Natural gas distribution system: A statistical analysis of accidents data

Accepted Manuscript Natural gas distribution system: A statistical analysis of accidents data Augusto Bianchini, Alessandro Guzzini, Marco Pellegrini, Cesare Saccani PII:

S0308-0161(17)30202-8

DOI:

10.1016/j.ijpvp.2018.09.003

Reference:

IPVP 3752

To appear in:

International Journal of Pressure Vessels and Piping

Received Date: 7 June 2017 Revised Date:

5 June 2018

Accepted Date: 3 September 2018

Please cite this article as: Bianchini A, Guzzini A, Pellegrini M, Saccani C, Natural gas distribution system: A statistical analysis of accidents data, International Journal of Pressure Vessels and Piping (2018), doi: 10.1016/j.ijpvp.2018.09.003. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Natural Gas distribution system: a statistical analysis of accidents data Augusto Bianchinib, Alessandro Guzzinia*, Marco Pellegrinib, Cesare Saccania

RI PT

a. Department of Industrial Engineering, University of Bologna, Viale Risorgimento 2, 40100 Bologna, Italy b. Department of Industrial Engineering, University of Bologna, Via Fontanelle 40, 47121 Forlì, Italy

M AN U

SC

Abstract Natural Gas (NG) distribution is obtained by a really complex system that must ensure safe conditions and avoid human or economic losses. This paper analyzes accidents that occurred between 2004 and 2015 in the United States distribution systems and were recorded in the Pipeline and Hazardous Material Safety Administration (PHMSA) database. Statistical trends are studied; number of accidents, injuries and fatalities are shown and risk indexes are proposed for different accident causes. An average value of 2.09 x 10-5 accidents/km are found in US distribution systems and it is shown that natural events pose the highest risk to distribution systems. Working conditions, such as pressure, pipe diameter and system age are considered in the study, finding that the low pressure and diameter systems account for the greatest number of injuries and fatalities in the case of failure. On the basis of the results, recommendations are given to sector stakeholders. Keywords: Natural Gas safety, Natural Gas distribution system, Natural Gas distribution accidents.

TE D

1. Introduction

EP

Natural Gas (NG) is a primary energy source although an increase of renewable energy production is necessary to reach the objectives set out in the Paris Agreement in 2015 to reduce the global average temperature to well below 2°C with respect to pre-industrial conditions and to limit the temperature increase to 1.5°C above pre-industrial levels. Among fossil fuels, the use of NG should be preferred for several reasons. First of all, the reduced environmental impact. In fact, considering the same level of efficiency, chemical energy can be converted into thermal energy by the combustion of NG ensuring the minimization of CO2 emissions (2.75 kg of CO2 per kg of fuel) thanks to the highest Lower Heat Value (LHV~50000 kJ/kg) and the lowest carbon-hydrogen ratio (i.e. C/H = 1/4 = 0.25) in the fuel composition as reported by [1]. Moreover, NG can also be transported between very long distances within pipelines or inside vessels in the form of Liquefied Natural Gas (LNG) with a cost that depends mainly on the distance and on the installation environment. A maximum value of 33 $/(km x Sm3) has been estimated for offshore transportation, and 11.3 $/(km x Sm3) for LNG transportation [2]. Because of the presence of pressure drops, several compressor stations are installed along the entire distance in order to maintain an average pressure between 70 – 100 bar in the onshore environment. In other words, the importance of NG is evident considering national energy balances despite the declining period due to the financial crisis and the implementation of solutions to improve energy efficiency in industrial and power generation sectors [3]. As reported in [4], NG was the second

AC C

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

*

Corresponding author. Tel: 00390512093403; Fax: 00390512093454. E-mail: [email protected].

ACCEPTED MANUSCRIPT

RI PT

M AN U

SC

Indigenous Russia Norway Algeria Qatar Changes Other Total net production in stocks balances supplies 1525.8 1227.0 1185.0 370.6 227.3 -72.8 -35.6 4427.2 EU-28 Table 1. Europe NG supplies from different sources for 2014, [4]. Values are in TWh/year.

TE D

However, the complexity of the system is due not only to the length of the pipelines but also to the millions of components installed in transportation and distribution systems to ensure correct operation. Failures during service life could imply high risks to human safety and economic damage to property; consequently, being responsible, operators have to plan inspections, maintenance and repair activities to reduce the risk due to failures of their systems [7]. Furthermore, when NG distribution systems are installed in highly populated areas, safety plans and monitoring systems are employed to respectively reduce the consequences and the probability of the failure, and thus the risk. In the case of failure, in fact, consequences such as jet fire, fire ball, Vapour Cloud Explosion (VCE) or gas dispersion depending on boundary conditions as described by several authors can be possible and threaten surrounding people and buildings [8–11]. However, the complexity of managing NG distribution system is due to two reasons: firstly, the highest concentration of buried pipeline is installed in very populated areas; secondly, failure conditions can be caused by several events that can reduce the resistance of the pipeline such as, but not limited to, corrosion, investigated by several studies [12–16], or phenomena that can increase the stresses on the pipeline such as, for example, earth movement [17–23]. It should also be noted that some failures are caused by human activities, for example poorly executed work or carelessness during installation or repair activities. These accidents can also be due to excavation work that, by damaging the pipe, can immediately cause a gas leakage or reduce the pressure capability of the pipe wall with the formation of a dent. Last but not least, material and equipment failures can lead to an unintentional release of NG if the device is defective or not properly used as reported in manufacturer’s datasheet. The causes mentioned above are only some of the possible ones that can threat NG infrastructures. To help in making technical and organizational decisions in order to reduce associated consequences, a historical analysis of past accidents can therefore be performed as reported by [24] in which 131 accidents involving NG transmission and distribution pipes recorded in the MHIDAS (Major Hazard Incident Data Service) database are analyzed. Even if further data are suggested, mechanical failure, explosion and fire are respectively the most frequent causes and consequences. Furthermore, the knowledge of the accident scenario probabilities and of the consequences of a

EP

58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87

source of energy consumed in Europe with 269.6 million tons of oil equivalent (Mtoe) in 2013 with a slight increase (+3.2%) with respect to the previous year while total energy consumption was estimated as equal to 1184 Mtoe with an increase of 0.1%. A similar trend was observed in the United States where a consumption of 701 million toe was estimated by [5], higher than 18% with respect to 2008 during the economic crisis. Only a small part of these consumptions is extracted and consumed locally; in fact, thousands of kilometers of pipelines are installed worldwide to ensure the transportation and distribution: for example, in Europe only 34.5% of the total, around 4427 TWh, is supplied from indigenous production, as reported in table 1, and very branched infrastructures are therefore necessary for the purpose. To ensure complete supply, European and American NG pipeline systems have thus reached a total length of 2,171,000 km [4] and 3,597,000 km [6] respectively, although it is reasonable to expect a further increase in the future.

AC C

45 46 47 48 49 50 51 52 53 54 55 56 57

ACCEPTED MANUSCRIPT

RI PT

SC

M AN U

TE D

Year 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 160 175 201 195 144 177 159 126 120 123 Number of accidents 18 16 19 15 15 14 9 14 15 11 Fatalities Table 2. Number of accidents and fatalities that occurred in Italian NG distribution sector recorded by CIG.

