Abstract
Managing the integrity of oil and gas, and utility infrastructure via non-destructive testing requires costly, labor-intensive inspections by highly trained personnel. For example, safely surveying hard-to-access pipes at a refinery commonly requires building scaffolding to enable measurements. In the case of ultrasonic testing of pipe wall thickness, it is often necessary to remove insulation that must later be repaired. When maintenance must be performed at regular intervals the costs and risks to personnel can rapidly escalate. These challenges indicate an unmet need for a contactless (over insulation and coatings), area-covering (insuring inclusion of the deepest part of the defect), long-life (up to 10 years), remote monitoring system (status is sent to a central office) that is easy to install. We introduce the WiSense (wireless sensing) NDT method to fill this gap. In the WiSense system, a low-power, battery-operated array of magnetometers is strapped onto the surface of a pipe or insulation, and periodic passive measurements of the magnetic field near the pipe are taken. A low-power wireless radio, adjacent to the magnetometer array, communicates the measurements to a remote analysis station where algorithms identify corrosion and erosion defects via the presence of specific patterns in the detected magnetic fields. This enables not only detecting and monitoring new areas of corrosion as they develop, but also tracking corrosion growth in problem areas. The system is capable of monitoring a single remote location for up to ten years on a single battery source. Alternatively, due to its simple installation, a WiSense array can be moved from place to place to survey large surface areas of pipe without the need to remove insulation prior to measurements. In this paper, we will review the physical principles behind the WiSense technique, including the physics of pipe magnetization, the nature of the defect signatures, and magnetic sensing. We will present an overview of the data processing techniques used to identify regions of corrosion and track its growth, with some examples. Finally, we will discuss how the WiSense technique fills an important gap in the NDT landscape not currently addressed by any other method and present specific use cases to demonstrate its benefits for utility, oil, and gas applications to provide improved situational awareness and operational flexibility at reduced cost and risk to personnel associated with traditional methods.
1. Introduction
Non-destructive testing (NDT) of energy infrastructure components is a key aspect of maintaining the integrity, safety, and long-term profitability of upstream, midstream, and downstream operations. NDT techniques typically exploit various mechanical or electromagnetic properties of the pipes used in the production, transport, and processing of oil and gas products. Common examples include ultrasound, eddy current, and magnetic flux leakage.
Ultrasound techniques(1) couple acoustic energy into the pipe, detect the reflected power, and analyse the signals to determine the thickness of the pipe wall. By scanning the probe across the area of interest, an image of the pipe wall thickness can be generated to identify areas of local corrosion with accuracies, in practice, of +/-0.5mm. Several limitations arise from the need to efficiently couple acoustic waves into the material being tested. First, direct contact must be made with the pipe, so a contactless measurement is not possible. As a result, insulation and coatings must be removed prior to the measurement and repaired afterward. Second, it is a challenge to maintain good acoustic coupling over many years of exposure to the elements, presenting a challenge to long-term, remote monitoring. Third, the acoustic power required to perform the active measurement imposes a practical limit on the length of time that a system can be operated on battery power, further limiting the potential scope for long-term unattended monitoring systems.
Eddy current techniques(2) apply a time-varying magnetic field near a metallic surface, measure the induced currents in the material, and analyse the collected data to identify flaws in the tested part. Discontinuities in the material lead to detectable alterations the eddy current signature. Several applied signal strategies exist, including single-frequency, multi-frequency, pulsed, and remote-field. The various schemes differ in the frequency content and/or geometrical configuration of the excite and detect coils to enable searching for defects at different depths in the material. Eddy current is best suited to detect sharp discontinuities (i.e. cracks) as opposed to wider areas of corrosion. The electrical power required to induce the eddy current response, along with the amount of data that must be transmitted to the processing location, limits its suitability for long-term unattended monitoring systems.
