Leak-before-break in cast iron mains: a failure analysis of a catastrophic pipe burst on Harris Street, Sydney

The fractured 500 mm diameter spun CI pipe was installed in 1961 with a nominal thickness of 15 mm, and the failure mode was identified as a longitudinal barrel fracture. The spun CI pipe was buried in a backfill sand trench with tar paint coating and factory applied cement lining. The pavement was 250 mm thickness bitumen covered concrete surface, and the total burial depth was approximately 1.6 m to the pipe crown. The traffic loads tend to cause low stress on buried pipes due to the existence of concrete pavement (Robert et al. 2016). The longitudinal barrel fracture is the predominant failure mode for large diameter CI pipes and is caused by local stress concentrations and hoop stresses produced by internal pressures (Makar 2000). Fatigue from cyclic internal pressures is considered as a mechanism to propagate existing cracks in pressurised pipe barrels (Cullin et al. 2014; Rathnayaka et al. 2016). Hence, the effects of internal pressures on fractured pipes requires detailed analysis.

In this article, metallurgical features and mechanical properties are reported for the failed CI pipe. Modelling processes and the results of fatigue crack growth and leak rates are specified as a part of the failure investigation. A significant difference between this study and previous research on large diameter water CI pipes is that a crack growth model by fatigue is proposed to predict the remaining life of leaking pressurised pipelines. A limited time window between leakage and catastrophic failure (LBB window) is identified that may be used to detect leakage incidents in order to predict or prevent future CI water pipe bursts.

The microstructure of the CI pipe material was observed using optical microscopy of a polished and etched specimen taken from the failed pipe covering the entire wall thickness. The microstructure consists of graphite flakes in a ferritic matrix. According to ASTM-A247-67 (1998), the features of graphite flakes, including type, form and size, were characterised (Table 1).

Scanning electron microscopy (SEM) - JEOL 7001 was used to produce SEM images on fracture surfaces of tensile coupon specimens.

Three dimensional (3-D) scanning (Pipecheck package) was conducted on remaining pipe pieces to quantify the effects of soil corrosion. The sand-blast in slow speed was undertaken in these pipe pieces to reveal the metal surface prior to 3-D scanning.

The mechanical properties of the CI pipe material were measured following the relevant ASTM standards (Table 2). Specimens were taken from bare piping metal without corrosion products. Mechanical tests were conducted in an Instron 4402 with a 50 kN load cell. The Aramis digital image correlation system was used to record the deformation of the specimen surface.

3-D finite element analyses were carried out using Abaqus 6.12 to evaluate the buried pipe stresses. The model was set up on the basis of the information gathered during the site investigation. The behaviour of CI pipes and soil was assumed to be elastic, and the parameters of piping material were obtained from the mechanical test results. The interaction between pipe and soil interface was simulated by Coulomb frictional model with a friction coefficient of 0.5. As can be seen from Figure 1, the effects of overlaid other pipelines, anchor block and existing crack were evaluated separately.

The images of two specimens from the fracture surface indicated that the crack may have been existing or initiated at the pipe spigot end where it was inserted in the bell joint. Most of the area of the fracture surface contained orange brown coloured corrosion products that were most likely generated after the pipe burst event. The existence of dual corrosion phases in failed CI pipes have also been observed in other studies, highlighting that fracture may have grown over a period of time prior to final catastrophic fracture (Makar 2000; Rajani et al. 2012; Cullin et al. 2014).

After sand-blasting and 3-D scanning, the remaining pipe section (except on the fracture surface discussed above) was found to be free of substantial corrosion damage. The maximum corrosion depth was less than 2.0 mm.

The ultimate tensile strength from eight coupon tests yielded test results in the range from 133 to 165 MPa with an average value of 148 MPa and a standard deviation of 8 MPa. The modulus of elasticity was estimated by the initial slope of the strain-stress curves. Single-edge notched bend specimens were adopted for both fracture toughness and FCGR tests. The results of mechanical tests from this study along with the values reported in literature are summarised in Table 2. These properties were used for subsequent FE analysis and fatigue crack growth modelling studies.

