A pipeline corrosion defect database construction method for artificial intelligence training

Abstract

The present application relates to the technical field of pipeline corrosion detection and data analysis, and provides a pipeline corrosion defect database construction method for artificial intelligence training, wherein a multi-scene corrosion experiment platform is constructed to simulate typical corrosion environments that the pipeline may encounter during service, real corrosion defects with different morphologies are prepared on pipeline steel samples through accelerated corrosion experiments; then high-precision three-dimensional measurement technology is adopted to obtain fine morphology data of the corrosion pits, and standard magnetic flux leakage detection equipment is used to collect magnetic flux leakage signals corresponding to the defects; finally, the corrosion environment parameters, corrosion pit geometric characteristic parameters and magnetic flux leakage detection signal data are uniformly and structurally processed to form a queryable and extensible associated database. The database can truly reflect the internal relationship between the corrosion morphology and the magnetic flux leakage signals, and provide key data support for subsequent quantitative evaluation of pipeline corrosion defects, optimization of detection models and research on intelligent inversion algorithms.

Description

Technical Field

[0001] This invention relates to the field of pipeline corrosion detection and data analysis technology, specifically to a method for constructing a pipeline corrosion defect database for artificial intelligence training. It is particularly suitable for establishing a quantitative correspondence between real corrosion morphology and magnetic flux leakage detection signals, providing a reliable data foundation for the intelligent interpretation of pipeline detection data. Background Technology

[0002] Oil and gas pipelines, urban underground pipe networks, and other infrastructure operate in complex environments for extended periods, and corrosion damage is one of the main causes of pipeline failure. Corrosion not only thins the pipeline wall but can also create defects such as pitting and cracking, seriously threatening the safe operation of pipelines. To promptly assess the corrosion status of pipelines, in-pipe inspection technologies have been widely applied. Among these, magnetic flux leakage (MFL) testing has become one of the most widely used in-pipe inspection methods due to its advantages such as fast detection speed, relatively low cost, and sensitivity to defects. The basic principle of MFL testing is to magnetize the pipeline wall to saturation using permanent magnets or electromagnets. When corrosion defects exist in the pipe wall, the change in magnetic permeability at the defect causes some magnetic lines of force to leak out of the pipe wall. By detecting the leaked magnetic field using a magnetic sensor placed on a probe, the location and approximate size of the defect can be determined.

[0003] However, existing magnetic flux leakage (MF) detection technology faces a key bottleneck in practical applications: MF signals can only indicate the presence and approximate extent of defects, but cannot accurately reflect the true morphology, depth, volume, and other key geometric parameters of corrosion pits, let alone directly predict corrosion development trends. The fundamental reason lies in the lack of a quantitative mapping relationship between MF signals and the morphology of corrosion defects. The amplitude and waveform characteristics of MF signals are influenced by various factors such as defect depth, length, width, shape, and edge sharpness, while the morphology of corrosion pits is complex and diverse. Limited empirical formulas or simplified simulation models cannot encompass the diversity of actual corrosion. Currently, the interpretation of MF detection data mainly relies on calibration experiments and numerical simulations. However, calibration experiments often use artificially processed standard defects (such as drilled holes and grooves), which differ significantly from the morphological characteristics of actual corrosion pits. While numerical simulations can generate a large amount of data, their accuracy depends on model assumptions and cannot fully reproduce the randomness and complexity of real corrosion. Therefore, although MF detection can detect defects, the detection results cannot be directly used for quantitative assessment of corrosion severity, nor can they provide reliable input parameters for predicting remaining life. Summary of the Invention

[0004] To address the aforementioned issues, there is an urgent need to establish a correlated database containing a large number of real corrosion defect samples and their corresponding magnetic flux leakage (MF) detection signals. This database would allow for a systematic study of the intrinsic relationships between corrosion morphology parameters and MF signal characteristics, thereby training data-driven inversion models and achieving a precise mapping from MF signals to corrosion pit geometric parameters. However, currently, no mature database of this kind exists. Existing research largely relies on a small number of experimental samples or single-type simulation data, resulting in limited sample size and narrow parameter coverage, making it difficult to support the training of high-precision models. Therefore, it is necessary to propose a systematic database construction method. This method involves preparing real corrosion defects with different morphologies through controlled corrosion experiments, combining high-precision three-dimensional measurement and MF detection technology to collect paired data of corrosion pit morphology and MF signals, and storing this data in a structured manner. Ultimately, this will establish a correlated database containing corrosion environment parameters, corrosion pit geometric feature parameters, and MF signal data, providing a reliable data foundation for pipeline corrosion detection and evaluation technologies.