In the magnetic flux leakage (MFL) method(3), large permanent magnets or electro-magnets placed near the pipe saturate its magnetization. Simultaneously, a magnetic sensor measures the fields emanating from the saturated pipe. Anomalies in the measured field due to pipe wall loss are identified via algorithms to estimate the remaining thickness of the pipe wall. Saturating the pipe magnetization ensures that the maximum signal-to-noise ratio is obtained, many inhomogeneities and environmental effects are reduced and helps to guarantee that the measurements are repeatable. However, generating the required magnetic field levels requires hardware with some combination of large size, weight, power, and cost. As a result, MFL is also not well-suited for long-term unattended monitoring systems.
The WiSense method of NDT for oil and gas pipe infrastructure finds its roots in MFL because it also measures the magnetic fields near pipes to identify regions of corrosion. However, WiSense is a passive magnetic system, eliminating the MFL step of active magnetization of the pipe. Instead, the magnetic field produced by the pipe is due to residual magnetization present in the pipe due to its past history, such as electromagnetic NDT performed after after fabrication. While the resulting fields are weaker than with MFL, they are strong enough for corrosion-related anomalies to be detected. Because there is no active magnetization source, WiSense system hardware can be designed to be light weight and use minimal power. Specifically, a battery-powered array of low-cost magnetometers is coupled with a radio and mounted to a pipe. Magnetic field measurements are taken on a regular basis (i.e. weekly), and the data is transmitted wirelessly to a data storage and analysis station. Algorithms process the data to determine if the measured magnetic field features match the signature of localized corrosion in the pipe wall. Such a system can be operated for up to a decade on a single battery, enabling low-cost, long-term management of oil and gas infrastructure to determine if new areas of corrosion are forming or to monitor known areas of corrosion for growth. This unique capability fills a crucial gap in the NDT landscape.
In this paper, we will describe the physics concepts that enable the WiSense technique. This includes a brief review of how the ferromagnetic properties of carbon steel pipes lead to measurable magnetic field signatures when wall loss occurs due to corrosion or erosion. We will show how the expected field patterns vary with defect dimensions, the geometry of the pipe, the layout of the magnetometer array, and the distance between the pipe and the array. We will demonstrate how computational models of defect magnetic signatures can be compared with measured field signatures to determine whether these models match the data to determine whether or not corrosion has occurred and provide information on defect size and growth rate. Finally, we will present use cases as to how WiSense fills an important gap in the NDT landscape to yield improved situational awareness at reduced cost and risk.
2. Physics Basis of WiSense
2.1 Pipe Magnetization
Within magnetic materials, the fundamental source of their magnetic properties are the electrons of the individual atoms. The magnetic properties of each atom are accurately described by the physics of magnetic dipoles. Macroscopic examples of objects that create dipole field patterns include a bar magnet and a circular loop of wire with area vector A (vector is oriented normal to surface) that carries a current I. In the case of the loop, the magnetic dipole d is described mathematically as d = A*I. Magnetic dipoles produce a characteristic magnetic field pattern, which is shown in Figure 1 for the cases of dipoles oriented along the X, Y and Z axis, respectively.
For the purposes of this section we will use the simplifying assumption that the defect is small as compared to the distance to the observing magnetometers. In these situations, the defect looks like an idealized dipole (an idealization that has magnetic field strength and direction, yet is a point source with no physical extent). When a defect is either larger, or the magnetometers are closer, the defect starts to look like a summation of laterally-spaced dipoles. These signatures, while more complex than that of a single dipole, follow the same principles as we will describe here.
In a material with magnetic properties (i.e., the potential to be magnetized), the atomic dipoles collectively combine to yield a bulk magnetization vector M, which is equal to the number of dipoles per unit volume and has units of amps per meter (A/m). When a magnetic material is placed in an external magnetic field with magnetic field strength H, the atomic dipoles experience forces that tend to orient them along H, which alters M. This is the process of magnetization (or de-magnetization).
In free space, the magnetic flux density B is related to the magnetic field strength H by the relationship B = μoH where μo is the permeability of free space and has a value of 4π x 10-7. Within magnetic materials, the reorientation of M under the influence of H alters the proportionality constant between B and H from μo to a material-dependent relative permeability μr. In magnetic materials such as carbon steel pipes used to transport oil and gas, the typical values of μr are hundreds of times higher than μo. In ferromagnetic materials, the value of μr is dependent on both H and the history of the pipe’s exposure to magnetic fields in the past. The complexity of this relationship is captured by a hysteresis curve, shown in Figure 2, which plots the relationship between B and H.
The MFL NDT technique is performed in the presence of a large H field. In this case, the pipe is “saturated”, meaning that the large majority of the atomic dipoles are aligned and further increases in H no longer increase B. This has the advantages that the strong pipe magnetization will produce a large signal that is repeatable in subsequent measurements. However, even in the absence of a saturating field there is an opportunity for leveraging pipe magnetization for NDT. The hysteresis curve shows that when H is ramped up from zero, then ramped back down, B in the material does not return to zero.
residual magnetic flux density of Br. This residual magnetization can be leveraged to detect pipe defects in a manner similar to the MFL technique. WiSense depends on the empirical observation that manufactured pipes have a residual magnetization that is sufficiently large to produce signals that can be detected by low-cost magnetometer integrated chips. Relying on Br for defect detection has the disadvantage that the signals are weaker than in MFL and Mr depends on an unknown history of the pipe’s exposure to magnetic fields. However, there is a tremendous advantage to the WiSense technique over the MFL technique in that it does not require expensive, heavy, power-hungry equipment to perform the measurement, which opens up an entirely new set of low-cost, long-term monitoring operational scenarios for infrastructure integrity management.
2.2 Defect Signatures
WiSense measures the magnetic field produced by the residual magnetization of a pipe to determine whether the pipe has lost wall thickness due to localized corrosion. Figure 3 shows the basic concept. Localized corrosion removes ferromagnetic material from the wall of the pipe. Conceptually, this is the equivalent to inserting ferromagnetic material (i.e. a bar magnet) with oppositely oriented M at the location of the defect. Note that unlike electric fields, magnetic fields can radiate from a defect on the inside surface of the pipe wall. Thus the externally mounted magnetometers are able to see both external and internal corrosion. As a result, the expected field pattern around the pipe will be a superposition of the field due to the localized defect and the field produced by the bulk pipe. Because the bulk pipe is very long, its magnetic field patterns vary gradually in the spatial coordinates. In contrast, the localized dipole-like fields vary rapidly in the spatial coordinates. As a result, the expected field in the region of a pipe defect is a dipole-like field pattern superimposed on a constant background with a small constant gradient. The constant background and gradient can be removed using a spatial high-pass filter to isolate the defect signature from the pipe magnetic fields. In practice, the pipe also produces magnetic noise with spatial frequencies similar to those of the defect, which requires further detection steps to separate the pipe noise from the defect signature.
The amplitude of a defect signal depends on both the residual magnetization Mr of the pipe and the amount of material removed. Recall that the magnetization has units of dipole-per-unit-volume, so the expected dipole moment due to a defect is equal to Mr* Vd, where Vd is the volume of the defect. For cases where the defect is larger than approximately the distance to the magnetic sensors, the defect no longer looks like a compact collection of dipoles with no lateral extent, but starts to have signals more complex than those in Figure 2. This more complex behaviour is dealt with in the same way as the compact collection of dipoles in that it is still modelled directly from physics, however it is a topic beyond the scope of this paper.
Once a defect has been identified (detected), the next goal is to characterize it. It is most useful to describe the defects extend. For defects that are small in lateral extent as compared to the distance to the magnetometers (a situation called the far field), the signal is proportional to the total volume of the defect. When the defect’s lateral dimensions are large compared to the distance to the magnetometers (a relationship called the near field), the spatial pattern of the field depends on the lateral dimensions of the defect, and the amplitude of the magnetic signal scales with the depth of the defect. However, because the signals are proportional to the product of the magnetization and defect dimensions, uncertainty in Mr translates to uncertainty in either of these sizing results. This uncertainty can be managed a few ways. First, the magnetic signals can be one-time calibrated for defect depth using a secondary NDT technology (e.g. ultrasound). Second, a corrosion engineer can apply past knowledge of typical corrosion properties such as the aspect ratio (ratio of lateral dimensions to depth) to help put bounds on the depth of the defect.
3. Corrosion detection methodology
3.1 Magnetic Sensing
The initial generation of WiSense hardware consists of a 2-dimensional array of three-axis magnetometers that conforms to the surface of a pipe. In the current hardware implementation, the individual magnetometers in the array are integrated circuits developed originally as a component for the cell phone market, so their cost is low (~ $1 US per sensor). Sixteen magnetometers are arranged along the edge of a printed circuit board with a spacing between the magnetometers of approximately one centimetre. Sixteen circuit boards are packaged together with a board-to-board spacing of approximately one centimetre. The result is a 16 x 16 array with a spacing of approximately one centimetre between the magnetometers. This physical layout allows the user to measure the spatial field patterns near the surface of a pipe.
The array packaging is designed such that the distance between the magnetometers and the surface of the pipe is the same for each sensor, so a single standoff distance (the distance from the pipe surface with the defect to the magnetometers) can describe the array. A constant standoff is achieved by introducing curvature to the array that conforms to the surface of the pipe. This curvature has been achieved in two different ways, as shown in Figure 4. While the adaptability of this hinged arrangement is an obvious advantage, its flexibility introduces additional uncertainty in the physical location of each sensor as compared to the rigid array design. This negatively impacts the ability to accurately capture the field and compare it to a physics model.
Prior to using the array, a calibration routine is followed prior to find the scale factor and orientation of each magnetometer in the array. This enables a precise understanding of the signals measured by the array, which improves the signal-to-noise ratio of the measurements. Following calibration, the patch is ready to be attached to the region of interest on a pipe. In a fielded operation, this could be a permanent installation for long-term monitoring of a particular location where corrosion is expected, or a temporary location if the patch will be moved to several locations as part of a large-area survey for defects. In either case, defects located underneath the array on either the outside or inside of the pipe produce detectable magnetic field features as described in Section 2.
3.2 Defect template matching
In the WiSense system defect detection is performed by looking for the unique signature of a missing-metal defect, based on physics modelling to create the field expected from an either compact or extended collection of dipoles. Because the magnetometer array geometry is known, accurate comparisons can be performed to compare modelled fields to those measured by the magnetometer array, resulting in determination of if a defect is on the pipe surface underneath the array. One algorithmic strategy is to compare the measured fields with the predicted fields for large number of defect “templates”. We have implemented one version of a template matching scheme in which the defects are assumed to have a uniform depth and an elliptical profile on the surface of the pipe, as shown in Figure 5. The defect is parameterized by Mr (magnitude and orientation), semi-major axis, semi-minor axis. As a reminder, the magnitude of Mr (|Mr|) the calculated fields depends on the product of depth and Mr. As a result, depth and |Mr| are interchangeable parameters and depth was not independently parameterized. The fields are calculated for thousands of possible defect templates and compared with the measured fields to identify the defect template that most closely resembles the measurement.
Figure 5 shows a case where a defect was created on a pipe by etching the external surface of the pipe with acid using a fixture with an elliptical surface shape, resulting in a defect semi-major axis of 3.2 cm and semi-minor axis of 2.9 cm. The measured field is shown in the first column of Figure 5b. The far right column shows the fields that best matched the measured field when a dipole physics model was used. In the center column the template model that has actual spatial extent was used. The best fit for the template model is shown in the center column. While not a perfect match to the measured field, it is seen that the template model is able to match more of the features of the measured data than the dipole model. This is due to the defect being in the near field and the template model is able to capture the effect of the dipoles being spread over a lateral extent, rather than the dipole’s model which is limited to only creating a field from a very compact collection of dipoles. The best match template model had semi-major and minor axes of xxcm and yycm, respectively, showing the detector provided an accurate estimate of the lateral dimensions of the 3.2cm x 29cm defect.