Table 2

Summary of mechanical testing results

Property .	Number of tests .	Average .	Standard deviation .	ASTM standard .	Literature reported .
Tensile strength, (MPa)	8	148.1	14.1	E8-09	137–212 (Conlin & Baker 1991)
Modulus of elasticity, E (GPa)	8	108.5	3.2	E8-09	103 (AIS 1941)
Fracture toughness, (MPa√m)	4	14.0	1.7	E1820-13	10.5–15.6 (Conlin & Baker 1991)
Paris constant, m	6	6.4	–	E647-13a	6.2–6.7 (Bulloch 1995)
Paris constant, C (m/cycle)	6	2.2 × 10−12	–		6.1 × 10−16–2.6 × 10−12 (Bulloch 1995)

Property .	Number of tests .	Average .	Standard deviation .	ASTM standard .	Literature reported .
Tensile strength, (MPa)	8	148.1	14.1	E8-09	137–212 (Conlin & Baker 1991)
Modulus of elasticity, E (GPa)	8	108.5	3.2	E8-09	103 (AIS 1941)
Fracture toughness, (MPa√m)	4	14.0	1.7	E1820-13	10.5–15.6 (Conlin & Baker 1991)
Paris constant, m	6	6.4	–	E647-13a	6.2–6.7 (Bulloch 1995)
Paris constant, C (m/cycle)	6	2.2 × 10−12	–		6.1 × 10−16–2.6 × 10−12 (Bulloch 1995)

View Large

The critical crack length was conservatively calculated as 300 mm, since the average tensile strength of the target pipe is 148 MPa. According to the measurement, the length inserted in bell joint was around 175 mm.

Finite element analyses also revealed that the impact of the overlaid pipes and the supporting anchor could be neglected for the stress analysis.

The current study reports failure analysis of a 500 mm CI water main which burst on the 10 th of August, 2013 in Harris Street and Ultimo Street, Sydney. A series of different analysis methods were used to investigate the possible causes of the catastrophic failure. Analysis indicates that an existing crack can grow due to cyclic loading (mainly water pressure in this case) over a period of time and can result in major-scale bursts even in environments with low corrosion potentials. Hence, the cause of failure was identified as a combination of initial (existing) material flaws and repetitive (i.e. fatigue) loading.

Metallurgical analyses were conducted to investigate the microstructure of the failed pipe. Results revealed that a pre-crack existed a long time before final fracture, which led to the eventual burst when the crack length increased to the critical size. Correlations between crack length and crack tip stresses were calculated using a 3-D finite element analyses to investigate the total length of the crack (300 mm in this study) that had caused to burst the pipe under high operating pressures.

Fatigue analyses were performed to identify the likely FCGR under cyclic internal water pressure. It is shown that the pre-crack may have stably propagated over a long period under the constantly low changes of internal pressure. The unstable crack growth was expected to occur when the total crack length exceeded 250 mm in the current study.

A window in time can be identified between significant leakage and pipe burst, verifying that the LBB concept can potentially be applied to pressurized CI water pipes. For a pressurised pipe with longitudinal cracks, leakage rates are strongly related to the crack lengths and internal pressures as was shown in the current study. During the period of stable crack growth, a leakage rate less than 0.03 L/s was calculated in the current study. With the unstable crack propagation, the leakage rate would rise dramatically and become detectable weeks before catastrophic failure of the CI pipe. However, an accurate LBB assessment that incorporates the effects of fatigue damage in deteriorated CI pipes, is required for better rehabilitation of aging pipes.

This work was supported by the Advanced Condition Assessment and Pipe Failure Prediction (ACAPFP) project funded by Sydney Water Corporation, Water Research Foundation of the USA, Melbourne Water, Water Corporation (WA), UK Water Industry Research Ltd, South Australia Water Corporation, South East Water, Hunter Water Corporation, Queensland Urban Utilities and City West Water. The authors would like to thank the supports from Monash Centre of Electron Microscopy (MCEM) and Mr. Yuxiang Wu.