A probabilistic model for a gas explosion due to leakages in the grey cast iron gas mains

doi:10.1016/j.ress.2013.06.034

Reliability Engineering & System Safety

Volume 119, November 2013, Pages 270-279

https://doi.org/10.1016/j.ress.2013.06.034 Get rights and content

Abstract

In this paper we propose a model for the probability of an explosion caused by a leakage form grey cast iron pipes in the city of Amsterdam as a function of pipeline and environmental characteristics. The parameters in the model are quantified, with uncertainty, using historical data and structured expert judgment, by use of the Classical Model. Eleven experts from Dutch distribution system operators (DSO) and Kiwa Gas Technology participated in the research. The model has to provide the overall probability of an explosion per year and a prioritization of pipes in terms of their potential contribution to the probability of explosion, which can help DSO's to prioritize their replacements.

Introduction

The Netherlands has the largest gas field in Europe and the tenth largest in the world [1]. These gas reserves enable the production of gas for the next 50 years. The Dutch transmission system operator (TSO) and distribution system operators (DSO) transport the natural gas from well to end user. This is done by transporting gas under pressure through pipelines. The total length of the gas infrastructure is approximately 130,000 km and is separated into a transport network and a distribution network.

TSO transports gas regionally in the Netherlands through the transport network, which is operated under a maximum pressure of 40–67 barg (above atmospheric). The transport network is used to supply gas to large industry (including power producers) and the regional gas distribution companies.

DSO distributes gas to Dutch households and smaller industry through the distribution network. Approximately 36,000 km of the distribution network is operated by Liander. The distribution network consists in general of two grids operating at different pressure levels; a high pressure grid and a low pressure grid. The high pressure grid is generally operated under a maximum pressure of 8 barg, although maxima of 1 and 4 barg are also possible. The high pressure grid is used to transport gas to the right district and also to supply gas to smaller industry. In a district station the pressure of the high pressure distribution grid is reduced to the low pressure distribution grid. The pressure in the low pressure distribution grid is operated under an overpressure of 100 mbarg or 30 mbarg. The low pressure distribution grid transports gas to the households and smaller industry.

Pipelines of the high pressure distribution grid are mainly made of steel or polyethylene (PE). The pipelines of the low pressure grid consist mainly of ductile Polyvinyl chloride (PVC), unplasticised PVC and PE. However, the distribution network of Liander still consists of 10% of grey cast iron pipes.

Grey cast iron is an alloy of iron, graphite (carbon), manganese and silicon, which is used for gas pipelines since the beginning of the 19th century. Grey cast iron is an irregular (heterogeneous) material by nature. This irregularity is caused by the flakes of graphite. As a result of these material properties grey cast iron has extremely high compressive strength, low tensile strength and a poor ductility.

Based on the material properties grey cast iron is a so called brittle material. Brittle materials tend to break if they are subjected to a bending stress, whereas other materials (like PE of PVC) deform before they break. Beside sudden rupture due to a bending stress (as a result of its breaking characteristics) grey cast iron pipes are vulnerable for two types of corrosion; pitting (a specific form of corrosion) and graphitization (a process of oxidizing of iron as a result of current resistance and pH value of the soil, from now on called uniform corrosion). Corrosion can accelerate the fracturing process or can be a direct cause of leakage.

Under the right conditions (gas concentration between 5% and 15% of air as well as a spark from an ignition source) gas can explode. Two explosions caused by (sudden) fracture in grey cast iron pipes happened in the Netherlands in the last ten years. There were no casualties and the effects were limited to serious injuries and material damage. The accident in Amsterdam on August 15th 2001 resulted in two heavily injured people. A second explosion in Amsterdam on March 9th 2008 resulted in four slightly injured people and material damage. However, in Great Britain (Larkhall 1999, four fatalities – Dundee 2002, two fatalities) and in France (Dyon 1999, 11 fatalities – Mulhouse 2004, 17 fatalities), there were multiple fatalities caused by an explosion of natural gas released from a fractured cast iron pipe.

Given the potential impact of an explosion in the grey cast iron gas mains we asked ourselves whether it was possible to model the probability of an explosion and to use this model to prioritize the replacement of grey cast iron mains. Therefore, it was, decided to build a quantitative model for the probability of an explosion caused by a leakage form grey cast iron pipes in the city of Amsterdam as a function of pipeline and environmental characteristics. The model has to provide the overall probability of an explosion per year and a prioritization of pipes in terms of their potential contribution to the probability of explosion so that the most dangerous pipes could be replaced first.