Having a defect detection pattern based on a dipole model is adequate for the far field defect detection because defects in the far field look like a dipole (having magnetic field strength and direction, yet no lateral extent as observed from a far field distance)..However, it is not so clear that a template that is an ellipsoidal cylinder is adequate for near field detection. The fact that the ellipsoidal template has lateral dimensions allows it to create many of the near-field magnetic field features seen in defects that have more complex shapes than that of a cylinder. As a detection tool the ellipsoidal cylinder is quite robust, but it has limitations for use as a sizing tool for irregularly shaped defects.
For such defects where the bottom of the defect is not flat and the overall shape is irregular, a more flexible sizing approach is needed. A profiling approach has been developed to provide more degrees of freedom in the search for a good fit that approximates the shape and depth profile of these more complex defects. Testing and evaluation of this profile approach to defect sizing continues.
4. Corrosion detection results
The viability of the WiSense approach is tested in a laboratory environment using manufactured defects. As shown in Figure 6, we have developed a method to chemically corrode pipes with acid using Teflon fixtures that are mounted to a pipe. The non-corrosive Teflon fixture has a well-defined geometry that holds the acid in a well-defined region on the surface of the pipe while it reacts with the pipe surface. The result is an etched defect with well-understood geometry that can be used to evaluate the effectiveness of defect detection strategies.
The template matching strategy has been tested using the manufactured defects produced in the lab. Figure 7 shows the results for two of the defects.
The defect in Figure 7a is a small elliptical defect with a semi-major axis of 1.5 cm and a semi-minor axis of 1.0 cm that was measured with a standoff of 1.25 cm. For this defect, a pure dipole field is a good match to the measured fields, showing that for defects with lateral dimensions that are not larger than the standoff, the far-field dipole approximate is sufficiently good to perform defect detection. It is significant that in spite of the presence of background noise due to the pipe, a defect template was still able to identify the defect.
In contrast, the defect in 7b has larger dimensions, a circular defect with radius = 2.5 cm, and smaller standoff, 0.7 cm. For this near-field case, an elliptical template defect model is a much better match to the observed field. Furthermore, the magnetization of this pipe was not oriented along the X (longitudinal), Y (circumferential) or Z (radial) axes of the pipe, so the defect does not resemble any of the dipole models shown in Figure 2. However, an appropriate parameter search over magnetization orientation finds a defect template that is a good statistical match to the data.
In one test cycle of twenty defects, the most statistically-likely template had dimensions that were equal to the actual dimensions of the defect with ± 0.5cm. In a fielded application these lateral dimensions can be used along with experience-based assumptions on defect aspect ratio to estimate the depth of the defect. Alternatively, when a magnetic field signature is measured that matches a defect template, a secondary NDT method such as UT can be performed once to calibrate the WiSense field-to-defect depth ratio, and subsequent WiSense measurements at that location can be performed to accurately model defect growth over time.
5. Benefits to Integrity Management
WiSense has a unique set of features that allow it to fill some gaps in the current capabilities of NDT systems used in the utility, and oil and gas domains. Among these are the ability to be simply installed by strapping it onto a pipe, and with no requirement for pipe contact, strapping it over insulation. Further, its wireless communication system means there is also no need to lay wires during installation. Once it is on place, it can provide regular observations of the area under the array for up to 10 years. The array provides area coverage, allowing direct determination of the lateral dimensions of the defect and sufficient coverage to monitor a defect site as the corrosion grows. While not providing direct measurement of defect depth, WiSense is very accurate in measuring the per cent change of the defect’s depth, and directly tracks the changes in lateral dimensions. Its depth measurement can be calibrated by a one-time use of an independent depth measurement, such as ultrasound, and this provides the depth calibration information WiSense uses for future measurements at this location.
