A method for quantitative analysis of pipeline defects

Abstract

The present invention provides a method for quantifying pipeline defects. First, the pipeline is magnetized to saturation by using a leakage magnetic field detection method; then, the sensor is placed at different lift-off heights to collect the radial component of the leakage magnetic field signal and extract its signal peak; the obtained signal peak value is fitted according to a given equation, and a neural network algorithm is used to construct the relationship between the defect geometric characteristic parameters and the fitting equation coefficients, thereby quantifying the defects. The present invention makes full use of the variation law of the leakage magnetic field signal with the lift-off space, and provides a new basic idea for defect quantification. The present invention has high accuracy, small absolute error and relative error, and high confidence.

Description

Technical Field

[0001] The invention relates to the technical field of magnetic flux leakage nondestructive testing, and is particularly applicable to a detection and quantitative analysis method for common corrosion pit defects in pipelines. Background Art

[0002] As one of the five traditional modes of transportation, pipeline transportation has very important applications in the energy, chemical, oil and gas industries, especially in the transportation of national strategic energy such as oil and natural gas. Among all kinds of pipes, ferromagnetic material pipes are widely used due to their abundant raw materials, simple manufacturing, easy molding, low cost, and pollution-free. However, since most oil and gas pipelines are buried deep underground or on the seabed, the working conditions are very harsh, which will cause defects such as corrosion pits and cracks, which will seriously affect the safety and service life of the pipelines, causing serious energy and economic losses and environmental pollution. Therefore, it is necessary to conduct regular safety assessments.

[0003] As one of the common methods in pipeline nondestructive testing, magnetic flux leakage testing has the advantages of low requirements for the pipeline environment, no coupling, wide application range, and low price. Most of the current methods for accurate quantitative analysis of defects directly construct the relationship between defect parameters and characteristic values ​​of magnetic flux leakage signals. Although magnetic flux leakage detection technology is very mature, there are still many problems to be solved in the accurate quantitative analysis of defects. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide an analysis method for quantifying defects through the attenuation law of the leakage magnetic field in space and the lift-off attenuation effect.

[0005] In order to solve the above technical problems, the pipeline defect quantitative analysis method provided by the present invention adopts the method of magnetic leakage detection, magnetizes the pipeline to saturation, and then detects to obtain the magnetic leakage signal at the defect; defines the distance between the extraction signal sensor position and the pipe wall as the lift-off value x, extracts the radial component of the magnetic leakage signal at the defect under different lift-off values ​​x (the radial component refers to the component perpendicular to the pipe wall and pointing to the center of the pipe) and extracts its signal peak value B x-max The peak values ​​obtained under different defect sizes are fitted to obtain the signal peak value B x-max The fitting parameters of the fitting equation vary with the lift-off value x; and a neural network algorithm is used to construct the relationship between the pit defect parameters and the fitting parameters, so as to quantify the defects.

[0006] The present invention specifically comprises the following steps:

[0007] Step 1: Construct a magnetizer model and adjust the magnetizer parameters to make the pipeline reach saturation state.

[0008] Step 2: Establish a three-dimensional pipeline model under different defect parameters to form a pipeline defect sample library.

[0009] Step 3: Use finite element simulation software to calculate the leakage magnetic field under different defect parameters, that is, simulation.

[0010] Step 4: Change the height of the extraction curve from the pipe wall to simulate the radial component distribution of the leakage magnetic signal under different lift-off values ​​x.

[0011] Step 5: Extract the radial component distribution of the leakage magnetic signal under different lift-off values ​​x.

[0012] Step 6: Extract the radial component peak B x-max , draw B x-max Scatter plot of the liftoff value x.

[0013] Step 7: Select a suitable fitting equation to fit the scatter plot obtained in Step 6 to obtain the fitting parameters.

[0014] Step 8: Take the fitting parameters of different defects as input and the defect parameters as output, and use the neural network algorithm for machine learning.

[0015] Step 9: Verify the results of Step 8 to ensure the learning effect.

[0016] Step 10: Repeat Step 2 to Step 6 to simulate and extract data for defects with different geometric parameters in the sample library, and draw B x-max Scatter plot as liftoff value x changes.

[0017] Step 11: Use the equation used in Step 7 to fit and obtain the fitting parameters.

[0018] Step 12: Bring the fitting parameters into the learning results, and the system automatically outputs the defect parameters to complete the defect quantification.