EP

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128

failure can be used to create a multi-attribute decision model for risk assessment in pipelines and for ranking sections of gas pipelines into a risk hierarchy in which environmental and economic factors are also considered in addition to injury and fatality, as proposed and evaluated for a case study by [25]. However, there are few studies in the literature that analyze statistical trends for NG distribution systems as made for the American NG transportation system [26]. It is thus very difficult for Distribution Systems Operators (DSOs) to decide where the resources should be allocated to improve the performance of the system. In 2015, Italian DSOs transported thirty billion Sm3 of gas to thirty-three million end users through the use of almost three hundred thousand kilometers of networks [27]. As described in the annual reports submitted by the Comitato Italiano Gas (CIG), the Italian NG distribution network is safe thanks to the inspection, monitoring and maintenance activities carried out by operators [28]. Data regarding NG accidents that occurred in Italy between 2007 and 2016 have been extracted from annual reports and listed in table 2; the data also include the failures that occurred in the customers’ domestic systems which are not the responsibility of the DSO. Consequently, it is difficult to identify the main causes of failures that occur in the Italian distribution system. More information about NG accidents that occurred in Italian NG networks are present in [29]: however, it should be highlighted that this information was obtained from newspapers and very little technical information is therefore available and so statistical analysis cannot be performed properly. It is therefore clear that no scientific approach can be achieved without data to understand and improve natural gas safety: in fact, to correctly perform a scientific study three stages are required: data analysis, development of mathematical models and experimentation/validation. Consequently, a systematic approach should be defined to record data both near misses and actual accidents to create a leakage database. This makes it possible to obtain useful results.

In 2017, Marcogaz, the Technical Association of the European Natural Gas Industry published two reports about the safety performance of the European NG distribution infrastructure [30,31] in which safety performance indicators and trends are respectively reported. The importance of data is also reported by the fact that in 2011 the Deutscher Verein des Gas-und Wasserfaches (DVGW) took up a recommendation by the joint national and regional committee to transform the damage and accident statistics which had been collected since 1980 into a code of engineering practice [32]. In the 2014 annual gas technical statistics edited by the Swiss Gas and Water Industry Association (SGWA), some data about NG pipeline length, material and accidents are reported even if it is impossible to evaluate the importance of each cause and the accidental conditions [33]. The European Gas Pipeline Data Group (EGIG) publishes a more accurate report of accidents that have occurred in European transmission systems between 1970 and 2016, but these data are only valid for transportation systems that operate at higher pressures than distribution ones [34]. Some data about the Deutschland NG distribution system are reported by the Paul Scherrer Institut (PSI), but

AC C

88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113

ACCEPTED MANUSCRIPT

RI PT

have not been updated since 2005 [35]. Statistical information is also reported for the US NG distribution system for the period 1999 – 2003 [9]. There is therefore a lack of information in the identification and classification of NG distribution systems accidents that have occurred in recent years. This paper aims to show how safety conditions in the NG sector have changed during these years. To achieve this objective, the main causes of a failure are analyzed and specific risk factors are proposed. This will allow operators to identify ways in which to efficiently improve existing performance [36]. In addition, NG stakeholders can use paper data to improve their maintenance programs and to evaluate the most critical parts of their system as done for other types of lifelines [37]. 2. Materials and method

SC

2.1 Data source, time interval and parts of the distribution system considered

M AN U

The source of the analyzed data is the open database of Pipeline and Hazardous Materials Safety Administration (PHMSA) [38] that follows the requirements in Part 191 of Title 49 of the Code of Federal Regulations. In the PHMSA database, accidents that occurred in the distribution system of the United States are recorded from 1970 but only the period from 1984 to 2015 is usually considered for statistical analysis [39]. This choice is due to the fact that PHMSA increased the minimum parameters necessary to record an accident after 1984, so a wrong interpretation of trends could be made if data before 1985 are used, as is demonstrated in Figure 1 where a high rate change of accidents reported is present between the years before and those after 1985.

152 153 154 155 156 157 158 159 160 161

AC C

EP

TE D

129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151

Figure 1. Total number of accidents from 1970 to 2015 and percentage of accidents that occurred in copper and iron pipelines.

In the case of an accident, the operator shall submit a written report to PHMSA after 30 days if one of these minimum criteria occurs: 1. An event that involves a release of gas from a pipeline, or of liquefied natural gas, liquefied petroleum gas, refrigerant gas, or gas from an LNG facility, and that results in one or more of the following consequences: (i) A death, or personal injury necessitating in-patient hospitalization;

ACCEPTED MANUSCRIPT

RI PT

(ii) Estimated property damage of $50,000 or more, including loss to the operator and others, or both, but excluding cost of gas lost; (iii) Unintentional estimated gas loss of three million cubic feet (i.e. 84951 m3) or more; 2. An event that results in an emergency shutdown of an LNG facility. Activation of an emergency shutdown system for reasons other than an actual emergency does not constitute an incident 3. An event that is significant in the judgment of the operator, even though it did not meet the criteria of paragraphs (1) or (2) of the list.

EP

TE D

M AN U

SC

As mentioned, the reporting format was changed three times by PHMSA in 1984, 2004 and 2010 respectively. However, while in 2004 and 2015 the formats can be analyzed together thanks to the similar information that is present, the data between 1984 and 2004 are instead less detailed and present some differences with respect to the other two periods. The American database can be used as a reference for other industrialized countries too, in particular for European countries. In fact, the technical rules used in the United States are almost the same as or similar to the ones used in Europe for the NG distribution system. The paper’s results can thus be qualitatively and quantitatively extended to European distribution systems; however, it should be underlined that some European standards are more conservative with respect to American one [40]. To make a more useful analysis of the data present in the database, only steel and plastic systems were considered. In fact, as reported in Figure 1 the percentage of accidents that occurred in copper and iron systems are less than 10% of the total; this could be explained by the fact that the length of steel and plastic distribution systems are two orders of magnitude larger than copper and iron ones, as represented in Figure 2. The same result was found also in Europe where old pipelines made from cast iron and other materials are going to be replaced by polyethylene and steel and represent only 11.4% of the total network length [30]. Furthermore, only mains and service lines are considered in the analysis; meter and pressure reducing stations and domestic meters are not taken into account in this paper but will be investigated in a future study.

AC C

162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190

191 192 193

Figure 2. Natural Gas distribution length for different materials. Mains and service lines are added together.

ACCEPTED MANUSCRIPT 2.2 Definition of safety indexes The definition of safety indexes is necessary to analyze and to comment on statistical trends. The first ones, defined in safety reports, are the total number of accidents, injuries and fatalities that occurred in a defined period of time, i.e. a year in this case. This parameter gives immediate information about the hazardousness of the NG distribution system. However, these values do not take into account the length variation of the system during the years and so do not correctly reflect the real improvements that occurred in the years. For this purpose, a failure rate is introduced in equation (1) as the ratio between the sum of all the accidents that occurred and the total length of the system for the year considered.

