You are on page 1of 8

Proceedings of the 2012 9th International Pipeline Conference IPC2012 September 24-28, 2012, Calgary, Alberta, Canada

IPC2012-90537
ENHANCED USE OF ILI DATA TO IMPROVE INTEGRITY DECISIONS
Collin Taylor Pipeline Integrity, Enbridge Pipelines Inc. Edmonton, Alberta, Canada Collin.taylor@enbridge.com Renkang (Rain) Zhu Pipeline Integrity, Enbridge Pipelines Inc. Edmonton, Alberta, Canada Rain.zhu@enbridge.com

ABSTRACT With the current generation of in-line inspection (ILI) tools capable of recording terabytes of data per inspection and obtaining millimeter resolution on features, integrity sciences are becoming awash in a sea of data. However, without proper alignment and relationships, all this data can be at best noise and at worst lead to erroneous assumptions regarding the integrity of a pipeline system. This paper will explore the benefits of a statistical alignment method utilizing joint characteristics, such as length, long seam orientation (LSO), wall thickness (WT) and girth weld (GW) counts to ensure precision data alignment between ILI inspections. By leveraging the fingerprint like morphology of a pipeline system many improvements to data and records systems become possible including but not limited to: Random ILI Tool performance errors can be detected and compensated for. Repair history and other records become rapidly searchable. New statistically accurate descriptions are created by leveraging the sensitivities of various ILI technologies. One area of material data improvement focused on within this paper relates to long seam type detection through ILI tools. Due to the differing threat susceptibility of various weld types, it is accordingly important to identify the long seam weld types for integrity management purposes. Construction records of older vintage lines do not always contain information down to the joint level; therefore, ILI tools may be leveraged to increase the accuracy of construction records down to this level. In this paper, the possibility of ILI tools, such as magnet flux leakage tools, ultrasonic crack tools, and ultrasonic metal loss tools, to distinguish different types of longitudinal seam welds is also discussed.. STATISTICAL METHOD For just over a decade Enbridge has had a standing policy within its ILI Vendor Contracts stating: It is important that a consistent GW number convention is used for many reasons, which all stems from the ability to identify a specific piece of pipe on consecutive ILI run data. It is also important that Enbridge eliminates excavating a previously repaired or re-coated GW which they confirm by a correlation of the present ILI data to past excavation sites; identified by GW. The enforcement of consistent and repeatable GW labeling between ILI inspections and vendors was the first step towards large scale data consistency and measurement repeatability. In the early years of this policy the alignment and validation of the GW numbering standard was a manual and time consuming process especially when considering pigable segments within the Enbridge system contain between 30,000 and 60,000 GWs. On average, the alignment between two ILI data sets could take several hours of manual work, at the end of which if an error was present, it was nearly impossible to determine if the discrepancy was within the current data set or the historic one. The only way to narrow down the source of error was to align a third dataset and look for consistent discrepancies where two thirds of the data sources agreed. The problem was this additional alignment also lead to additional hours of manual work. That was until the development of an Excel based alignment method that was implemented in early 2008. This method is being improved continuously. The creation of the GW correlation macro reduced the statistical assessment of ILI datasets from hours to minutes. By leveraging the fact that all ILI vendors were required to conform to the GW numbering system the alignment method was reduced to a simple lookup and differential system. The fingerprint-like characteristics of a pipeline lead to a highly

Copyright 2012 by ASME

accurate and repeatable measurement of the system with any preceding errors in measurement becoming eliminated at the boundary of each new joint. This method, by default, aligns three datasets thus allowing for a statistical approach and an enhanced ability to focus in on run specific ILI performance issues. When correlating data sets, Enbridge typically allows for joint measurement tolerances of length = +/- 10% (or 1 foot), LSO = +/- 15 degrees and WT = +/- 10%. All deviations beyond the tolerances are investigated and, when required, ILI vendors are contacted for explanations or corrections and reissues. By employing a form of noise cancelation, known internally as Data Democracy where the majority measurement between the three data sets is considered to be most likely, we are able to eliminate random errors and identify the specific data set with potential performance issues. The results of this process are then compiled into a single source of pipeline data called the Baseline PipeBook. This has proven to be exceptionally valuable as it provides statistically accurate and repeatable data which can be quickly used across all ILI runs through that segment. Since the adoption of this method several interesting ILI tool performance issues have been observed and compensated for prior to becoming injurious to the overall data quality. EXAMPLES OF ILI PREFORMANCE ISSUES 1. Odometer damage resulting in 2% length loss past point of incident

