Methods and systems for metal object feature mapping

Abstract

A method includes: obtaining first data for an object of interest, the first data having a first modality or domain, obtaining second data for the object of interest, the second data having a different second modality or domain, converting the first data into the modality or domain of the second data, aligning the converted first data with the second data such that a physical feature of the first data is aligned with a corresponding physical feature of the second data, determining a transformation applied to the converted first data or to the second data to achieve the aligning, applying the transformation to the first data or the second data to obtain aligned first or second data, and determining a feature map of the object of interest, based on the aligned first data and the second data or the first data and the aligned second data, using a reconstruction model.

Description

Attorney Docket No.: 474060-123Patent ApplicationMETHODS AND SYSTEMS FOR METAL OBJECT FEATURE MAPPINGTECHNICAL FIELD

[0001] Embodiments described herein relate to methods and systems for metal object feature mapping.BACKGROUND

[0002] Metal object features may need to be mapped, for example, to determine the geometries or locations of defects. In some cases, it may be difficult to access the metal object to inspect it, for example, in a pipeline that may be underground, underwater, or otherwise partially or totally inaccessible. Identifying anomalies, e.g., defects, such as corrosion, rust, pitting, or other physical damage, is important to be able to take preventive measures before a failure occurs.

[0003] Magnetic flux leakage (MFL) is a magnetic non-destructive testing (NDT) method that may be used to inspect a metal object. Axial MFL (“MFL- A” or “aMFL”) and circumferential MFL (“MFL-C” or “cMFL”) are two common setups of MFL tools. The inspection capability of magnetic flux leakage (MFL) is subject to the angle between its magnetic field and the anomaly. To get a comprehensive assessment on corrosion anomalies, more and more pipelines are inspected with two MFL techniques with perpendicular magnetic fields, for example, MFL-A and circumferential MFL-C, e.g., with complementary detection capabilities. Currently, inspection data from each MFL tool are analyzed separately, and two inspection reports are generated respectively. Having two separate inspection reports adds to the time and cost of a project.

[0004] Accordingly, there is a need for methods and systems for metal object feature mapping that do not require a separate inspection report for each of multiple inspection techniques.Attorney Docket No.: 474060-123Patent ApplicationSUMMARY

[0005] This disclosure pertains to methods and systems for metal object feature mapping.

[0006] A first aspect of this disclosure pertains to a method, including: obtaining first data for an object of interest, the first data having a first modality or domain, the first data being associated with a first non-destructive inspection technique, obtaining second data for the object of interest, the second data having a second modality or domain different from the first modality or domain, the second data being associated with a second non-destructive inspection technique different from the first non-destructive inspection technique, converting the first data into the modality or domain of the second data, aligning the converted first data with the second data such that a physical feature of the first data is aligned with a corresponding physical feature of the second data, determining a transformation that was applied to the converted first data or to the second data to achieve the aligning, applying the transformation to the first data or the second data to obtain aligned first data or aligned second data, and determining a feature map of the object of interest, based on the aligned first data and the second data or the first data and the aligned second data, using a reconstruction model.

[0007] A second aspect pertains to the method of the first aspect, wherein the aligning the converted first data with the second data includes performing an image registration operation in which a data area of the converted first data is compared to a same-sized data area of the second data.

[0008] A third aspect pertains to the method of the first aspect, wherein the aligning the converted first data with the second data includes performing a template matching operation in which a data area of the converted first data is compared to a different-sized data area of the second data, such that: the data area of the converted first data is smaller than the data area of the second data so that all of the converted first data overlaps with a portion of the second data,Attorney Docket No.: 474060-123 Patent Application or the data area of the second data is smaller than the data area of the converted first data so that all of the second data overlaps with a portion of the converted first data.

[0009] A fourth aspect pertains to the method of the first aspect, wherein the determining a transformation that was applied includes calculating a shift of the converted first data or of the second data.

[0010] A fifth aspect pertains to the method of the first aspect, wherein the aligning the converted first data with the second data is performed automatically by iteratively adjusting a two-dimensional (2D) window of data among the converted first data to match with a 2D window of the second data to within 10 mm of misalignment.

[0011] A sixth aspect pertains to the method of the first aspect, wherein the converting the first data into the modality or domain of the second data includes converting the first data via an encoder-decoder neural network that inputs the first data into a first plurality of convolutional layers, then transforms the output of the first plurality of convolutional layer through a plurality of residual blocks, then transforms the output of the plurality of residual blocks through a second plurality of convolutional layers to generate the converted first data.

[0012] A seventh aspect pertains to the method of the first aspect, wherein: the first non-destructive inspection technique includes an axial magnetic flux leakage (MFL-A) technique, and the second non-destructive inspection technique includes a circumferential magnetic flux leakage (MFL-C) technique.

[0013] An eighth aspect pertains to the method of the first aspect, wherein the first and second non-destructive inspection techniques have complementary detection capabilities.

[0014] A ninth aspect pertains to the method of the first aspect, wherein: the obtaining the first data includes operating a first inspection tool on the object of interest, the first inspection tool collecting the first data using the first non-destructive inspection technique, and the obtaining the second data includes operating a second inspection tool on the object ofAttorney Docket No.: 474060-123 Patent Application interest, the second inspection tool collecting the second data using the second non-destructive inspection technique.

[0015] A tenth aspect pertains to the method of the ninth aspect, wherein: the first nondestructive inspection technique includes an axial magnetic flux leakage (MFL-A) technique, and the second non-destructive inspection technique includes a circumferential magnetic flux leakage (MFL-C) technique.

[0016] An eleventh aspect of this disclosure pertains to a system, including: a data obtaining unit configured to: obtain first data for an object of interest, the first data having a first modality or domain, the first data being associated with a first non-destructive inspection technique, and obtain second data for the object of interest, the second data having a second modality or domain different from the first modality or domain, the second data being associated with a second non-destructive inspection technique different from the first non-destructive inspection technique, and a computing system including: one or more processors, and a memory system including one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations, the operations including: converting the first data into the modality or domain of the second data, aligning the converted first data with the second data such that a physical feature of the first data is aligned with a corresponding physical feature of the second data, determining a transformation that was applied to the converted first data or to the second data to achieve the aligning, applying the transformation to the first data or the second data to obtain aligned first data or aligned second data, and determining a feature map of the object of interest, based on the aligned first data and the second data or the first data and the aligned second data, using a reconstruction model.

[0017] A twelfth aspect pertains to the method of the eleventh aspect, wherein the aligning the converted first data with the second data includes performing an image registrationAttorney Docket No.: 474060-123 Patent Application operation in which a data area of the converted first data is compared to a same-sized data area of the second data.

[0018] A thirteenth aspect pertains to the method of the eleventh aspect, wherein the aligning the converted first data with the second data includes performing a template matching operation in which a data area of the converted first data is compared to a different-sized data area of the second data, such that: the data area of the converted first data is smaller than the data area of the second data so that all of the converted first data overlaps with a portion of the second data, or the data area of the second data is smaller than the data area of the converted first data so that all of the second data overlaps with a portion of the converted first data.