RI PT

194 195 196 197 198 199 200 201 202 203 204

208 209 210 211 212 213 214

M AN U

207

Where: • X is the accident failure rate, expressed in [#/km]; • xz is the z-th accident that occurred in the year considered, [#]; • L is the total length of the NG distribution system in the year considered, [km]. To evaluate how the dangerousness of the single events has evolved over time, injury and fatality rates are calculated; this index is defined as the ratio between the number of injuries and fatalities that occurred in the year considered and the total number of accidents in the same year as reported in equation (2):

TE D

205 206

218 219 220 221 222 223 224 225 226 227 228

,

Where: • K is the injury (j=1) and fatality (j=2) rate, expressed in [#/accident]; • ij,z is the z-th injury or fatality that occurred in the year considered, [#]; • xz is the z-th accident that occurred in the year considered, [#].

AC C

217

(2)

EP

=

215 216

SC

=

(1)

To set out a correct failure root analysis, different types of cause are introduced: “External corrosion”, “Natural forces damage”, “Incorrect operations”, “Excavation damage”, “Material and equipment failures”. External corrosion is divided into galvanic corrosion, microbiological corrosion and other types as proposed in the PHMSA report; however, with respect to the PHMSA, accidents due to stray current and improper cathodic protection are included in galvanic corrosion. Another cause, that is defined as “other cause”, is not considered in the study. In fact, this term includes car accidents, truck accidents, vandalism, etc. which is not useful in proposing solutions for integrity management plans.

ACCEPTED MANUSCRIPT

Y =

244 245 246

Where: • Yy is the injury, fatality rate for the y-th cause, expressed in [#/accident]; • yy,z is the z-th injury or fatality caused by the y-th cause that occurred in the year considered, [#]; • x , is the z-th accident that occurred for the y-th cause in the year considered, [#]. Furthermore, to correctly analyze an accident it is necessary to know not only the number of injuries or fatalities that occurred in the event but also the number of events in the defined time period, i.e. the frequency. To take this into account, a specific failure rate is defined in accordance with equation 4 as the ratio between the z-th accident due to the y-th cause and the total length of the system in the year considered.

254 255 256 257 258 259 260 261 262 263

264

AC C

EP

247 248 249 250 251 252 253

M AN U

243

,

(3)

TE D

241 242

x

,

RI PT

And to have even more useful information the analysis was performed only for the period between 2004 and 2015; this makes it possible to obtain information more representative of the state of the art. The information obtained is important in allowing operators to identify the impact of each cause; in fact, it is always more difficult and expensive for NG stakeholders to improve safety because some solutions are characterized by a high actuation cost that it is not followed by the expected results. It is therefore important to show where and how to intervene to optimize safety actions. In order to propose and justify safety management plans and technical rules, injury and fatality rates are thus calculated for each cause. This term is defined in equation 3 as the ratio between the number of injuries or fatalities that occurred in the defined year for the y-th cause and the total number of accidents for the y-th cause.

SC

229 230 231 232 233 234 235 236 237 238 239 240

=

(4) ,

Where: • Xy is the accident failure rate factor for the y-th cause, expressed in [accident/km]; • xy,z is the z-th accident that occurred for the y-th cause in the year considered, [#]; • L is the total length of the NG distribution system in the year considered, [km]. Finally, to introduce the risk of each cause, a risk factor is defined as the product between the consequences of the event, i.e. the injury or fatality rate as defined in equation (3), and the frequency, i.e. the failure rate as defined in equation (4). =Y ×

(5)

ACCEPTED MANUSCRIPT

295 296 297 298 299 300 301 302

2.3 System parameters performance

SC

RI PT

The implementation of technical solutions to avoid failures needs careful system analysis. As for some introduced indexes, this is done considering only the accidents that occurred between the 2004 and 2015. Accidental pressure is the first parameter which is investigated. For this purpose, the pressure in the analysis is the estimated pressure at the point and time of the incident in order to evaluate any trends as a function of network conditions. To perform the study, pressure is divided into three groups: < 10 psig, [10; 50[ psig, ≥ 50 psig1. The number of accidents, injuries, fatalities, ignitions, break outs and explosions following an accident are divided within the three groups. The same is done also for pipeline diameter. For this purpose, Nominal Pipe Size (NPS) was considered. In this case, five different ranges are proposed to evaluate the presence of trends: < 1.5 inch, [1.5; 2[ inch, [2; 3[ inch, ≥ 3 inch and unknown. These ranges are defined in order to classify the networks as a function of the volumetric flowrate that can be transported in the pipeline and so, indirectly, the hazard that could be induced. After gas leakage, ignition can occur if the required conditions are present, i.e. the concentration of air and gas are within the flammable limits. In particular, ignition probability is studied as a function of the product between accidental pressure, p, and the square of the pipeline diameter, d2, i.e. pd2 [39]. To perform the calculation for each analyzed accident the product pd2 was taken in account. Accidents were then divided into groups, paying attention that the same (or a similar) number of accidents were present in each group. Accidents for which pressure or diameter is unknown are not considered for the probability calculation. For each group the following have been calculated: 1. The average value of the products pd2 as in equation (6):

M AN U

268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294

TE D

267

EP

266

Where: • Ry is the risk factor for each cause, [#/km]; • Yy is the injury, fatality rate for the y-th cause, expressed in [#/accident]; • Xy is the accident failure rate factor for the y-th cause, expressed in [accident/km].

AC C

265

=

∑

(6)

Where: • pdi is the product of pressure and square diameter of the i-th accident in the group considered; • N is the total number of accidents of the considered group; 2. The probability of ignition as in equation (7) i.e. the ratio between the number of ignition and the number of accidents of the group. != 1

"#$ % &$

(7)

1 psig corresponds to almost 0.069 bar g. Therefore 10 psig and 50 psig are equivalent to 0.69 bar g and 3.45 bar g.

ACCEPTED MANUSCRIPT

307 308 309 310 311 312 313 314 315 316 317 318 319

Finally, for each group, the probability of ignition P is plotted as a function of the average value of the products pd2 of the group. The period of operation of failed systems was also studied. Failed pipelines with unknown installation data are not considered, nor are failures due to excavation, incorrect operation and other causes, since these failures are not related to the degradation produced by time and work. Years of activity were divided into twenty groups of five years. For each group the number of accidents that occurred in systems with fewer or the same years of activity is defined, and an obsolescence failure rate is calculated as in equation (8); obsolescence failure rate is defined as the ratio between the number of accidents that occurred in two consecutive groups and the time difference between the two.

RI PT

306

SC

305

Where: • P is the probability of ignition of the considered group; • NIgnition is the number of the ignitions that occurred in the considered group; • N is the total number of accidents of the considered group

F() =

324 325 326 327 328 329 330 331 332 333 334 335 336 337

3. Results and discussion

TE D

323

(8)

EP

322

∑ *) − ∑ *), ∆.)

Where: • F(/ is the obsolescence failure rate for the group k, [Failure/year]; • ∑ F/ is the sum of accidents that occurred in the k groups, [Failure]; • ∑ F/, is the sum of accidents that occurred in the k-1 groups, [Failure]; • ∆tk is the number of years between the groups k+1 and k and it is equal to 5, [year]; • k is the group number, k=1, .., 20.