indicative of run specific odometer slips or other trivial wanderings. Notice however, in the second half of the run there is an obvious and systematic change in how the current tool is reporting the joint lengths as compared to two historic data sets. Upon further investigation between Enbridge and the vendor, it was discovered that the above behavior shift was due to damage experienced by the odometer wheel as it passed a valve site. The actual damage to the odometer was nothing more than a single tooth chipped in the gear system which resulted in the under-recorded length and larger statistical spread in measurements averaging around 2% length loss. Typically, even an entire ILI run with less than 2% is considered well within specification; however, due to the obvious nature of an event occurring within in the ILI Run the vendor was requested to compensate for the length discrepancy within any of their length dependent feature calculations (rupture pressure calculations: modified B31.G and effective area for example). In this particular instance, when recalculating features to compensate for lost length, several features were found to meet Enbridges excavation criterion where prior to the recalculation they had not. 2. Out of spec tool rolls around sharp corners or rapid elevation changes During a recent ILI run, a new tool was used which required a preferential orientation and was therefore bottom weighted. Typically, tools are axially neutral and allowed to roll throughout a run and are not subject to out of specification rolling. In this particular case however, the statistical GW correlation exposed LSOs wandering away from and then back to nominal orientation over several joints. This behavior seemed odd until the roll information was overlaid onto a Google Earth plot and the local pipe elevation profile taken into account.

Figure 1: Statistic Length delta plot between a primary data set and 2 historic counterparts. Blue = Primary History(1) Red = Primary History(2). Solitary points in either blue or red = random error in historic data set. Overlaid points = likely error in Primary data set.

Figure 1 is an example of a length delta plot over roughly 40,000 GWs between the primary tool and two historic data sets. Each joint in the data set is directly compared to its historic counterparts producing a full run delta plot where random errors are seen to oscillate around zero and averaged out over the run. The random spread in the first half is

Figure 2: An exaggerated elevation profile of a pipeline river crossing which is color coded for Long seam deviation away from nominal as an ILI Tool swings in the line, Green = within 10 degrees noise, red = peak deviation of 48 degrees.

As can be seen from Figure 2, the bottom weighted ILI tool experienced a roll due to rapid direction / elevation change through the river crossing. Surprisingly, the vendor was not

Copyright 2012 by ASME

aware of this rolling prior to the statistical alignment as the inertial pendulum on board the tool was also experiencing centripetal acceleration pulling it down. Only after the area was highlighted to the vendor did we get confirmation that indeed the LSOs appeared to be twisting, indicating a tool roll. 3. Human and computer errors in above ground marker (AGM) placements The same method of statistical alignment and GW repeatability has been applied to other data collection techniques with great success. AGM placements, for instance, during an ILI run can be a rather hectic activity while chasing a tool to ensure it does not become stuck in the line and those pump stations are bypassed. Occasionally markers have been placed on the opposite side of a highway without proper documentation which has resulted in unnecessary road excavations. In an effort to improve data quality and reduce the effects of human error a more statistical approach was employed. By leveraging the GW repeatability between ILI runs and the theory that consistently places AGMs will occur at roughly the same range within the data then errors of any kind will stand out as deviations from the average. Not surprisingly, this method when first utilized, was quite effective at flagging human placement errors; however, the method was also sensitive enough to detect systemic errors based within an apparent clock synchronization error for many 2010 runs.

method required reissue which had a significant impact to these programs. In light of the susceptibility for errors within the AGM placement process and the impact to excavation programs Enbridge is now generating statistical marker lists for all lines. These statistical data sets provide a readily reusable listing of accurate markers within the system and have reduced the dependence on field accuracy. 4. Additional benefits to overall data management There have been other gains within pipeline integritys overall data structure by including the GW number on all pertinent records related to an excavation program from dig packages to non- destructive examination (NDE) records. By coupling the GW number as a primary key in related excavation records, NDE reports or pressure restrictions, the process for history research prior to executing a dig program has become more efficient by orders of magnitude. In the past, records research could take several hours or days of manual records retrieval and verification. Now, by enforcing all new records to conform to a strict filing structure and GW based naming convention, records research and alignment to ILI feature listings requires just minutes of work for the rapid, accurate, and automatic alignment for thousands of records directly to the target joints. Also, by employing a few tricks with network path mapping an active hyperlink can be included within the Excel based ILI feature listings allowing for one click access to the records.

Figure 3: Example of statistical AGM placements between ILI Runs.


