A method and system for pipe thickness measurement that is resistant to interference from a multi-layer thermal insulation medium

Abstract

The present application relates to the technical field of electromagnetic nondestructive testing, in particular to a pipeline thickness measurement method and system resisting interference of multi-layer heat preservation medium, wherein the method comprises the following steps: using an EMD algorithm improved by a particle swarm algorithm to filter pulse eddy current signals of pipelines with different thicknesses under different heat preservation media to obtain pulse eddy current filtered signals; extracting medium-term signal intercept eigenvalues, power spectral density peak eigenvalues and magnetic flux eigenvalues of the pulse eddy current filtered signals of the pipelines with different thicknesses under different heat preservation media; performing linear compensation on the extracted medium-term signal intercept eigenvalues, power spectral density peak eigenvalues and magnetic flux eigenvalues of the pulse eddy current signals; and establishing a pipeline thickness inversion model of multi-dimensional feature fusion by using the linearly compensated eigenvalues. The present application performs linear compensation on the extracted power spectral density peak, magnetic flux and medium-term signal intercept under different heat preservation media to eliminate the interference of different heat preservation media on thickness measurement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electromagnetic nondestructive testing, and more particularly to a pipeline thickness measurement method and system resistant to multi-layer thermal insulation medium interference. BACKGROUND

[0002] In the fields of petroleum, chemical industry and energy, pipelines are the core carriers for transporting media such as crude oil, natural gas and high-temperature steam, and are long-term in extreme working conditions of high pressure, high temperature and strong corrosion. Most of these pipelines need to be coated with multi-layer thermal insulation medium, and the inner layer is usually made of thermal insulation materials such as magnesium aluminum silicate, aluminum silicate, polyurethane and foam glass, and the outer layer is finished with a metal protective layer to achieve heat insulation, corrosion protection and mechanical protection. However, once defects such as wall thickness thinning and local corrosion occur in such pipelines, they may cause media leakage, explosion and other serious accidents. Traditional detection methods, such as ultrasonic thickness gauges, have obvious limitations. However, the pulsed eddy current detection technology has unique advantages in this scenario: it can achieve continuous scanning without removing the thermal insulation layer, and can complete the detection within the continuous production cycle of oil transportation stations and chemical plants.

[0003] The multi-layer thermal insulation medium interference problem of petroleum, chemical and energy pipelines is more complex. The thermal insulation layer may be partially compacted or hollow due to long-term vibration, and the high-temperature environment of energy pipelines may also cause fluctuations in the dielectric constant of the thermal insulation layer. Different thermal insulation media have different effects on pulsed eddy current signals. How to resist the interference of different types of thermal insulation media to achieve accurate measurement of pipeline thickness has become a top priority for pulsed eddy current detection. Currently, many scholars mostly use a single eigenvalue for thickness inversion, but for different thermal insulation medium interference, a single eigenvalue cannot achieve accurate measurement. Therefore, how to resist multi-layer thermal insulation medium interference when measuring pipeline thickness has become a technical problem that needs to be solved by technical personnel in the field. SUMMARY

[0004] Therefore, the present application provides a pipeline thickness measurement method and system resistant to multi-layer thermal insulation medium interference, which can eliminate the interference of different thermal insulation media on pipeline thickness measurement.

[0005] To achieve the above purpose, the present application adopts the following technical solutions:

[0006] In a first aspect, the present application provides a pipeline thickness measurement method resistant to multi-layer thermal insulation medium interference, comprising the following steps:

[0007] S1: using an EMD algorithm improved by a particle swarm algorithm to filter the pulsed eddy current signals of pipelines of different thicknesses under different thermal insulation media, to obtain pulsed eddy current filtered signals;

[0008] S2: extract the mid-term signal intercept characteristic value, power spectral density peak characteristic value and magnetic flux characteristic value of the pulse eddy current filtered signal of the pipeline with different thickness under different heat preservation medium;

[0009] S3: linearly compensate the mid-term signal intercept characteristic value, power spectral density peak characteristic value and magnetic flux characteristic value of the extracted pulse eddy current signal;

[0010] S4: establish a pipeline thickness inversion model of multi-dimensional feature fusion by using the linearly compensated characteristic values.

[0011] Further, S1 comprises:

[0012] S101: fix the basic parameters of EMD algorithm, set the parameters and optimization variables of particle swarm algorithm, and the optimization variables are the screening stop threshold, the effective IMF starting layer and the effective IMF ending layer;

[0013] S102: initialize the particle swarm and calculate the initial fitness, randomly generate n groups of parameters in the constraint range, integerize the IMF layer number and ensure to meet the constraint; perform EMD decomposition for each group of parameters, accumulate the components from the effective IMF starting layer to the effective IMF ending layer to reconstruct the signal, calculate the signal-to-noise ratio (SNR) of the reconstructed signal and the original signal as the initial fitness; record the individual optimal and global optimal;

[0014] S103: update the fitness and optimal solution in iteration, linearly decrease the inertia weight in each iteration, update the speed and position according to the current position, individual optimal and global optimal of the particle, and perform constraint processing on the new position parameters; recalculate the SNR of the new parameters as the fitness, and update the individual optimal and global optimal position;

[0015] S104: when the iteration number reaches m times or the global optimal is not updated for k consecutive generations, stop optimization, output the parameters corresponding to the global optimal, use the optimal parameters to perform EMD decomposition on the original pulse eddy current signal under different heat preservation medium, reconstruct the corresponding interval IMF component, and obtain the final pulse eddy current filtered signal.