3.1 Accidents, injuries and fatalities trends

AC C

320 321

M AN U

303 304

The number of accidents that occurred between 1984 and 2015 decreased, reaching an asymptotic trend in the last five years, as can be seen in Figure 3 in the United States NG distribution system. Figure 4 and Figure 5 show that the number of injuries and fatalities that occurred in the period presents a different trend, with some peaks that depend on the single event; however, for injuries a trend can be observed with a minimum occurred in 2004.

RI PT

ACCEPTED MANUSCRIPT

SC

Figure 3. Number of accidents that occurred between Figure 4. Number of injuries between 1984 and 2015 in 1984 and 2015 in the United States NG distribution the United States NG distribution system. system.

EP

Figure 5. Number of fatalities between 1984 and 2015 in the United States NG distribution system.

To compare performance trends in different countries, suitable factors should be introduced. For this purpose, accident failure rate, injury failure rate and fatality failure rate were selected. As regards accidents, a lower reduction of accident failure rate than previous years is present in US distribution networks. In 1984 to 2015, in fact, a higher rate of reduction is present as shown in Figure 6, demonstrating that it is always more difficult to improve safety conditions. The mathematical interpolation of the data is shown in the figure and it is valid for the period between 1984 and 2014; in fact, the accident failure rate tends to + infinitive for x vanishing (not physically acceptable) and tends to zero for x that tends to + infinitive. This condition is not in accordance with the data because it seems that an asymptotical trend was reached with almost 1.6 x 10-5 accidents/km. It is interesting to note that, with the current trend, more than 250 years are necessary to reach the goal of 1 x 10-5 accidents/km. To reduce this time interval, technical and economic efforts have to be made in all NG distribution phases. A similar trend is also reported by [32] from 1981 to 2013, even if higher values for accident failure rates are reported. This difference can be justified by the different conditions required to report an accident; in fact, considering immediately reported accidents in accordance with the DVGW G 410 (A), a similar trend is present as reported

AC C

339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357

TE D

M AN U

338

ACCEPTED MANUSCRIPT

RI PT

in Figure 6, particularly for the last years of the analyzed period; care should, however, be taken in using these data because of the presence of accidents that occurred in the German transportation system. [31] reports a decreasing trend between 2006 and 2015. Since the conditions required to record an accident are similar to those defined by the PHMSA, it is interesting to compare the values. This report shows an average value of 3.8 x 10-5 accident/km, slightly higher with respect to the US distribution system, although the US and European systems present similar deviations. Furthermore, [35] also shows a similar behavior between 1981 and 2002 even if higher values are present; this conditions could be justified by difference in the conditions required to record an accident. However, the decreasing trends reported by the available sources demonstrate the implementation of similar improvements and safety solutions in the NG distribution market.

EP

Figure 6. Accident failure rate from 1984 to 2015 considering data from the following source: [38], [31], [32] and [35].

With respect to accident failure rates, different considerations should be made for injury and fatality rates as shown in Figure 7 and Figure 8; in fact, it is not possible to define a correlation because of the high variability year by year due to the presence of peaks. Considering a moving average analysis, more stabilized values are found for injury and fatality rates. In a period of ten years for fatalities, the rate is between 0.078 and 0.102 fatalities per accident. For injuries, a five year period was used, finding a minimum in 2009 (about 0.2 injuries per accident) and a subsequent slight increase, as reported in Figure 7. From the figure, it is shown that [31] reports a higher number of injuries due to accidents with respect to US distribution networks as reported by the five years average values. Only in 2014 is a similar value present but this was due to a very severe event in the US distribution network. However, average values seem to be slightly higher with respect to US values from the data reported by [31]. Consequently, an investigation should be performed to understand the reasons. In other words, a more accurate analysis is not possible from the available data so improvements in data recording should be performed for this purpose. Considering fatality rates, similar behaviors were detected in PHMSA, [31] and [35]. It is interesting to identify a higher rate in [35], especially in the last recorded years. Similar values can be considered, instead, between PHMSA data and [31].

AC C

369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388

TE D

M AN U

SC

358 359 360 361 362 363 364 365 366 367 368

ACCEPTED MANUSCRIPT

RI PT

Therefore, an improvement in European NG distribution networks can be identified for a safety point of view even if a high value is present in 2014; solutions implemented to improve safety have had a positive impact reducing the number of accidents but have not been sufficient to avoid injuries and fatalities. In particular, it seems that new solutions should be proposed to contrast the recent increase of injury rate. It should, however, be highlighted that it is difficult to understand the trends from values reported in [31] because of the lack of information. In fact, the total number of injuries and fatalities is reported only from 2010. Last, no data are present in [32] in order to calculate injury and fatality rate.

Figure 7. Injury rate in the years between 1984-2015 considering data from the following source: [38], [31], [32] and [35].

AC C

EP

398 399 400 401

TE D

M AN U

SC

389 390 391 392 393 394 395 396 397

402 403 404

Figure 8. Fatality rate in the years between 1984-2015 considering data from the following source: [38], [31] and [32].

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

Figure 9 shows accident failure rate factors for different causes in the period between 2004 and 2015. The higher number of accidents is caused by excavation in US distribution networks (48.9%), followed by material or equipment failure (9.9%), natural forces (8.6%), incorrect operations (5.5%) and corrosion (3.0%). It is also interesting to note that in the last ten years a reduction of excavation damage occurred thanks to the improvement of procedures and care during excavation activities. As regards the European context, [31], [32] and [35] analyze the causes of failure. Only the main causes are considered. The reported percentages refer to the total number of accidents recorded. In [31], excavation is considered the main cause (41.5%) followed by incorrect operations (15.8%) and natural force (4.0%). [35] reports that incorrect operations (40.2%) is the main cause of failure followed by excavation damage (25.6%), natural force (19.4%), material or equipment failure (8.4%) and corrosion (5%). For excavation damage and incorrect operations, an average accident failure rate has been reported respectively of 2.1 x 10-5 and 0.7 x 10-5, slightly higher than the US situation. Even if the difference is small for excavation damage, efforts should be ensured to reduce incorrect operation errors. Finally, in [32] excavation is reported as the first cause of immediately reportable accidents (39%) even if corrosion is mentioned in the case of metallic connections. Corrosion has the lowest impact thanks to the technological advancement in cathodic protection systems and coating applications; as reported in Figure 10, the main cause is galvanic corrosion with nineteen accidents, i.e. 76% of the total number. Galvanic corrosion has occurred principally in not protected systems, as shown in Figure 11. These systems are usually part of old installations that were installed when cathodic protection was not mandatory and also no coating was present. Moreover, even if cathodic protection is introduced in the systems, some failures still occur; in this case the number of failed coated systems is higher than bare (not coated) systems. This condition could be explained because the damage of coating could cause the phenomenon of shielding of cathodic protection defined as disbandment, reducing the protection current and thus the protection capability [41,42]; this is found in the case of polyethylene base coatings. For other cases, such as for coal tar coatings, which are not affected by shielding phenomenon, cathodic protection current may not alone be sufficient to protect the system after coating damage. Furthermore, corrosion failure produces holes and a lower gas flowrate that could be identified during inspection activities before an accident can occur. To effectively identify the presence of corrosion, as mentioned by [32], data from inspection activities should also be recorded by DSOs in order to ensure a more accurate statistical data analysis.

AC C

405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438

Figure 9. Accident rate factors for different causes of accident. Years between 2004 and 2015.

EP

TE D

M AN U

439 440 441

SC

RI PT

ACCEPTED MANUSCRIPT

442 443 444 445 446 447 448 449 450 451 452 453 454 455

AC C

Figure 10. Number of accidents due to different causes Figure 11. Characteristics of failed systems due to of corrosion between 2004 and 2015. galvanic corrosion between 2004 and 2015.

In table 3 and table 4 fatality and injury rate factors are reported. The analysis shows that a small reduction of fatalities rate factors for excavation occurred in the first years of the period analyzed in accordance with accident rate factors, while injury rate factor remained almost constant with a slight increase between 2010 and 2013. Another interesting factor is that natural force damage gives a lower value of the average injury and fatality rate factors with respect to excavation damage, but is responsible for the highest value. These results should justify the installation of devices able to immediately stop gas supply when a high flow rate escapes outside the conduit, as occurs for natural events or in case of the rupture of the pipeline damaged during excavation activities. For example, in 2016 PHMSA changed requirements (49 Code of Federal Regulations (CFR) part 192 requirement) requiring the installation of Excess Flow Valves (EFV) for new or repaired branched service lines. Accidents in service pipes were 40.2% of the total that occurred after 2004 in main and service networks (i.e. 1100 accidents). Considering the definition of fatality rate and of the injury rate respectively, 0.15 fatality/accident and 0.56 injury/accident were calculated. That means

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

that only those accidents that are responsible for large amounts of leakage should be considered for the installation of EFV. Therefore, only accidents that cause rupture of the pipeline are considered for the purpose of the analysis. From the analysis of the data reported, the leakage of NG in 26.9% of the accidents that occurred in service lines are considered as avoidable. In these cases, a fatality rate equal to 0.15 and an injury rate equal to 0.29 were calculated. From the analysis it appears that EFV can reduce the number of accidents each year by almost 2.5%. Therefore, the implementation of these devices would not ensure that the desired failure rate of 1.0 x 10-5 could be reached. Furthermore, it seems that EFV would not have a big impact in the reduction of injury and fatality rates. The installation of EFV, however, should be considered as one of the technical solutions necessary to improve natural gas safety even if there are no standardized technical rules for installation. Furthermore, operators are still inexperienced in their use; technical issues arise from the identification of the protectable length and from the need to minimize false alarms. Research activities and technical committees are therefore needed to overcome these obstacles. While EFV or other similar devices including protective equipment are needed, preventive solutions such as pipeline monitoring systems should also be installed. These systems could be installed in those areas where high risks are present. Several types of sensors have been studied in the literature such as fiber optic ones that, without the need for electrical supply, are exempt from ATEX requirements; however, a cost analysis of the technology should be performed as done to analyze the installation of ultrasonic devices in existing and new pipelines to justify the implementation [43]. Corrosion seems to be a low impact agent on fatality rate of gas distribution and in some years (2007 and 2015) no accidents occurred. In these cases, the definition of a fatality and injury rate factor is not applicable (NA). The same occurs for incorrect operations in 2004 and 2007 that can however have a serious impact both for injuries and fatalities when an accident occurs. Similarly, material or equipment failures do not have a great impact on the fatality rate factor. Considering injury rate, instead, material or equipment failure is characterized by a constant trend. Therefore, technical improvements should be required to avoid this type of failure or to improve the performances of the devices on the market. Natural force damage is characterized by the highest average fatality rate index. Furthermore, it should be highlighted that natural forces can be responsible for very severe events, as occurred in 2014 when both injury and fatality rate were characterized by the highest values for the analyzed period. This information should be considered by DSOs to improve monitoring systems in those areas where natural events such as soil movements (settlement, landslide or other) can occur and high population density is present. Particular attention should also be paid to those cases in which NG pipelines are installed near other lifelines such as sewage systems, electric cables or water pipes. In fact, a hazard distance from the leak of 20 m is proposed by [44] as the distance in which fires and explosion due to gas leakage are most likely to occur. However, in the presence of other buried lifelines, soil movement can damage them too and allow the NG to enter and flow for longer distances than that proposed. For example, in the case of soil movement, monitoring systems such as buried fiber optic cables for new networks or GPS tracking systems for existing ones [45] could be considered by DSOs in order to carry out preventive actions if abnormal conditions are identified.

AC C

456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498

ACCEPTED MANUSCRIPT 499

0.2

0.08

0

0

0.33

1

0.095

0

0

2007

NA

0

0.04

0.13

NA

2008

0

0

0.03

0

0

2009

0

0.2

0.02

0

2010

0

0

0.09

0.17

2011

0

0

0.03

0

2012 2013

0

0

0.06

0

0

0

0.15

0

2014

0

2.67

0.03

0

2015

NA

0

0

Excavation

2004

0.25

0

2005

0

2006

SC

Natural Force

RI PT

Incorrect operation NA

Corrosion

0

0

0.33 0 0 0

0.25 0

M AN U

0.06 0.13* 0.05 0.05 0.06 Average Standard 0.12 0.30* 0.04 0.08 0.12 deviation * 2014 is not considered being the calculated yearly injury rate influenced by a very severe event. Table 3. Fatality rate factor for different causes of accidents between 2004 and 2015 [Fatalities/Accident].

Corrosion

Natural Force damage

Excavation

Material or equipment failure

Incorrect operation

2004

0

0.25

0.09

0.25

NA

2005

0

0.4

0.21

0.14

0.75

2006

0.33

0.25

0.31

0

0.75

2007

NA

0.5

0.18

0.13

NA

2008

0.33

0

0.30

1

0.5

2009

0

0.4

0.17

0.14

0.8

2010

0.33

0

0.14

1.17

0.83

2011

0.67

0

0.55

1

0.75

2012 2013

0.5

0.33

0.56

0.5

4.5**

0

0

0.56

0.2

0

0

16**

0.5

0.25

0.75

EP

TE D

Year

AC C

500 501 502

0

Material or equipment failure 0.25

Year

2014 2015

503 504 505 506 507 508 509

NA 0 0.22 0.5 1 0.22 0.19* 0.32 0.44 0.68** Average Standard 0.24 0.20* 0.17 0.38 0.29** deviation * 2014 is not considered since the calculated yearly injury rate is influenced by a very severe event. ** 2012 is not considered since the calculated yearly injury rate is influenced by a very severe event. Table 4. Injury rate factor for different causes of accidents between 2004 and 2015 [Injuries/Accident].

