A method for optimizing threshold value of recurrent graph and extracting feature parameters based on particle swarm algorithm

Abstract

The application discloses a kind of based on particle swarm optimization recursive graph threshold and feature parameter extraction method, it belongs to the field of nondestructive testing.The method is through a set of ultrasonic signal acquisition system, acquires ultrasonic backscattering signal and carries out phase space reconstruction, calculates the Euclidean distance between any two time vectors in phase space.Set particle swarm parameter and convergence condition, using particle swarm optimization algorithm to optimize recursive graph threshold.Recursive graph is drawn to ultrasonic backscattering signal using optimized threshold.Recursive quantitative analysis method is used to quantitatively calculate recursive graph, extract feature parameters, finally construct the mapping relationship between feature parameters and material degradation damage grade, realize the evaluation of different material degradation grade.The application overcomes the deficiency of only setting recursive threshold artificially, can accurately and comprehensively analyze ultrasonic backscattering signal, realize the accurate extraction of recursive feature parameters, and has important significance for the development of acoustic evaluation method of material degradation damage.

Description

A method for optimizing threshold and feature parameter extraction of recursive graphs based on particle swarm optimization algorithm Technical Field

[0001] This invention discloses a method for optimizing the threshold and feature parameter extraction of recursive graphs based on the particle swarm optimization algorithm. This method can optimize the selection of recursive graph thresholds and belongs to the field of non-destructive evaluation. Therefore, the extracted and optimized recursive feature parameters can be used to accurately characterize the material degradation damage level. Background Technology

[0002] High-temperature, pressure-bearing hot-walled pipelines are prone to material degradation damage during service. Common metallic materials include carbon steel and low-alloy steel. Material degradation causes changes in the microstructure of the metallic material, including material texture, grain size, and grain shape. Ultrasonic backscattered emission (UARE) signals are sensitive to changes in the microstructure of materials, and the material degradation level can be effectively evaluated by extracting characteristic parameters from UARE signals. However, because UARE signals are nonlinear and non-stationary, extracting their time-domain and frequency-domain characteristic parameters is difficult. Recursion graphs, on the other hand, can effectively identify the characteristics of UARE signals and evaluate the material degradation level of metallic materials to ensure safe equipment operation. Therefore, plotting UARE signals of different material degradation levels using recursion graphs and extracting recursive characteristic parameters using recursive analysis methods is of great significance for studying the extraction of characteristic parameters from recursion graphs and developing acoustic evaluation methods for material degradation levels.

[0003] Currently, most methods for constructing recursive graphs of ultrasound backscattered signals rely on manually set thresholds, using recursive analysis to extract feature parameters. However, in practice, due to inaccurate human judgment and a lack of effective threshold settings for ultrasound backscattered signals, effective extraction of recursive feature parameters is not possible during calculation.

[0004] This invention proposes a method for optimizing the threshold and extracting feature parameters of a recursive graph based on particle swarm optimization (PSO). Backscattered signals are acquired using an ultrasonic testing device. The embedding dimension and delay time are determined using the spurious neighbor method and the average mutual information method. The distance between any two ultrasonic backscattered signal vectors in the phase space is calculated. PSO parameters and convergence conditions are set, and the recursive graph threshold is optimized using the PSO optimization algorithm. The optimized threshold is then used to construct the recursive graph of the ultrasonic backscattered signals. A recursive quantitative analysis method is used to quantitatively calculate the recursive graph, extract feature parameters, and finally construct a mapping relationship between the feature parameters and the material degradation damage level. This invention, based on the traditional recursive graph, introduces the PSO optimization algorithm to optimize the recursive graph threshold, which is of great significance in researching recursive graph threshold selection, recursive feature parameter extraction, and developing acoustic evaluation methods for material degradation levels. Summary of the Invention

[0005] The purpose of this invention is to propose a method for optimizing the threshold and feature parameter extraction of recursive graphs based on particle swarm optimization (PSO). This method combines PSO with the selection of threshold values ​​for different material degradation levels, enabling accurate extraction of recursive feature parameters under varying degradation levels. This overcomes the shortcomings of manually setting recursive thresholds alone, allowing for accurate and comprehensive analysis of ultrasonic backscattered signals and precise extraction of recursive feature parameters. This is of great significance for developing acoustic evaluation methods for material degradation damage.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for optimizing recursion graph thresholds and extracting feature parameters based on particle swarm optimization (PSO) algorithm is proposed. This method utilizes an ultrasonic signal acquisition system consisting of an ultrasonic testing device, an oscilloscope, and a digital signal acquisition card to acquire ultrasonic backscattered signals, which are time-series signals. The time-series signals are then reconstructed in phase space, and the Euclidean distance between any two ultrasonic backscattered signal vectors in the phase space is calculated. The difference between the Euclidean distance and the recursion threshold is calculated, and the difference is transformed into a two-dimensional matrix, i.e., the recursion graph matrix, using the Headside function. PSO parameters and convergence conditions are set, and the recursion graph threshold is optimized using the PSO algorithm. The optimized recursion graph threshold is then used to plot the recursion graph of the ultrasonic backscattered signals. A recursive quantitative analysis method is used to quantitatively calculate the recursion graph, extract feature parameters, and finally, the correlation between the feature parameters and the material degradation damage level is constructed to achieve the assessment of different material degradation levels.