[0019] The following is a description of some of the steps:

[0020] For Step 1, the method to determine whether the pipeline is saturated is:

[0021] The simulation calculation is carried out based on the perfect model, the magnetic flux density value at the center of the pipeline is extracted and compared with the BH curve (magnetization curve) of the pipeline material. When the magnetic flux density value at the center reaches the saturation point of the BH curve of the pipe (that is, the B value corresponding to when the curve starts to rise as a linear function), the magnetization of the pipeline reaches saturation.

[0022] The adjustable magnetizer parameters are: magnet coercivity, magnet length, magnet width, magnet thickness and pole spacing.

[0023] For Step 2, the prefabricated defect is a common ellipsoidal pit defect, and the defect characteristic parameters are the defect diameter L and the defect depth D.

[0024] For Step 6, calculate the peak value B of the radial component of the magnetic flux density x-max When the average value is used, the positive peak-to-peak value B of the radial component distribution curve of the magnetic flux density is extracted respectively. x-max_1 and negative peak-to-peak value B x-max_2 , use formula (1) to calculate the peak value B of the radial component of magnetic flux density x-max .

[0025]

[0026] For Step 7, exponential fitting, power function fitting or quadratic fitting can all achieve a higher degree of fit, and the fitted curve can well reflect the B x-max As the lift-off value increases, the fitting equation is selected as one of the following:

[0027] Exponential fit:

[0028] y=A+Bexp(C·x) (2)

[0029] Power function fitting:

[0030] y=Ax B (3)

[0031] Quadratic fit:

[0032] y=A+Bx+Cx 2 (4)

[0033] Where A, B and C are fitting parameters, y is the peak value of the radial component of the magnetic flux density B x-max .

[0034] For Step 8, the appropriate number of neural network layers and sample distribution (i.e., the sample size used for training, testing, and validation) should be selected. When performing machine learning, appropriately increasing the sample size can better improve the effect of machine learning.

[0035] The beneficial effects of the present invention are: making full use of the distribution law of the leakage magnetic field in space, proposing a method for analyzing the quantification of corrosion pit defects based on the attenuation law of the leakage magnetic field in space, and providing new ideas and methods for defect inversion. The present invention determines the fitting parameters of the lift-off value attenuation equation under different defect parameters by changing the size of the lift-off value, and constructs a mapping relationship between the fitting parameters and the defect parameters, thereby providing ideas for defect identification and inversion of leakage magnetic detection. The present invention has high accuracy, small absolute error and relative error, and high confidence. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of the present invention;

[0037] Figure 2 It is a schematic diagram of defect parameters (defect diameter L and defect depth D);

[0038] Figure 3 It is the radial component distribution diagram of the leakage magnetic signal under different lift-off values ​​at the defect;

[0039] Figure 4 It is B x-max Scatter plot with lift-off value and its exponential fitting result graph;

[0040] Figure 5 It is a comparison chart between the defect quantification value obtained by machine learning using the neural network algorithm and the actual defect parameter value. DETAILED DESCRIPTION

[0041] In order to more clearly illustrate the purpose, scheme and advantages of the present invention, a specific example is given below to further explain the present invention in detail. In addition, the specific embodiment described here is only used to explain the present invention and is not used to limit the present invention.

[0042] like Figure 1 As shown, the steps of the present invention include: Step 1: constructing a magnetizer model, and adjusting various parameters of the magnetizer so that the pipeline reaches a saturated state.

[0043] The magnet material of the magnetizer is permanent magnet, and the pipeline is made of 1008 steel (Q235 steel). According to its BH curve, the saturation magnetization intensity of the pipeline is about 1750mT.

[0044] Adjust various parameters of the magnetizer, wherein the adjustable parameters include coercive force, magnet length, magnet width, magnetic pole spacing and magnet thickness.

[0045] At this time, the average magnetic induction intensity inside the pipeline is about 1754mT, reaching saturation state, and obvious leakage magnetic signals can be detected.

[0046] Step 2: According to Figure 2 In the given example, a 3D model of a pipeline with defects is constructed.

[0047] Properly increasing the sample size can significantly improve the learning effect of the neural network algorithm. In this case, a total of 81 groups of defects were designed: the defect diameter was taken from 10mm to 50mm, and the value was taken every 5mm, and a total of 9 groups of values ​​were obtained; the defect depth was taken from 1mm to 9mm, and the value was taken every 1mm, and a total of 9 groups of values ​​were obtained. Combining the two, a total of 81 groups of defect samples were obtained.

[0048] Step 3: Use finite element simulation software to calculate all prefabrication defects and obtain the following Figure 3 The result graph (part) is shown.

[0049] Step 4: Change the height of the extraction curve from the pipe wall to simulate the radial component distribution of the leakage magnetic signal under different lift-off values.