Figure 12 and Figure 13 show fatality and injury risk factors. Excavation is characterized by high values of the two factors, even if a decreasing trend for the fatality factor can be observed since 2008; after that the value remained almost constant except for 2013 when more fatalities occurred with respect to previous years; for the injury risk factor the trend seems to be constant for the period examined. Material and equipment failure has almost the same fatality risk factor of corrosion, that

ACCEPTED MANUSCRIPT

Cause

529 530

Corrosion

M AN U

SC

RI PT

instead has the lowest injury risk factor. The risk caused by incorrect operations can also be discussed, as reported for fatality in Figure 12. As regards injury, in some cases incorrect operations had the same risk as excavation damage and so should not be underestimated during Quality, Health, Safety and Environment (QHSE) procedure definitions. Being characterized by a high injury risk factor, incorrect operation should be treated with particular care. In fact, operators should improve Personal Protective Equipment (PPE) supplied to employees and increase disciplinary sanctions if it is proved that these are not used during work. For the same reason, operators should give bonuses to anyone who notices and communicates the possibility of an impending danger. Only with these actions is it possible to reduce the impact of incorrect operations in safety statistics. Risk factors can be used to identify more hazardous causes for NG distribution networks and to propose preventive solutions. Parameters should be evaluated for each cause as descriptive of NG network (for example for corrosion coating thickness and corrosion velocity); current loads should also be analyzed together in accordance with superposition effects. This will make it possible to identify threshold values for which hazard conditions would occur with a defined probability. The proposed idea will be presented in a future study. Table 5 and table 6 report fatality and injury risk factors; for the calculation of the risk, however, very severe and low frequency events are also considered in order to have immediate information about the possible hazard.

Fatality risk factor, [Fatalities / km] Natural forces Excavation Material and equipment failures 3.45 x 10-7 5.77 x 10-7 1.04 x 10-7

Incorrect operation 8.78 x 10-8

6.29 x 10-8 Average Standard 1.26 x 10-7 6.79 x 10-7 5.21 x 10-7 1.98 x 10-7 1.88 x 10-7 deviation 3.23 x 10-7 2.29 x 10-6 1.58 x 10-6 6.45 x 10-7 5.9 x 10-7 Maximum Table 5. Fatality Risk factor for different causes. Average value and maximum value between 2004 and 2015.

TE D

510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528

Injury risk factor, [Injuries / km]

Corrosion

Natural forces

EP

Cause

Material and equipment failures 7.18 x 10-7

Incorrect operation 1.21 x 10-6

1.78 x 10-7 1.45 x 10-6 3.1 x 10-6 Average Standard 1.96 x 10-7 3.72 x 10-6 1.36 x 10-6 6.08 x 10-7 1.41 x 10-6 deviation 5.86 x 10-7 1.37 x 10-5 5.49 x 10-6 2.07 x 10-6 4.1 x 10-6 Maximum Table 6. Injury Risk factor for different causes. Average value and maximum value between 2004 and 2015.

AC C

531 532

Excavation

M AN U

Figure 12. Fatality risk factor for different type of causes between 2004 and 2015.

536 537 538 539 540 541 542 543 544 545 546

AC C

EP

TE D

533 534 535

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 13. Injury Risk factor for different types of causes between 2004 and 2015.

3.2 Accident characterization

The highest number of accidents occurred for the pressure range [10; 50[ psig, and diameter ≥ 3 inches as reported in Figure 14 and Figure 16; this condition is explained by the fact that the increase of pressure and diameter produces an increase of mechanical stresses. However, the highest number of fatalities and injuries per accident occurred for low pressure accidents, i.e. pressure less than 10 psig and diameter in the range [1; 2[ inch as reported in Figure 15 and Figure 17. These types of systems, i.e. low pressures and low diameters, are usually installed in high

ACCEPTED MANUSCRIPT density areas. So as shown by the results, the most hazardous parts of the system are those installed near the end-users.

Figure 15. Number of injuries and fatalities per accident for different pressures.

M AN U

Figure 14. Number of accidents for different pressures.

SC

RI PT

547 548 549

EP

TE D

550

551 552 553 554 555 556 557 558 559 560 561

AC C

Figure 16. Number of accidents for different range of diameters.

Figure 17. Number of injuries and fatalities per accident for different diameters.

Figure 18 shows the relationship between the product pd2 and the probability of the occurrence of ignition. The probability of ignition is greater for low values of the product pd2. This condition is due to two considerations: the first is that since low pressure systems are installed in highly populated areas (Class 3 and Class 4) several ignition sources which can supply the necessary energy to start the combustion of the leaked gas are present. The second reason is that for high pressures and large diameter systems, a bigger flowrate can be expected for a rupture and so it is possible that air and gas are outside the flammable limits in the zone of the leakage, so the combustion cannot start. So, in low pressure accidents, where a high probability of ignition is present, a lot of energy is released as thermal radiation to the surrounding population [7].

SC

Probability of ignition 0.73 0.88 0.63 0.43 0.60 0.48 0.50 0.22 0.12

TE D

M AN U

Group Average pd2, [psig x in2] Standard deviation, [psig x in2] 6.91 4.66 1 23.07 6.63 2 53.68 12.99 3 107.25 21.47 4 184.87 19.99 5 368.39 107.32 6 931.01 247.70 7 2876.27 1129.93 8 21745.73 15615.20 9 Table 7. Probability of ignition for different values of pressure x diameter^(2).

Finally, the analysis of failures caused by age gave different results for steel and plastic systems. It was found that the amount of damage that occurred for age in steel systems is 70% greater than in plastic ones. Nevertheless, as reported in Figure 19 for steel systems, the highest failure rate is present between 40 and 60 years of life. Consequently, integrity management plans should pay particular attention to steel systems being in service for longer than forty years.

AC C

565 566 567 568 569 570 571 572

Figure 18. Probability of ignition for different values of pressure x diameter^(2).

EP

562 563 564

RI PT

ACCEPTED MANUSCRIPT

573 574 575

Figure 19. Percent of failures for different years of activity in steel systems.

ACCEPTED MANUSCRIPT For plastic, having fewer data, the trend is not so clear as for steel pipelines, even if it was found that the highest failure rate is present between 25 and 35 years of activity, as reported in Figure 20. This information can suggest the conclusion that plastic pipelines are characterized by a lower number of accidents with respect to steel, but that a peak of failures occurs earlier, if compared to steel systems. However, the plastic pipelines that are considered in this study do not have the same performance as current plastic materials, so the result should be applied only to old plastic systems. Table 8 reports obsolescence failure rates for steel and plastic systems.

Figure 20. Percent of failures for different years of activity in plastic systems.

Years of 5 10 15 20 25 30 activity Steel failure 0.6 0.4 0.6 0.6 0.2 0.6 rates [Failure/year] Plastic failure 1.6 0.8 0.6 0.6 1.2 1.6 rates [Failure/year] Table 8. Failure rates trend for steel and plastic systems.

