Z.L. Chou and J.J.R. Cheng, University of Alberta, Edmonton, Alberta;

and Joe Zhou, TransCanada Pipelines Ltd., Calgary, Alberta

As exploration of energy resources develops further into the remote Canadian north, pipeline construction is being pushed further into the unknown. The new pipeline construction in the north has generated a great need for understanding and predicting the behavior of pipelines under harsh northern environmental conditions. Continuous real-time monitoring technology using distributed strain sensors has become a possible method for monitoring the performance of these pipelines in the field.

Recently, an investigation was performed to find the correlation between the distributed strains along the line pipes and the local buckling (wrinkling) of these pipes, and to study the contribution of these distributed strains to the detection of initial wrinkle of buried pipelines. Both experiments and finite element analyses (FEA) led to important findings. Conventional strain gauges and advanced Brillouin scattering fiber optic sensors (BSFOSs) were employed in the experimental programs.

These sensors can measure the distributed strains spacing as close as 50 mm along the line pipes, so that these sensors can detect the wrinkle location along the monitored pipeline. The goal here is to describe the distributed strains measured from the sensors and also from conventional strain gauges. From this, we have proposed a methodology for detecting the initiation of pipe wrinkling and finding the optimal positions of installation of the distributed sensors.

Operating in extreme environments

Increased energy demand along with depletion of conventional oil and natural gas has led to increased activities in the remote Canadian north. Buried pipelines have proven to be an effective method of transporting these resources to their market. However, these pipelines are exposed to some of the most extreme environments, such as freeze and thaw, discontinuous permafrost and soil movement and settlement. These exposures may result in severe loading conditions on the pipes, which could lead to pipe buckling or even fracture. There is a great need to be able to monitor and predict pipe wrinkling when pipelines are being constructed further into these regions.

For thousands of kilometers of buried pipelines in Canadian north, it is very difficult to detect their damage. In the past, to prevent the buried pipeline from buckling failure, engineers would need to inspect the pipeline conditions in the field periodically, and then determine the schedule to relieve the accumulated strains on the pipeline through excavation. The inspecting procedure was mainly based on engineers’ judgement with aid of some field information, such as soil properties, geotechnical information from slope indicators, or information from in-line inspection. In addition, this procedure did not have any active warning system for possible failures between inspections.

These conditions pointed to the need for a real-time automatic system that has the ability to monitor and predict pipe wrinkling. Among various monitoring techniques, fiber optic sensors have shown some potential and have gradually gained attention from pipeline industry as a viable tool in monitoring line pipes. However, even though fiber optic sensing technology has been rapidly progressing, it can only provide limited data at a few specific locations. In general, these discrete measurements cannot effectively reveal the local behavior of critical pipe segments of buried pipelines unless the locations of wrinkling are instrumented. Distributed strain sensors have become an excellent candidate to be used in the real-time warning system for pipe wrinkling due to its unique ability to measure indiscriminate distributed strains along the pipelines.

Research objective

The objective of this research was to study the potential of using distributed strain sensors in predicting the development of wrinkles in a pipe. Brillouin scattering fiber optic sensors were used to measure the distributed strains in a full-scale test under combined internal pressure and bending load. The finite element method was used to develop pipe buckling models based on the full-scale test results. A study of patterns (or signatures) of the strain distributions of pipe buckling of a pipe under combined loading is conducted using finite element analyses, and was verified by the test results. A prediction algorithm for pipe wrinkling using distributed strain patterns and their development was proposed. The ultimate goal of this project was to provide pipeline operators with reliable real-time prediction of pipe wrinkling by using distributed strain sensors, so that preventive measures can be taken to prevent pipe wrinkling.

Test program

A total of 19 full-scale pipe tests were conducted in a test program. The pipe dimensions, materials, and test conditions are summarized in Table 1. A photo and schematic test set-up are shown in Figure 1. In the experiments, in addition to conventional instrumentation, such as strain gauges, linear variable displacement transducers (LVDTs), and rotational meters, a laser profiling system was used to provide deformation profiles along the pipes. Brillouin scattering fiber optic sensors were used in one of the 19 tests (Pipe#14) to measure the distributed strains along the test specimen under combined loading.

The sensors used in Pipe#14 can provide strain distributions as close as 50 mm along a line pipe. The Brillouin scattering fiber-optic sensing system is based on the Brillouin loss technology in which two counter-propagating laser beams, a pulse and a continuous wave, exchange energy through an induced acoustic field. The distributed strains can be obtained through the frequency measured along the fiber-optic sensors attached on a pipeline.

