A pulse vortex tube pipeline wall thickness detection method based on RSBO optimized LSTM network

Abstract

The application provides a pulse eddy current pipe wall thickness detection method based on an RSBO optimized LSTM network, pulse eddy current signals are used to estimate the thickness of the pipe, a data sample length evaluation criterion based on a least square fitting algorithm is used to determine the interval of the retained measurement signals, the processed signals are input into an LSTM model to detect the thickness of the pipe, and a combination optimization based on random search and Bayesian optimization is used to set the hyperparameters of the LSTM model.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of PEC detection, pipeline thickness measurement and the like, and particularly relates to a pulse eddy current pipeline wall thickness detection method based on RSBO optimized LSTM network. BACKGROUND

[0002] Corrosion of pressure vessels, particularly of steel pressure vessels, poses a serious threat to the safety and reliability of industrial operations, and as a common pipeline structure in pressure vessels, effective detection methods are urgently needed. PEC detection technology has the characteristics of non-contact, deep penetration depth and wide frequency spectrum, and has been widely used in thickness measurement of metal samples, especially in detection under high thickness and high lift-off conditions. For the detection of pipeline structure, the coil or magnetic field sensor can receive the eddy current transient diffusion signal affected by the performance of the plate material, wherein the change of the sample pipe thickness will be reflected through the signal.

[0003] The prior art has achieved estimating thickness from received signals by investigating the relationship between signal characteristics and thickness and solving the inverse problem by optimization methods. Y. Shin et al. proposed estimating sample thickness and conductivity parameters from the peak point and zero-crossing time characteristics of the received signal (see Shin, Y.-K.; Choi, D.-M.; Kim, Y.-J.; Lee, S.-S., Signal characteristics of differential-pulsed eddy current sensors in the evaluation of plate thickness. NDT&E International 2009, 42, (3), 215-221.), however this method is affected by the lift-off height between the sensor probe and the sample. M. Fan et al. proposed that the phase spectrum of the converted received signal into the frequency domain is not affected by the lift-off effect (see Fan, M.; Cao, B.; Sunny, A. I.; Li, W.; Tian, G.; Ye, B., Pulsed eddy current thickness measurement using phase features immune to liftoff effect. Ndt&E International 2017, 86, 123-131.). This also applies to the phase spectrum of multi-frequency eddy current testing if the change in lift-off is not significant (see Yin, W.; Peyton, A. J., Thickness measurement of metallic plates with an electromagnetic sensor using phase signature analysis. IEEE Transactions on Instrumentation and Measurement 2008, 57, (8), 1803-1807.).A frequency component mixing method combining the features of each frequency was proposed to obtain a linear relationship between the features and the sample thickness (see Ge, J.; Yusa, N.; Fan, M., Frequency component mixing of pulsed or multi-frequency eddy current testing for nonferromagnetic plate thickness measurement using a multi-gene genetic programming algorithm. NDT&E International 2021, 120, 102423.).

[0004] In recent years, data-driven deep learning (DL) methods have been frequently applied to eddy current testing research. DL methods learn the nonlinear mapping between measurements and physical properties of the sample, providing rich data samples that can be obtained from intensive measurements and numerical simulations (see Zhu, P.; Cheng, Y.; Banerjee, P.; Tamburrino, A.; Deng, Y., A novel machine learning model for eddy current testing with uncertainty. ndt&e International 2019, 101, 104-112. and Kucheryavskiy, S.; Egorov, A.; Polyakov, V., Coupling analytical models and machine learning methods for fast and reliable resolution of effects in multifrequency eddy-current sensors. Sensors 2021, 21, (2), 618.). H. Seo et al. implemented a deep neural network for pipe thickness estimation, with a dataset including 720 samples (see Seo, H. G.; Jun, J. H.; Park, D. G.; Shin, J. W., Pipe Thickness Estimation by Deep Learning of Pulsed Eddy Current Time-Series Data. Journal of the Korean Society for Nondestructive Testing 2021, 41, (3), 164-171.).M. Tian et al. embedded a convolutional neural network into a custom PEC system for estimating the thickness of aluminum and low carbon steel plates (see Meng, T, Xiong, L, Zheng, X, et al. Real-time automatic thickness recognition using pulse eddy current with deep learning. 2023 IEEE International Instrumentation and Measurement Technology Conference (I2MTC). IEEE, 2023: 1-6.). The dataset consists of over 10k samples with various PEC measurement configurations and is available online. Due to the abundance of PEC measurement samples, it helps to achieve accurate and efficient sample thickness estimation.