[0016] Further, in S2, the extraction step of the mid-term signal intercept characteristic value comprises:

[0017] S201: in the double logarithmic coordinate system, intercept the 10-1000 mV part of each pulse eddy current filtered signal, iterate the length and position of the intercepted signal segment, and the initial position and step of iteration are 10 mV, to find the best signal segment for calculating the mid-term signal intercept characteristic value;

[0018] S202: compare the mid-term signal intercepts of different signal segments in S21, when the characteristic values of different thickness pipelines under the same medium lift are equal, the corresponding signal segment is the best signal segment;

[0019] S203: Linear fitting is performed on the optimal signal segment of S22 to obtain the y-axis intercept of the fitting straight line as the mid-term signal intercept characteristic value of the pulsed eddy current filtered signal.

[0020] Further, in S2, the step of extracting the power spectral density peak characteristic value comprises:

[0021] S204: The pulsed eddy current filtered signal of the pipe with the maximum thickness is selected as the reference signal, and the pulsed eddy current filtered signals of the pipes with other thicknesses are respectively subtracted from the reference signal to obtain differential signals of the pipes with different thicknesses;

[0022] S205: The differential signals are intercepted at 0.01-1000 mV, and the power spectral density peaks of the differential signals at different lengths and different positions are iteratively calculated, and the initial position and step length of iteration are both 10 mV;

[0023] S206: The power spectral density peaks of the differential signals of the pipes with different thicknesses are compared, and the power spectral density peak monotonically decreases with the increase of the pipe thickness, and when the power spectral density peaks of the pipes with different thicknesses have the maximum difference, the optimal differential signal segment is obtained;

[0024] S207: The power spectral density peak of the optimal differential signal segment is calculated using the Welch method as the power spectral density peak characteristic value of the pulsed eddy current filtered signal.

[0025] Further, in S2, the step of extracting the magnetic flux characteristic value comprises:

[0026] S208: The 0.01-1000 mV part of each pulsed eddy current filtered signal is intercepted, and the signal is divided into signal segments of different lengths, and the initial position and step length are both 10 mV;

[0027] S209: The magnetic flux is calculated according to the signal segments divided in S208, and the signal segment with the maximum difference in magnetic flux of the pipes with different thicknesses is taken as the optimal signal segment;

[0028] S210: The magnetic flux of the optimal signal segment is calculated as the magnetic flux characteristic value of the pulsed eddy current filtered signal.

[0029] Further, in S3, the step of linearly compensating the power spectral density peak and the magnetic flux characteristic value comprises:

[0030] S301: The power spectral density peak characteristic values of the pipes with different thicknesses under air are taken as standard values, the power spectral density peak characteristic values of the pipes with different thicknesses under different thermal insulation media are taken as inputs, and the y=ax+b is used to fit them to obtain the compensation coefficients a and b of the power spectral density peak characteristic values under different thermal insulation media, respectively;

[0031] The peak value of the power spectrum density of the pipeline with different thicknesses under different heat preservation media and the compensation coefficients a and b are used to calculate the peak value of the power spectrum density of the pipeline with different thicknesses under different heat preservation media after compensation.

[0032] In S302, the magnetic flux characteristic value of the pipeline with different thicknesses under air is taken as a standard value, and the magnetic flux characteristic value of the pipeline with different thicknesses under different heat preservation media is taken as an input, which is fitted using y=cx+d, to obtain the compensation coefficients c and d of the magnetic flux characteristic value under different heat preservation media.

[0033] The magnetic flux characteristic value of the pipeline with different thicknesses under different heat preservation media and the compensation coefficients c and d are used to calculate the magnetic flux characteristic value of the pipeline with different thicknesses under different heat preservation media after compensation.

[0034] Further, in S3, the mutual compensation relationship of the peak value of the power spectrum density and the magnetic flux characteristic value is as follows:

[0035] The peak value of the power spectrum density and the magnetic flux characteristic value are normalized, when the pipeline thickness is 1.5-5 mm, the weight of the peak value of the power spectrum density in the thickness inversion model is increased, and the weight of the magnetic flux characteristic value in the thickness inversion model is reduced; when the pipeline thickness is 6-12 mm, the weight of the peak value of the power spectrum density in the thickness inversion model is increased, and the weight of the magnetic flux characteristic value in the thickness inversion model is reduced; and the characteristic value with high sensitivity to thickness is always taken as the dominant characteristic value of thickness inversion.

[0036] Further, in S3, the medium signal intercept characteristic value is used to replace the medium lift-off condition, and the medium signal intercept characteristic value is used as a criterion to select different compensation coefficients for the peak value of the power spectrum density and the magnetic flux characteristic value under different medium signal intercept characteristic values.