It is noted that for conventional instrumentation, such as strain gauges, LVDTs, and even the point fiber optic sensors – such as Bragg grating fiber optic sensors – the gauge length always affects the measuring accuracy. But, for this kind of distributed sensor, the measurement accuracy depends on the spatial resolution of the sensing region. For the sensor technology used in this program, the measurement accuracy can be controlled under 100 micro-strains. Closer spacing provides higher accuracy, but requires more time for data acquisition.

Monitoring for pipe wrinkles

In the test program, a total of 20 meters of optical fiber was installed longitudinally at 10 positions around the pipe circumference. The optical fiber was spaced at 45° intervals circumferentially, as shown in Figure 2, to monitor the buckling or overstress of the test specimen. Conventional strain gauges were also used to validate the test data acquired from the sensor system. The experimental instrumentation is shown in Figure 2. The results acquired from the BSFOSs were compared with those obtained from strain gauges; and the sensor data was compared with the results from finite element analyses as well.

One of the benefits of using distributed strain sensors is that they can detect the strain anomaly far before the pipe wrinkling occurs. Figure 3(a) shows the deformation profiles measured from the laser profiling system and Figure 3(b) shows the strain profile from the sensor system. In the Figure 3, location represents the longitudinal distance measured from the end of a pipe. These two graphs clearly demonstrate the advantage of the BSFOS system. It is found that even with the laser profiling system, the occurrence of wrinkling cannot be positively identified until the loads are very close to the peak moment, which corresponds to the 1.32% critical buckling strain shown in Figure 3 (a).

In addition, the distributed strains measured with the BSFOSs reveal the same wrinkle location in a much earlier loading stage, i.e. 75% of the peak moment, as shown in Figure 3 (b). On the other hand, when moment applied up to 75% of peak moment, which corresponds to the 0.41% compression strain, there is no sign indicating the wrinkle location in the deformation profile obtained from the laser profiling system. In addition, the laser profiling system would be difficult, if not impossible, to be used on buried pipelines in the field. However, the BSFOS system can be attached continuously on a buried pipe to provide real-time monitoring and warning of the wrinkling of buried line pipes.

Prediction of pipe wrinkles

Since the loading on a buried pipeline can be very complex and unpredictable, monitoring a structural behavior should be based on the qualitative behavioral signature change rather than the exact magnitude of pipe response. For example, unless the complete load history on a line pipe is known, the measured strains represent only the relative measurements rather than the true accumulated strains. Therefore, the strain distributions obtained from field measurements are not necessary to give the true accumulated strain distributions at the critical location.

Consequently, the measured magnitude of strains along the pipe does not sufficiently represent the current strain status of the buried pipeline, and cannot compare directly with the critical buckling strain of the pipe. However, if the behavioral pattern (or signatures) of a wrinkled pipe can be identified through distributed strains along the line pipe, the pipe wrinkle can be detected by monitoring the changes in the distributed strain patterns. The philosophy of the prediction of pipe wrinkle using distributed strain sensors is explained below.

Prediction methodology

It is observed from both tests and analyses, as shown in Figures 3 and 4, that when wrinkling initiates in a pipe, a localized strain increases at the wrinkle location. However, the increasingly localized strain can also be caused by local damages, such as corrosion or a dent. Therefore, in order to distinguish the wrinkling event from other events for a buried pipeline, it is necessary to consider not only the strain localization but also the distributed strain patterns that are unique to the pipe wrinkling. It is also important to monitor the development of the distributed strain patterns in real time to be able to predict the development of winkling before it happens.

The methodology for predicting the pipe wrinkling can be illustrated by a pressurized pipe under bending load. Figure 5 shows that using the developed FE model (Pipe#14 is used here) to simulate the distributed strains on compression face (regarded as zero degree position) of the pipe at different loading stages. Each load stage shown in Figure 5 represents a specific time of loading on the pipe. The figure clearly shows distinct strain distribution development for a buckling pipe.

Through monitoring the behavioral signature changes of a buried pipe and comparing it with the established behavioral patterns of the pipe buckling, the pipe wrinkling can be predicted. It can be seen that the phenomenon of localized strain concentrated around the wrinkle location (at 170 cm in Figure 5) gradually becomes apparent as the load approaches to the peak moment. As shown in Figure 5, the compression strains at 75% of the peak moment remain relatively constant, but when the moment reaches 86% of the peak moment, the strain distribution shows wave shape distribution, along with one dominating strain localization. Carefully observing the maximum strain on each loading stage, the location of the maximum strain is becoming more definitive with increase of bending moment. However, before the dominating localized strain becomes apparent, the pipe exhibits wave shape of strain distribution. This wave pattern of strain distribution can become a warning sign for the pipe wrinkling.