[0019] A fourteenth aspect pertains to the method of the eleventh aspect, wherein the determining a transformation that was applied includes calculating a shift of the converted first data or of the second data.

[0020] A fifteenth aspect pertains to the method of the eleventh aspect, wherein the aligning the converted first data with the second data is performed automatically by iteratively adjusting a two-dimensional (2D) window of data among the converted first data to match with a 2D window of the second data to within 10 mm of misalignment.

[0021] A sixteenth aspect pertains to the method of the eleventh aspect, wherein the converting the first data into the modality or domain of the second data includes converting the first data via an encoder-decoder neural network configured to: input the first data into a first plurality of convolutional layers, transform the output of the first plurality of convolutional layer through a plurality of residual blocks, and transform the output of the plurality of residual blocks through a second plurality of convolutional layers to generate the converted first data.

[0022] A seventeenth aspect pertains to the method of the eleventh aspect, wherein: the first non-destructive inspection technique includes an axial magnetic flux leakage (MFL-A)Attorney Docket No.: 474060-123 Patent Application technique, and the second non-destructive inspection technique includes a circumferential magnetic flux leakage (MFL-C) technique.

[0023] An eighteenth aspect pertains to the method of the eleventh aspect, wherein the first and second non-destructive inspection techniques have complementary detection capabilities.

[0024] A nineteenth aspect pertains to the method of the eleventh aspect, the data obtaining unit includes: a first inspection tool configured to collect the first data using the first non-destructive inspection technique, and a second inspection tool configured to collect the second data using the second non-destructive inspection technique.

[0025] A twentieth aspect pertains to the method of the nineteenth aspect, wherein: the first non-destructive inspection technique includes an axial magnetic flux leakage (MFL-A) technique, and the second non-destructive inspection technique includes a circumferential magnetic flux leakage (MFL-C) technique.

[0026] Other aspects of the embodiments will become apparent by consideration of the detailed description and accompanying drawings.

[0027] This summary is provided to introduce a selection of concepts that are further described below in the detailed description. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used as an aid in limiting the scope of the claimed subject matter.

[0028] Additional features and advantages of embodiments of the disclosure will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of such embodiments. The features and advantages of such embodiments may be realized and obtained by means of the instruments and combinations particularly pointed out in the appended claims. These and other features will become moreAttorney Docket No.: 474060-123 Patent Application fully apparent from the following description and appended claims or may be learned by the practice of such embodiments as set forth hereinafter.BRIEF DESCRIPTION OF THE DRAWINGS

[0029] To describe the manner in which the above-recited and other features of the disclosure can be obtained, a more particular description will be rendered by reference to specific implementations thereof which are illustrated in the appended drawings. For better understanding, the like elements have been designated by like reference numbers throughout the various accompanying figures. While some of the drawings may be schematic or exaggerated representations of concepts, at least some of the drawings may be drawn to scale. Understanding that the drawings depict some example implementations, the implementations will be described and explained with additional specificity and detail through the use of the accompanying drawings in which:

[0030] FIG. l is a flowchart for a method of mapping defects in a pipeline in accordance with an example embodiment of the present disclosure.

[0031] FIG. 2 is flowchart for a method for signal alignment in accordance with an example embodiment of the present disclosure.

[0032] FIG. 3 is a set of graphs showing MFL-A and MFL-C signals simulated from a laser scan.

[0033] FIG. 4 is a set of graphs showing experimental results of data fusion on simulated data in accordance with an example embodiment of the present disclosure.

[0034] FIG. 5 is a graph showing experimental results of a depth comparison on simulated data in accordance with an example embodiment of the present disclosure.

[0035] FIG. 6 is a set of graphs showing experimental results of data fusion on measured data in accordance with an example embodiment of the present disclosure.Attorney Docket No.: 474060-123 Patent Application

[0036] FIG. 7 is a graph showing experimental results of a depth comparison on measured data in accordance with an example embodiment of the present disclosure.

[0037] FIG. 8 is a graph showing experimental results of a depth comparison on a finetuned model in accordance with an example embodiment of the present disclosure.

[0038] FIG. 9 is a set of graphs showing Remaining Strength (RSTRENG) and Plausible Profiles models for box data versus laser scans.

[0039] FIG. 10 is a set of graphs showing RSTRENG and Psqr models for data fusion in accordance with an example embodiment of the present disclosure versus laser scans.

[0040] FIG. 11 is a flowchart for a three-dimensional (3D) reconstruction in accordance with an example embodiment of the present disclosure.

[0041] FIG. 12 is a flowchart for generating a defect profile in accordance with an example embodiment of the present disclosure.

[0042] FIG. 13 is an encoder-decoder network in accordance with an example embodiment of the present disclosure.

[0043] FIG. 14 is a set of views showing two image matching options.

[0044] FIG. 15 is a block diagram of a neural network trained for a reconstruction model in accordance with an example embodiment of the present disclosure.

[0045] FIG. 16 illustrates certain components that may be included within a computer system according to an example embodiment of the present disclosure.

[0046] Before explaining the disclosed embodiment of this disclosure in detail, it is to be understood that the invention is not limited in its application to the details of the particular arrangement shown, as the invention is capable of other embodiments. Example embodiments are illustrated in referenced figures of the drawings. It is intended that the embodiments and figures disclosed herein are to be considered illustrative rather than limiting. Also, the terminology used herein is for the purpose of description and not of limitation.Attorney Docket No.: 474060-123Patent ApplicationDETAILED DESCRIPTION

[0047] Before any embodiments are explained in detail, it is to be understood that the embodiments are not limited in application to the details of the configuration and arrangement of components set forth in the following description or illustrated in the accompanying drawings. The embodiments are capable of being practiced or of being carried out in various ways. Also, it is to be understood that the phraseology and terminology used herein are for the purpose of description and should not be regarded as limiting. The use of “including,” “comprising,” or “having” and variations thereof are meant to encompass the items listed thereafter and equivalents thereof as well as additional items. Unless specified or limited otherwise, the terms “mounted,” “connected,” “supported,” and “coupled” and variations thereof are used broadly and encompass both direct and indirect mountings, connections, supports, and couplings.

[0048] In addition, it should be understood that embodiments may include hardware, software, and electronic components or modules that, for purposes of discussion, may be illustrated and described as if the majority of the components were implemented solely in hardware. However, one of ordinary skill in the art, and based on a reading of this detailed description, would recognize that, in at least one embodiment, the electronic-based aspects may be implemented in software (e.g., stored on non-transitory computer-readable medium) executable by one or more electronic processors, such as a microprocessor and / or application specific integrated circuits (“ASICs”). As such, it should be noted that a plurality of hardware and software-based devices, as well as a plurality of different structural components, may be utilized to implement the embodiments. For example, “servers,” “computing devices,” “controllers,” “processors,” etc., described in the specification can include one or more electronic processors, one or more computer-readable medium modules, one or moreAttorney Docket No.: 474060-123 Patent Application input / output interfaces, and various connections (e.g., a system bus) connecting the components.

