A Methodology for Risk-Based Inspection Planning of Oil and Gas 308

Pipes Based on Fuzzy Logic Framework. Engineering Failure Analysis, Vol. 16, 2009, pp. 309

2098-2113. 310

9. Doremami, N., A. Afshar, A. D. Mohammadi. Hierarchical Risk Assessment in Gas Pipelines 311

Based on Fuzzy Aggregation. Proceedings of 2nd International Conference on Reliability, 312

Safety & Hazard, 2010. 313

10. Mozzola A.. A Probabilistic Methodology for the Assessment of Safety from Dropped Loads 314

in Offshore Engineering. Risk Analysis, Vol. 20, No. 3, 2000, 327-337. 315

11. Yan S. W., and Y. H. Tian. Analysis of Pipeline Damage to Impact Load by Dropped Objects. 316

Transactions of Tianjin University, Vol. 12, 2006, pp. 138-141. 317

12. Risk Assessment of Pipelines Protection. Publication DNV-RP-F107. Det Norske Veritas 318

(DNV), Norway, 2010. 319

13. Bai, Y., and Q. Bai. Subsea Pipelines and Risers. Elsevier Science Ltd., 2005. 320

14. Dong, Y. H., and D. T. Yu. Estimation of Failure Probability of oil and gas transmission 321

pipelines by fuzzy fault tree analysis. Journal of Loss Prevention in the Process Industries, 322

Vol. 18, 2005, pp.83–88. 323

15. Wang Q., and J. P. Zhao. Fuzzy fault tree analysis method on submarine pipelines fault by the 324

third-party damage. Nature Gas Industry, Vol. 28, No. 3, 2008, pp.109-111. 325

16. J. T. Francisco, and H. Manfred. Learning a Causal Model from Household Survey Data by 326

Using a Bayesian Belief Network. In Journal of Transportation Research Board: 327

Transportation Research Record , No. 1836, Transportation Research Board of the 328

National Academies, Washington, D.C., 2003, pp. 29-36. 329

17. Ni D. H., and J. D. Leonard Ⅱ. Markov Chain Monte Carlo Multiple Imputation Using 330

Bayesian Networks for Incomplete Intelligent Transportation Systems Data. In Transportation 331

Research Record: Journal of the Transportation Research Board, No.1935, 2005, pp. 57-67. 332


Yutao Liu 13

18. Bang, S., R. J. Taylor, J. Yu, H. T. Kim. Analysis of anchor mooring lines in cohesive seafloor. 333

In Journal of Transportation Research Board: Transportation Research Record, No. 1526, 334

1996, pp. 47-56. 335

19. Inference between Trawl Gear and pipelines. Publication DNV-RP-F111. Det Norske 336

Veritas, Norway, 2006. 337

20. Khakzad, N., F. Khan, and P. Amyotte. Safety analysis in Process Facilities: Comparison of 338

Fault Tree and Bayesian network approaches. Reliability Engineering and System Safety, Vol. 339

96, 2011, pp.925-932. 340

21. Wang, Y. F., M. Xie, K. M. Ng, Habibullah M S. Probability analysis of offshore fire by 341

incorporating human and organizational factor. Ocean Engineering. Vol. 38, 2011, pp. 342

2042-2055. 343

22. Ship-Modu Collision Frequency, Technica, London, 1987. 344

23. HUGIN Expert software version 7.6, www.hugin.com, 2012. 345

24. Sun, D. J., L. Elefteriadou. Research and implementation of lane-changing model based on 346

driver behavior, In Journal of Transportation Research Board: Transportation Research 347

Record 2161, 2010, pp. 1-10. 348

25. Perrin, J., Jr., and R. Dwivedi. Need for Culvert Asset Management. In Journal of 349

Transportation Research Board: Transportation Research Record, No. 1957, 2006, pp. 350

8-15. 351


probability analysis of damage to offshore pipeline …docs.trb.org/prp/13-0923.pdfyutao liu 1 1...

Documents