4. Conclusion

AC C

588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603

35

40

45

50

55

60

65

0.2

0.8

2.8

2.6

2.2

1.2

0.8

1.6

1

0.4

0.2

0.2

0

0

EP

TE D

584 585 586 587

M AN U

SC

RI PT

576 577 578 579 580 581 582 583

Analyzing the accidents that occurred in United States NG distribution systems, we can state the following: 1. It was found that technical and procedural solutions implemented in the last decades ensured a reduction of failures from 1984 to 2015; it is not, however, possible to define a trend for injury and fatality rates because of the high annual variability even if a slight increase is observed for injury rate in the last five years. 2. With respect to European networks, similar decreasing trends were seen even if values for identified indexes are slightly higher for the European context. In other words, the European NG association should improve the information recorded in available reports or create dedicated databases to ensure more detailed analysis; 3. Excavation is the first cause of failure but in the last years a reduction has been observed; injury and fatality rate factors were almost constant in the period analyzed. Corrosion,

ACCEPTED MANUSCRIPT

M AN U

SC

RI PT

equipment failure and incorrect operation are the second cause of failure thanks to technical and procedural improvements. Natural forces lead to a low number of failures but can have high impacts in terms of injuries and fatalities when they occur. Monitoring systems should thus be considered by DSOs where the consequences of the failure of the pipeline are the highest. 4. Natural forces are also responsible for the highest fatality risk factor with 2.29 x 10-6 fatalities/km, while incorrect operations and excavation damages have the highest average values respectively of 6.47 x 10-7 fatalities/km and 5.77 x 10-7 fatalities/km. The same was found also for injury risk factor with values of 1.37 x 10-5, 3.1 x 10-6 and 1.21 x 10-6 injuries/km. 5. The highest number of accidents were seen for pressure between [10;50[ psig and for diameter ≥3”; however, low pressure and small diameter systems were responsible for the highest number of injuries and fatalities, usually being installed in highly populated areas. 6. Steel systems have the highest failure rate with a maximum value of 2.8 failure/year between 40 and 45 years of service life. For plastic systems, instead, the maximum failure rate was observed between 25 and 35 years from the installation.

TE D

Several improvements have been identified for implementation in distribution systems: 1. The installation of protective devices such as EFV able to immediately interrupt gas leakage when a rupture occurs as one of the possible technical solutions to reduce the number of accidents. However, other solutions should be considered to reach the milestone of 1 x 10-5 accident/km. 2. Implementation of monitoring systems able to trigger alarms in the case of system thresholds being exceeded; however, a technical and cost analysis should be performed in order to justify the investment. 3. Operators should increase prevention policies rewarding employees based on the reduction of accidents due to incorrect procedures.

EP

The importance of a well-designed and updated database for safety statistical analysis is evident; Italy and Europe should also develop an open database, like the PHMSA one, to ensure analysis of safety indexes. This database could be useful to define and promote safety policies, such as the definition of safety rating indexes for operators. These indexes should be used to reward operators with economic incentives for their ability to guarantee a high quality supply service and to stimulate safety investments, especially for low and medium enterprises [46]. Instead, in the case of low safety ratings, operators should be fined or obliged to ensure greater safety performances improvement. This would ensure true competition between NG operators that work in the same context.

AC C

604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648

References [1] M.J. Economides, D.A. Wood, The state of natural gas, J. Nat. Gas Sci. Eng. 1 (2009) 1–13. doi:10.1016/j.jngse.2009.03.005. [2] M.J. Economides, S. Mokhatab, Compressed Natural Gas: Monetizing Stranded Gas | Gas Natural, (2007). https://naturalgas.wordpress.com/2007/10/23/compressed-natural-gasmonetizing-stranded-gas/ (accessed May 7, 2018). [3] D. Jones, M. Dufour, J. Gaventa, Europe’s declining gas demand trends and fact on European

ACCEPTED MANUSCRIPT

[9]

[10]

[11] [12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

RI PT

[8]

SC

[7]

M AN U

[6]

TE D

[5]

EP

[4]

gas consumption, 2015. http://www.energypost.eu/gazprom-deals-deepen-eu-gasdilemma/%5Cnpapers3://publication/uuid/6B0A6BA6-7CF4-410E-ABA1-B2C32636D258. Eurogas, Statistical Report 2015, Brussels, 2016. http://www.eurogas.org/uploads/2016/flipbook/statistical-report-2015/eurogas-statisticalreport-2015_LR.pdf. British Petroleum, BP Statistical Review of World Energy 2017, Br. Pet. (2017) 1–52. doi:http://www.bp.com/content/dam/bp/en/corporate/pdf/energy-economics/statisticalreview-2017/bp-statistical-review-of-world-energy-2017-full-report.pdf. PHMSA, Oracle BI Interactive Dashboards - Public Reports, (n.d.). https://hip.phmsa.dot.gov/analyticsSOAP/saw.dll?Portalpages (accessed April 30, 2018). A. Bianchini, F. Donini, A. Guzzini, M. Pellegrini, C. Saccani, Natural gas pipelines distribution: Analysis of risk, design and maintenance to improve the safety performance, in: XX Summer Sch. “Francesco Turco” - Ind. Syst. Eng., Naples, 2015: pp. 243–248. S. Sklavounos, F. Rigas, Estimation of safety distances in the vicinity of fuel gas pipelines, J. Loss Prev. Process Ind. 19 (2006) 24–31. doi:10.1016/j.jlp.2005.05.002. C.J. Trench, Safety Incidents on Natural Gas Distribution Systems: Understanding the Hazards, New York, 2005. https://primis.phmsa.dot.gov/dimp/docs/SafetyIncidentsOnNaturalGasDistributionSystems_ UnderstandingHazards.pdf. R. Alzbutas, T. Iesmantas, M. Povilaitis, J. Vitkute, Risk and uncertainty analysis of gas pipeline failure and gas combustion consequence, Stoch. Environ. Res. Risk Assess. 28 (2014) 1431–1446. doi:10.1007/s00477-013-0845-4. P.L. Metropolo, A.E.P. Brown, Natural gas pipeline accident consequence analysis, Process Saf. Prog. 23 (2004) 307–310. doi:10.1002/prs.10054. I.S. Cole, D. Marney, The science of pipe corrosion: A review of the literature on the corrosion of ferrous metals in soils, Corros. Sci. 56 (2012) 5–16. doi:10.1016/j.corsci.2011.12.001. M. Ahammed, Probabilistic estimation of remaining life of a pipeline in the presence of active corrosion defects, Int. J. Press. Vessel. Pip. 75 (1998) 321–329. doi:10.1016/S03080161(98)00006-4. A. Valor, F. Caleyo, J.M. Hallen, J.C. Velázquez, Reliability assessment of buried pipelines based on different corrosion rate models, Corros. Sci. 66 (2013) 78–87. doi:10.1016/j.corsci.2012.09.005. A.P. Teixeira, C. Guedes Soares, T.A. Netto, S.F. Estefen, Reliability of pipelines with corrosion defects, Int. J. Press. Vessel. Pip. 85 (2008) 228–237. doi:10.1016/j.ijpvp.2007.09.002. F. Caleyo, J.C. Velázquez, A. Valor, J.M. Hallen, Markov chain modelling of pitting corrosion in underground pipelines, Corros. Sci. 51 (2009) 2197–2207. doi:10.1016/j.corsci.2009.06.014. S. Esposito, S. Giovinazzi, L. Elefante, I. Iervolino, Performance of the L’Aquila (central Italy) gas distribution network in the 2009 (Mw6.3) earthquake, Bull. Earthq. Eng. 11 (2013) 2447–2466. doi:10.1007/s10518-013-9478-8. T.D. O’Rourke, S.S. Jeon, S. Toprak, M. Cubrinovski, M. Hughes, S. Van Ballegooy, D. Bouziou, Earthquake response of underground pipeline networks in Christchurch, NZ, Earthq. Spectra. 30 (2014) 183–204. doi:10.1193/030413EQS062M. X. Liu, M.J. O’Rourke, Behaviour of continuous pipeline subject to transverse PGD, Earthq. Eng. Struct. Dyn. 26 (1997) 989–1003. doi:10.1002/(SICI)10969845(199710)26:10<989::AID-EQE688>3.0.CO;2-P. M.J. O’Rourke, X. Liu, R. Flores-Berrones, Steel Pipe Wrinkling due to Longitudinal Permanent Ground Deformation, J. Transp. Eng. 121 (1995) 443–451. doi:10.1061/(ASCE)0733-947X(1995)121:5(443).