[0049] Relative terminology, such as, for example, “about,” “approximately,” “substantially,” etc., used in connection with a quantity or condition would be understood by those of ordinary skill to be inclusive of the stated value and has the meaning dictated by the context (e.g., the term includes at least the degree of error associated with the measurement accuracy, tolerances [e.g., manufacturing, assembly, use, etc.] associated with the particular value, etc.). Such terminology should also be considered as disclosing the range defined by the absolute values of the two endpoints. For example, the expression “from about 2 to about 4” also discloses the range “from 2 to 4.” The relative terminology may refer to plus or minus a percentage (e.g., 1%, 5%, 10%, or more) of an indicated value.

[0050] Functionality described herein as being performed by one component may be performed by multiple components in a distributed manner. Likewise, functionality performed by multiple components may be consolidated and performed by a single component. Similarly, a component described as performing particular functionality may also perform additional functionality not described herein. For example, a device or structure that is “configured” in a certain way is configured in at least that way but may also be configured in ways that are not explicitly listed.

[0051] One or more specific embodiments of the present disclosure are described herein. These described embodiments are examples of the presently disclosed techniques. Additionally, in an effort to provide a concise description of these embodiments, not all features of an actual embodiment may be described in the specification. It should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous embodiment-specific decisions will be made to achieve the developers’ specific goals, such as compliance with system-related and business-related constraints, which mayAttorney Docket No.: 474060-123 Patent Application vary from one embodiment to another. Moreover, it should be appreciated that such a development effort might be complex and time consuming, but would nevertheless be a routine undertaking of design, fabrication, and manufacture for those of ordinary skill having the benefit of this disclosure.

[0052] The articles “a,” “an,” and “the” are intended to mean that there are one or more of the elements in the preceding descriptions. The terms “comprising,” “including,” and “having” are intended to be inclusive and mean that there may be additional elements other than the listed elements. Additionally, it should be understood that references to “one embodiment” or “an embodiment” of the present disclosure are not intended to be interpreted as excluding the existence of additional embodiments that also incorporate the recited features. For example, any element described in relation to an embodiment herein may be combinable with any element of any other embodiment described herein. Numbers, percentages, ratios, or other values stated herein are intended to include that value, and also other values that are “about” or “approximately” the stated value, as would be appreciated by one of ordinary skill in the art encompassed by embodiments of the present disclosure. A stated value should therefore be interpreted broadly enough to encompass values that are at least close enough to the stated value to perform a desired function or achieve a desired result. The stated values include at least the variation to be expected in a suitable manufacturing or production process, and may include values that are within 5%, within 1%, within 0.1%, or within 0.01% of a stated value.

[0053] A person having ordinary skill in the art should realize in view of the present disclosure that equivalent constructions do not depart from the spirit and scope of the present disclosure, and that various changes, substitutions, and alterations may be made to embodiments disclosed herein without departing from the spirit and scope of the present disclosure. Equivalent constructions, including functional “means-plus-function” clauses areAttorney Docket No.: 474060-123 Patent Application intended to cover the structures described herein as performing the recited function, including both structural equivalents that operate in the same manner, and equivalent structures that provide the same function. It is the express intention of the applicant not to invoke means-plus- function or other functional claiming for any claim except for those in which the words ‘means for’ appear together with an associated function. Each addition, deletion, and modification to the embodiments that falls within the meaning and scope of the claims is to be embraced by the claims.

[0054] The terms “approximately,” “about,” and “substantially” as used herein represent an amount close to the stated amount that still performs a desired function or achieves a desired result. For example, the terms “approximately,” “about,” and “substantially” may refer to an amount that is within less than 5% of, within less than 1% of, within less than 0.1% of, and within less than 0.01% of a stated amount. Further, it should be understood that any directions or reference frames in the preceding description are merely relative directions or movements. For example, any references to “up” and “down” or “above” or “below” are merely descriptive of the relative position or movement of the related elements.

[0055] Magnetic flux leakage (MFL) inline inspection tools are one the most utilized inspection methods to detect and size corrosion on a pipeline. The MFL technologies can be run, for example, with the magnetic field aligned axially (MFL-A) or circumferentially (MFL-C) in the pipeline. MFL-A tools are typically used for circumferential, pinhole, pitting, and general corrosion features, whereas the MFL-C focus more on the axially-oriented corrosion anomalies. These tools are complementary; therefore, they are often run together to provide improved confidence in detection and sizing of a wide range of corrosion morphologies on the pipeline.

[0056] To gain the most benefit of running the tools concurrently, the results of both the MFL-A and MFL-C tool runs are typically evaluated together by a highly experiencedAttorney Docket No.: 474060-123 Patent Application human evaluation expert. With deep knowledge of both tools, a human evaluator may use the signals to infer the best length, width, and depth sizing of the corrosion anomalies. This specialized evaluation is extremely labor-intensive, and the sizing process is subjective.

[0057] Evaluation of MFL technologies, whether individually or together, is typically presented as a resulting box around the corrosion area. This cuboid box of length, width, and depth is a simplified way of describing the corrosion anomalies. Although it may be a convenient way to receive corrosion details in a tabular report, much of the information may be lost about the corrosion shape, and conservatism is inherently introduced around the failure pressure calculations.

[0058] Direct field analysis (DFA) may have benefits of moving away from calculation of failure pressures based on clustering of boxed anomalies to calculations on the actual metal loss depth maps. The results of DFA are a step change in presenting data from the two MFL tools as one combined three-dimensional (3D) depth map, as opposed to an evaluated box based on human expert evaluation. The computation of the 3D shape allows direct leverage of the strength of both MFL tools for accurately evaluating a wide range of morphologies on the pipeline, as well as a more accurate calculation of failure pressure.

[0059] However, combining of signals of the two MFL tools into one 3D depth map using a machine learning data fusion model according to example embodiments of the present disclosure is a further evolution of known techniques for metal object feature mapping, e.g., pipeline defect evaluation. An advantage of the methodology used in example embodiments is that, once the tools have been adequately aligned, a pre-trained convolutional neural network (CNN) can be applied effectively to all of the corrosion anomalies on a pipeline. The improvements of computational speed allow the evaluation of every corrosion feature on a pipeline with a full high resolution 3D depth map; like that of an in-ditch laser scan. TheAttorney Docket No.: 474060-123 Patent Application advantages in morphology, depth, and failure pressure calculations will be presented herein through a case study with a tape-coated line, which has a challenging morphology.

[0060] - Technology Outline

[0061] As described above, the combining of two signals is typically a manual process involving a human expert. It requires a very experienced analyst (or team) with expertise in both MFL-A and MFL-C technologies. They must understand the both the strengths and the limitations of each tool to glean the most pertinent information from the tools to properly characterize the feature. This analysis, at best however, only results in a cuboid representation of the feature itself. The method of data fusion on MFL-A and MFL-C outlined herein is the fusion of two input magnet signals into a CNN and outputting one depth map. This image-to- image translation negates the need for experts to attempt to combine individual MFL signals, and it fully describes the feature, not only in depth, but also its entire 3D morphology.