[0037] Further, S4 includes:

[0038] S401: A database of the linearly compensated peak value of the power spectrum density, the magnetic flux characteristic value and the medium signal intercept characteristic value of the pipeline with different thicknesses under different heat preservation media is established;

[0039] S402: The database is divided into a training set and a test set, a pipeline thickness inversion model based on a random forest algorithm is constructed, and training and testing are performed;

[0040] S403: The model is hyperparameter-optimized using a cross-validation grid search method;

[0041] S404: 5-fold cross-validation is performed on different parameter combinations one by one, and the parameter combination with the minimum negative mean square error is selected as the optimal solution;

[0042] S405: updating the model parameters according to the optimal solution to obtain a final pipe thickness inversion model; the input of the pipe thickness inversion model is the power spectral density peak eigenvalue, the magnetic flux eigenvalue and the intermediate signal intercept eigenvalue of the pulse eddy current signal, and the output is the pipe thickness.

[0043] In a second aspect, the present application provides a pipe thickness measurement system resistant to interference of multi-layer thermal insulation medium, which adopts the method as described above, comprising:

[0044] A signal acquisition module is configured to acquire pulse eddy current signals of pipes with different thicknesses under different thermal insulation media.

[0045] A signal filtering module is configured to filter the pulse eddy current signals of pipes with different thicknesses under different thermal insulation media using the EMD algorithm improved by the particle swarm algorithm to obtain pulse eddy current filtered signals.

[0046] A signal eigenvalue extraction module is configured to extract the intermediate signal intercept eigenvalue, the power spectral density peak eigenvalue and the magnetic flux eigenvalue of the pulse eddy current filtered signals of pipes with different thicknesses under different thermal insulation media.

[0047] A compensation module is configured to linearly compensate the intermediate signal intercept eigenvalue, the power spectral density peak eigenvalue and the magnetic flux eigenvalue of the extracted pulse eddy current signals.

[0048] A pipe thickness inversion module is configured to establish a multi-dimensional feature fusion pipe thickness inversion model using the linearly compensated eigenvalues.

[0049] According to the technical solution described above, compared with the prior art, the present application has the following beneficial effects:

[0050] 1. The EMD algorithm improved by the particle swarm algorithm is used for filtering in the present application, which can reduce noise and improve the signal-to-noise ratio.

[0051] 2. The power spectral density peak and the magnetic flux are used to compensate each other in terms of thickness and noise in the present application. The power spectral density peak has higher sensitivity at small thickness, while the magnetic flux has higher sensitivity at large thickness, and the sensitivity to thickness is always high in different thickness ranges. The two are extracted by different methods, which can reduce different types of noise.

[0052] 3. The characteristic quantities extracted under different thermal insulation media are linearly compensated in the present application, which effectively resists the interference of different thermal insulation media and improves the inversion accuracy.

[0053] 4. The intermediate signal intercept is used to replace the lift-off condition in the present application, which can reduce the influence of lift-off on inversion accuracy. The power spectral density peak, the magnetic flux and the intermediate signal intercept are used for multi-eigenvalue fusion to train the inversion model, which effectively resists the interference of different thermal insulation media and improves the inversion accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description only constitute a part of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of the provided drawings.

[0055] Figure 1 The flow chart of the pipeline thickness measurement method provided by the present application against multi-layer thermal insulation medium interference;

[0056] Figure 2 The structural schematic diagram of the pipeline with different thickness to be measured provided by the present application;

[0057] Figure 3 The structural schematic diagram of different thermal insulation media provided by the present application;

[0058] Figure 4 The pulse eddy current signal diagram of the pipeline with a thickness of 1.5-12 mm after filtering the air 50 mm away provided by the present application; DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort belong to the scope of protection of the present application.

[0060] The embodiments of the present application disclose a pipeline thickness measurement method against multi-layer thermal insulation medium interference, comprising the following steps:

[0061] S1: using the EMD algorithm improved by the particle swarm algorithm to filter and process the pulse eddy current signals of the pipelines with different thickness under different thermal insulation media, to obtain pulse eddy current filtered signals;

[0062] S2: extracting the medium-term signal intercept characteristic values, power spectral density peak characteristic values and magnetic flux characteristic values of the pulse eddy current filtered signals of the pipelines with different thickness under different thermal insulation media;

[0063] S3: performing linear compensation on the medium-term signal intercept characteristic values, power spectral density peak characteristic values and magnetic flux characteristic values of the extracted pulse eddy current signals;

[0064] S4: establishing a multi-dimensional feature fusion pipeline thickness inversion model by using the linearly compensated characteristic values.

[0065] AsFigure 2 As shown, the pipe to be tested provided in this embodiment of the invention is a 20# ferromagnetic steel pipe, wherein the length of the pipe to be tested is 250 mm, the inner diameter is 50 mm, and the thicknesses are 1.5, 2, 2.5, 3, 4, 5, 6, 7, 8, 10 and 12 mm respectively.

[0066] like Figure 3 As shown, the thermal insulation media provided in the embodiments of the present invention are magnesium aluminum silicate, aluminum silicate, polyurethane and foam glass, wherein each thermal insulation medium has a length of 250 mm, a width of 250 mm and a thickness of 50 mm.

[0067] The following is combined Figure 1 Further explanation is provided for each of the above steps S1-S4.

[0068] S1: The pulse eddy current signals of pipes with different thicknesses under different insulation media are filtered using the EMD algorithm improved by particle swarm optimization, resulting in a pulse eddy current filtered signal, specifically including:

[0069] S101: Fix the basic parameters of the EMD algorithm, set the parameters and optimization variables of the particle swarm optimization algorithm. The optimization variables are the screening stop threshold, the effective IMF start layer, and the effective IMF end layer. Among them, the screening stop threshold is constrained to a range of 0.15-0.45 with a step size of 0.05, the effective IMF start layer ranges to 2-5 with a step size of 1, and the effective IMF end layer ranges to 4-8 with a step size of 1. It must also satisfy that the effective IMF end layer ≥ the effective IMF start layer + 1. The particle swarm optimization parameters are a particle size of 20, a maximum number of iterations of 50, learning factors c1 and c2 are both 2.0, and the inertia weight decreases linearly from 0.8 to 0.3. The fitness function is the signal-to-noise ratio (SNR) of the filtered signal and the original signal, and the objective is to maximize the SNR.

[0070] S102: Initialize the particle swarm and calculate the initial fitness. Randomly generate n sets of parameters within the constraints, integerize the number of IMF layers and ensure that the constraints are met, where n=20; perform EMD decomposition for each set of parameters, accumulate the reconstructed signals from the effective IMF start layer to the effective IMF end layer, calculate the signal-to-noise ratio (SNR) of the reconstructed signal to the original signal, and use it as the initial fitness; record the individual optimum (the highest SNR of each particle and its corresponding parameters) and the global optimum (the highest SNR among all particles and its corresponding parameters).

[0071] S103: Iterative optimization updates fitness and optimal solution. In each iteration, the inertia weight is linearly decreased. The velocity and position are updated based on the particle's current position, individual optimal and global optimal. Constraints are applied to the new position parameters. The signal-to-noise ratio (SNR) corresponding to the new parameters is recalculated as the fitness, and the individual optimal and global optimal positions are updated.

[0072] S104: when the number of iterations reaches m times or the global optimum is not updated for k generations in succession, stop the optimization, output the parameter corresponding to the global optimum, use the optimal parameter to perform EMD decomposition on the original pulse eddy current signals of different heat preservation media, reconstruct the corresponding interval IMF component, and obtain the final pulse eddy current filtered signal; wherein m = 50 and k = 10.

[0073] After obtaining the induced voltage signals of the pipelines with different thicknesses under different heat preservation media, the induced voltage signals are logarithmically processed. The late signal segments of the induced voltage signals of the pipelines with different thicknesses are extremely weak, usually in the order of mV. This leads to the fact that the signals of different thicknesses almost coincide in the Cartesian coordinate system, and the specific curves corresponding to the thicknesses cannot be identified. Then, the EMD algorithm improved by the particle swarm is used to optimize the filtering parameters. In this process, the parameters of the particle swarm optimization algorithm are set, the optimization variables are the screening stop threshold, the effective IMF starting layer and the effective IMF ending layer of the EMD algorithm, the screening stop threshold is constrained in the range of 0.15-0.45 with a step of 0.05, the effective IMF starting layer is in the range of 2-5 with a step of 1, the effective IMF ending layer is in the range of 4-8 with a step of 1, the fitness function is the signal-to-noise ratio (SNR) of the filtered signal and the original signal, and maximizing the SNR is taken as the optimization target of the fitness function. The size of the particle swarm is 20, the maximum number of iterations is 50, the learning factors c1 and c2 are both 2.0, the initial value of the inertia weight is 0.8, and the inertia weight is linearly decreased to 0.3 according to the number of iterations. In this embodiment, the optimal screening stop threshold obtained by the particle swarm algorithm is 0.25, the optimal effective IMF starting layer is 3, and the optimal effective IMF ending layer is 6. Therefore, the EMD algorithm with the screening stop threshold of 0.25, the effective IMF starting layer of 3 and the effective IMF ending layer of 6 is used to filter the eddy current signals, and the result is shown in FIG. 4. Figure 4

[0074] On the basis of step S1, step S2 is further performed to extract the mid-term signal intercept characteristic value, the power spectral density peak characteristic value and the magnetic flux characteristic value of the pulse eddy current filtered signals of the pipelines with different thicknesses under different heat preservation media. The extraction steps of the mid-term signal intercept characteristic value include the following steps.

[0075] S201: in the double logarithmic coordinate system, the 10-1000 mV part of each pulse eddy current filtered signal is intercepted, the length and position of the intercepted signal segment are iterated, the initial position and step of the iteration are 10 mV, and the optimal signal segment for obtaining the mid-term signal intercept characteristic value is found.

[0076] S202: the mid-term signal intercepts of the different signal segments in S21 are compared, and when the characteristic values of the pipelines with different thicknesses under the same medium lift are equal, the corresponding signal segment is the optimal signal segment.

[0077] ​S203: Linear fitting is performed on the optimal signal segment of S22 to obtain the y-axis intercept of the fitting straight line as the mid-term signal intercept characteristic value of the pulse eddy current filtered signal.