[0005] This invention provides a method for constructing a pipeline corrosion defect database for artificial intelligence training, aiming to address the lack of data on the correspondence between realistic corrosion morphology and signals in existing magnetic flux leakage (MFL) detection technologies. The method constructs a multi-scenario corrosion experimental platform to simulate typical corrosion environments that pipelines may encounter during service. Real corrosion defects with different morphologies are prepared on pipeline steel samples through accelerated corrosion experiments. Then, high-precision three-dimensional measurement technology is used to acquire detailed morphological data of the corrosion pits, and MFL signals corresponding to the defects are collected using standard MFL detection equipment. Finally, corrosion environment parameters, corrosion pit geometric feature parameters, and MFL detection signal data are uniformly structured to form a queryable and scalable relational database. This database can realistically reflect the intrinsic relationship between corrosion morphology and MFL signals, providing crucial data support for subsequent quantitative assessment of pipeline corrosion defects, optimization of detection models, and research on intelligent inversion algorithms.

[0006] This invention provides a method for constructing a pipeline corrosion defect database for artificial intelligence training, comprising the following steps: Step 1: Constructing the corrosion experimental environment A corrosion testing platform capable of simulating the service environment of pipelines was constructed. Based on the actual environmental type of the pipeline, multiple corrosion testing modules were set up, including but not limited to soil corrosion, stray current corrosion, atmospheric corrosion, and aquatic environment corrosion modules. Each module is equipped with environmental parameter control and monitoring devices, allowing for the precise control of key parameters to reproduce typical corrosion scenarios under laboratory conditions.

[0007] Specifically: Soil corrosion module: A soil tank with controllable resistivity, moisture content, pH, and microbial activity is used. The soil resistivity is adjustable from 10 to 1000 Ω·m, the moisture content from 5% to 30%, and the pH from 4 to 9. Pipe steel samples can be buried within the soil tank, and a cathodic protection system is provided to simulate corrosion under different cathodic protection conditions.

[0008] Stray current corrosion module: Constructs an AC / DC interference simulation circuit, applying a controllable stray current to the sample via DC and AC power supplies, with the current density adjustable from 0 to 100 A / m. 2 It can simulate the corrosive effects of interference sources such as DC electrified railways and AC transmission lines on pipelines.

[0009] Atmospheric corrosion module: It adopts a climate chamber with controllable temperature, humidity and pollutant concentration. The temperature adjustment range is -20~60℃, the relative humidity adjustment range is 30%~95%, and corrosive gases such as SO2 and NO2 can be introduced to simulate the industrial atmospheric environment.

[0010] Aquatic environment corrosion module: Designed freshwater / seawater circulating tank, equipped with dissolved oxygen sensor, flow meter and temperature sensor, dissolved oxygen concentration adjustment range is 0~10 mg / L, flow rate adjustment range is 0~2 m / s, can simulate corrosion process under different water quality conditions.

[0011] All environmental parameters are monitored and recorded in real time through a computer control system to ensure the stability and repeatability of experimental conditions.

[0012] Step 2: Prepare corrosion defect samples In the constructed corrosion test environment, accelerated corrosion experiments are conducted on pipeline steel samples to obtain corrosion defects with different morphological characteristics. The sample material should be consistent with the actual pipeline material, typically using common pipeline steels such as X42, X52, X60, X65, X70, and X80. The sample size can be designed according to the requirements of the experimental equipment; for example, plate samples can be 200 mm × 100 mm × (8~15) mm, or pipe section samples can be 300 mm long and 100 mm in diameter. The sample surface needs to be standardized, such as grinding to a certain roughness, cleaning, and degreasing. Accelerated corrosion experiments can employ constant potential polarization, constant current polarization, wet-dry alternating cycle methods, and salt spray tests, etc., to accelerate the corrosion process by controlling factors such as potential, current, temperature, humidity, and corrosive medium concentration. The experimental cycle is set according to the target corrosion degree and can range from several hours to several weeks. During the experiment, changes in environmental parameters, including potential, current, temperature, pH value, and dissolved oxygen concentration, are recorded in real time. After the experiment, remove the sample, remove the loose corrosion products on the surface (chemical cleaning or mechanical methods can be used), and retain the morphology of the corrosion pits for subsequent measurements.