AC C

649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

[21] D. Ha, T.H. Abdoun, M.J. O’Rourke, M.D. Symans, T.D. O’Rourke, M.C. Palmer, H.E. Stewart, Centrifuge Modeling of Earthquake Effects on Buried High-Density Polyethylene (HDPE) Pipelines Crossing Fault Zones, J. Geotech. Geoenvironmental Eng. 134 (2008) 1501–1515. doi:10.1061/(ASCE)1090-0241(2008)134:10(1501). [22] T.H. Abdoun, D. Ha, M.J. O’Rourke, M.D. Symans, T.D. O’Rourke, M.C. Palmer, H.E. Stewart, Factors influencing the behavior of buried pipelines subjected to earthquake faulting, Soil Dyn. Earthq. Eng. 29 (2009) 415–427. doi:10.1016/j.soildyn.2008.04.006. [23] S.A. Naeini, E. Mahmoudi, M.M. Shojaedin, M. Misaghian, Mechanical response of buried High-Density Polyethylene pipelines under normal fault motions, KSCE J. Civ. Eng. 20 (2016) 2253–2261. doi:10.1007/s12205-015-0695-3. [24] H. Montiel, J.A. Vílchez, J. Arnaldos, J. Casal, Historical analysis of accidents in the transportation of natural gas, J. Hazard. Mater. 51 (1996) 77–92. doi:10.1016/S03043894(96)01819-5. [25] A.J. Brito, A.T. de Almeida, Multi-attribute risk assessment for risk ranking of natural gas pipelines, Reliab. Eng. Syst. Saf. 94 (2009) 187–198. doi:10.1016/j.ress.2008.02.014. [26] C. Lam, W. Zhou, Statistical analyses of incidents on onshore gas transmission pipelines based on PHMSA database, Int. J. Press. Vessel. Pip. 145 (2016) 29–40. doi:10.1016/j.ijpvp.2016.06.003. [27] ARERA, ARERA - Dati statistici, (n.d.). https://www.arera.it/it/dati/elenco_dati.htm (accessed April 30, 2018). [28] Comitato Italiano Gas (CIG), Incidenti Da Gas Combustibili 2016, Milan, 2017. [29] V.U. Rocca, Incidenti da gas, (2006). http://www.resit.website/wpcontent/uploads/2014/04/incidenti3.pdf. [30] Marcogaz, Performance Indicators Report 2008 – 2014, 2017. http://www.marcogaz.org/index.php/gas-infrastructure2. [31] Marcogaz, Report on European Gas Safety , Gas Distribution ( EGAS B ), 2017. http://www.marcogaz.org/index.php/gas-infrastructure2. [32] F. Dietzsch, Integrated safety concept of DVGW in terms of statistical verification of incidents, Pipeline Technol. J. (2017) 16–21. www.pipeline-journal.net. [33] SVGW, Statistique technique annuelle Gaz 2015, 2016. http://www.svgw.ch/fileadmin/resources/svgw/web/Shop-Boutique/download/05_GFachinformationen/SVGW_Shop_G15001_f_2016.pdf. [34] EGIG, 10th Report of the European Gas Pipeline Incident Data Group (period 1970 – 2016), 2018. [35] P. Burgherr, S. Hirschberg, Comprehensive Assessment of Energy Systems (GaBE): Comparative Assessment of Natural Gas Accident Risks, (2005) 53. [36] W. Girsberger, Safety Performance indicators in gas distribution: a door opener for public acceptance and guideline for risk-based improvements, n.d. http://www.iapg.org.ar/WGC09/admin/archivosNew/Special Projects/4. IGU Best Practices/4. IGU Best Practices FINAL - CD contents/4. Distribution - Safety performance indicators in gas distribution.pdf. [37] H.A. Kishawy, H.A. Gabbar, Review of pipeline integrity management practices, Int. J. Press. Vessel. Pip. 87 (2010) 373–380. doi:10.1016/j.ijpvp.2010.04.003. [38] PHMSA, Pipeline Incident 20 Year Trends | PHMSA, (n.d.). https://www.phmsa.dot.gov/data-and-statistics/pipeline/pipeline-incident-20-year-trends (accessed April 30, 2018). [39] C. Lam, S. Wenxing Zhou, Statistical Analyses of Historical Pipeline Incident Data with Application to the Risk Assessment of Onshore Natural Gas Transmission Pipelines, (2015) 86. http://ir.lib.uwo.ca/etd. [40] G. Goodfellow, IPC2006-10507, in: Proc. IPC2006 6th Int. Pipeline Conf., Calgary, 2006: pp. 0–12.

AC C

700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750

ACCEPTED MANUSCRIPT

EP

TE D

M AN U

SC

RI PT

[41] I. Thompson, J.R. Saithala, Review of pipeline coating systems from an operator’s perspective, Corros. Eng. Sci. Technol. Int. J. Corros. Process. Corros. Control. 51 (2016) 118–135. doi:10.1179/1743278215Y.0000000038. [42] K. Riggs Larsen, Using Pipeline Coatings with Cathodic Protection, Mater. Perform. 55 (2016). [43] A. Bianchini, A. Guzzini, M. Pellegrini, C. Saccani, Natural gas distribution system: Overview of leak detection Systems, in: XXI Summer Sch. “Francesco Turco” - Ind. Syst. Eng., Naples, 2016: pp. 123–127. [44] D.W. Heldenbrand, A Review of Soil Gas Migration in Natural Gas Incidents, in: West. Reg. Gas Assoc. Oper. Conf., 2017. http://www.westernregionalgas.org/2017/presentations/Gas Soil Migration Study Hildebrand WRGC 2017.pdf. [45] A. Bianchini, A. Guzzini, M. Pellegrini, C. Saccani, Earthquake and earth movement monitoring: The possibility to use Natural Gas distribution infrastructure, in: XXII Summer Sch. “Francesco Turco” - Ind. Syst. Eng., Palermo, 2017: pp. 143–149. [46] A. Bianchini, M. Pellegrini, D. Peta, C. Saccani, Economic Evaluation of Investments for Workplace Safety, 36 (2014) 49–54. doi:10.3303/CET1436009.

AC C

751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767

ACCEPTED MANUSCRIPT Highlights Statistical analysis of accidents occurred in Natural Gas distribution systems in the United States between 2004 and 2015

• •

Analysis of accidents as a function of causes. Definition and evaluation of safety factors

•

Analysis of system performances as a function of material, pressure, size

AC C

EP

TE D

M AN U

SC

RI PT

•