[0078] The extraction step of the power spectrum density peak characteristic value includes:

[0079] S204: The pulse eddy current filtered signal of the pipe with the maximum thickness is selected as the reference signal, and the pulse eddy current filtered signals of the pipes with other thicknesses are respectively subtracted from the reference signal to obtain differential signals of the pipes with different thicknesses.

[0080] S205: The differential signal is intercepted at 0.01-1000 mV, and the power spectrum density peak of the differential signal at different lengths and different positions is iteratively calculated, and the initial position and step length of iteration are both 10 mV.

[0081] S206: The power spectrum density peaks of the differential signals of the pipes with different thicknesses are compared, and the power spectrum density peak monotonously decreases with the increase of the pipe thickness. When the power spectrum density peaks of the pipes with different thicknesses have the maximum difference, the optimal differential signal segment is obtained.

[0082] S207: The power spectrum density peak of the optimal differential signal segment is calculated using the Welch method as the power spectrum density peak characteristic value of the pulse eddy current filtered signal.

[0083] The extraction step of the magnetic flux characteristic value includes:

[0084] S208: The 0.01-1000 mV part of each pulse eddy current filtered signal is intercepted, and the signal is divided into signal segments with different lengths, and the initial position and step length are both 10 mV.

[0085] S209: The magnetic flux is calculated according to the signal segments divided in S208, and the signal segment with the maximum difference in the magnetic flux of the pipes with different thicknesses is taken as the optimal signal segment.

[0086] S210: The magnetic flux of the optimal signal segment is calculated as the magnetic flux characteristic value of the pulse eddy current filtered signal.

[0087] After the eddy current signals of the pipes with different thicknesses are filtered, in order to extract the mid-term signal intercept characteristic value, the voltage threshold of the eddy current signal segment is found. In the double logarithmic coordinate system, the mid-term signal intercepts of the pipes with different thicknesses under the same lift-off are equal. In this embodiment, the approximate range of the mid-term signal segment is 10-1000 mV, and the signal segment length and position are iterated with 10 mV as the step length and 10 mV as the initial value to obtain the optimal signal segment.

[0088] In order to extract the power spectrum density peak characteristic value, the eddy current signal segment voltage threshold is found, the power spectrum density peak characteristic value is extracted in the late signal segment, and monotonically decreases with the increase of the pipeline thickness. In this embodiment, the approximate range of the late signal segment is 0.01-1000 mV, the maximum difference of the power spectrum density peak characteristic value of different thicknesses is taken as the target, the signal segment length and position are taken as the step of 10 mV, the initial value is 10 mV, and iteration is performed to obtain the best differential signal segment.

[0089] In order to extract the magnetic flux characteristic value, the eddy current signal segment voltage threshold is found, the magnetic flux characteristic value is extracted in the late signal segment, and monotonically increases with the increase of the pipeline thickness. In this embodiment, the approximate range of the late signal segment is 0.01-1000 mV, the maximum difference of the magnetic flux characteristic value of different thicknesses is taken as the target, the signal segment length and position are taken as the step of 10 mV, the initial value is 10 mV, and iteration is performed to obtain the best differential signal segment.

[0090] Next, step S3 is performed, and the mid-term signal intercept characteristic value, the power spectrum density peak characteristic value and the magnetic flux characteristic value of the extracted pulsed eddy current signal are linearly compensated. Among them, the power spectrum density peak and the magnetic flux characteristic value under different heat preservation media are basically equal to the power spectrum density peak and the magnetic flux characteristic value under air after linear compensation, resisting the interference of different heat preservation media on thickness measurement; the mid-term signal intercept allows the power spectrum density peak and the magnetic flux to select different compensation coefficients under different lift-offs, improves the compensation accuracy, and further resists the interference of different heat preservation media.

[0091] Specifically, the steps of linearly compensating the power spectrum density peak and the magnetic flux characteristic value include:

[0092] S301: taking the power spectrum density peak characteristic value of the pipeline under air at different thicknesses as a standard value, taking the power spectrum density peak characteristic value of the pipeline under different heat preservation media at different thicknesses as an input, using y=ax+b to fit it, and respectively obtaining the compensation coefficients a and b of the power spectrum density peak characteristic value under different heat preservation media;

[0093] The power spectrum density peak characteristic value of the pipeline under different heat preservation media at different thicknesses after compensation is calculated by using the power spectrum density peak and the compensation coefficients a and b of the pipeline under different heat preservation media at different thicknesses.

[0094] Specifically, first, in the air environment, the power spectrum density peak characteristic values corresponding to a series of different thickness pipes are measured, and these standard values are taken as x in the fitting formula. Then, the pipes are placed in different thermal insulation media, and the power spectrum density peak characteristic values corresponding to different thickness pipes are measured, and these values are taken as y in the fitting formula. For each specific thermal insulation medium, an independent fitting is performed, and the standard values x measured in the air and the corresponding values y measured in the thermal insulation medium under different pipe thicknesses are taken as data points, and a linear model y=ax+b is used to fit these data points, and then the compensation coefficients a and b of the power spectrum density peak characteristic values under the thermal insulation medium are obtained.