These unique features of WiSense lead to a set of use cases where WiSense’s strengths provide an advantage or fill a gap in the NDT tool box.
1) For situations where the corrosion location is difficult to precisely locate, or the deepest area may migrate as the corrosion evolves, the area coverage and long-duration-observing capability of WiSense is an advantage. Some examples of these situations are preferential weld corrosion, onset of corrosion at a stainless-to-carbon-steel transition point
2) In dangerous and difficult to access locations, whether due to the height of the corrosion region or a hazardous atmosphere or other dangerous conditions, WiSense’s ability to be installed easily and quickly and its ability to provide information for up to 10 years provides a safety benefit. Units with overheads or significant heights, such as crude distillers, Texas towers, or drain lines on a rig are potential targets for this system. Areas within a sour field or a flare tower represent extreme hazardous areas that a WiSense would reduce the need for periodic entry for inspection.
3) For areas where there is the potential for an unexpected process or operational change that could accelerate corrosion, this system’s ability to both sense small percent changes in defect depth and to provide frequent reports can allow early warnings of any unexpected change in the local corrosion rate. This is of value to not only units that are either upstream or at the input end of a refinery, but also to locations that are unmanned. Examples include the application of inhibitors, tar sands operations, production well top-side piping for wells near the end of their life, and the crude distillers.
4) WiSense’s ability to detect and monitor corrosion through coatings and insulation make it capable of being applied to CUI and some areas of LNG operations.
6. Conclusions
WiSense takes a passive magnetometry approach to NDT for corrosion and erosion detection and monitoring. By using magnetometry, the WiSense magnetometer array can be mounted on top of coatings or insulation, requiring no direct contact with the pipe to see both internal and external corrosion. By being passive, no energy is applied to the pipe to induce stronger magnetic fields, resulting in the WiSense system being low power, so able to have a n operational life of up to 10 years with a single, on-board battery. The inclusion of a low-power wireless system completes the system, allowing simple installations of a detection and monitoring system that can be remotely operated for 10 years.
While exploiting the fact that pipes get magnetized during their production and quality testing, the need for large magnets or power is removed. However, this means that the magnetic field signals that WiSense operates with are much smaller. These small signals have more interfering and confusing signals from the pipe and local environment, requiring a more sophisticated capability to find defect signals within all of this noise. Extensive studies have been performed in simulation and with etched defects on real pipes to understand the behaviour of the magnetic fields produced by defects. Parameterized, physics-based templates of what a defect field should look like in the far field (approximated by a dipole) and the near field (approximated by an elliptical cylinder) have been shown to be robust in detecting a range of defects.
An intrinsic limitation of magnetometry-based systems is that they cannot distinguish between highly magnetized pipes vs. deep defects. While there are many use cases where simply the existence or non-existence of a defect is actionable information, magnetometry systems, such as WiSense, need a one-time secondary measurement of defect depth to provide a depth-to field strength calibration. With or without this calibration, the WiSense system provides precise information of the relative growth of a defect’s depth. In contrast, the system does measure directly the lateral extent of a defect, which is also useful information to a corrosion expert in that this provides information on the nature of the corrosion and a means to estimate the defect’s depth.
WiSense is a gap-filling technology. It is not a tool that necessarily replaces current capabilities in the NDT tool box, but is an adjunct tool that allows detection and monitoring to be performed in situations that were not viable previously. These include:
1) Situations with difficult or hazardous access,
2) Locations that require long-duration monitoring, such as unmanned and places with unpredictable product or environmental conditions, and
3) Situations where corrosion detection and monitoring through coatings and insulation is desirable.
Comments