Valve sites in green are directly visible to the ILI tools and have negligible offsets of only a few centimeters. Human errors in red vary randomly and are obvious standouts. Systemic computer or algorithmic errors present as consistent shifts throughout the run

The case shown in Figure 3 serves as an excellent example of human error, algorithmic error, and true repeatable markers within a single snapshot. The human error significantly differs from standard deviation of the data set whereas the consistent fixed point valves experience almost no shifting. The algorithmic error is seen as a consistent 30 meter shift between nearly all AGM sites throughout the run. The source of the 30 meter shift was traced back to an algorithmic error in the post process alignment method used by the ILI vendor to place the AGM points within the data. Through further investigation it was revealed that nearly all of the 2010 ILI runs using this

Copyright 2012 by ASME

POSSIBILITY OF LONG SEAM IDENTIFICATION THROUGH ILI

WELD

TYPE

Long seam failures have been seriously considered due to the defects inherent to the seam welding process, which can cause a lack of fusion, burned metal defects, stitched welds, cold welds, fatigue cracks, cracks in the hard heat affected zone (HAZ), surface breaking hook cracks near the weld, and selective seam (axial grooving) corrosion. Different weld types show different resistant abilities, and as a result, the factor of different seam welds for internal design pressure is also different. (1)

In 1812 Osborne developed the first machine to realize hammer lap-welding, many long seam weld types have been developed since then, such as low-frequency (LF) electric resistance weld (ERW) in 1924 and high-frequency (HF) ERW in 1960s, submerged arc weld (SAW) seam line pipe in 1930 and double submerged arc weld (DSAW), and seamless line pipe as early as in 1840. The reliability of a longseam weld keeps improving till now. (2) Due to the various types of long seam welds used in the pipeline industry (3), it is a challenge for pipeline integrity professionals to distinguish weld types on each individual joint amongst the thousands of miles of pipelines. The ability of ILI tools to classify different long weld types is investigated in this work. The methods to identify a long seam weld type are developed based on ILI data and NDE pictures. 1. MFL and AFD tools MFL technology is the most commonly used ILI in order to detect internal or external corrosion. The MFL tool uses a circumferential array of MFL detectors embodying strong permanent magnets to magnetize the pipe wall to near saturation flux density. (4,5) Anomalies in the pipe wall, such as corrosion pits, generate a magnetic flux leakage near the pipes surface; this flux is detectable by probes on the MFL tool.

Figure 4: Screen shots of MFL signal at Flash Weld (FW) (a), DSAW (b) and ERW (c) long seam welds. The upper one of each pair is the weld images and the lower plot is the MFL signal lines. (Possible weld geometries were illustrated at left side.)

The weld region on the pipe could be regarded as on axial long defect on pipe, which could be identified by the MFL tool. Figure 4 shows MFL ILI images of various weld types on line pipes. The long seam weld seems much clearer on DSAW pipe than on ERW pipe or flash weld (FW) pipe. On the line plot figures, about 3 lines are related to the weld region (zigzag line at 10:30) on FW weld region. There are about 8 zigzag lines related to the DSAW weld region and the two interfaces (base metal-filler material) could be seen clearly. However, for the ERW weld, the signal related to the weld region is quite difficult to observe (Figures 4a and 4c). The DSAW long seam is manufactured with filler material to bond two edges together as illustrated in Figure 4b. There should be one small gap with filler material existing along the weld. If the filler material is not as magnetically sensitive as iron, it appears as a crack and can be distinguished by the MFL tool. If the filler material is magnetically sensitive, the interface between the pipe material and filler material may result in a slight difference of the MFL signal, which may be identified by an experienced data analyst. On the other hand, ERW long seams are manufactured using electric heating to bond two edges together; there is no filler material existing at the weld. The slit between two edges is so small that it is very difficult to see with the naked eye. Although the interface between two edges still exists, the difference between the bulk pipe and the interface is too weak for magnetic tool to clearly identify and the long seam cannot be clearly seen as illustrated in Figure 4c. Therefore, it is possible for the MFL tool to distinguish SAW/DSAW from ERW or FW. Since the manufacturing process of FW is very similar to that of ERW, it could be expected that it is very difficult for MFL tools to distinguish these two weld types. There is almost

Copyright 2012 by ASME

no chance for an MFL tool to distinguish HF-ERW from LFERW/DC-ERW; although, the HAZ on HF-ERW welds is smaller than that on LF-ERW or DC-ERW welds. Prior to 1960, all ERW materials and EFW pipe were made by means of DC or LF-ERW (frequency<306 Hz), and starting in 1960, mills began to convert LF-ERW to HF-ERW (Frequency at 450,000 Hz). After 1978, it is believed that few, if any, LF-ERW were still being used. (6) In absence of concrete information, pipe vintage may be used to venture a differentiation between these two difference welds. Circumferential MFL tools (CMFL) differ from conventional MFL tools in that the magnetic field direction is perpendicular, not parallel, to the pipe. The purpose of this is an increased ability to detect axially oriented defects on the pipe wall, such as selective seam corrosion (SSC). It is clear that the long seam, a feature on pipe wall and in an axial direction, could be detected more clearly with a CMFL than with a conventional MFL. However, due to limited data this paper does not investigate the effectiveness of CMFL tools in differentiating between these three long seam weld types.