The big challenge in building such a model was in combining an available data which is neither complete nor collected in one unified format. The model must be as complex as necessary to satisfy the goals of the project and such that it can be quantified with existing data accompanied by expert judgment input where necessary. In Section 2 we introduce the available data and discuss assumptions concerning model complexity motivated by discoveries observed in the data. Section 3 contains the description of the model and its quantification with data. The quantification of the model has been assisted with expert judgment. We summarize the expert judgment study in Section 4. Section 5 contains model results and presents short sensitivity analysis concerning assumptions made in the model. Finally in the last section conclusions and recommendations are given.

Section snippets

Data analysis

The first step of the model building process was to discuss with experts the most important influences leading to gas accident. Liander employees were a valuable source of information in this respect. To extend our knowledge of the problem a literature study has been conducted. We found that the fracture in pipes is the most common cause of leakages and it might occur as a result of ground loading [2] or 3rd party damage [3]. Corrosion is also regarded as one of the main causes of pipeline

Model and quantification

In Table 1 we see that leakages can be mainly from pipes and joints and they can be caused by corrosion, third party damage and ground movement. Joint failures are most common but also the least is known about their location, exposure to corrosion and ground movement. Hence after discussion with Liander's experts we decided that only four causes of leakages will be taken into account in the model: leakages from joints (irrespectively of the cause), from pipes due to corrosion, caused by third

Expert judgment

This section presents brief summary of expert judgment study which was conducted in order to obtain missing parameters of the model. The Classical method [12], [13], [14] was used. 11 experts took part in the elicitation session. They were employees of the Dutch DSOs Liander (7) and Enexis (2) and Kiwa Gas Technology (2), a consultancy and engineering firm for gas technology services. We asked them to provide their assessments for questions in form of three quantiles of their uncertainty

Results

The goal of the research was to calculate the probability of explosion in grey cast iron pipelines in Amsterdam. Results were obtained for different pipes’ sub-populations. Main outcomes are presented in this section.

Conclusion

The research shows that it is possible to model the probability of an explosion in the grey cast iron gas mains. The result obtained for the whole population in Amsterdam seems to be plausible taking into account that two explosions took place in the past 10 years. The model takes into account four major failure mechanisms and two preconditions (Factor 1 and Factor 2) which have to be met before a leakages result in an explosion.

Factor 1 and Factor 2 are determined by Expert Judgment, therefore

References (14)

RM Cooke et al.
TU Delft expert judgement database
Reliability Engineering & System Safety
(2008)
Whaley J. The Groningen gas field GeoExpro magazine. 〈http://www.geoexpro.com/history/groningen〉, accessed on August 2...
Rosemary McAll, Balancing Safety, Investment and Maintenance in Gas Distribution, Advantica Ltd, United...
RM Cooke et al.
A probabilistic model for the failure frequency of underground gas pipelines
Risk Analysis
(1998)
Dutch Standardization Institute (1984), Standards for gas and water supply (In...
Arcadis (March 2004), Analysis of factors influencing pipes in Amsterdam (In...
Cast Iron foundry “Fansuld” (In Polish)...

There are more references available in the full text version of this article.

Cited by (6)