[0013] Step 3: Obtain the three-dimensional morphology of the corrosion pits High-precision 3D surface morphology measurement technology is used to scan corrosion defect samples and acquire 3D point cloud data of corrosion pits. Optional measurement equipment includes laser 3D scanners, white light interferometers, confocal microscopes, and structured light 3D measuring instruments, with a measurement accuracy better than 0.01 mm. Before measurement, the sample must be cleaned and fixed to ensure no vibration interference during the measurement process. The scanning covers the entire corrosion area to acquire dense 3D coordinate points. For larger samples, multi-field stitching technology can be used. The acquired point cloud data needs to be preprocessed, including noise reduction, filtering, and coordinate alignment, before reconstructing the 3D morphology of the corrosion pits. Based on the reconstructed morphology data, the geometric feature parameters of the corrosion pits can be further extracted.

[0014] Step 4: Perform magnetic flux leakage detection Magnetic flux leakage (MF) testing is performed on corrosion defect samples whose morphology has been measured to obtain corresponding MF signal data. MF testing devices can be commercial pipe-mounted detectors or laboratory MF scanning platforms, typically including a magnetization unit (permanent magnet or electromagnet) and a magnetic sensor array (Hall element, magnetoresistive sensor, etc.). During testing, the sample is placed in the magnetization circuit to achieve local magnetic saturation. The sensor scans the sample surface at a fixed lift height (usually 1-3 mm), acquiring the triaxial components (axial, radial, and circumferential) of the leakage magnetic field. The scanning step is set according to the defect size, typically 0.5-2 mm, and the sampling frequency should be high enough to capture signal details. The position coordinates and corresponding magnetic flux density values ​​of each scanning point are recorded during the testing process. MF signal data includes axial, radial, and circumferential MF signals, as well as a composite magnetic flux density signal. Simultaneously, parameters such as the magnetization intensity, lift height, and scanning speed of the detection device are recorded to ensure data repeatability.

[0015] Step 5: Extract geometric parameters of corrosion defects Based on the 3D point cloud data obtained in step 3, the geometric feature parameters of the corrosion pits are extracted using image processing algorithms or specialized software. First, the original surface of the sample (uncorroded area) is compared with the corroded area through point cloud registration to determine the boundaries of the corrosion pits. Then, the following key geometric parameters are calculated: Maximum corrosion depth: The vertical distance between the lowest point of the corrosion pit area and the original surface; Corrosion pit diameter: The maximum equivalent diameter passing through the centroid of the corrosion pit within the plane of the pit opening; Corrosion area: The projected area of ​​the corrosion pit on the original surface plane; Corrosion volume: The three-dimensional spatial volume occupied by the corrosion pit, which can be obtained by the difference between the integrated point cloud and the original surface; Pit shape factor: Parameters describing the morphology of corrosion pits, such as depth to diameter ratio, pit ellipticity, pit bottom roughness, etc.

[0016] These parameters comprehensively characterize the geometric features of the corrosion pits, providing quantitative indicators for subsequent correlation analysis.

[0017] Step 6: Establish a relational database All data obtained in steps 1-5 are processed into a unified structure to construct a correlation database between corrosion defects and magnetic flux leakage detection signals. The database is stored using a relational database (such as MySQL or PostgreSQL) or a non-relational database (such as MongoDB). Each data record corresponds to a corrosion defect sample and contains the following three types of fields: Corrosion environment parameters: soil resistivity, water content, pH value, oxygen content, applied potential, stray current density, temperature, humidity, experimental period, etc. Geometric parameters of corrosion defects: maximum corrosion depth, corrosion pit diameter, corrosion area, corrosion volume, pit shape factor, etc. Parameters of magnetic flux leakage detection signal: axial magnetic flux leakage signal waveform (stored in array form), radial magnetic flux leakage signal waveform, circumferential magnetic flux leakage signal waveform, magnetic flux density amplitude, signal peak value, signal width, signal gradient, etc.

[0018] In addition, the database should include metadata such as unique sample identifiers, sample material grades, sample dimensions, and testing dates. Data storage must ensure the accuracy and consistency of each field and include indexes for fast retrieval. The database should support data export, statistical analysis, and the training and validation of machine learning models.

[0019] Database structure: The database of this invention adopts a standardized structural design to ensure data integrity, readability, and scalability. The database mainly includes the following data fields: I. Corrosion Environment Parameters Experiment ID: A unique identifier for each experiment.

[0020] Environment_Type: Identifies the type of corrosive environment, such as soil, stray current, atmosphere, water environment, etc.

[0021] Soil resistivity: The unit is Ω·m. Record the set value and the measured value of soil resistivity.