each bounce. Echo loss is frequently observed on DSAW long seams. Figure 5 shows the USCD images of three weld types. Figure 5a shows the signal response on FW weld region. As expected, the signal intensity is quite strong. The width at circumference is about 4 degrees out of 360. Figure 5b shows the SAW signal; also quite strong. The width at circumference is about 8 degrees out of 360, which is much larger than that observed on FW. Figure 5c shows the ERW signal; it seems that the detected signal density on this image is less than that either on FW or SAW. The width at circumference is about 4 degrees out of 360, which is similar to FW but much smaller than that of SAW.

2. UltraSonic Crack Detection (USCD) tools In a USCD inspection, the ultrasonic waves are sent through the pipe wall at an angle close to 45o. (6) Any reflector will send a part of the wave back to the transducer, where the signal will be recorded as an echo. If there are no cracks or other reflectors in the pipe wall, no echo could be observed. In terms of a reflector the long seam could be regarded as a long axial crack, regardless of the weld type. Accordingly, long seams are visible to ultrasonic technology. Because FW and ERW weld types, have similar processes it is difficult to differentiate them based on this USCD signals. However, if there is some geometrical difference or some metallurgic interface difference between these two long seam types, it is still possible to differentiate them. Since LF-ERW is a different manufacturing process than HF-ERW, and the HAZ of LFERW is larger than that of HF-ERW, the metallurgic interface is also possibly different from base metal to HAZ, and from HAZ to weld. It is difficult, but these differences may be identified by ultrasonic technology. When an ultrasound wave hits a SAW interface, and if the weld interface is V-shaped (Figure 5b), a strong response is expected. Due to the bigger weld region on circumferential orientation, the signal region on the whole circumferential image is often quite larger than that on a FW or ERW. Echo loss is rare on both SAW and FW/ERW long seams. However, when the ultrasound wave hits a DSAW interface (normally X-shaped as illustrated by Figure 6b), the resulting signal is weak since the propagation wave in the pipe wall is reflected more times than in either FW or SAW long seams and energy is lost on

Figure 5: USCD signals at FW (a), SAW(b) and ERW (c) weld types

Figure 6 shows the USCD response on SAW (6a) and DSAW (6b). The detected signal density on DSAW is much less than on SAW. Also the circumferential width of the DSAW is about 3 degrees out of 360, which is much smaller than the SAW. Using signal strength and circumferential width it is possible for a USCD tool to differentiate between a SAW and a DSAW.

Figure 6: USCD signals at SAW (a) and DSAW (b) weld types

Copyright 2012 by ASME

3. UltraSonic Wall thickness Measurement (USWM) tools USWM is quite similar to USCD, but the wave incident angle is 0 degrees. The wall thickness is derived from the time it takes between an ultrasonic wave being produced and its reflection received.

Figure 7: Ultrasonic wall measurement signals at different long seam weld types. a: FW; b: ERW; c: DSAW and d: SAW.

Figure 7a shows the C-scan plot of the USWM tool on FW pipe. Three continuous pixel lines (signal) can be observed, a small portion of the whole circumferential area. No echo loss (green pixels) is observed. Figure 7b shows ERW pipe. One continuous pixel line (signal) can be observed and the signal region is smaller than that observed on either FW or DSAW/SAW. Figure 7c shows a DSAW long seam and Figure 7d a SAW long seam. Three continuous pixel lines can be observed on both welds; this is similar to FW. Also, echo loss is observed on both weld regions. This information may be used to identify the different weld types, such as FW out of ERW, or DSAW/SAW out of ERW. However, it seems that it is difficult to distinguish between FW, DSAW or SAW. The echo loss could be used to distinguish DSAW/SAW with other welds. However, echo loss can be caused by other factors and this means of differentiation should be employed with caution. The stand-off signal for different weld types was also investigated and the results are shown in Figure 8. The standoff signal on both ERW (Figures 8a) and FW (Figures 8b) is very similar to the signal on base metal. However, a slight negative shift of USWM signal could be observed on DSAW/SAW (Figures 8c and 8d), which is distinguishable by the tool. However, due to the strong echo loss at DSAW/SAW, sometimes the stand-off signal at these two types of welds could be a positive shift from the base metals signal, which could be paid attention to during the analysis. By stand-off signal analysis, the USWM tool can differentiate DSAW/SAW from FW.