Modeling and performance analysis of gas leakage emergency disposal process in gas transmission station based on Stochastic Petri nets
2022, Reliability Engineering and System Safety
Citation Excerpt :
However, the uncertainty of rescue time prediction makes the performance analysis difficult. On the other hand, the subjectivity of expert during modeling and prediction is inevitable [32]. How to integrate different expert opinions on rescue time assignment when evaluating the performance of emergency disposal process has been the main concern.
Efficient emergency disposal on site after gas leakage is crucial to avoid accident escalatio and expansion. However, current performance analysis studies of emerency disposal process are insufficient due to the uncertainty of rescue time prediction. In this paper, we aimed to explore the integration of triangular membership and regional center method to SPN for the modeling and evaluation of gas leakage emergency rescue process in gas transmission station. The stochastic Petri nets model was established by considering the interrelation of each execution stage of emergency disposal. Considering the subjectivity of expert analysis, triangular membership function and regional center method were combined to obtain the values of firing delays. Then isomorphic Markov chain was built from the SPN model and the stability probability linear equations were obtained. By calculating the transition utilization rate and place busy probability, the critical places and transitions were obtained. The list of critical emergency disposal stages and actions was generated, based on which countermeasures were presented for the risk reduction of gas leakage emergency disposal process in gas transmission stations.
Expert judgments calibration and combination for assessment of river levee failure probability
2019, Reliability Engineering and System Safety
Citation Excerpt :
The mathematical calibration models such as the classical model of Cooke [18] allow determining a performance weight for expert proportional to the product of statistical accuracy and informativeness. The classical model has been used in several applications [15,29,30], such as nuclear, banking, volcanoes, underground gas pipelines reliability [31–34], to quantify internal erosion process of dams [35] and levees failure frequency in Netherland [36]. One of the objectives of the calibration models is to aggregate expert opinions onto a single opinion, however several authors recommend combining and aggregating expert opinions using a simple average of the expert elicitated values, which does not require calibration [37,38].
In civil engineering, the utilisation of expert judgement is important for evaluating the reliability of structures whose failure behaviour is little known and difficult to quantify. This article presents the application of the approaches of elicitation, calibration and aggregation of expert opinions to evaluate the reliability of river levees. These approaches have seldom been used in the field of French hydraulic structures whereas there is growing interest for them in other countries, concerning structures reliability and other fields such as aerospace industry, nuclear industry, hydrology, statistics, economics, psychology etc. This article proposes a quantitative approach to the use of expert judgement for evaluating river levee failure probability. An application to the case of an existing river levee is presented. The approach developed assesses the expert opinions according to their calibration (statistical accuracy) and their informativeness, then aggregates them in a single calibrated opinion. Applying the proposed weighting and aggregation procedure reduces variability in the probability of failure estimate, then it provides a better probabilistic distribution of aggregated expert opinion than of individual expert opinions. The results allow identifying a trend of aggregated expert opinions that point towards over- or lack of confidence.
Modeling and prediction of daily gas concentration variation at a mining face based on the elliptic orbit model: A case study
2015, International Journal of Mining Science and Technology
Citation Excerpt :
Addressing the formation process of outburst shock waves and gas flow during coal and gas outburst, numerical models for simulating the real-time propagation of outburst gas flow and the gas transport process were proposed [16]. Based on historical data in the city of Amsterdam, a probability model for evaluating a gas explosion with respect to a leakage from gray cast iron pipes was presented [17]. Actually, daily gas concentration variation is also a time-series.
Monitoring and analysis of daily gas concentrations at a mining face is a vital task on safety production and security management in the coal-mining industry. This study addresses modeling and prediction of daily gas concentration variations based on the elliptic orbit model. The model describes the hourly variation in daily gas concentration by mapping its time-series into the polar coordinates to create its elliptic orbit trace for further analysis. Experiments show workability of the proposed method that daily gas concentration variation at a mining face of one coal mine in China is well described by the elliptic orbit model. Result analysis and performance comparison of the proposed elliptic orbit model with the classical AR model on the same prediction tasks indicate potentiality of the proposed elliptic orbit model, which presents a vivid approach for modeling and forecasting daily gas concentration variations in an intuitive and concise way.
Research on the effect of gas nitriding treatment on the wear resistance of ball seat used in multistage fracturing
2015, Materials and Design
Citation Excerpt :
In Fig. 11(a), the irreversible plastic deformation of material is observed on the surface of untreated specimens. The appearance of plastic deformation results from the low tensile strength, low hardness and poor ductility of surface [31]. The EDS analysis results of debris as shown in Fig. 11(c) also confirm the adhesion mechanism of the untreated specimens.
With the further development of unconventional oil and gas reservoirs, multistage fracturing technique has played an important role in completion process. Wear problem has severely restricted the development of this technique. However, surface modification of fracturing tools has not been studied widely. In this study, ball seats which act as key component in multistage fracturing were nitrided in ammonia (NH₃) and the nitride layer characterizations were investigated. Tribological tests were carried out to study the wear resistance of specimens. The test results show that microhardness and wear resistance are improved significantly by gas nitriding treatment. Based on the comprehensive analyses of the wear morphologies and debris, wear mechanisms are identified. The on-site fracturing experiment results indicate that the ball seat shows better wear resistance after gas nitriding treatment.
Development of an expert judgement debiasing method for assessing levee failure probabilities
2023, European Journal of Environmental and Civil Engineering
Predicting leaks in natural gas distribution networks using generalized linear models
2017, Congress on Computing in Civil Engineering, Proceedings

View full text