[0022] Soil moisture content (Water_Content): The unit is %, which records the soil moisture content.

[0023] pH value: Records the pH value of the corrosive medium.

[0024] Dissolved Oxygen Concentration: The unit is mg / L, which records the dissolved oxygen concentration in the aquatic environment.

[0025] Applied Potential: The unit is V, and it is the potential value relative to the reference electrode.

[0026] Stray current density (Stray_Current_Density): Unit is A / m 2 Record the applied DC or AC current density.

[0027] Temperature: The unit is ℃, which records the average temperature during the experiment.

[0028] Humidity: The unit is %, which records the relative humidity of the atmospheric corrosion experiment.

[0029] Test Duration (Test_Duration): The unit is h or d, which records the duration of the corrosion experiment.

[0030] II. Geometric parameters of corrosion defects Sample ID: A unique identifier for each corrosion defect sample.

[0031] Maximum corrosion depth (Max_Depth): in mm, the maximum depth of the corrosion pit.

[0032] Pit Diameter: The unit is mm, which is the equivalent diameter of the pit opening.

[0033] Corrosion area (Pit_Area): unit is mm 2 The projected area of ​​the corrosion pit on the original surface.

[0034] Corrosion volume (Pit_Volume): in mm 3 The volume of the corrosion pit.

[0035] Shape Factor 1: Depth to Diameter ratio.

[0036] Shape Factor 2: Ovality of the pit opening (major axis / minor axis).

[0037] Bottom Roughness: The unit is μm, and it is a roughness parameter that reflects the micro-morphology of the pit bottom.

[0038] III. Parameters of Magnetic Flux Leakage Detection Signal Detection Number (MFL_ID): A unique identifier for each magnetic flux leakage test.

[0039] Axial leakage magnetic field signal (MFL_Axial): Stores the axial magnetic flux density values ​​along the scanning direction in array form, in mT or Gauss.

[0040] Radial leakage magnetic field signal (MFL_Radial): The radial magnetic flux density values ​​are stored in array form.

[0041] Circumferential leakage signal (MFL_Circumferential): Stores circumferential magnetic flux density values ​​in array form.

[0042] Magnetic flux density amplitude (Peak_Flux): Peak value of leakage magnetic signal, in mT.

[0043] Signal Width: Half-width of leakage magnetic signal, in mm.

[0044] Signal Gradient: The maximum slope of the rising / falling edge of the leakage magnetic signal.

[0045] Lift_Off: The distance between the detection probe and the sample surface, in mm.

[0046] Scan speed (Scan_Speed): Unit: mm / s.

[0047] Each corrosion defect sample corresponds to a complete set of data records, with all fields strictly defined and stored. The database supports multi-condition queries based on environmental parameters, geometric parameters, or signal characteristics, facilitating researchers to extract data subsets as needed.

[0048] This invention provides a method for constructing a pipeline corrosion defect database for artificial intelligence training. It establishes a real corrosion defect database: corrosion defect samples prepared through physical experiments have morphological characteristics derived from real electrochemical corrosion processes. They contain the randomness and complexity of actual corrosion and are more representative than artificial defects or simulation data. This provides a reliable data source for studying the relationship between corrosion morphology and magnetic flux leakage signals.

[0049] A precise correspondence between corrosion morphology and magnetic flux leakage signal was obtained: high-precision three-dimensional measurement and magnetic flux leakage detection were performed simultaneously on the same corrosion defect, realizing the spatiotemporal alignment of the geometric features of corrosion pits and the waveform of magnetic flux leakage signal, providing high-quality training data for establishing a morphology-signal mapping model.

[0050] Provides a data foundation for pipeline corrosion detection and defect assessment: This database can be directly used to train and validate magnetic flux leakage signal inversion models, helping to achieve quantitative inversion from detection signals to key parameters such as corrosion pit depth and volume, and improving the quantitative assessment capability of pipeline internal detection.

[0051] It improves the quantitative assessment capability of pipeline corrosion detection technology: the database-trained model can more accurately assess the severity of corrosion defects, providing a scientific basis for pipeline remaining life prediction and maintenance decisions, thereby improving the level of pipeline integrity management.

[0052] The database has a standardized structure and strong scalability: it adopts a unified data structure for storage, which facilitates data sharing and subsequent expansion. It can incorporate data on more corrosion types and detection technologies to form a comprehensive pipeline corrosion database.

[0053] Advantages of this invention: Authenticity and reliability: The real corrosion defects prepared through physical experiments have morphological characteristics such as pit shape, pit bottom roughness, and edge transition that are closer to actual pipeline corrosion, avoiding deviations caused by artificial defects or simulation data, and ensuring the authenticity and reliability of the data.

[0054] High-precision morphology characterization: Using precision measurement technologies such as laser 3D scanning and white light interferometry, 3D point cloud data of corrosion pits are obtained, which can accurately extract geometric parameters such as depth, diameter, area, volume, and shape factor, providing fine input for establishing a quantitative relationship between morphology and signal.

[0055] Multi-parameter correspondence: The database simultaneously records corrosion environment parameters, corrosion pit geometric parameters, and leakage magnetic signal parameters, forming a complete data chain, which facilitates the analysis of the influence of environmental factors on corrosion morphology and the mapping law between morphology and signal.

[0056] Structured and Scalable: Data is stored using a unified database structure, with each record corresponding to a corrosion defect sample, which facilitates data retrieval, analysis, and subsequent expansion, and supports the introduction of more samples or AI-generated data.

[0057] High application value: This database can be directly used to train leakage magnetic field signal inversion models, improve the interpretation capability of internal detection data, realize quantitative inversion from signal to corrosion pit depth and volume, and provide a scientific basis for pipeline integrity management. Attached Figure Description

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments: Figure 1 Technical flowchart (database construction process); Figure 2 Schematic diagram of corrosion test platform; Figure 3 Schematic diagram of the three-dimensional morphology of the corrosion pit; Figure 4 Schematic diagram of magnetic flux leakage detection principle; Figure 5 Graph of magnetic flux leakage signal; Figure 6 Database structure diagram. Detailed Implementation

[0059] The present invention will be further explained below with reference to specific implementation schemes, but it is not limited to the present invention. The structures, proportions, sizes, etc. shown in the accompanying drawings are only used to complement the content disclosed in the specification, so as to enable those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modification of the structure, change of the proportion relationship or adjustment of the size, without affecting the effect and purpose that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0060] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to a typical embodiment. This embodiment takes the corrosion defects of pipeline steel in a soil environment as the object and fully describes the entire process from experimental design to database construction.

[0061] 1. Setting Corrosion Experiment Parameters In this embodiment, X70 pipeline steel was selected as the sample material, with sample dimensions of 200 mm × 100 mm × 10 mm. The sample surface was mechanically ground to a roughness Ra ≤ 0.8 μm, cleaned with acetone to remove oil, dried, and weighed for later use. The corrosion experiment was conducted in a soil corrosion module, which contains an adjustable soil tank filled with artificially prepared simulated soil. The basic soil components were quartz sand, kaolin, humus, etc., and resistivity and ion concentration were adjusted by adding salts such as NaCl and Na2SO4. This embodiment sets four sets of key parameters: soil resistivity is set at three levels: 50 Ω•m, 100 Ω•m, and 200 Ω•m; moisture content is set at three levels: 10%, 20%, and 30%; pH value is adjusted to three levels: 5, 7, and 9 by adding acid or alkali; and the applied potential is applied using a potentiostat and set to -0.85 V (cathodic protection state), -0.65 V (near the natural corrosion potential), and -0.45 V (anodic polarization state) relative to a saturated calomel electrode. Using an orthogonal experimental design method, a total of 3×3×3×3 = 81 parameter combinations were obtained. Three parallel samples were prepared for each combination, for a total of 243 samples. During the experiment, the temperature inside the soil tank was controlled at 25±2℃, and all parameters were monitored and recorded in real time by a computer.

[0062] 2. Formation process of corrosion defects The prepared samples were buried in the soil trench, ensuring close contact between the samples and the soil at a uniform depth. A potentiostat and an external potential electrode were connected, and the samples were polarized according to the set potential. The experimental period was uniformly set to 14 days. Parameters such as actual potential, current density, and soil resistivity were recorded every 24 hours. After the experiment, the samples were removed, rinsed with deionized water, then soaked in an acid pickling solution (such as dilute hydrochloric acid containing corrosion inhibitors) to remove corrosion products, followed by rinsing with clean water and then alcohol, and finally dried with cold air. Visible corrosion pits formed on the dried sample surface. The morphology of these pits varied depending on the parameter combinations, including shallow dish-shaped, deep pore-shaped, and elliptical shapes.

[0063] 3. Three-dimensional topography measurement process The corrosion samples were scanned using a Keyence VR-5000 series 3D white light interferometry instrument. Before measurement, the sample was placed on the measurement platform and leveled to ensure that the measurement area included all corrosion pits. The scanning resolution was set to 0.5 μm, and the measurement range covered the entire sample surface. For each sample, 3D point cloud data (approximately 5 million points) of the original surface was acquired. The instrument's built-in analysis software was used to perform noise filtering and reference plane correction on the point cloud data, and then the boundaries of the corrosion pits were determined by comparing them with uncorroded areas. Parameters such as the maximum depth, equivalent diameter, projected area, and volume of the corrosion pits were automatically extracted, and shape factors such as depth-to-diameter ratio and ellipticity were calculated. All data were exported in CSV format and associated with the sample number.

[0064] 4. Magnetic flux leakage detection process Magnetic flux leakage detection utilizes a self-developed laboratory scanning platform, which consists of an electromagnet magnetizer, a triaxial Hall sensor array, a precision motion control system, and a data acquisition card. The magnetizer provides a magnetization field of approximately 1.5 T, the sensor lift-off height is fixed at 2 mm, the scanning step is 1 mm, and the sampling frequency is 500 Hz. The corrosion sample is placed on the scanning platform, ensuring the magnetizer poles contact the two ends of the sample to form a closed magnetic circuit. The sensor array scans uniformly along the sample length, acquiring axial, radial, and circumferential magnetic flux leakage signals. Each sample is scanned three times and averaged to ensure signal stability. The acquired raw signals are amplified, filtered, and converted into digital signals for storage. Post-processing extracts signal peak values, full width at half maximum (FWHM), gradients, and other features, and the waveform data is saved as an array.

[0065] 5. Data preparation and database construction process After all data is aggregated, it is stored in a MySQL database. First, a database table named "Corrosion_MFL_Data" is created with the following fields: Sample_ID (primary key), Material, Environment_Type, Resistivity, Water_Content, pH_Value, Applied_Potential, Test_Duration, Max_Depth, Pit_Diameter, Pit_Area, Pit_Volume, Depth_Diameter_Ratio, Ellipticity, MFL_Axial (BLOB type, storing waveform arrays), MFL_Radial (BLOB), MFL_Circumferential (BLOB), Peak_Flux, Signal_Width, Signal_Gradient, Lift_Off, Scan_Speed, Scan_Date, etc. The experimental data for 243 samples are entered into the database one by one, ensuring that each record corresponds to a corrosion defect sample. After data entry, a data integrity check was performed to remove abnormal samples (such as those missing due to operational errors), ultimately yielding approximately 230 valid samples. The database was indexed for rapid retrieval by parameter range; for example, all samples with "depth > 2 mm and resistivity = 50 Ω•m" could be searched. This database can later be used to train a neural network model to predict the depth of corrosion pits from leakage magnetic signals.

[0066] This embodiment is merely a typical implementation of the present invention and is not intended to limit the scope of protection of the present invention. Based on the technical concept of the present invention, those skilled in the art can adjust the type of corrosive environment, sample material, measuring equipment, etc., to construct a database suitable for different scenarios. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention. Matters not covered in this invention are common knowledge.