[0050] Step 5: Extract the radial component distribution of the leakage magnetic signal under different lift-off values.

[0051] Step 6: Extract the radial component peak B x-max , draw B x-max Scatter plot with varying lift-off values.

[0052] Step 7: In this case, the exponential fitting method is used to fit the scatter plot obtained in Step 6 to obtain the fitting parameters.

[0053] The curve is fitted exponentially and optimized using the Levenberg-Marquardt algorithm (LM algorithm, a least squares optimization algorithm in nonlinear regression). Figure 4 Results (partial) are shown.

[0054] Step 8: Take the fitting parameters of different defects as input and the defect parameters as output, and use the neural network algorithm for machine learning. After learning, the system automatically outputs the defect parameters such as Figure 5 shown.

[0055] Step 9: Verify the learning results to ensure the learning effect.

[0056] Step 10: Prefabricate defects of other parameters and perform simulation calculations.

[0057] Step 11: Use the equation used in Step 7 to fit and obtain the fitting parameters.

[0058] Step 12: Bring the fitting parameters into the learning results, and the system automatically outputs the defect parameters to complete the defect quantification.

[0059] For Step 10 to Step 12, this case randomly selected 5 groups of defects and brought them into the training results for inversion. The results are shown in Table 1 below.

[0060] Table 1 Random defect quantitative analysis table

[0061]

[0062] It can be seen from Table 1 that this defect quantification method has high accuracy, and both the absolute error and the relative error are relatively small, which means that this method has high confidence and is feasible.

[0063] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art of the present invention, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. A pipeline defect quantitative analysis method, characterized by: The following steps are involved: Step 1: Build a magnetizer model and adjust the magnetizer parameters to saturate the pipeline; Step 2: Simulate the leakage magnetic field distribution under different parameter defects, extract the radial component of the magnetic flux density under different lift-off values ​​x, and extract the radial component peak value B x-max ; Step 3: The radial component peak value B obtained in Step 2 x-max Fitting is performed on the scatter plot with the change of lift-off value x to obtain fitting parameters; Step 4: Use neural network algorithm to construct the relationship between defect parameters and fitting parameters for quantitative analysis; In Step 1, the method for judging whether the pipeline is magnetized to saturation is as follows: perform simulation calculation based on the defect-free model, extract the magnetic flux density value at the center of the pipeline and compare it with the BH curve of the pipeline material; when the magnetic flux density value at the center reaches the saturation point of the BH curve of the pipeline material, the pipeline magnetization reaches saturation; In Step 3, the method of obtaining the fitting curve parameters is performed according to the scatter plot of the leakage magnetic signal changing with the lift-off value. The specific steps are as follows: Step 3.1: For a certain defect, make B x-max Scatter plot with variation of lift-off value x; Step 3.2: Using the exponential fitting method, the fitting equation is: y=A+Bexp(C*x) A, B, C are fitting parameters, y is the peak value of the radial component of magnetic flux density B x-max ; In Step 4, a neural network algorithm is used to construct the relationship between defect parameters and fitting parameters. The specific steps are as follows: Step 4.1: Take the fitting parameters as input and the defect parameters as output; Step 4.2: Use neural network algorithm to train the input sample data; Step 4.3: After the training is completed, the original data is input as input, and the output results after machine learning are obtained. The learning effect is verified by comparing with the actual output results; Step 4.4: Randomly give defect parameters for simulation and obtain fitting parameters. Input the data as input and use machine learning to obtain output results. Compare the output results with the actual output results to conduct defect quantitative analysis. In Step 2, the prefabricated defect is a semi-elliptical corrosion pit, and the pit diameter L and pit depth D are used to characterize the defect parameters; In Step 3.1, the radial component curve of magnetic flux density is two peaks, one positive and one negative, which are symmetrical. The positive peak value B is extracted. x-max_1 and negative peak B x-max_2 , and find the average value to calculate B x-max : The fitting method also includes power function fitting or quadratic fitting, wherein the power function fitting formula is: y=Ax B The quadratic fitting formula is: y=A+Bx+Cx 2 ; In Step 1, the magnetizer parameters include: magnet coercivity, magnet length, magnet width, magnet thickness and magnetic pole spacing; By changing the size of the lift-off value, the fitting parameters of the lift-off value attenuation equation under different defect parameters are determined, and the mapping relationship between the fitting parameters and the defect parameters is constructed, thereby providing ideas for defect identification and inversion of magnetic flux leakage detection. The accuracy is high, the absolute error and relative error are relatively small, and the confidence level is high.

Images