Parallel topology was identified as the best feasible configuration. It consists of three or more parallel sensors, installed along the entire length of the pipeline, parallel to its axis. Minimum three sensors are needed to fully describe spatial deformation of the pipe and provide for sufficient spatial sensitivity to damage. Information on spatial deformation of the pipeline serves as an early warning about potentially high stresses in the pipeline. High spatial sensitivity to damage is needed as the location and direction of the permanent ground displacement are not known before the earthquake. The capabilities of parallel topology installed onto the pipeline (i) to early identify stressed zones of the pipeline (i.e., before the damage occurred), and (ii) to detect and localize the damage, were validated and confirmed through large scale testing.
Additional parallel topology consisting of one or more sensors can be simply embedded in the soil. This topology cannot provide with spatial deformation or information about the damage to the pipe, but it can detect and localize the movements in the soil that can potentially imperil the pipe, and point to endangered locations before the damage occurred. In the event of the damage, the leaked fluid will locally change the thermal properties of the soil that are then detected and localized by the sensors. The capability of parallel topology embedded in the soil to detect and localize the zones of the soil displacement was validated and confirmed through large scale testing.
Three sensors were evaluated: two commercially available sensors – “Tape” sensor and “Profile” sensor, and one novel sensor – “Cord” sensor. The performances of the sensors were assessed and verified through large scale testing. Tape sensor demonstrated excellent measurement performance if installed onto the pipeline, but it is rather expensive and to some extent difficult to handle on-site; also it gets damaged for higher levels of damage to pipe. Profile sensor is less accurate, but also less expensive; its performance for damage detection on pipeline was not sufficient, however it demonstrated very good performance in detection of soil movements. Finally, the novel Cord sensor is inexpensive, but also very inaccurate, and less sensitive to small damage; nevertheless, it demonstrated ability to detect both damage to pipeline and the displacement of soil, for higher levels of damage and for bigger soil displacements; also it demonstrated exceptional mechanical robustness and was not damaged during tests, even for very high levels of damage. As a final recommendation, combination or separate use of Tape and Cord sensors is proposed for monitoring the pipeline, while combination or separate use of Cord and Profile sensor is proposed for monitoring the soil. These sensors provide with reliable means for detection and localization of:
(i) ground displacement,
(ii) pre-damage stressing,
(iii) small damage (early damage detection),
(iv) large damage and damage propagation.
In order to provide good strain transfer from the pipeline to the sensor, the latter must be glued to the former. Clamping of the sensor was also considered, but this idea is abandoned as it would be technically challenging (the sensors would have to be pre-tensioned between the clamps), time consuming, and expensive. The adhesive was carefully selected for each type of the sensor in order to match with materials to be bonded. As the minimum bending radius of the fiber optic sensors is around 30 mm, the installation procedure should overcome abrupt changes in pipeline geometry, especially at the bell-and-spigot joints where abrupt (90-degree) gap of several centimeters is present between the outer surface of the bell and the body of the jointed pipe segment. Initial solution consisted of installing plastic “bridges” that connect the bell of a pipe segment on one side and the body of the jointed pipe segment on the other side, and allow for “smooth” transition of sensor between two sides. This solution was unsuccessful as the axial compression applied to pipeline moved spigots into the bells and consequently break the sensors by shearing. Two other solutions were identified: (i) to pre-cast the openings in the bell of the joint so the sensor can pass through it and remain non-bended, and (ii) to leave the sensors loose when bridging the gap created by joints and carefully fill the gap with the soil. The first solution was fully proven in the validation test, but it required modification of the bell of the joint. The second solution was less good and it is not recommended: tape sensor was broken due to soil movements, while Cord sensor detected the damage, but results were ambiguous and difficult to interpret.
The method validation testing was performed at The Cornell Large-Scale Lifelines Testing Facility, the NEES site at Cornell University (Cornell NEES Site). A segmented concrete pipe specimen consisted of five segments assembled by grouted bell-and-spigot joints. It was placed in the testing basin, which was 3.40-m wide, 13.40-m long, and 2.0-m deep. The basin had two parts: the movable north part and the fixed south part. The movable north part of the test basin was attached to four hydraulic actuators which were anchored in massive concrete counter bearings. During the tests, the hydraulic actuators were used for controlled movement of the movable part of the basin. The joint between the north and the south part was designed in that way, that a transverse fault oriented 50 degree relative to the longitudinal length of the basin can be simulated. Two tests were performed with different types of the pipeline, and they helped to gather the knowledge on the real pipelines’ structural behavior, assess the performance of the sensors, improve installation procedures, and the most importantly, to validate the developed method.
The data was analyzed at three levels: (i) the raw data collected by fiber optic sensors was cleansed, processed, and analyzed manually in order to assess and evaluate the capabilities of the monitoring system to detect and localize the damage; (ii) the processed data was compared with the data obtained by NEESR Award in order to evaluate its sensitivity and measurement performance, as well as to further validate the results of measurements; and (iii) simple, but effective algorithms based on thresholds “in space” (comparison of strain values along the pipeline) and “in time” (comparison of strain value at each point in time) were applied and demonstrated ability to automate damage detection and localization. These algorithms are based on classical statistical thresholds (multiples of the standard deviation observed over a window of recent measurements, or the standard deviation of current measurements over a local region of sensors). Another simple statistical method that is applicable is modified “Z-score” method. This derives statistical thresholds based on the root-squared deviation from the median (i.e., rather than the mean). Combination of two algorithms is recommended in order to improve the reliability of automatic damage detection and localization. The successful application of algorithms represents the final step and completion of research and development of the method for real-time, automatic or on-demand, assessment of health condition of buried pipelines after the earthquake.
Data from this project is archived and made available to the public through the NEES data repository at: