A Quantitative Evaluation Method for Multiple Types of Defects in Oil and Gas Pipelines Based on Magnetic Flux Leakage Signals

Abstract

A quantitative evaluation method for multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals, comprising the following steps: measuring magnetic flux leakage signals of pipelines with different defect characteristic parameters through double probes; establishing a non-linear fast calculation model, establishing a particle swarm algorithm to optimize a support vector machine network to achieve accurate identification of metal loss defect types; using double probes to perform magnetic flux leakage detection on in-service pipelines to obtain measurement signals; inputting the measurement signals with clear defect position information into the FCM algorithm to output defect types; determining an initial estimated value of the size parameters of the defect to be measured; solving the target signal of the corresponding measurement signal; judging whether the loss function value meets the preset accuracy; using a particle swarm intelligent optimization algorithm to update the defect characteristic parameters to achieve the evaluation of optimal defect characteristic parameters; the magnetic flux leakage signals collected by the double probes of the present invention can intelligently output defect position, type and size information, and realize the detection and identification of different types of metal loss defects in pipelines.

Description

Technical Field

[0001] The present invention relates to the technical field of non-destructive testing, and particularly to a method for quantitatively evaluating various types of defects in oil and gas pipelines based on magnetic flux leakage signals. Background Art

[0002] Oil and gas long-distance pipelines, as a transportation equipment, are widely used in industries such as petroleum, petrochemical, and chemical industries. With the increase of operation time, problems in pipeline design, manufacturing, installation, and operation management are gradually exposed, resulting in pipeline accidents from time to time, posing threats to people's lives and property safety, social stability, and industrial production. Regularly using non-destructive testing technology for pipeline inspection, timely finding and repairing pipeline defects can eliminate potential safety hazards during pipeline operation, reduce the occurrence of accidents, and extend the pipeline life.

[0003] Due to reasons such as the pipeline serving in a relatively harsh environment for a long time and bearing alternating repeated loads for a long time, different types of metal loss defects often appear inside and outside the pipeline. Referring to the metal loss types given in the national standard (GBT27699-2011) Steel Pipeline In-line Inspection Technical Specification, according to the length and width characteristics of the defects, the pipeline metal loss type defects are divided into the following 6 types: pinhole, horizontal groove, horizontal concave groove, tangential groove, tangential concave groove, and pit-like defect, as shown in Figure 1 the attachment.

[0004] The magnetic flux leakage detection technology is widely used in the defect detection of oil and gas pipelines. In actual detection conditions, the defect positions and types of the pipelines to be tested are unknown. However, in the existing defect evaluation methods of magnetic flux leakage detection technology, it is usually assumed that the defect sizes are quantitatively evaluated under the condition of known defect positions and types. This default assumption scheme seriously affects the accuracy of defect parameter evaluation and poses a great challenge to the quantitative evaluation of defect sizes. Summary of the Invention

[0005] In order to overcome the above defects existing in the prior art, the purpose of the present invention is to provide a method for quantitatively evaluating various types of defects in oil and gas pipelines based on magnetic flux leakage signals. Through the magnetic flux leakage signals collected by double probes, the defect positions, types, and size information can be intelligently output, realizing the detection and identification of different types of metal loss defects in pipelines.

[0006] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0007] A method for quantitatively evaluating various types of defects in oil and gas pipelines based on magnetic flux leakage signals, comprising the following steps;

[0008] Step 1: Fabricate a reduced-scale pipeline with different defect characteristic parameters;

[0009] Step 2: Use dual probes to measure the magnetic flux leakage signals of pipelines with different defect characteristic parameters, and obtain a dataset of magnetic flux leakage signals collected by the detection probes at different liftoff values for different defect characteristic parameters;

[0010] Step 3: Based on the neural network algorithm, establish a non-linear fast calculation model for the relationship between different defect characteristic parameters, different liftoff values t 1 and the magnetic flux leakage detection signals;

[0011] Step 4: Based on the dataset established in Step 2 considering the relationship between defect characteristic parameters and liftoff value t 1 and the magnetic flux leakage signals, establish a particle swarm optimization support vector machine network (PSO-SVM) for locating the defect position, and accurately predict the defect positions on the inner and outer walls of the pipeline through the collected magnetic flux leakage signals;

[0012] Step 5: Based on the dataset established in Step 2 considering the relationship between defect characteristic parameters and liftoff value t 1 and the magnetic flux leakage signals, establish a fuzzy clustering algorithm (FCM) that can classify multiple types of pipeline defects, and accurately identify the types of metal loss defects through the collected magnetic flux leakage signals;

[0013] Step 6: Use dual probes to perform magnetic flux leakage detection on the in-service pipeline to obtain the measurement signals;

[0014] Step 7: Input the measurement signals obtained in Step 6 into the PSO-SVM algorithm established in Step 4 to obtain the defect position information; then input the measurement signals with clear defect position information into the FCM algorithm established in Step 5 to output the defect type;

[0015] Step 8: Determine the initial estimated values of the size parameters of the defect to be measured; only consider the defect width w, defect depth d, defect length l, and defect angle θ;

[0016] Step 9: Based on the non-linear fast calculation model established in Step 3, solve the target signal corresponding to the measurement signal; compare the target signal with the measurement signal and calculate the loss function value;

[0017] Step 10: Determine whether the loss function value meets the preset accuracy (5%); use whether the loss function value meets the preset accuracy as an index to evaluate the optimization effect of the optimization algorithm. If the loss function value does not meet the accuracy requirement (5%), then go to Step 11 to further optimize the defect characteristic parameters; if it meets, end the iteration and output the optimal defect characteristic parameters for evaluation;

[0018] Step 11: Use the particle swarm intelligent optimization algorithm to update the defect characteristic parameters to achieve the evaluation of the optimal defect characteristic parameters;

[0019] Step Twelve: Output the coordinates of the defect position, the defect type, and the defect size parameters output in the final iteration step.

[0020] Through the above steps, the system can continuously evaluate and optimize each link in the defect detection and evaluation process to ensure that the final output result can meet the expected accuracy and precision requirements.

[0021] In the first step, the defect feature parameters are the defect width, defect depth, defect length, defect angle, defect planar position, and defect type.

[0022] The second step is specifically: Establish a data set that describes the corresponding relationship between the defect feature parameters and the lift-off value t 1 and the magnetic flux leakage signal; the dual-probe for collecting the magnetic flux leakage signal is composed of sensor 1 and sensor 2 vertically stacked, where the lift-off value of sensor 1 from the surface of the object to be measured is t 1 ; the lift-off value of sensor 2 from the surface of the object to be measured is t 2 ; the lift-off interval between sensor 1 and sensor 2 is ω.

[0023] The third step is specifically:

[0024] The established non-linear fast calculation model is as shown in formula (1).

[0025]

[0026] Where w is the defect width; d is the defect depth; l is the defect length; θ is the defect angle; (x,y) is the defect position coordinate; F c is the defect category, c = 1, 2, 3,... 6, respectively representing pinhole, horizontal groove, horizontal concave groove, tangential groove, tangential concave groove, and pit-shaped defect; B 1 is the target signal obtained by sensor 1 in the dual-probe; B 2 is the target signal obtained by sensor 2 in the dual-probe; f represents the non-linear function of the relationship between the defect feature parameters, the lift-off value, and the magnetic flux leakage signal.

[0027] The measurement signals in the sixth step are and Where is the measured magnetic flux leakage signal obtained by sensor 1 in the dual-probe, is the measured magnetic flux leakage signal obtained by sensor 2 in the dual-probe.

[0028] The loss function in the ninth step is:

[0029]

[0030] Where, is when the defect parameter is and the target signal corresponding to the liftoff value t of sensor 1 1 when; is when the defect parameter is and the liftoff value of sensor 2 is t 2 when the corresponding target signal; is the measurement signal measured by sensor 1, is the measurement signal measured by sensor 2.

[0031] In the eleventh step, by repeating the analysis through the iterative steps nine and ten until the loss function value meets the accuracy requirements, through the above dynamic evaluation and feedback process, continuously improve the accuracy and accuracy of the proposed method in the optimization process, ensure that the adopted optimization algorithm can converge to the optimal solution that meets the accuracy requirements within an appropriate number of iterations, and realize the optimal defect characteristic parameter evaluation.

[0032] Advantages of the present invention:

[0033] The present invention provides a combined scheme for evaluating the defect position, defect type and defect size parameters. The proposed scheme realizes the accurate identification of the positions, types and sizes of various types of pipeline defects by combining multiple intelligent algorithms with classification and regression functions, improves the detection speed and efficiency of various types of pipeline defects, and reduces the detection time and labor costs.

[0034] In a method for quantitatively evaluating various types of pipeline defects based on magnetic flux leakage signals provided by the present invention, a non-linear fast calculation model considering various characteristic parameters of pipeline metal loss defects (irregular defect size, angle, position and defect category) is established, realizing the prediction of magnetic flux leakage signals of double probes under any defect characteristic parameters, and improving the evaluation accuracy of defect information.

[0035] Compared with other manual discrimination or intelligent algorithms, the present invention realizes the effective identification of internal and external wall defects of pipelines through the particle swarm optimization support vector machine network (PSO-SVM), realizes the accurate distinction of various types of defects through the fuzzy clustering algorithm (FCM), and realizes the quantitative evaluation of defect size parameters through the particle swarm algorithm. It provides richer and more comprehensive defect characteristic information, which helps to formulate more effective maintenance plans and measures in the future. Description of the drawings

[0036] Figure 1 Classification standard for metal loss type defects. (A refers to the geometric parameters of the defect, L represents the defect length, W represents the defect width. If the normal wall thickness t of the pipeline < 10 mm, A = 10 mm; if t ≥ 10 mm, A = t.)

[0037] Figure 2 Flow chart for establishing a method for detecting and intelligently evaluating magnetic flux leakage signals of different types of metal loss defects in pipelines.

[0038] Figure 3 Schematic diagram of dual probes.

[0039] Figure 4 Describe the defect characteristic parameters and the lift-off value t 1 And the non-linear function of the relationship between the magnetic flux leakage signals.

[0040] Figure 5 Operation execution flowchart of the magnetic flux leakage signal detection and intelligent evaluation method for different types of metal loss defects in pipelines. Specific implementation manner

[0041] The present invention will be further described in detail below with reference to the accompanying drawings.

[0042] A quantitative evaluation method for multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals provided by the present invention, as shown in the attached Figure 2 , The specific implementation manner of the establishment process of this method is as follows:

[0043] Step 1: Fabricate a scaled-down pipeline with different defect characteristic parameters (defect width, defect depth, defect length, defect angle, defect planar position, and defect type);

[0044] Step 2: Measure the magnetic flux leakage signals of pipelines with different defect characteristic parameters through dual probes to obtain the magnetic flux leakage signals of different defect parameters at different lift-off values t 1 ;

[0045] Establish a data set describing the corresponding relationship between the defect characteristic parameters and the lift-off value t 1 And the magnetic flux leakage signals. The schematic diagram of the dual probes used to collect the magnetic flux leakage signals is shown in the attached Figure 3 , which is composed of sensor 1 and sensor 2 stacked vertically. The lift-off value of sensor 1 from the surface of the object to be measured is t 1 ; The lift-off value of sensor 2 from the surface of the object to be measured is t 2 ; The lift-off interval between sensor 1 and sensor 2 is ω.

[0046] Step 3: Based on the neural network algorithm, establish a non-linear fast calculation model for the relationship between different defect characteristic parameters, different lift-off values t 1 And the magnetic flux leakage detection signals.

[0047] The established non-linear fast calculation model is as shown in formula (1).

[0048]

[0049] Where w is the defect width; d is the defect depth; l is the defect length; θ is the defect angle; (x, y) is the defect position coordinate; F cLet \(c = 1, 2, 3, \cdots, 6\) be the defect category, representing pinhole, horizontal groove, horizontal concave groove, tangential groove, tangential concave groove, and pit-shaped defect respectively; \(B\) 1 is the target signal obtained by sensor 1 in the dual-probe; \(B\) 2 is the target signal obtained by sensor 2 in the dual-probe; \(f\) represents the non-linear function of the relationship between the defect characteristic parameter, lift-off value, and magnetic flux leakage signal, as shown in the appendix Figure 4 shown

[0050] Step 4: Based on the data set established in Step 2 considering the relationship between the defect characteristic parameter, lift-off value \(t\) 1 and the magnetic flux leakage signal, establish a PSO-SVM algorithm that can locate the defect position, and accurately predict the defect positions on the inner and outer walls of the pipeline through the collected magnetic flux leakage signals.

[0051] The specific establishment process is as follows:

[0052] In the process of using SVM, selecting an appropriate kernel function is an important step. Its commonly used kernel functions mainly include polynomial function, radial basis function (RBF), and sigmoid function. Here, the RBF radial basis function is taken as an example to establish an SVM classifier. The penalty factor \(C\) and kernel parameter \(\sigma\) in the kernel function affect the accuracy of the classifier. Here, the PSO optimization algorithm is used to find the optimal solutions of \(C\) and \(\sigma\).

[0053] The specific process is as follows:

[0054] Step ①: Initialize the PSO algorithm parameters. Set the maximum number of iteration steps \(iter\) max = 10; determine the size of the particle swarm \(pop = 10\); set the inertia weight \(weight\) as a function that linearly decreases with time. The functional form of the inertia weight is: \(weight = weight\) max -(weight max -weight min )×(k / iter max ), where the initial inertia weight \(weight\) max = 0.9, the final inertia weight \(weight\) min = 0.2, \(k\) is the \(k\)th iteration; the learning factor \(c\) 1 = c 2 = 2.

[0055] Step ②: Encode the SVM network parameters. Encode the kernel parameter \(\sigma\) and penalty factor \(C\) of SVM as the position of the particle. Here, a two-dimensional feature space is considered. Randomly generate the population position \(X\), where \(X=(X 1 ,X 2 ,...,X j ,..,X N ), \(X\)j represents the position of the j-th particle, which consists of the penalty factor C and the RBF parameter with two components. Randomly generate the velocity v of the particle, where v ∈ [-v max , v max . Set the ranges of C and б.

[0056] Step ③: Build the SVM network. Input the encoded SVM parameters obtained based on the training set into the SVM network. Here, is the input vector obtained from the validation set, where the subscripts x and y represent the horizontal and vertical components of the magnetic flux leakage signal, and the superscripts t 1 and t 2 represent the liftoff values t 1 and t 1 +ω; is the support vector obtained from the training set, where i = 1,..., M, M is the total number of samples, and i is the i-th sample. The output is the sample label, where the outer wall defect label is 1 and the inner wall defect label is 2.

[0057] Step ④: Calculate the initial fitness of each particle. By inputting the defect location information obtained in Step ③ into where fitness represents the fitness function used to determine the accuracy of the classifier. M is the number of data samples, y(i) is the accuracy of the i-th prediction of the defect category by the support vector machine, and y i is the accuracy of the true defect category.

[0058] Step ⑤: Update the velocity and position of the particle.

[0059] Step ⑥: Recalculate the fitness of each particle.

[0060] Step ⑦: Determine whether the termination condition is satisfied. If it is satisfied, output the global optimal solution. If not, go back to Step ⑤ to continue the optimization.

[0061] Step ⑧: Output the best SVM parameters C and б, and obtain the trained PSO - SVM network architecture.

[0062] Step ⑨: Network verification. Input the test set into the trained PSO - SVM network and output the defect location situation.

[0063] Step five: Based on the data set established in Step two considering the defect characteristic parameters and the relationship between the liftoff value t 1 and the magnetic flux leakage signal, establish the FCM algorithm that can classify multiple types of pipeline defects, and accurately identify the types of metal loss defects through the collected magnetic flux leakage signals.

[0064] The specific establishment process is as follows:

[0065] (1) Build the FCM network. Given a training set of data containing M prediction signals: where i = 1, 2, 3,..., M. In this case study, the number of clustering types C = 6, and the cluster centers of the C classes are [c 1 , c 2 ,..., c j ,..., c c , j = 1, 2, 3,..., C. B ij is the j-th sample attribute of B i .

[0066] Step ①: Determine the number of cluster centers C, the fuzzy factor m, and the iteration stop threshold ε. In the provided case, C = 6, m is taken as 2 according to the empirical value, and the stop threshold ε = 10 -6 .

[0067] Step ②: Initialize the membership degree u ij between the random values [0, 1]. where u ij is the membership degree of the i-th data B i belonging to the cluster center c j .

[0068] Step ③: Calculate the cluster center c ij according to the membership degree u i .

[0069] Update the membership degree: where l is the classification level of the cluster center after iteration, l = 1, 2, 3,..., C.

[0070] Calculate the cluster center:

[0071] Step ④: Calculate the objective function J, where

[0072] Step ⑤: Compare J(k) and J(k - 1), where k is the k-th iteration; if ||J(k) - J(k - 1)|| ≤ ε, that is, the iteration termination condition is satisfied, execute Step ⑥ and stop the iteration; otherwise, set k = k + 1, then return to Step ③ and continue the iteration.

[0073] Step ⑥: Output the defect classification result.

[0074] So far, the trained FCM network architecture is obtained.

[0075] (2) Verify the network effect. Input the measurement signal into the above-trained FCM algorithm and output the defect classification situation.

[0076] Step 6: Use dual probes to perform magnetic flux leakage detection on the pipeline in the actual service condition to obtain the measurement signal. The measurement signal is and Evaluate the defect characteristics based on the measured signal collected, where is the measured magnetic flux leakage signal obtained by sensor 1 in the dual probes, is the measured magnetic flux leakage signal obtained by sensor 2 in the dual probes.

[0077] Step 7: Determine the initial estimated values of the defect size parameters to be measured, only considering the defect width w, defect depth d, defect length l, and defect angle θ.

[0078] Step 8: Based on the non-linear fast calculation model established in Step 3, solve the target signal corresponding to the measured signal. Compare the target signal with the measured signal and calculate the loss function value.

[0079] The loss function is:

[0080]

[0081] where, is the target signal corresponding to when the defect parameter is and the lift-off value of sensor 1 is t 1 ; is the target signal corresponding to when the defect parameter is and the lift-off value of sensor 2 is t 2 ; is the measured signal under the measurement of sensor 1, is the measured signal under the measurement of sensor 2.

[0082] Step 9: If the loss function value does not meet the accuracy requirement (5%), then go to Step 10; if it meets, then go to Step 11 to end the iteration.

[0083] Step 10: Use the particle swarm algorithm to update the defect characteristic parameters. Repeat the analysis by iterating from Step 8 to Step 10 until the loss function value meets the accuracy requirement.

[0084] The specific process steps of the particle swarm algorithm are:

[0085] Step ①: Initialize the PSO parameters. Set the maximum number of iteration steps iter max = 40; Determine the size of the particle swarm pop = 50; The maximum speed of the particle is v max = 1; Learning factor c 1 = 2.5 and learning factor c 2 = 3; Set the inertia weight weight as an exponential function, and the functional form of the inertia weight is: where the initial inertia weight weightmax = 0.9, the final inertia weight weight min = 0.4.

[0086] Step ②: Initialize the positions and velocities of the particles. Randomly generate the population positions X = (X 1 , X 2 ,..., X j ,.., X N ), representing the position of the j-th particle, where X j is composed of four components: defect length, defect depth, defect width, and lift-off value t 1 . Set the limit ranges of the four components respectively.