[0005] Optimization algorithms can be used to solve the least square problem of the received signal between measurement and calculation, and the sample thickness is obtained by iteration. X. Chen et al. studied the optimization method for PEC inverse problem and simultaneous estimation of thickness and electromagnetic properties (see Chen, X.; Li, J.; Wang, Z., Inversion method in pulsed eddy current testing for wall thickness of ferromagnetic pipes. IEEE Transactions on Instrumentation and Measurement 2020, 69, (12), 9766-9773.). Without prior knowledge, it is still difficult to provide a reasonable initial point for optimization, which affects the estimation results to some extent. Ulapane, N. et al. proposed that the decay rate of the received signal is not affected by the lift-off, but only depends on the diffusion of eddy current in the sample (see Ulapane, N.; Alempijevic, A.; Miro, J. V.; Vidal-Calleja, T., Non-destructive evaluation of ferromagnetic material thickness using pulsed eddy current sensor detector coil voltage decay rate. NDT&E International 2018, 100, 108-114.). However, for small size samples, the position of the probe will also affect the signal of eddy current diffusion, which brings difficulties to the estimation of the properties of the tube, for example, when the probe is at the edge of the sample, the edge effect will be produced. It is very complex to accurately simulate various conditions of PEC test with an analytical model due to the complex boundary conditions. SUMMARY

[0006] Therefore, aiming at the defects and deficiencies existing in the prior art, the present application provides a pulse eddy current pipe wall thickness detection method based on RSBO optimized LSTM network, which is a deep learning prediction method, a Long and Short Term Memory (LSTM) network based on random searching and Bayesian optimization (RSBO) combined optimization, and uses a pulsed eddy current (PEC) signal to estimate the pipe thickness. In view of the contradiction between signal data redundancy and feature retention, a data sample length evaluation criterion based on a least square fitting algorithm is introduced to improve the operation efficiency and accuracy of the model. In addition, for the LSTM model, the RSBO optimization method is used for hyperparameter setting, and an estimation scheme is designed to automatically adjust the hyperparameters of the LSTM network, avoiding the manual parameter setting process, and having good estimation performance.

[0007] The technical solutions adopted by the present application to solve its technical problems are:

[0008] A pulse eddy current pipe wall thickness detection method based on RSBO optimized LSTM network: using a pulsed eddy current signal to estimate the pipe thickness, using a data sample length evaluation criterion based on a least square fitting algorithm to determine the interval of the retained measurement signal, inputting the processed signal into an LSTM model to detect the pipe thickness, and using a combination of random search and Bayesian optimization to optimize the hyperparameters of the LSTM model.

[0009] Further, the data sample length evaluation criterion based on the least square fitting algorithm is specifically: after fitting the measurement curve using the least square method, the slope curve is calculated, the standard deviation of the fitting matrix is used to evaluate the overall data fluctuation of the intercept section, the average value of each cutting segment is calculated to evaluate the slope change of the overall data curve, and the standard deviation of the fitting matrix is calculated, and the average value of the slope is plotted on the same graph, so that the intersection of the two curves is used as the basis for determining the sample interval.

[0010] Further, the LSTM model is composed of an input layer, multiple stacked hidden layers, a fully connected layer and a regression layer, the output of each time step of each layer is used as the input of the next time step, the feature extraction and feature transformation of the pulsed eddy current measurement signal are completed through the input layer and the hidden layer, and the final thickness prediction value is output through the nonlinear regression of the fully connected layer.

[0011] Further, the training of the LSTM model adopts a learning rate decay mechanism.

[0012] Further, the hyperparameters include the number of hidden layers, the number of hidden neurons Num, the initial learning rate of the network learnRate0, the number of iterations Inter, and the training batch BatchSize.