[0062] - Fusion Process

[0063] FIG. l is a flowchart for a method of mapping defects in a pipeline in accordance with an example embodiment of the present disclosure.

[0064] A method 100 includes several operations outlined in FIG. 1. In the example shown in FIG. 1, the method 100 starts with both MFL-A and MFL-C tools being run in a pipeline segment. However, it should be appreciated that the method 100 may be applied to other types of inspection tools than MFL-A and MFL-C. For example, ultrasonic testing (UT) data may be obtained and aligned, e.g., with MFL-A and / or MFL-C data. Furthermore, an example method 100 may include actually operating multiple inspection tools to obtain inspection data, or may begin with receiving inspection data obtained using the multiple inspection tools. It should also be appreciated that the example method 200 may convert more than two different modalities or domains of inspection data from one type to another, and more than two types of inspection data may be aligned to form a 3D reconstructed depth map.Attorney Docket No.: 474060-123 Patent Application

[0065] With further reference to FIG. 1, the MFL-A tool receives an MLF-A signal 110, and the MFL-C tool receives an MLF-C signal 120. The tools may be run as in any traditional service, and no special coordination of the tools may be required. It may not be necessary for the tools to be run on the same (or nearly same) date. However, it should be appreciated that, the further apart the inspection dates of the two tools, the more one should consider the effects of corrosion growth or additional defects being introduced between runs in the results.

[0066] Next is a pre-processing operation for each respective input type, e.g., for each of the MFL-A and MFL-C inspection run results, individually to remove background magnetization and noise in the signal. The MFL-A signal 110 undergoes a first data preprocessing operation 130, and the MLF-C signal 120 undergoes a second data pre-processing operation 140. Then, a signal alignment operation 150 is performed on the pre-processed data from both MFL signals to align the physical location of the features of the pipeline to one another. The signal alignment operation 150 may include a standard pipeline signal alignment to match each joint of the pipeline between the two runs. A fine alignment may then be done on each individual joint to ensure a high alignment of the anomalies from the two tools. The aligned anomalies are then processed by the data fusion model in operation 160, which may produce a highly detailed 3D metal loss depth map in operation 170.

[0067] - Data Pre-Processing

[0068] Example embodiments of the data pre-processing 130, 140 may each include two pre-processing steps performed on the raw signal data. When the tools are MFL tools, the first step is magnetic normalization to compensate for individual tool effects on background magnetization. The next step is to remove individual artifacts in the signal data. The aim is to have robust quality input data for the data fusion model without compromising the signal of the anomalies themselves.Attorney Docket No.: 474060-123Patent Application

[0069] - Alignment

[0070] The initial pipeline alignment may be performed by a standard pipeline process, e.g., aligning the joints of the pipeline, which may be done routinely within the inspection analysis process. However it is a standard process, so is not discussed further here. After the general alignment of the features detected by both MFL runs in a pipe is performed, a detailed alignment of the signals is performed in accordance with an example embodiment of the present disclosure.

[0071] FIG. 2 is flowchart for a method for signal alignment in accordance with an example embodiment of the present disclosure.

[0072] The signal alignment of the two tools is an important step in the process of data fusion. If the signals are not accurately aligned, then fusion of the two signals will be adversely affected. FIG. 2 shows a method 200 used for alignment of the two MFL signals. The data in the images shown in FIG. 2 was derived from experimental results using a method 200 according to an example embodiment.

[0073] The difference of multi-modal MFL signals makes it difficult to conduct matching directly. The signal from one MFL can be translated to the other domain where the mono-modal matching can be achieved. Therefore, the method 200 includes conversion operations 230 and 240 for converting MFL- A signal data into the MFL-C domain or modality, and vice versa. With this conversion, defects in the pipeline or other object of interest that would register in both of the axial and circumferential directions can be directly compared to aid in alignment of the data sets.

[0074] First, using neural networks trained for conversion, measured MFL-A signals 210 were converted in conversion operation 230 into converted MFL-C signals 250, and measured MFL-C signals 230 were converted in conversion operation 240 into converted MFL-A signals 260. The training of the neural network(s) to perform the conversionAttorney Docket No.: 474060-123Patent Application operations 230 and 240 for respectively converting the MFL-A signals 210 into converted MFL-C signals 250 and converting the MFL-C signals 230 into the converted MFL-A signals 260 is done by using MFL-A and MFL-C data obtained through simulation of both MFL-A and MFL-C measurement data, each for a set of different geometries. The geometries are either measured or artificially created so that the geometries capture a wide range of possible effects that MFL-A and / or MFL-C could measure. Thus, the conversion operations 230 and 240 necessarily require the use of a technological solution and cannot be practically performed in the human mind.

[0075] Then, the newly converted signals were matched to the respective measured signals. In other words, the original MFL-A data 210 that was converted into MFL-C data 250 was matched to the original MFL-C data 220, and the original MFL-C data 220 that was converted into MFL-A data 260 was matched to the original MFL-A data 210. The matching (or aligning) of the converted data with the original data in the same modality or domain may be performed, e.g., automatically, by iteratively adjusting a two-dimensional (2D) window of data among the converted data to match with a 2D window of the original data to within about 10 mm of misalignment. This may be performed, for example, using one or more algorithms generally known to be suitable to this task, such as template matching. A two-channel template matching may be used to align the two matched signals sets together. The image of the converted MFL-C data 250 may be automatically aligned to the original MFL-C data 220 in alignment operation 270 to find where the prominent features match to identify a physical location alignment. Also, the image of the converted MFL-A data 260 may be automatically aligned to the original MFL-A data 210 in alignment operation 280 to find where the prominent features match to identify a physical location alignment. Two-channel template matching, in accordance with an example embodiment, allows for performing the matching operations in parallel in both directions (e.g., MFL-A to MFL-C and MFL-C to MFL-A) by matching twoAttorney Docket No.: 474060-123 Patent Application superposed images, both of which include two channels. In the FIG. 2 example, the alignment operation 280 is illustrated as having the first channel of the first superposed image containing the original MFL-A data 210, and the first channel of the second superposed image containing the converted MFL-A data 260. Also in the FIG. 2 example, the alignment operation 270 is illustrated as having the second channel of the first superposed image containing the converted MFL-C data 250, and the second channel of the second superposed image containing the original MLF-C data 220. The parallel matching can achieve better matching results than only matching in one direction.

[0076] The alignment operations 270, 280 may be performed, in accordance with an example embodiment, by identifying peaks in the converted data and matching those locations to the peaks in the original data in the same domain or modality. Once the converted data is aligned with the original data in the same domain or modality, a determination may be made as to what transformation was used to achieve the alignment, e.g., how much the pixels were shifted, for example, in the x-direction (e.g., an axial direction) and the y-direction (e.g., a circumferential direction) in the converted data or the original data to make the physical features in the data line up. For example, a difference between a start position and an end position may be used to identify a shift vector for the alignment. The transformation, e.g., shift, can then be applied to the original data of the type that was not shifted in the alignment operation so that all of the original data of the aligned type may then be aligned with the original data of the other type. The alignment may be performed with either of the converted data types, or may be performed for both, e.g., in parallel, which may help validate and increase accuracy of the alignment.