[0095] S302: The magnetic flux characteristic values of the pipes of different thicknesses in the air are taken as standard values, the magnetic flux characteristic values of the pipes of different thicknesses in different thermal insulation media are taken as inputs, and y=cx+d is used to fit them, and the compensation coefficients c and d of the magnetic flux characteristic values in different thermal insulation media are obtained respectively.

[0096] The magnetic flux characteristic values of the pipes of different thicknesses in different thermal insulation media after compensation are calculated by using the magnetic flux characteristic values of the pipes of different thicknesses in different thermal insulation media and the compensation coefficients c and d.

[0097] Different thermal insulation media interfere with eddy current signals differently, and the extracted characteristic values will also differ in value, which needs to be compensated. The power spectrum density peak value is extracted from the differential signal, and the differential signal amplitude of the maximum thickness pipe is always 0, and the extracted power spectrum density peak value is also 0. Only linear compensation is required for other thickness pipes. Table 1 is the power spectrum density peak characteristic value compensation coefficient under different thermal insulation media, and the magnetic flux is integrated on the eddy current signal rather than the differential signal, and there is no case where the characteristic value of a certain thickness pipe is always 0. Linear compensation is required for the magnetic flux characteristic values of all thickness pipes. Table 2 is the magnetic flux characteristic value compensation coefficient under different thermal insulation media.

[0098] Table 1: Power spectrum density peak characteristic value compensation coefficient under different thermal insulation media

[0099]

[0100] Table 2: Magnetic flux characteristic value compensation coefficient under different thermal insulation media

[0101]

[0102] Further, the mutual compensation relationship between the power spectrum density peak characteristic value and the magnetic flux characteristic value is as follows:

[0103] The power spectrum density peak characteristic value and the magnetic flux characteristic value are normalized. When the pipe thickness is 1.5-5 mm, the weight of the power spectrum density peak characteristic value in the thickness inversion model is increased, and the weight of the magnetic flux characteristic value in the thickness inversion model is reduced. Specifically, the weight of the power spectrum density peak characteristic value in the thickness inversion model can be adjusted to 70%, and the weight of the magnetic flux in the thickness inversion model can be adjusted to 30%.

[0104] When the pipe thickness is 6-12 mm, the weight of the power spectrum density peak characteristic value in the thickness inversion model is increased, and the weight of the magnetic flux characteristic value in the thickness inversion model is reduced. The weight of the power spectrum density peak in the thickness inversion model is adjusted to 30%, and the weight of the magnetic flux in the thickness inversion model is adjusted to 70%. The characteristic value with high thickness sensitivity is always used as the dominant characteristic value for thickness inversion.

[0105] The intermediate signal intercept characteristic value is related to the medium lift-off condition and is independent of the thickness. The intermediate signal intercept is used instead of the lift-off, and then the intermediate signal intercept characteristic value is used as the criterion. Different compensation coefficients are selected for the power spectrum density peak characteristic value and the magnetic flux characteristic value under different intermediate signal intercept characteristic values, so that the influence of the lift-off can be eliminated. The power spectrum density peak has high sensitivity at small thickness, and the magnetic flux has high sensitivity at large thickness. When the pipe thickness is small, the weight of the power spectrum density peak in the thickness inversion model is adjusted to be high, and the weight of the magnetic flux in the thickness inversion model is adjusted to be low. When the pipe thickness is large, the weight of the power spectrum density peak in the thickness inversion model is adjusted to be low, and the weight of the magnetic flux in the thickness inversion model is adjusted to be high. The characteristic value with high thickness sensitivity is always used as the dominant characteristic value for thickness inversion. The magnetic flux, the power spectrum density peak and the intermediate signal intercept multi-characteristic values are fused, the complex multi-layer insulation medium interference is decomposed into a sub-problem that can be independently processed, different correction methods are selected for different interferences to ensure the consistency of the inversion results.

[0106] S4: a pipeline thickness inversion model of multi-dimensional feature fusion is established by using the linearly compensated characteristic values, including:

[0107] S401: a database of linearly compensated power spectrum density peak characteristic values, magnetic flux characteristic values and intermediate signal intercept characteristic values of pipes with different thicknesses under different insulation media is established;

[0108] S402: the database is divided into a training set and a test set, a pipeline thickness inversion model based on a random forest algorithm is constructed, the pipeline thickness inversion model is trained by using the training set, and the pipeline thickness inversion model is verified by using the test set;

[0109] S403: Hyperparameter tuning is performed on the random forest model using the cross-validation grid search method, and the hyperparameters include the number of decision trees, the maximum depth of the decision tree, and the minimum number of samples required for a leaf node;

[0110] S404: The training set is evenly divided into 5 parts, 4 of which are used to train the model, and the remaining 1 part is used to verify the model performance; 5-fold cross-validation (GridSearchCV) is performed on different parameter combinations one by one, and the parameter combination with the minimum negative mean square error is selected as the optimal solution;

[0111] S405: The model parameters are updated according to the optimal solution to obtain the final pipe thickness inversion model; based on the final pipe thickness inversion model, the thickness of the pipe to be measured is inverted, and the input of the pipe thickness inversion model is the power spectral density peak eigenvalue, the magnetic flux eigenvalue and the intermediate signal intercept eigenvalue of the pulse eddy current signal, and the output is the pipe thickness.