[0067] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a pipeline corrosion defect database for artificial intelligence training, characterized in that: Includes the following steps: Step 1: Constructing the corrosion experimental environment A corrosion test platform capable of simulating the service environment of pipelines is constructed; multiple corrosion test modules are set up according to the actual environmental type of the pipeline, including but not limited to soil corrosion module, stray current corrosion module, atmospheric corrosion module and water environment corrosion module; each module is equipped with environmental parameter control and monitoring devices, and typical corrosion scenarios are reproduced under laboratory conditions by precisely controlling key parameters. Step 2: Prepare corrosion defect samples In the corrosion test environment constructed above, accelerated corrosion tests were conducted on the pipe steel samples to obtain corrosion defects with different morphological characteristics. Step 3: Obtain the three-dimensional morphology of the corrosion pits High-precision three-dimensional surface morphology measurement technology is used to scan corrosion defect samples and obtain three-dimensional point cloud data of corrosion pits. Optional measurement equipment includes laser three-dimensional scanner, white light interferometer, confocal microscope, and structured light three-dimensional measuring instrument, with a measurement accuracy better than 0.01 mm. Before measurement, the sample is cleaned and fixed to ensure no vibration interference during the measurement process; during scanning, the entire corrosion area is covered to obtain dense three-dimensional coordinate points; for larger samples, multi-field stitching technology can be used; the acquired point cloud data needs to be preprocessed, including noise reduction, filtering, and coordinate alignment operations, and then the three-dimensional morphology of the corrosion pit is reconstructed; based on the reconstructed morphology data, the geometric feature parameters of the corrosion pit are further extracted. Step 4: Perform magnetic flux leakage detection Magnetic flux leakage (MF) detection is performed on corrosion defect samples for which morphology measurements have been completed to obtain corresponding MF signal data. The MF detection device uses a commercial pipe in-situ detector or a laboratory MF scanning platform, including a magnetization unit and a magnetic sensor array. Step 5: Extract geometric parameters of corrosion defects Based on the 3D point cloud data obtained in step 3, the geometric feature parameters of the corrosion pits are extracted using image processing algorithms or specialized software. First, the original surface of the sample (i.e., the uncorroded area) is compared with the corroded area through point cloud registration to determine the boundaries of the corrosion pits. Then, key geometric parameters are calculated. Step 6: Establish a relational database All data obtained in steps 1-5 are processed into a unified structure to construct a correlation database between corrosion defects and magnetic flux leakage detection signals. The database is stored using either a relational or non-relational database. Each data record corresponds to a corrosion defect sample and contains the following three types of fields: Corrosion environment parameters: soil resistivity, water content, pH value, oxygen content, applied potential, stray current density, temperature, humidity, experimental period, etc. Geometric parameters of corrosion defects: maximum corrosion depth, corrosion pit diameter, corrosion area, corrosion volume, pit shape factor; Parameters of magnetic flux leakage detection signal: axial magnetic flux leakage signal waveform, radial magnetic flux leakage signal waveform, circumferential magnetic flux leakage signal waveform, magnetic flux density amplitude, signal peak value, signal width, and signal gradient; In addition, the database also contains metadata such as unique sample identifiers, sample material grades, sample dimensions, and testing dates; data storage must ensure the accuracy and consistency of each field and establish indexes for fast querying; the database can support data export, statistical analysis, and training and validation of machine learning models.

2. The method for constructing a pipeline corrosion defect database for artificial intelligence training according to claim 1, characterized in that: Step 1, constructing the corrosion experimental environment, specifically includes the following: Soil corrosion module: A soil tank with controllable resistivity, water content, pH value, and microbial activity is used. The soil resistivity adjustment range is set to 10~1000 Ω•m, the water content adjustment range is 5%~30%, and the pH value adjustment range is 4~9. Pipe steel samples can be buried in the soil tank and equipped with a cathodic protection system to simulate corrosion under different cathodic protection conditions. Stray current corrosion module: Constructs an AC / DC interference simulation circuit, applying a controllable stray current to the sample via DC and AC power supplies, with the current density adjustable from 0 to 100 A / m. 2 It can simulate the corrosive effects of interference sources such as DC electrified railways and AC transmission lines on pipelines; Atmospheric corrosion module: It adopts a climate chamber with controllable temperature, humidity and pollutant concentration. The temperature adjustment range is -20~60℃, the relative humidity adjustment range is 30%~95%, and it can introduce corrosive gases such as SO2 and NO2 to simulate the industrial atmospheric environment. Aquatic environment corrosion module: Design a freshwater / seawater circulating tank, equipped with dissolved oxygen sensor, flow meter and temperature sensor. The dissolved oxygen concentration adjustment range is 0~10 mg / L, and the flow rate adjustment range is 0~2 m / s, which can simulate the corrosion process under different water quality conditions. All environmental parameters are monitored and recorded in real time through a computer control system to ensure the stability and repeatability of experimental conditions.

3. The method for constructing a pipeline corrosion defect database for artificial intelligence training according to claim 1, characterized in that: Step 2 involves preparing corrosion defect samples. The sample surface needs to undergo standardized treatment, including grinding to a certain roughness, cleaning, and degreasing. Accelerated corrosion experiments can employ constant potential polarization, constant current polarization, wet-dry alternating cycle, or salt spray testing. By controlling factors such as potential, current, temperature, humidity, and concentration of the corrosive medium, the corrosion process is accelerated. The experimental cycle is set according to the target corrosion level and can range from several hours to several weeks. During the experiment, changes in environmental parameters, including potential, current, temperature, pH value, and dissolved oxygen concentration, are recorded in real time. After the experiment, the sample is removed, and loose corrosion products on the surface are removed using chemical cleaning or mechanical methods, while the morphology of the corrosion pits is preserved for subsequent measurements.