[0077] In the experiment as conducted, to ensure a high quality of matching, every matched set was reviewed manually and adjusted. The experimental results were then fed into the data fusion model, e.g., as used in operation 160 of FIG. 1.Attorney Docket No.: 474060-123 Patent Application

[0078] - Training of the Data Fusion Model

[0079] FIG. 3 is a set of graphs showing MFL-A and MFL-C signals simulated from a laser scan.

[0080] The data fusion model may have a U-Net architecture. U-Nets are fully convolutional neural networks that use skip-connections in addition to a simple encoderdecoder structure. The training of the network uses a supervised learning method in which paired input data and desired output data are used. The training data may include laser scans. Simulations of those laser scans as shown in FIG. 3. In FIG. 3, part (a) is the MFL-A signal data, part (b) is the MFL-C signal data, and part (c) is laser scan data. One training pair may include the MFL-A and the MFL-C signal stacked on together forming the input and one laser scan patch as the desired output. Thus, the training of the data fusion model necessarily requires the use of a technological solution and cannot be practically performed in the human mind.

[0081] - Fine-Tuning Option

[0082] If an operator has historical laser scans on a given pipeline, it may be possible to carry out a fine-tuning of the data fusion model. The initial or pre-trained model may be trained with an entire original training set. A fine-tuning of the model may start with the pretrained model, and may further train using the historical laser scans. This extra training may tailor the model to the corrosion on a particular pipeline of interest.

[0083] - Validation of the Training

[0084] FIG. 4 is a set of graphs showing experimental results of data fusion on simulated data in accordance with an example embodiment of the present disclosure. FIG. 5 is a graph showing experimental results of a depth comparison on simulated data in accordance with an example embodiment of the present disclosure.

[0085] To validate the training of the model in the experiment, the magnetic responses of MFL-A and MFL-C were simulated on additional laser scans not used in the training, andAttorney Docket No.: 474060-123 Patent Application were processed with the data fusion model. A data fusion example is given in FIG. 4, which shows the high accuracy of image reconstruction. In FIG. 4, the MFL-A signal is shown in part (a), the MFL-C signal is shown in part (b), the resultant data fusion 3D depth is shown in part (c), and the corresponding validation laser scan is shown in part (d). A pixel-by-pixel comparison on depth was then carried out between the original laser scan patch and the resultant 3D prediction image. A unity plot comparison is shown in FIG. 5, in which the mean absolute error (MAE) is 0.26% in depth. As such, the results demonstrate that the model is adequately trained. A case study was conducted to the accuracy of the data fusion on real pipeline data, which is discussed below.

[0086] - Case Study

[0087] To demonstrate the effectiveness of the data fusion technique in accordance with an example embodiment of the present disclosure, a real pipeline validation was carried out. Two inspections were run in a tape-coated segment of a pipeline: an MFL-A Ultra (high resolution MFL-A) inspection in 2019 and an MFL-C inspection in 2021. Twenty laser scans were provided for depth and failure pressure validation of the data fusion model and an additional nineteen laser scans were provided for fine tuning of the model. Depth and failure pressure validation of the model was carried out for both the pre-trained model and the finetuned model.

[0088] - Pre-trained Depth Validation

[0089] FIG. 6 is a set of graphs showing experimental results of data fusion on measured data in accordance with an example embodiment of the present disclosure. FIG. 7 is a graph showing experimental results of a depth comparison on measured data in accordance with an example embodiment of the present disclosure.

[0090] The pre-trained data fusion model was trained on laser scans and simulations from fifty different pipeline segments. This model was used to fuse the MFL-A Ultra andAttorney Docket No.: 474060-123 Patent ApplicationMFL-C signals into 3D metal loss depth maps. FIG. 6 shows an example of the data used for the fusion in which the MFL-A signal is shown in part (a), the MFL-C signal is shown in part (b), the resultant data fusion 3D depth is shown in part (c), and the corresponding validation laser scan is shown in part (d). The similarity of the morphology is immediately apparent in a visual comparison of the data fusion 3D depth map and the laser scan. As such, it was validated that data fusion is effective at reproducing all the fine details of the corrosion structure.

[0091] A comparison of the deepest point in the laser scan anomalies and the data fusion anomalies is shown in FIG 7. The comparison of the depth was performed by boxing the validation laser scan data with a preconfigured boxing routine. These boxes were projected onto the data fusion 3D depth map in the same region and the deepest points of each boxed anomaly was compared. Boxing of anomalies on the twenty validation laser scans provided 5,570 individual boxed anomalies for feature validation.

[0092] The comparison of the deepest points shows 97.5% of anomalies within the ±10% depth band of the unity plot with an associated 4.42% mean absolute error (MAE). Qualitatively, the results of the validation show a slight under-call in the depth of anomalies under 20%wt and a tendency of slight overcalling of some anomalies above about 40%wt.

[0093] Investigating these cases, it was observed that many of the laser scan results had areas of data quality inconsistencies. This is thought to be due to the stitching together of the images in the software for the handheld laser scanners. This may have in part contributed to some of the under-calling of anomalies under 20% depth.

[0094] Some of the deeper anomalies appear to have a prediction that are slightly narrower than actual anomalies, which is likely due to training on feature types that are different from the ones seen in this segment. It may also be due to relative fewer training anomaliesAttorney Docket No.: 474060-123 Patent Application greater than 40%wt. To address this issue, a fine-tuning of the model was performed to tailor the model to the specific of this segment.

[0095] - Fine-Tuned Depth Validation

[0096] FIG. 8 is a graph showing experimental results of a depth comparison on a finetuned model in accordance with an example embodiment of the present disclosure.

[0097] The pre-trained fusion model was fine-tuned using an additional nineteen laser scans that were not included in the original training. The validation of depth was carried out on the same set of twenty laser scans and is shown in FIG. 8. The results of the fine-tuning increased the number of anomalies in the ±10% depth band to 98.3% with an MAE of 3.32%. The comparison of depth shows a tightening of the distribution around the unity line. The previously observed over-calls were reduced, however there were a few minor under calls. In general, the results improved with the fine-tuning, which demonstrates that it is beneficial to integrate previous knowledge of the pipeline feature characteristics, e.g., historical data, into the data fusion model, if available.