With increasing load, the imperfection in pipe triggers deformation localization, and results in strain concentration, and eventual wrinkling, at the imperfection location. Even though the localization of pipe deformation can be detected through monitoring the distributed strains, it may not be sufficient to rely simply on these distributed strains, which are measured along a pipe, and on the compression face (critical buckling position) to predict pipe wrinkling. The longitudinally distributed strains at other circumferential positions can be used to further validate the development of pipe wrinkling.

Measuring maximum strains

Practically, the distributed sensors can only be installed in a few specific positions, and hence the maximum strains measured on these sensors may not be the true maximum strains on the monitored pipe. In Figure 6, it can be seen that the sensors are installed along a line pipe at 8 different circumferential positions, numbered from 1 to 8, but the maximum strains are located between the 6 and 7 positions circumferentially. In this circumstance, the relation of longitudinal strains along the pipe at different circumferential positions need be studied to insure that the important behavioral signatures will not be missed due to the limitation of the measured positions.

To study the effect of sensor positions, the 22.5° and 45° positions, circumferentially away from the maximum strain position, are selected for studying the strain correlation in different circumferential positions. In addition, a range of loading stages from 70% to 90% of the peak moment was selected, as shown in Table 2. By selecting a specific loading stage, the strain ratios between the selected position and the critical buckling position can be obtained. The relation of strains at different positions with the critical buckling position is shown in Table 2. In this table, the strain ratio represents the ratio of maximum longitudinal strain at the 22.5° or 45° position to the true maximum strain, which occurred at 0° position. The average strain ratios and the corresponding standard deviations are also presented in Table 2.

As can be seen in Table 2, the strain ratios are similar between different test specimens that have different pipe geometry, material properties, and loading conditions. The strain ratios at a specific position relative to the critical buckling position remain relatively constant regardless of the stage of the loading. As expected, as the specific position closer to the critical buckling position, the strain ratios are approaching unity. The table also indicates that the better prediction will be obtained if the specific position is closer to the critical buckling position, i.e. the strain ratios at 22.5° position yielded better agreement than at 45° position for all test specimens at selected loading stages. Table 2 also shows good agreement between sensors and FEA results.

Based on the above findings, a distributed strain monitoring system can be installed along the line pipe at eight different positions with a 45° circumferentially interval spacing, as shown in Figures 6 and 7. The 45° interval spacing represents the possible maximum deviated position to the critical buckling position of 22.5°. Therefore, along the designated measured positions, even if the monitoring system cannot exactly capture the strain distribution in the true maximum strain direction, 90% of the strain ratios to the true maximum strain distribution are sufficient to be used in predicting the pipeline buckling.

Prediction procedures

The procedures for predicting pipe wrinkling using the distributed sensing system can be summarized as follows:

  1. Install distributed strain sensors along a buried pipeline at eight different positions around the circumference of the pipe, as shown in Figure 7. A 45° interval between the positions should be used.
  2. Before the pipeline starts operation, the initial distributed strains are recorded as the reference strain distributions.
  3. A finite element model is established, and the model is used to establish the patterns or signatures of the potential wrinkling of the pipeline under different loading conditions.
  4. A real-time distributed strain monitoring system is installed to produce the true distributed strain curves to be checked against the established signature curves for pipe buckling.
  5. An automatic damage detection algorithm can then be established based on the established buckling behavioral patterns and the measured distributed strains from the distributed sensors at different positions. The buckling behavioral patterns should include the early detection capability, such as wave shape recognition and development of strain distributions at wrinkle location. The ability of the detection algorithm to distinguish the distributed strain patterns of pipe wrinkling from other possible events should also be developed.
  6. Once the wrinkle initiation has been detected, a warning system will inform the pipeline operator to take appropriate action.

Conclusion

The potential of using distributed strain sensing systems to detect pipe wrinkling in buried pipelines was studied in this project using both experimental investigation and numerical modeling. The monitoring of pipe wrinkle using the BSFOS system has been proven feasible and reliable in the test program. The advantage of using distributed strain sensors is that they allow monitoring a pipeline at distributed locations axially and around different positions circumferentially on a continuous real-time basis. The indiscriminate strain distributions along the distributed sensors allow us to establish distinct behavioral pattern or signature changes for pipe buckling. The system can accurately predict the pipe wrinkling at the applied moment as low as 86% of the peak moment.

Acknowledgment

Based on a paper presented at ASME’s 6th Annual International Pipeline Conference, held in Calgary, Alberta, Canada