[0112] In other embodiments, the present application also provides a pipe thickness measurement system resistant to interference of multi-layer thermal insulation medium, which adopts the above method, comprising:

[0113] A signal acquisition module is configured to acquire pulse eddy current signals of pipes with different thicknesses under different thermal insulation media;

[0114] A signal filtering module is configured to filter the pulse eddy current signals of pipes with different thicknesses under different thermal insulation media using an EMD algorithm improved by a particle swarm algorithm to obtain pulse eddy current filtered signals;

[0115] A signal eigenvalue extraction module is configured to extract the intermediate signal intercept eigenvalue, the power spectral density peak eigenvalue and the magnetic flux eigenvalue of the pulse eddy current filtered signals of pipes with different thicknesses under different thermal insulation media;

[0116] A compensation module is configured to perform linear compensation on the intermediate signal intercept eigenvalue, the power spectral density peak eigenvalue and the magnetic flux eigenvalue of the extracted pulse eddy current signals;

[0117] A pipe thickness inversion module is configured to establish a multi-dimensional feature fusion pipe thickness inversion model using the linearly compensated eigenvalues.

[0118] At this point, the entire process of the pipe thickness measurement method and system resistant to interference of multi-layer thermal insulation medium provided by the present application is completed. It can be understood that the various numbers involved in the embodiments of the present application are only for differentiation for description, and are not used to limit the scope of the embodiments of the present application.

[0119] The various embodiments described in this specification are implemented in a progressive manner, each embodiment focusing on the differences from other embodiments, and the same or similar parts between embodiments can be mutually referred to. For the apparatus disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0120] The above description of disclosed embodiments enables one of ordinary skill in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those of ordinary skill in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of pipe thickness measurement that is resistant to interference from a multi-layered thermal insulation medium, characterized in that, The method comprises the following steps: S1: using the EMD algorithm improved by the particle swarm algorithm to filter the pulse eddy current signals of the pipelines with different thicknesses under different heat preservation media, to obtain pulse eddy current filtered signals; S2: extracting the medium-term signal intercept characteristic values, power spectral density peak characteristic values and magnetic flux characteristic values of the pulse eddy current filtered signals of the pipelines with different thicknesses under different heat preservation media; S3: performing linear compensation on the medium-term signal intercept characteristic values, power spectral density peak characteristic values and magnetic flux characteristic values of the extracted pulse eddy current signals; The step of performing linear compensation on the power spectral density peak and magnetic flux characteristic values comprises: S301: taking the power spectral density peak characteristic values of the pipelines with different thicknesses under air as standard values, taking the power spectral density peak characteristic values of the pipelines with different thicknesses under different heat preservation media as input, and using y=ax+b to fit them, to obtain the compensation coefficients a and b of the power spectral density peak characteristic values under different heat preservation media respectively; Using the power spectral density peaks and the compensation coefficients a and b of the pipelines with different thicknesses under different heat preservation media, the power spectral density peak characteristic values of the pipelines with different thicknesses under different heat preservation media after compensation are calculated; S302: taking the magnetic flux characteristic values of the pipelines with different thicknesses under air as standard values, taking the magnetic flux characteristic values of the pipelines with different thicknesses under different heat preservation media as input, and using y=cx+d to fit them, to obtain the compensation coefficients c and d of the magnetic flux characteristic values under different heat preservation media respectively; Using the magnetic flux characteristic values and the compensation coefficients c and d of the pipelines with different thicknesses under different heat preservation media, the magnetic flux characteristic values of the pipelines with different thicknesses under different heat preservation media after compensation are calculated; S4: using the linearly compensated characteristic values to establish a pipeline thickness inversion model based on multi-dimensional feature fusion; S4 comprises: S401: establishing a database of the linearly compensated power spectral density peak characteristic values, magnetic flux characteristic values and medium-term signal intercept characteristic values of the pipelines with different thicknesses under different heat preservation media; S402: dividing the database into a training set and a test set, constructing a pipeline thickness inversion model based on a random forest algorithm, and training and testing the model; S403: using a cross-validation grid search method to optimize the hyperparameters of the model; S404: performing 5-fold cross-validation on different parameter combinations one by one, and selecting the parameter combination with the minimum negative mean square error as the optimal solution; S405: updating the model parameters according to the optimal solution, to obtain a final pipeline thickness inversion model; the input of the pipeline thickness inversion model is the power spectral density peak characteristic values, magnetic flux characteristic values and medium-term signal intercept characteristic values of the pulse eddy current signals, and the output is the pipeline thickness.