[0087] Step ③: Calculate the initial fitness of the particles. The fitness function is where, is the predicted signal corresponding to when the defect parameters are and the lift-off value of sensor 1 or sensor 2 is t, and are the magnetic flux leakage signals of the x-direction components obtained through sensors 1 and 2 respectively, and are the magnetic flux leakage signals of the y-direction components obtained through magnetic sensors 1 and 2 respectively; is the measured signal, and are the measured signals of the x-direction components collected through sensors 1 and 2 respectively, and are the measured signals of the y-direction components collected through sensors 1 and 2 respectively.

[0088] Step ④: Update the velocities and positions.

[0089] Step ⑤: Recalculate the fitness of each particle.

[0090] Step ⑥: Judge the termination condition. Judge whether the fitness all reaches the threshold ε. If it is satisfied, output the global optimal solution. If not, go back to Step ④ to continue the optimization.

[0091] Step ⑦: Output the obtained defect characteristics.

[0092] Step Eleven: Output the defect position coordinates, defect type, and defect size parameters identified in the final iteration step.

[0093] A magnetic flux leakage signal detection and intelligent evaluation method based on different types of metal loss defects in pipelines provided by the present invention, as shown in the appendix Figure 5 , the specific implementation manner of the operation and execution process of this method is as follows:

[0094] Step 1: Use a dual-probe to perform magnetic flux leakage detection on the in-service pipeline to obtain measurement signals. The measurement signals are and where is the measured magnetic flux leakage signal obtained by sensor 1 in the dual-probe, is the measured magnetic flux leakage signal obtained by sensor 2 in the dual-probe.

[0095] Step 2: Input the measurement signals obtained in Step 1 into the PSO-SVM algorithm to obtain the defect position coordinates.

[0096] Step 3: Input the measurement signals obtained in Step 1 into the FCM algorithm to obtain the defect type.

[0097] Step 4: Determine the initial estimated values of the size parameters of the defect to be measured, only considering the defect length, width, depth, and angle.

[0098] Step 5: Input the initial estimated values of the size parameters of the defect to be measured, the measurement signals obtained in Step 1, the defect position coordinates obtained in Step 2, and the defect type information obtained in Step 3 into the non-linear fast calculation model to solve for the target signal.

[0099] Step 6: Compare the target signal with the measurement signal and calculate the loss function value.

[0100] Step 7: Determine whether the loss function meets the minimum error (5%). If so, execute Step 9 and the iteration ends; if not, then execute Step 8.

[0101] Step 8: Use the particle swarm algorithm to update the defect feature parameters and re-iterate Steps 6 - 8 until the loss function meets the minimum error requirement.

[0102] Step 9: Output the defect position coordinates, defect type, and defect feature parameters.

[0103] So far, the present invention has solved the integrated and complete evaluation of pipeline metal loss defects, including defect position location, defect type differentiation, and quantitative identification of defect size, improving the detection speed and efficiency of pipeline metal loss defects, providing richer and more comprehensive defect information, and helping to formulate more effective maintenance plans and measures in the future.