[0098] - Failure Pressure Validation

[0099] In the pipeline industry, two failure pressure calculation methods that are commonly used are Remaining Strength (RSTRENG) and increasing Plausible Profiles (“Psqr” or “P2”), both of which rely on two-dimensional (2D) river bottom profiles in their calculations. MFL technologies traditionally only provide boxed data sets in the reports. Therefore, generated river bottom profiles used for RSTRENG and Psqr calculations are conventionally derived from box data by the projection of a path through the box pseudo-3D landscape. Using boxed anomalies for river bottom profiles has an inherent conservatism because the boxed anomalies are represented by the max depth of the corrosion anomaly. The volume loss of feature or cluster is overestimated and results in a conservative failure pressure calculation.Attorney Docket No.: 474060-123 Patent Application

[0100] FIG. 9 is a set of graphs showing RSTRENG and Psqr models for box data versus laser scans. FIG. 10 is a set of graphs showing RSTRENG and Psqr models for data fusion in accordance with an example embodiment of the present disclosure versus laser scans.

[0101] To rigorously demonstrate the conservatism of the boxed data sets vs the detailed 3D depth maps the validation laser scans were sectioned into 180 patches of corrosion and boxed. RSTRENG and Psqr were calculated on each of these patches, for 2D profiles derived from the boxed data sets and the actual laser scans. The result of the comparison is given in FIG. 9. In FIG. 9, part (a) is the RSTRENG calculations and part (b) is the Psqr calculations. It can be seen in FIG. 9 that, for anomalies with a lower failure pressure, there is a conservatism that results from boxing the data. This is consistent for both RSTRENG and Psqr calculations. Therefore, 3D depth maps derived from the fusion should also eliminate this element of conservatism and should yield more accurate results. The results of both RSTRENG and Psqr calculation of a data fusion in accordance with an example embodiment of the present disclosure versus the laser scan are shown in FIG. 10. In FIG. 10, part (a) is the RSTRENG calculations and part (b) is the Psqr calculations.

[0102] FIG. 10 shows two important experimental results. First, failure pressure calculations of the fusion results correlate strongly to the failure pressure calculations of the lasers scan in both RSTRENG and Psqr model calculations. Second, MFL technologies can provide an accurate 3D metal loss depth map using data fusion, which can reduce or eliminate the conservatism traditionally inherent in the failure pressure calculations of MFL technologies.

[0103] In experiments, there was no noticeable difference between the failure pressure results of the pre-trained model and the fine-tuned model. As such, only the pre-trained results are shown and discussed herein for that purpose.Attorney Docket No.: 474060-123 Patent Application

[0104] It has been shown that the data fusion technology can accurately predict the corrosion depth maps using MFL-A and MFL-C technologies together. The depth profiling from data fusion shows accurate morphologies of the corrosion anomalies, as well as a strong correlation in the maximum depth of the anomalies. The ability to reduce or even eliminate the principal conservatism in failure pressure calculations by reducing or eliminating the need to box anomalies allows for much more accurate failure pressure calculations with MFL. While the fine-tuning with laser scans can slightly improve maximum depth values, it may have less influence on pressure predictions.

[0105] FIG. 11 is a flowchart for a 3D reconstruction in accordance with an example embodiment of the present disclosure.

[0106] As shown in FIG. 11, a method 1100 may include inputting data collected with a first non-destructive inspection tool, e.g., MFL-A data 1110 collected from an MFL-A tool, and data collected with a second non-destructive inspection tool, e.g., MFL-C data 1120 collected from an MFL-C tool. The MFL-A data 1110 and the MFL-C data 1120 may be passed through a trained reconstruction model 1130 to generate a 3D profile 1140.

[0107] FIG. 12 is a flowchart for generating a defect profile in accordance with an example embodiment of the present disclosure. FIG. 13 is an encoder-decoder network in accordance with an example embodiment of the present disclosure.

[0108] As shown in FIG. 12, a method 1200 may include combining signals from at least two non-destructive investigation tool types, e.g., MFL-A and MFL-C, that were used to inspect an object of interest, e.g., a pipeline. The signals from the two MFL systems may be processed to eliminate the measurement variations. For example, the MFL-A signal data 1210 may be denoised in operation 1230, and the MFL-C signal data 1220 may be denoised in operation 1240. Then the denoised signals may be matched in operation 1250. FIG. 13 shows an example of an encoder-decoder network 1300, e.g., a neural network, that may be used toAttorney Docket No.: 474060-123Patent Application match the denoised signals. An input 1310 to the encoder-decoder network 1300 may be the data from the first non-destructive inspection technique, and an output 1320 may be data that has been converted (or transferred) into the data corresponding to the second non-destructive inspection technique. The data may be converted via a plurality of convolutional layers, e.g., convolutional layer 1330, and residual blocks, e.g., residual block 1340, as may be used in a U-Net convolutional neural network (CNN). For example, MFL-A data may be input into the encoder-decoder network 1300, and MFL-C data may be output by the encoder-decoder network 1300. The reconstruction model may then be used on this matched data set in operation 1270 to generate a feature map 1280 (e.g., a defect profile), for example, which shows features in the object of interest, e.g., to detect a defect in a pipeline.

[0109] FIG. 14 is a set of views showing two image matching options.

[0110] Two different techniques for aligning the data were used in an experiment to determine which technique is superior. The setup for each is shown in FIG. 14, in which image registration with MFL-A data is shown in part (a), image registration with MFL-C data is shown in part (b), and template matching with MFL-A data is shown in part (c). As can be seen in FIG. 14, part (a), image registration uses two input frames with identical size, but it can be seen that there is some area that does not overlap, which leads to less overall available matching data. As can be seen in FIG. 14, part (c), in a template matching operation a larger image area from the original tool data may be used as a baseline, and then a smaller window area of the converted tool data may be “slid” to find the corresponding physical area of the original tool such that all parts of the converted tool data overlap with some part of the original tool data. It should be appreciated that the reverse template matching operation may also be employed, e.g., “sliding” a smaller original data area over a larger converted data area such that all parts of the original tool data overlap with some part of the converted tool data.Attorney Docket No.: 474060-123 Patent Application[OHl] The training data set for the simulated data was built using 121 patches with size of 400 mm x 800 mm, which were divided into training data (107 patches) and test data (14 patches). Then, each patch was cropped into three overlapping 400 mm x 400 mm patches. For the test data, corresponding measured data was obtained and manually aligned with the simulated data.

[0112] In the experiment, a good alignment was considered to be within about ±10mm. The image registration technique could accurately match the simulated data, but failed on the measured data due to the existence of measurement variation. The template matching technique could achieve higher matching accuracy on the measured data. All fourteen test data samples using the template matching with a 400 mm x 800 mm patch had misalignments of within 10 mm, and had an average misalignment of 3.15 mm. The matching accuracy of the template matching technique could be further improved by enlarging the image size.

[0113] FIG. 15 is a block diagram of a neural network trained for a reconstruction model in accordance with an example embodiment of the present disclosure.