Figure 8: Screen-shots of 4 different long seam weld types on pipelines with Standoff readings showing weld morphology. A: FW; b: ERW; c: DSAW; and d: SAW.

Copyright 2012 by ASME

4. NDE pictures for long seam weld identification Three different weld types in NDE reports were reviewed and the typical ones are showed in Figure 9. The large square extrude long seam of FW long seam is unique to A.O. Smith products (Figure 9a), which is helpful for NDE crews to differentiate FW from other weld types.
Group 1 2 Welds

Table 1:
List of possibility of identifying long seam welds by different tools MFL USWM USCD P Y Y P P Y Y NDE Pictures P Y Y N N Y Y

ERW-FW N P ERW vs. P Y SAW/DSAW 3 FW vs. Y P SAW/DSAW 4 SAW vs. N Un DSAW 5 HF-ERW vs. N N LF-ERW 6 Spiral Y Y 7 Seamless Y Y *N = No; Y = Yes; P = Possible; and Un = Unknown.

Figure 9: Pictures of 3 different long seam weld types on line pipes. A: FW; b: SAW/DSAW; and c: ERW.

Normally, SAW and DSAW have the similar round weld top-shape, as shown in Figure 9b. This feature allows NDE crews to differentiate them from FW or ERW. The geometry of ERW is very similar to FW weld. It is flat on the top and most of them seem not to extrude out of the surface. Possibly, it is dependent on how it was trimmed in the pipeline mills. There are two possible methods used in ERW: HF ERW and LF ERW. No visible difference was found on the NDE pictures between HF ERW and LF ERW. CONCLUSION: By shifting the focus of ILI data positioning away from above ground references and instead aligning data based on GW characteristics new levels of data precision become possible down to the millimeter scale. Additionally, records management, data alignment, complex threat integration and material trending also benefit from the statistical alignment method and bring analysis closer to the nano-integrity scale. It is also possible to use ILI data in long seam weld identification. Based on the above research and theoretical analysis, several conclusions could be drawn which were summarized in the following table regarding the detectability of long seam weld types through ILI data (Table 1):

THE FUTURE OF STATISTICAL ILI ANALYSIS: There are other areas within the pipeline integrity sciences which are beginning to leverage the GW repeatability providing greater levels of data integration accuracy and data mining. Below is a quick list of several examples from within Enbridge: Multi-level data threat integration. All dents, corrosion and cracking features typically detected by separate ILI technologies are now post processed and aligned for a more complete picture of the threats present on the system. This alignment process is accurate down to the feature level and is used to flag any potential areas of complex feature interactions which can affect growth rates or reliability. Material trending and proactive data mining for prediction of feature growth and susceptibility on a joint by joint and feature by feature level. Construction records for new lines are now detailed enough to tie directly to each joint within subsequent ILI data sets. Heat number, batch, roll number, material records, and weld x-ray records. New levels of causative analysis are possible by quickly searching out a joints family within the system to determine or rule out construction, manufacturer or other potentially systemic material problems.

Copyright 2012 by ASME

REFERENCES [1] ASME, Pipeline Transportation Systems for Liquid Hydrocarbons and Other Liquids, ASME B31.4-2009, 2010, P15 [2] J. F. Kiefner and E. B. Clark, History of Line Pipe Manufacturing in North America (ASME Research Report), CRTD-Vol. 43, P294 [3] Enbridge Pipeline Inc., Main Line Route Maps. Pipeline Integrity, Enbridge Pipeline Inc. 2011 [4] M. Beller, A. Barbian and D. Strack, Combined In-Line Inspection of Pipelines for Metal Loss and Cracks. Hydrocarbon World, 2007

[5] Robert Bricherstaff, Mark Vaughn, Gerald Stoker, et al. Review of Sensor Technologies for In-line Inspection of Natural Gas Pipeline. Pipes and Pipelines International, 2002, 48 (1), P25-28 [Error! Reference source not found.] Michael Baker Jr., Low Frequency ERW and Lap Welded Longitudinal Seam Evaulation (Final Report), Rev. 3, Integrity Management Program Delivery Order DTRS56-02-D-70036, TTO Number 5. 2004 [6] A. Barbian, M. Beller and U. Schneider, Advanced Pipeline Inspection Utilizing Ultrasound: Special Applications. Pipeline Technology Conference, Hannover, Germany, 2007

Copyright 2012 by ASME

You might also like