Claims

1. A quantitative evaluation method for multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals, characterized in that: The steps include: Step 1: Make proportionally reduced pipelines with different defect characteristic parameters; Step 2: Measure the leakage magnetic signal of the pipeline with different defect characteristic parameters through dual probes to obtain a leakage magnetic signal data set of different defect characteristic parameters collected by the detection probe at different lift-off values; Step 3: Based on the neural network algorithm, a nonlinear fast calculation model of the relationship between different defect characteristic parameters and different lift-off values ​​and magnetic flux leakage detection signals is established; Step 4: Based on the data set established in step 2 that considers the relationship between the defect characteristic parameters and the lift-off value and the magnetic leakage signal, a particle swarm algorithm optimized support vector machine network PSO-SVM for locating the defect position is established, and the defect position of the inner and outer walls of the pipeline is accurately predicted through the collected magnetic leakage signal; Step 5: Based on the data set established in step 2 that considers the relationship between defect characteristic parameters and lift-off values ​​and magnetic leakage signals, a fuzzy clustering algorithm FCM for classifying multiple types of pipeline defects is established, and accurate identification of metal loss defect types is achieved through the collected magnetic leakage signals; Step 6: Use dual probes to perform magnetic flux leakage detection on the pipeline in service to obtain measurement signals; Step 7: Input the measurement signal obtained in step 6 into the PSO-SVM algorithm established in step 4 to obtain the defect location information; input the measurement signal with clear defect location information into the FCM algorithm established in step 5 to output the defect type; Step 8: Determine the initial estimated value of the defect size parameter to be tested; Step nine: based on the nonlinear fast calculation model established in step three, solve the target signal corresponding to the measured signal; compare the target signal with the measured signal and calculate the loss function value; Step 10: Determine whether the loss function value meets the preset accuracy of 5%; determine whether the loss function value meets the preset accuracy as an indicator for evaluating the optimization effect of the optimization algorithm. If the loss function value does not meet the accuracy requirement of 5%, go to step 11 to further optimize the defect feature parameters; if it does, end the iteration and output the optimal defect feature parameters evaluated; Step 11: Use the particle swarm intelligent optimization algorithm to update the defect characteristic parameters and achieve the optimal defect characteristic parameter evaluation; Step 12: Output the defect position coordinates, defect type and defect size parameters output in the final iteration step; The step 2 is specifically as follows: establishing a data set describing the corresponding relationship between the defect characteristic parameters and the lift-off value and the magnetic flux leakage signal; the dual probe for collecting the magnetic flux leakage signal is composed of a vertical stack of sensor 1 and sensor 2, wherein the lift-off value of sensor 1 from the surface of the object to be measured is t1; the lift-off value of sensor 2 from the surface of the object to be measured is t2; and the lift-off interval between sensor 1 and sensor 2 is ω; The step three is specifically as follows: The established nonlinear fast calculation model is shown in formula (1): Where w is the defect width; d is the defect depth; l is the defect length; θ is the defect angle; (x, y) is the defect position coordinate; F c is the defect category, c = 1, 2, 3, ... 6, representing pinhole, horizontal groove, horizontal groove, tangential groove, tangential groove and pit defect respectively; B1 is the target signal obtained by sensor 1 in the dual probe; B2 is the target signal obtained by sensor 2 in the dual probe; f is the nonlinear function of the relationship between the defect characteristic parameters, lift-off value and leakage magnetic signal.

2. The method for quantitatively evaluating multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals according to claim 1 is characterized in that: In the step 1, the defect characteristic parameters are defect width, defect depth, defect length, defect angle, defect plane position and defect type.

3. The method for quantitatively evaluating multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals according to claim 1 is characterized in that: The measurement signal in step six is: and in is the measured magnetic flux leakage signal obtained by sensor 1 in the dual probe, It is the measured magnetic flux leakage signal obtained by sensor 2 in the dual probe.

4. The method for quantitatively evaluating multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals according to claim 3 is characterized in that: The loss function of step nine is: in, When the defect parameter is and the target signal corresponding to the lift-off value of sensor 1 being t1; When the defect parameter is And the target signal corresponding to the lift-off value of sensor 2 is t2.

5. The method for quantitatively evaluating multiple types of defects in oil and gas pipelines based on magnetic flux leakage signals according to claim 1 is characterized in that: In the step eleven, the analysis is repeated by iterating steps nine and ten until the loss function value meets the accuracy requirement. Through the above-mentioned dynamic evaluation and feedback process, the accuracy and precision of the proposed method in the optimization process are continuously improved, ensuring that the adopted optimization algorithm can converge to the optimal solution that meets the accuracy requirement within an appropriate number of iterations, thereby achieving the optimal defect characteristic parameter evaluation.

Images