[0013] Further, the hyperparameter tuning of the LSTM model by using the combined optimization based on random search and Bayesian optimization specifically includes:

[0014] Randomly sampling the combination of hyperparameters by random search to find the optimal hyperparameters;

[0015] Using Bayesian optimization to further optimize the hyperparameters after random search by constructing a probability model of the objective function;

[0016] Recording the hyperparameter set with the minimum target loss after the iteration is completed, and saving the model hyperparameters when the thickness error in the validation set reaches the minimum value.

[0017] Further, the further optimization calculation of the hyperparameters after random search by using Bayesian optimization by constructing a probability model of the objective function specifically includes the following steps:

[0018] Step S1: Select a set of initial parameters, train and evaluate the LSTM model, and take the optimal result as the initial observation value of Bayesian optimization;

[0019] Step S2: Train the Gaussian process model using the initial observation value, and use the trained Gaussian process model to predict the objective function value and uncertainty of unobserved points;

[0020] Step S3: Determine the next sampling point using the acquisition function;

[0021] The acquisition function is a function that maps from the input space , the observation space and the hyperparameter space to the real number space ; the posterior distribution obtained from the observation data set is constructed, and the next sampling point is selected by maximizing the acquisition function :

[0022]

[0023] Where s represents the sampling point.

[0024] Step S4: Select a new combination of hyperparameters for evaluation according to the result of the acquisition function, and update the Gaussian process model;

[0025] Step S5: Repeat steps S1-S4 until the iteration stopping condition is met, and output the optimal hyperparameter result.

[0026] And an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method for detecting the wall thickness of the pulse vortex pipe based on the RSBO optimized LSTM network as described above when executing the program.

[0027] A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the method for detecting the wall thickness of the pulse vortex pipe based on the RSBO optimized LSTM network as described above.

[0028] In view of the defects and deficiencies of the prior art, the main design points and advantages of the present application and its preferred schemes are at least embodied in:

[0029] 1. In view of the contradiction between PEC signal data redundancy and feature retention, a data sample length evaluation criterion based on a least squares fitting algorithm is introduced, which improves the operation efficiency and accuracy of the model.

[0030] 2. In view of the large amount of existing sample data and multiple optimization targets, such as the time-consuming of using traversal optimization algorithm, a parameter optimization method combining RS and BO is proposed.

[0031] 3. For the LSTM model, the RSBO optimization method is used for hyperparameter setting, which avoids the manual parameter setting process, improves the accuracy of pipe thickness detection, and has good estimation performance. BRIEF DESCRIPTION OF DRAWINGS

[0032] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0033] Figure 1 The sensor measurement position schematic diagram for the embodiment of the present application;

[0034] Figure 2 The measurement result graph for the embodiment of the present application with the sensor at the center position, the lift is 6mm, and the thickness is 20mm;

[0035] Figure 3 The slope standard deviation and mean change graph for different sample intervals of the embodiment of the present application;

[0036] Figure 4 The LSTM network structure graph for thickness estimation of the embodiment of the present application;

[0037] Figure 5 The basic LSTM network structure graph;

[0038] Figure 6 The hyperparameter tuning flowchart of the embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the features and advantages of the present application more apparent, specific examples are described in detail below, as follows:

[0040] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0041] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments in accordance with the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0042] The embodiment of the present application proposes a pulse vortex pipe wall thickness detection based on RSBO optimized LSTM, which realizes a fast and accurate detection method for sample pipe thickness, and the specific implementation steps are as follows:

[0043] Step one: PEC feature data acquisition and preprocessing

[0044] As shown in Figure 1 Experiments are performed on four positions {center, edge, corner, random} of the sample, six lift-off distances l∈{0,3,…,15} mm, and nine thicknesses d∈{20,25,…,60} mm. Each type of experiment is performed 50 times of PEC detection, a total of 10800 data sets are collected, each data set contains 270 data, and the diversity of detection thickness, position and lift-off height brings some differences in pulse response. However, some pulse response curves are relatively similar. In order to improve the operation efficiency and reduce data redundancy, data enhancement is needed, and the main features in the signal curve are intercepted, and as much signal features as possible are contained in the shortest possible data length.

[0045] The main features of PEC signal curve include jump, attenuation and flat, among which the jump part is short in time and contains the information of the measured sample pipe, so the jump of the whole first half part should be included in the intercepted data segment. The latter half of the pulse vortex signal curve is flat to the process of exponential decay, and this transition is caused by vortex diffusion, so the data in this part should also be included.