[0114] With reference to FIG. 15, a neural network 1500 may combine signals from at least two non-destructive investigation tool types, e.g., MFL-A data 1510 and MFL-C data 1520. In an experiment training a neural network, the input was 32 x 32 patches from aligned MFL-A and MFL-C data. The whole data set included 143 defects, the training set included 114 defects, and the test set included twenty-nine (29) defects. The output was compared to a laser scan of an example pipeline. A series of convolutional layers may convert the data from one form to the other using, for example, a plurality of fully connected neuron layers 1530, to generate a 3D reconstruction 1540, which shows a corrosion depth map at one location. Thus, the training of the reconstruction model necessarily requires the use of a technological solution and cannot be practically performed in the human mind.Attorney Docket No.: 474060-123 Patent Application

[0115] Example embodiments of the present disclosure may include a model that aligns the data from two magnetic field orientations and fuses the respective signals a single inspection result to achieve a 3D metal loss depth map with laser-like precision. The alignment of the signals is achieved through conversion into the same modality, e.g., MFL-A converted to MFL-C and vice versa. The fusion model is a neural network trained on historical MFL and laser scan data. It takes the aligned MFL-A and MFL-C signal data as the input and produces 3D metal loss depth maps with high resolution. In a case study, a fusion model in accordance with an example embodiment was validated on the field data from an operational pipeline. The depth comparison of the derived 3D metal loss depth maps versus actual 3D laser scans was similar in the experimental result. The 3D metal loss depth maps are also used for deriving 2D profiles as inputs to RSTRENG and P2methodologies. The fusion derived results, compared to the box geometry, allow for more accurate estimation of pipeline burst pressure.

[0116] FIG. 16 illustrates certain components that may be included within a computer system according to an example embodiment of the present disclosure.

[0117] FIG. 16 illustrates certain components that may be included within a computer system 1600, which may be used to control the methods, techniques, and neural networks described herein. One or more computer systems 1600 may be used to implement the various devices, components, and systems described herein.

[0118] The computer system 1600 includes a processor 1601. The processor 1601 may be a general-purpose single- or multi-chip microprocessor (e.g., an Advanced RISC (Reduced Instruction Set Computer) Machine (ARM)), a special-purpose microprocessor (e.g., a digital signal processor (DSP)), a microcontroller, a programmable gate array, etc. The processor 1601 may be referred to as a central processing unit (CPU). Although just a single processor 1601 is shown in the computer system 1600 of FIG. 16, in an alternative configuration, a combination of processors (e.g., an ARM and DSP) could be used. In one or more embodiments, theAttorney Docket No.: 474060-123 Patent Application computer system 1600 further includes one or more graphics processing units (GPUs), which can provide processing services related to both entity classification and graph generation.

[0119] The computer system 1600 also includes memory 1603 in electronic communication with the processor 1601. The memory 1603 may be any electronic component capable of storing electronic information. For example, the memory 1603 may be embodied as random access memory (RAM), read-only memory (ROM), magnetic disk storage media, optical storage media, flash memory devices in RAM, on-board memory included with the processor, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) memory, registers, and so forth, including combinations thereof.

[0120] Instructions 1605 and data 1607 may be stored in the memory 1603. The instructions 1605 may be executable by the processor 1601 to implement some or all of the functionality disclosed herein. Executing the instructions 1605 may involve the use of the data 1607 that is stored in the memory 1603. Any of the various examples of modules and components described herein may be implemented, partially or wholly, as instructions 1605 stored in memory 1603 and executed by the processor 1601. Any of the various examples of data described herein may be among the data 1607 that is stored in memory 1603 and used during execution of the instructions 1605 by the processor 1601.

[0121] A computer system 1600 may also include one or more communication interfaces 1609 for communicating with other electronic devices. The communication interface(s) 1609 may be based on wired communication technology, wireless communication technology, or both. Some examples of communication interfaces 1609 include a Universal Serial Bus (USB), an Ethernet adapter, a wireless adapter that operates in accordance with an Institute of Electrical and Electronics Engineers (IEEE) 802.11 wireless communicationAttorney Docket No.: 474060-123 Patent Application protocol, a Bluetooth® wireless communication adapter, and an infrared (IR) communication port.

[0122] A computer system 1600 may also include one or more input devices 1611 and one or more output devices 1613. Some examples of input devices 1611 include a keyboard, mouse, microphone, remote control device, buttonjoystick, trackball, touchpad, and lightpen. Some examples of output devices 1613 include a speaker and a printer. One specific type of output device that is typically included in a computer system 1600 is a display device 1615. Display devices 1615 used with embodiments disclosed herein may utilize any suitable image projection technology, such as liquid crystal display (LCD), light-emitting diode (LED), gas plasma, electroluminescence, or the like. A display controller 1617 may also be provided, for converting data 1607 stored in the memory 1603 into text, graphics, and / or moving images (as appropriate) shown on the display device 1615.

[0123] The various components of the computer system 1600 may be coupled together by one or more buses, which may include a power bus, a control signal bus, a status signal bus, a data bus, etc. For the sake of clarity, the various buses are illustrated in FIG. 16 as a bus system 1619.

[0124] Systems and software, e.g., implemented on a non-transitory computer-readable medium, for performing the methods discussed herein are also within the scope of embodiments of the present disclosure.

[0125] Embodiments of the present disclosure may thus utilize a special purpose or general-purpose computing system including computer hardware, such as, for example, one or more processors and system memory. Embodiments within the scope of the present disclosure also include physical and other computer-readable media for carrying or storing computerexecutable instructions and / or data structures, including applications, tables, data, libraries, or other modules used to execute particular functions or direct selection or execution of otherAttorney Docket No.: 474060-123 Patent Application modules. Such computer-readable media can be any available media that can be accessed by a general purpose or special purpose computer system. Computer-readable media that store computer-executable instructions (or software instructions) are physical storage media. Computer-readable media that carry computer-executable instructions are transmission media. Thus, by way of example, and not limitation, embodiments of the present disclosure can include at least two distinctly different kinds of computer-readable media, namely physical storage media or transmission media. Combinations of physical storage media and transmission media should also be included within the scope of computer-readable media.

[0126] Both physical storage media and transmission media may be used to temporarily store or carry, software instructions in the form of computer readable program code that allows performance of embodiments of the present disclosure. Physical storage media may further be used to persistently or permanently store such software instructions. Examples of physical storage media include physical memory (e.g., RAM, ROM, EPROM, EEPROM, etc.), optical disk storage (e.g., CD, DVD, HDDVD, Blu-ray, etc.), storage devices (e.g., magnetic disk storage, tape storage, diskette, etc.), flash or other solid-state storage or memory, or any other non-transmission medium which can be used to store program code in the form of computerexecutable instructions or data structures and which can be accessed by a general purpose or special purpose computer, whether such program code is stored as or in software, hardware, firmware, or combinations thereof.

[0127] A “network” or “communications network” may generally be defined as one or more data links that enable the transport of electronic data between computer systems and / or modules, engines, and / or other electronic devices. When information is transferred or provided over a communication network or another communications connection (either hardwired, wireless, or a combination of hardwired or wireless) to a computing device, the computing device properly views the connection as a transmission medium. Transmission media canAttorney Docket No.: 474060-123 Patent Application include a communication network and / or data links, carrier waves, wireless signals, and the like, which can be used to carry desired program or template code means or instructions in the form of computer-executable instructions or data structures and which can be accessed by a general purpose or special purpose computer.