2. The method of claim 1, wherein, S1 comprises: S101: fixing the basic parameters of the EMD algorithm, setting the parameters and optimization variables of the particle swarm algorithm, and the optimization variables are the screening stop threshold, the effective IMF starting layer and the effective IMF ending layer; S102: initialize the particle swarm and calculate the initial fitness, randomly generate n sets of parameters within the constraint range, integerize the IMF layer number and ensure that the constraints are met; perform EMD decomposition for each set of parameters, accumulate the effective IMF starting layer to the effective IMF ending layer component to reconstruct the signal, calculate the signal-to-noise ratio SNR of the reconstructed signal and the original signal as the initial fitness; Record the individual optimal and global optimal; S103: update the fitness and optimal solution in each iteration, linearly decrease the inertia weight, update the velocity and position according to the current position of the particle, the individual optimal and the global optimal, and perform constraint processing on the new position parameters; recalculate the SNR of the new parameters as the fitness, and update the individual optimal and global optimal position; S104: when the number of iterations reaches m times or the global optimal is not updated for k consecutive generations, stop optimization, output the parameters corresponding to the global optimal, and use the optimal parameters to perform EMD decomposition on the original pulse eddy current signals of different heat preservation media, reconstruct the corresponding interval IMF components, and obtain the final pulse eddy current filtered signal.

3. The method of claim 1, wherein, In S2, the extraction step of the intermediate signal intercept characteristic value includes: S201: In the double logarithmic coordinate system, intercept the 10-1000 mV part of each pulse eddy current filtered signal, and iterate the length and position of the intercepted signal segment, the initial position and step of iteration are 10 mV, and the best signal segment for seeking the intermediate signal intercept characteristic value is found; S202: Compare the intermediate signal intercepts of different signal segments in S21, when the characteristic values of different thickness pipes under the same medium lift are equal, the corresponding signal segment is the best signal segment; S203: Linearly fit the best signal segment of S22 to obtain the y-intercept of the fitted straight line as the intermediate signal intercept characteristic value of the pulse eddy current filtered signal.

4. The method of claim 3, wherein, In S2, the extraction step of the power spectral density peak value characteristic value includes: S204: Select the pulse eddy current filtered signal of the maximum thickness pipe as the reference signal, and subtract the pulse eddy current filtered signals of other thickness pipes from the reference signal to obtain the differential signals of different thickness pipes; S205: Interpolate the 0.01-1000 mV part of the differential signal, and iteratively calculate the power spectral density peak value of the differential signal at different lengths and different positions, the initial position and step of iteration are both 10 mV; S206: Compare the power spectral density peak values of the differential signals of different thickness pipes, the power spectral density peak value monotonically decreases with the increase of the pipe thickness, when the power spectral density peak values of different thickness pipes have the maximum difference, it is the best differential signal segment; S207: Calculate the power spectral density peak value of the best differential signal segment using the Welch method as the power spectral density peak value characteristic value of the pulse eddy current filtered signal.

5. The method of claim 4, wherein, In S2, the extraction step of the magnetic flux characteristic value includes: S208: Interpolate the 0.01-1000 mV part of each pulse eddy current filtered signal, and divide it into signal segments of different lengths, the initial position and step are both 10 mV; S209: Calculate the magnetic flux according to the signal segments divided in S208, and take the signal segment with the largest magnetic flux difference of different thickness pipes as the best signal segment; S210: Calculate the magnetic flux of the optimal signal segment as the magnetic flux characteristic value of the pulse eddy current filtered signal.

6. The method of claim 1, wherein, In S3, the mutual compensation relationship of the power spectrum density peak characteristic value and the magnetic flux characteristic value is as follows: The power spectrum density peak characteristic value and the magnetic flux characteristic value are normalized, when the pipe thickness is 1.5-5 mm, the weight of the power spectrum density peak characteristic value in the thickness inversion model is increased, and the weight of the magnetic flux characteristic value in the thickness inversion model is reduced; when the pipe thickness is 6-12 mm, the weight of the power spectrum density peak characteristic value in the thickness inversion model is increased, and the weight of the magnetic flux characteristic value in the thickness inversion model is reduced; the characteristic value with higher thickness sensitivity is always used as the dominant characteristic value for thickness inversion.

7. The method of claim 1, wherein, In S3, the medium signal intercept characteristic value is used instead of the medium lift condition, and the medium signal intercept characteristic value is used as the criterion to select different compensation coefficients for the power spectrum density peak characteristic value and the magnetic flux characteristic value under different medium signal intercept characteristic values.

8. A pipe thickness measurement system resistant to interference from a multi-layered thermal insulation medium, characterized by, It uses the method of any one of claims 1-7, comprising: a signal acquisition module for acquiring pulse eddy current signals of pipes with different thicknesses under different heat preservation media; a signal filtering module for filtering the pulse eddy current signals of pipes with different thicknesses under different heat preservation media using an EMD algorithm improved by a particle swarm algorithm to obtain pulse eddy current filtered signals; a signal characteristic value extraction module for extracting the medium signal intercept characteristic value, the power spectrum density peak characteristic value and the magnetic flux characteristic value of the pulse eddy current filtered signals of pipes with different thicknesses under different heat preservation media; a compensation module for linearly compensating the extracted medium signal intercept characteristic value, power spectrum density peak characteristic value and magnetic flux characteristic value of the pulse eddy current signals; a pipe thickness inversion module for establishing a multi-dimensional feature fusion pipe thickness inversion model using the linearly compensated characteristic values.

Images