4. The method for constructing a pipeline corrosion defect database for artificial intelligence training according to claim 1, characterized in that: Step 4 involves performing magnetic flux leakage detection. During detection, the sample is placed in a magnetization circuit to achieve local magnetic saturation. The sensor is lifted at a fixed height of 1-3 mm to scan the sample surface and collect the triaxial components of the leakage magnetic field. The scanning step is set according to the defect size, ranging from 0.5 to 2 mm, and the sampling frequency should be high enough to capture signal details. The position coordinates and corresponding magnetic induction intensity values ​​of each scanning point are recorded during the detection process. The magnetic flux leakage signal data includes axial magnetic flux leakage signal, radial magnetic flux leakage signal, circumferential magnetic flux leakage signal, and composite magnetic flux density signal. Simultaneously, the magnetization intensity, lift-off height, and scanning speed parameters of the detection device are recorded to ensure data repeatability.

5. The method for constructing a pipeline corrosion defect database for artificial intelligence training according to claim 1, characterized in that: Step 5: After extracting the geometric parameters of the corrosion defects and determining the boundaries of the corrosion pits, calculate the following key geometric parameters: Maximum corrosion depth: The vertical distance between the lowest point of the corrosion pit area and the original surface; Corrosion pit diameter: The maximum equivalent diameter passing through the centroid of the corrosion pit within the plane of the pit opening; Corrosion area: The projected area of ​​the corrosion pit on the original surface plane; Corrosion volume: The three-dimensional spatial volume occupied by the corrosion pit, which can be obtained by the difference between the integrated point cloud and the original surface; Pit shape factor: Parameters describing the morphology of corrosion pits, such as depth to diameter ratio, pit ellipticity, and pit bottom roughness; These parameters comprehensively characterize the geometric features of the corrosion pits, providing quantitative indicators for subsequent correlation analysis.

6. The method for constructing a pipeline corrosion defect database for artificial intelligence training according to claim 1, characterized in that: The database structure described: The database employs a standardized structural design to ensure data integrity, readability, and scalability; the database primarily contains the following data fields: I. Corrosion Environment Parameters Experiment number: A unique identifier for each experiment; Environment type: Identify the type of corrosive environment, such as soil, stray current, atmosphere, water environment, etc. Soil resistivity: The unit is Ω•m. Record the set value and the measured value of soil resistivity. Soil moisture content: The unit is %, which is recorded as soil moisture content; pH value: Records the pH value of the corrosive medium; Dissolved oxygen concentration: The unit is mg / L. Record the dissolved oxygen concentration in the aquatic environment. Applied potential: The unit is V, and it is the potential value relative to the reference electrode; Stray current density: unit is A / m 2 Record the applied DC or AC current density; Temperature: The unit is ℃. Record the average temperature during the experiment. Humidity: The unit is %, which records the relative humidity of the atmospheric corrosion experiment; Experiment duration: in hours or days, record the duration of the corrosion experiment; II. Geometric parameters of corrosion defects Sample Number: A unique identifier for each corrosion defect sample; Maximum corrosion depth: in mm, the maximum depth of the corrosion pit; Corrosion pit diameter: in mm, equivalent diameter of the corrosion pit opening; Corrosion area: in mm 2 The projected area of ​​the corrosion pit on the original surface; Corrosion volume: in mm 3 The volume of the corrosion pit; Shape factor 1: depth to diameter ratio; Shape factor 2: Ovality of pit opening (major axis / minor axis); Pit bottom roughness: The unit is μm, which is a roughness parameter that reflects the micro-morphology of the pit bottom; III. Parameters of Magnetic Flux Leakage Detection Signal Test number: A unique identifier for each magnetic flux leakage test; Axial leakage magnetic field signal: The axial magnetic flux density value along the scanning direction is stored in array form, in mT or Gauss; Radial leakage magnetic field signal: The radial magnetic flux density values ​​are stored in array form; Circumferential leakage magnetic signal: The circumferential magnetic induction intensity values ​​are stored in array form; Magnetic flux density amplitude: peak value of leakage magnetic signal, in mT; Signal width: Half-width at half-maximum of leakage magnetic signal, in mm; Signal gradient: The maximum slope of the rising / falling edge of the leakage magnetic signal; Lift-off height: The distance between the detection probe and the sample surface, in mm; Scanning speed: unit mm / s; Each corrosion defect sample corresponds to a complete set of data records, and all fields are strictly defined and stored. The database supports multi-condition queries based on environmental parameters, geometric parameters, or signal characteristics, making it convenient for researchers to extract data subsets as needed.

Images