[0128] Further, upon reaching various computer system components, program code in the form of computer-executable instructions or data structures can be transferred automatically or manually from transmission media to physical storage media (or vice versa). For example, computer-executable instructions or data structures received over a network or data link can be buffered in memory (e.g., RAM) within a network interface module (NIC), and then eventually transferred to computer system RAM and / or to less volatile physical storage media at a computer system. Thus, it should be understood that physical storage media can be included in computer system components that also (or even primarily) utilize transmission media.

[0129] The present disclosure may be embodied in other specific forms without departing from its spirit or characteristics. The described embodiments are to be considered as illustrative and not restrictive. The scope of the disclosure is, therefore, indicated by the appended claims rather than by the foregoing description. Changes that come within the meaning and range of equivalency of the claims are to be embraced within their scope.

Claims

Attorney Docket No.: 474060-123Patent ApplicationCLAIMSWhat is claimed is:

1. A method, comprising: obtaining first data for an object of interest, the first data having a first modality or domain, the first data being associated with a first non-destructive inspection technique; obtaining second data for the object of interest, the second data having a second modality or domain different from the first modality or domain, the second data being associated with a second non-destructive inspection technique different from the first non-destructive inspection technique; converting the first data into the modality or domain of the second data; aligning the converted first data with the second data such that a physical feature of the first data is aligned with a corresponding physical feature of the second data; determining a transformation that was applied to the converted first data or to the second data to achieve the aligning; applying the transformation to the first data or the second data to obtain aligned first data or aligned second data; and determining a feature map of the object of interest, based on the aligned first data and the second data or the first data and the aligned second data, using a reconstruction model.

2. The method of claim 1, wherein the aligning the converted first data with the second data comprises performing an image registration operation in which a data area of the converted first data is compared to a same-sized data area of the second data.Attorney Docket No.: 474060-123 Patent Application3. The method of claim 1, wherein the aligning the converted first data with the second data comprises performing a template matching operation in which a data area of the converted first data is compared to a different-sized data area of the second data, such that: the data area of the converted first data is smaller than the data area of the second data so that all of the converted first data overlaps with a portion of the second data; or the data area of the second data is smaller than the data area of the converted first data so that all of the second data overlaps with a portion of the converted first data.

4. The method of claim 1, wherein the determining a transformation that was applied comprises calculating a shift of the converted first data or of the second data.

5. The method of claim 1, wherein the aligning the converted first data with the second data is performed automatically by iteratively adjusting a two-dimensional (2D) window of data among the converted first data to match with a 2D window of the second data to within 10 mm of misalignment.

6. The method of claim 1 , wherein the converting the first data into the modality or domain of the second data comprises converting the first data via an encoder-decoder neural network that inputs the first data into a first plurality of convolutional layers, then transforms the output of the first plurality of convolutional layer through a plurality of residual blocks, then transforms the output of the plurality of residual blocks through a second plurality of convolutional layers to generate the converted first data.Attorney Docket No.: 474060-123Patent Application7. The method of claim 1, wherein: the first non-destructive inspection technique comprises an axial magnetic flux leakage (MFL-A) technique; and the second non-destructive inspection technique comprises a circumferential magnetic flux leakage (MFL-C) technique.

8. The method of claim 1, wherein the first and second non-destructive inspection techniques have complementary detection capabilities.

9. The method of claim 1, wherein: the obtaining the first data comprises operating a first inspection tool on the object of interest, the first inspection tool collecting the first data using the first non-destructive inspection technique; and the obtaining the second data comprises operating a second inspection tool on the object of interest, the second inspection tool collecting the second data using the second nondestructive inspection technique.

10. The method of claim 9, wherein: the first non-destructive inspection technique comprises an axial magnetic flux leakage (MFL-A) technique; and the second non-destructive inspection technique comprises a circumferential magnetic flux leakage (MFL-C) technique.Attorney Docket No.: 474060-123Patent Application11. A system, comprising: a data obtaining unit configured to: obtain first data for an object of interest, the first data having a first modality or domain, the first data being associated with a first non-destructive inspection technique; and obtain second data for the object of interest, the second data having a second modality or domain different from the first modality or domain, the second data being associated with a second non-destructive inspection technique different from the first non-destructive inspection technique; and a computing system comprising: one or more processors; and a memory system comprising one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations, the operations comprising: converting the first data into the modality or domain of the second data; aligning the converted first data with the second data such that a physical feature of the first data is aligned with a corresponding physical feature of the second data; determining a transformation that was applied to the converted first data or to the second data to achieve the aligning; applying the transformation to the first data or the second data to obtain aligned first data or aligned second data; andAttorney Docket No.: 474060-123Patent Application determining a feature map of the object of interest, based on the aligned first data and the second data or the first data and the aligned second data, using a reconstruction model.

12. The system of claim 11, wherein the aligning the converted first data with the second data comprises performing an image registration operation in which a data area of the converted first data is compared to a same-sized data area of the second data.

13. The system of claim 11, wherein the aligning the converted first data with the second data comprises performing a template matching operation in which a data area of the converted first data is compared to a different-sized data area of the second data, such that: the data area of the converted first data is smaller than the data area of the second data so that all of the converted first data overlaps with a portion of the second data; or the data area of the second data is smaller than the data area of the converted first data so that all of the second data overlaps with a portion of the converted first data.

14. The system of claim 11, wherein the determining a transformation that was applied comprises calculating a shift of the converted first data or of the second data.

15. The system of claim 11, wherein the aligning the converted first data with the second data is performed automatically by iteratively adjusting a two-dimensional (2D) window of data among the converted first data to match with a 2D window of the second data to within 10 mm of misalignment.Attorney Docket No.: 474060-123 Patent Application16. The system of claim 11, wherein the converting the first data into the modality or domain of the second data comprises converting the first data via an encoder-decoder neural network configured to: input the first data into a first plurality of convolutional layers; transform the output of the first plurality of convolutional layer through a plurality of residual blocks; and transform the output of the plurality of residual blocks through a second plurality of convolutional layers to generate the converted first data.

17. The system of claim 11, wherein: the first non-destructive inspection technique comprises an axial magnetic flux leakage (MFL-A) technique; and the second non-destructive inspection technique comprises a circumferential magnetic flux leakage (MFL-C) technique.

18. The system of claim 11, wherein the first and second non-destructive inspection techniques have complementary detection capabilities.

19. The system of claim 11, wherein the data obtaining unit comprises: a first inspection tool configured to collect the first data using the first non-destructive inspection technique; and a second inspection tool configured to collect the second data using the second nondestructive inspection technique.Attorney Docket No.: 474060-123 Patent Application20. The system of claim 19, wherein: the first non-destructive inspection technique comprises an axial magnetic flux leakage(MFL-A) technique; and the second non-destructive inspection technique comprises a circumferential magnetic flux leakage (MFL-C) technique.

Images