[0046] In this embodiment, 54 sets of experimental data of the sensor placed in the center position, lifted by a distance l e {0, 3, …, 15} mm, and a thickness d e {20, 25, …, 60} mm are taken as representative data. In a sampling period, the numerical values between the curves are different in the degree of decay rate, the slope of the curve and the diffusion rate of the eddy current on the surface are corresponding, and the degree of slope fluctuation is considered as the judgment of the difference degree between the curves at different time points. Therefore, after fitting the measurement curve by using the least square method, the slope curve is selected for calculation, the fitted measurement curve is used to evaluate the overall change of the slope of the data, and the standard deviation of the fitting matrix is used to evaluate the overall data fluctuation of the intercept section (for specific implementation, refer to Feng Ding, "Least squares parameter estimation and multi-innovation least squares methods for linear fitting problems from noisy data", Journal of Computational and Applied Mathematics, vol. 426, pp. 115107, 2023.).

[0047] From Figure 2 It can be seen that the curve slowly falls back after 0.4-1s. Therefore, this embodiment selects a time span of 0.37-0.9s (corresponding to data points 100-250), and performs least square fitting on it, with a fitting step of 10, 15 segments, and then obtains a fitting matrix [54x15]. The average of each intercept segment is calculated to evaluate the slope change of the overall data curve, and the standard deviation of the fitting matrix is calculated, and the two curves are plotted in the same graph, as shown in Figure 3 .

[0048] It can be seen that, as time goes on, the standard deviation of the slope gradually increases, the data fluctuation is slow, and the maximum value tends to be stable; and the average value of the slope becomes smaller and tends to be the minimum value. The intersection of the two curves is taken as the sample interval 2, at this time the change between the curves and the value is more balanced, and the corresponding length of the measurement signal is reserved, that is, the time period of 0.407-0.444s is intercepted. In order to make the intercepted data more contain the characteristics of the back-off response process, this embodiment takes 0.444s (120 data points) as the maximum value of the sample interval.

[0049] Step two: establish an LSTM model based on RSBO

[0050] The deep learning LSTM model proposed by the application for pipeline thickness detection is composed of an input layer, a plurality of stacked hidden layers, a full connection layer and a regression layer, and the network structure is as shown in Figure 4The output of each time step of each layer is taken as the input of the next time step of the next layer, so the feature extraction and transformation of the eddy current measurement signal are completed through the input layer and the hidden layer, and the final thickness prediction value is output through the nonlinear regression of the fully connected layer.

[0051] The multi-hidden layer LSTM model is composed of the most basic LSTM unit, and its structure mainly consists of a forget gate, an input gate and an output gate, as shown in the formula (1). Figure 5 The forget gate decides what information should be forgotten by the neuron, and the input x t and the previous time h t-1 are weighted and operated, and the formula is as follows:

[0052] (1)

[0053] The neuron updates its state vector through the input gate layer and the tanh layer to save the relevant information of the input sequence. The input gate layer uses the Sigmoid function to decide the value to be updated, and the tanh layer obtains the candidate value g t to be added to the state of the neuron through nonlinear transformation. The two steps are combined to produce an updated state vector, so as to realize the state update of information, and the formula is as follows:

[0054]

[0055] Through the transformation of the input layer and the neuron state by the Sigmoid and tanh functions, the neuron can filter out the unnecessary information and retain the information needed for output, and the output h t at this time will be passed to the next neuron for information update. The output of the last layer of the network is subjected to weight and bias operation and then output as the final result. The formula is as follows:

[0056]

[0057] where W and b are the weight and bias of the corresponding layer respectively, and the Sigmoid operation.

[0058] Selecting appropriate hyperparameters can significantly improve the performance of the LSTM model. In the present application, the hyperparameters required for the LSTM neural network structure include the number of network hidden layers, the number of hidden neurons Num, the initial learning rate of the network learnRate0, the number of iterations Inter and the training batch BatchSize. In order to make the deep learning network reach the convergence state in the later training stage, the learning rate decay mechanism should be used:

[0059] (8)

[0060] The specific process of hyperparameter tuning of the embodiment of the application is as shown in Figure 6 Compared with grid search, it is not to traverse all possible hyperparameter combinations, but to randomly extract a certain number of combinations in the hyperparameter space for evaluation. Compared with swarm intelligence optimization algorithm, it has smaller calculation intensity and can quickly find better hyperparameter combination in limited time.

[0061] Therefore, the embodiment first adopts RS to optimize and select the LSTM hyperparameter combination, divides the data set into a training set and a validation set at a ratio of 4:1, adopts the Adam gradient descent algorithm for training, the network input is 120 and the output is 1, and the mean squared error (MSE) is used as the target loss function. The parameter configuration is shown in Table 2. The hyperparameter set with the minimum target loss of the parameter variable setting after recording the completion of iteration is recorded, and the model hyperparameter is saved when the thickness error in the validation set reaches the minimum value.

[0062] Then, BayesOPT is used to further optimize the hyperparameters after random search to improve the thickness estimation accuracy. For BO, the global optimal solution can be obtained under a small amount of experiments by constructing a probability model of the target function, which is suitable for neural network parameter optimization with large amount of calculation.

[0063] The specific steps are as follows:

[0064] 1. Select a set of initial parameters, train and evaluate the model, and take the optimal result as the initial observation value of Bayesian optimization;

[0065] 2. Train the Gaussian process model using these observation data, and use the model to predict the target function value and uncertainty of the unobserved point;

[0066] 3. Use a sampling function to determine the next sampling point, which balances the exploration and utilization trade-off.

[0067] The sampling function maps from the input space , the observation space and the hyperparameter space to the function in the real number space. The function is constructed by the posterior distribution obtained from the observation data set , and the next evaluation point is selected by maximizing it:

[0068] (9)

[0069] 4. Select a new hyperparameter combination for evaluation according to the result of the sampling function, and update the Gaussian process model.

[0070] 5. Repeat steps (1)-(4) until the iteration stopping condition is met, and output the optimal hyperparameter result.

[0071] In one test instance, different hyperparameter sets were tested under different hidden layers and neurons. Based on RS optimization, when the hyperparameters were set as Inter = 2000, nInter = 8, learnRate0 = 0.01, a = 0.6, Layer = 3, BatchSize = 1000, and Num = 150 for each hidden layer, the network had the best estimation effect. Under this setting, the model had a better effect than other methods, with the MAPE improved to 5.53% and the MAE value improved to 1.98 mm of thickness estimation.

[0072] Table 1 Estimation results using different neural networks

[0073]

[0074] Table 2 Thickness estimation results of the training set

[0075] Thickness (mm) 20 25 30 35 40 45 50 55 60 Number 944 884 943 960 944 979 943 950 953 MAPE 0.0605 0.0970 0.0466 0.0529 0.0427 0.0416 0.0385 0.0370 0.0333

[0076] Table 3 Thickness estimation results of the test set

[0077] Thickness (mm) 20 25 30 35 40 45 50 55 60 Number 260 267 264 244 263 227 265 256 254 MAPE 0.0638 0.1071 0.0510 0.0591 0.0415 0.0418 0.0409 0.0373 0.0359

[0078] Those skilled in the art will understand that embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage media, etc.) having computer-usable program code embodied therein.

[0079] The present application is described with reference to the flowcharts and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in a flow or multiple flows and / or blocks Figure 1 The functions specified in a flow or multiple flows and / or blocks

[0080] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flow or block Figure 1 one or more flow or block

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions executed on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flow or block Figure 1 one or more flow or block

[0082] It should be noted that unless otherwise defined, technical terms or scientific terms used in the present application shall have the common meaning to those skilled in the art. The terms "first", "second" and similar terms used in the present application do not denote any order, quantity or importance, but are used to distinguish different components. The terms "include" or "contain" and similar terms mean that the elements or objects before the terms encompass the elements or objects listed after the terms and their equivalents, and do not exclude other elements or objects. The terms "connect" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "upper", "lower", "left", "right" and the like are used only to indicate relative positional relationships, and when the absolute positions of the described objects are changed, the relative positional relationships can also be changed accordingly.

[0083] The above description is only the preferred embodiment of the present application, and is not intended to limit the other forms of the present application. Any skilled person in the art can modify or change the above-mentioned disclosed technical content into equivalent embodiments with equivalent changes. However, any simple modification, equivalent change and modification of the above-mentioned embodiments made without departing from the technical solution of the present application, and in accordance with the technical essence of the present application, shall fall within the protection scope of the present application.

[0084] The present application is not limited to the above-mentioned preferred embodiments, and anyone can derive other various forms of a pulse vortex pipe wall thickness detection method based on an RSBO optimized LSTM network under the inspiration of the present application. Any equivalent changes and modifications made in accordance with the scope of the present application shall fall within the scope of the present application.

Claims

1. A pulse vortex tube pipeline wall thickness detection method based on RSBO optimized LSTM network, characterized by: The pipe thickness is estimated by using a pulse eddy current signal, a data sample length evaluation criterion based on a least square fitting algorithm is used to determine the interval of the reserved measurement signal, the processed signal is input into an LSTM model to detect the pipe thickness, and a combination optimization based on random search and Bayesian optimization is used to set the hyperparameters of the LSTM model; The data sample length evaluation criterion based on the least square fitting algorithm specifically comprises: after fitting the measurement curve by using the least square method, the slope curve is calculated, the overall data fluctuation of the intercept section is evaluated by using the standard deviation of the fitting matrix, the slope change of the overall data curve is evaluated by calculating the mean value of each cutting segment, and the standard deviation of the fitting matrix is calculated, which is plotted in the same graph as the average slope to take the intersection of the two curves as the determination basis of the sample interval; The combination optimization based on random search and Bayesian optimization for setting the hyperparameters of the LSTM model specifically comprises: Randomly sampling the combination of hyperparameters by random search to find the optimal hyperparameters; Further optimization calculation is performed on the hyperparameters after random search by constructing a probability model of the objective function through Bayesian optimization; The parameter variable setting target loss minimum hyperparameter set after the iteration is completed is recorded, and when the thickness error in the validation set reaches a minimum value, the model hyperparameters are saved; The specific steps of further optimization calculation on the hyperparameters after random search by constructing a probability model of the objective function through Bayesian optimization are as follows: Step S1: select a group of initial parameters, train and evaluate the LSTM model, and take the optimal result as the initial observation value of Bayesian optimization; Step S2: train the Gaussian process model using the initial observation value, and predict the objective function value and uncertainty of the unobserved point using the trained Gaussian process model; Step S3: determine the next sampling point using the acquisition function; The acquisition function is a function mapping from the input space , the observation space and the hyperparameter space to the real number space ; the posterior distribution is constructed from the observation dataset and the next acquisition point is chosen by maximizing the acquisition function : Where s represents the acquisition point; Step S4: select a new combination of hyperparameters for evaluation according to the result of the acquisition function, and update the Gaussian process model; Step S5: repeat steps S1-S4 until the iteration stop condition is met, and output the optimal hyperparameter result.

2. The method of claim 1, wherein the method is based on an RSBO-optimized LSTM network for pulsed-vortex tube pipe wall thickness detection. The LSTM model is composed of an input layer, multiple stacked hidden layers, a fully connected layer and a regression layer, the output of each time step of each layer is taken as the input of the next time step of the next layer, the feature extraction and transformation of the pulse eddy current measurement signal are completed through the input layer and the hidden layer, and the final thickness prediction value is output through the nonlinear regression of the fully connected layer.

3. The method of claim 2, wherein the method is based on an RSBO-optimized LSTM network for pulsed-vortex tube pipe wall thickness detection. The training of the LSTM model adopts a learning rate decay mechanism.

4. The method of claim 1, wherein the method is based on an RSBO-optimized LSTM network for pulsed-vortex tube pipe wall thickness detection. The hyperparameters include the number of hidden layers, the number of hidden neurons Num, the initial learning rate of the network learnRate0, the number of iterations Inter and the training batch size BatchSize.

5. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the pulse eddy current pipe wall thickness detection method based on RSBO optimized LSTM network according to any one of claims 1-4.

6. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of a method for detecting the thickness of a pipeline wall of a pulse vortex tube based on an RSBO optimized LSTM network according to any one of claims 1-4.

Images