[0008] The specific steps are as follows:

[0009] Step 1: Using an ultrasonic signal acquisition system consisting of ultrasonic testing equipment, an oscilloscope, and a digital signal acquisition card, ultrasonic backscatter signals are acquired from high-temperature pressure-bearing hot-walled pipes of different materials that have deteriorated. These ultrasonic backscatter signals will serve as the time series for subsequent research.

[0010] Step 2: Calculate the points that tend to stabilize and the positions corresponding to the first local minimum by using the spurious neighbor method and the average mutual information method. These positions are used as the final values ​​of the embedding dimension m and the delay time τ.

[0011] The spurious neighbor method reconstructs a time series to obtain an m-dimensional phase space, denoted by phase point X. n The nearest neighbor is X η(n) The distance between a phase point and its nearest neighbor is calculated as follows:

[0012] X η(n) -X n (m) (1)

[0013] After the phase space is transformed into m+l dimensions, the distance between phase points is expressed as:

[0014]

[0015] Calculate the ratio B of the distance between phase points in the (m+1)-dimensional phase space and the distance between phase points in the m-dimensional phase space. Define a threshold Br. If B > Br, then the point is a false neighbor. m is the optimal embedding dimension.

[0016] The average mutual information method selects the τ corresponding to the first local minimum of the mutual information function, and the expression for the mutual information function is:

[0017]

[0018] Using the embedding dimension m and delay time τ obtained above, the ultrasonic backscattered signal is reconstructed according to the phase space reconstruction technique based on Takens' embedding theorem, as shown in Equation (4):

[0019]

[0020] In the formula, X i Let m be the embedding dimension of the ultrasound backscattered signal, and τ be the delay time of the ultrasound backscattered signal. The number of reconstructed phase space vectors is M = n - (m - 1), where n is the sampling point of the ultrasound backscattered signal. The rows in the matrix are vectors reconstructed by adding the embedding dimension m, and the columns in the matrix are vectors reconstructed by adding the delay time τ.

[0021] The ultrasonic backscattered signal vector X obtained by phase space reconstruction i Each vector represents a point in the phase space. The distance between any two ultrasonic backscatter signal vectors in the phase space is calculated, and the recursion matrix is ​​shown in equations (5) and (6) below:

[0022] R i,j (ε)=Θ(ε-||x i -x j ||), i,j=1,...,N (5)

[0023]

[0024] In the formula, ε is the recursive graph threshold, Θ(x) is the Heaviside function, and x i x j It is the reconstructed ultrasonic backscatter signal vector.

[0025] Step 3: Set the initial parameters for the recursion graph threshold particle swarm optimization of the high-temperature pressure hot wall pipeline, including the number of particles in the particle swarm, the particle swarm calculation speed, the number of iterations of the recursion graph threshold, and the recursion graph threshold range.

[0026] Based on the above parameter settings, the global optimum and local optimum in the population are calculated according to the fitness function, and the velocity and position of the particles are calculated, as shown in formulas (7) and (8):

[0027] v id (t+1)=w·v id (t)+c1·rand()·(pbest(t)-x id (t))+c2·rand()·(gbest-x id (t)) (7)

[0028] x id (t+1)=v id (t+1)+x id (t) (8)

[0029] Calculate the fitness function values ​​of all particles at their new positions, and compare and update the global optimum. Select a local optimum from the particle swarm by comparing the current fitness function value with previous fitness function values.

[0030] Repeat the above steps until the objective function reaches its minimum value when the threshold iteration number is met, thus completing the optimized selection of the recursive graph threshold.

[0031] Step 4: Use the optimized recursion graph threshold to draw a recursion graph of the ultrasonic backscatter signal.

[0032] Step 5: Use recursive quantitative analysis to quantitatively calculate the recursion graph, extract feature parameters, and finally construct a mapping relationship between feature parameters and material degradation damage level. Polynomial fitting is used, and the mathematical model is as follows: Where x i For material degradation level, y i These are recursive feature parameters.

[0033] The present invention can achieve the following beneficial effects:

[0034] Based on traditional recursive graphs, a particle swarm optimization algorithm is introduced to optimize the threshold of the recursive graph, achieving precise selection of the threshold. The optimized threshold is then reset for recursive feature parameter extraction, constructing a comprehensive mapping model between the recursive feature parameters and different levels of material degradation, thus enabling acoustic evaluation of material degradation levels. Attached Figure Description

[0035] Figure 1 is a flowchart provided by the present invention.

[0036] Figure 2 shows an ultrasonic backscatter signal provided by the present invention.

[0037] Figure 3 shows the iteration curve of an algorithm provided by this invention.

[0038] Figure 4 shows an optimized threshold recursion graph provided by the present invention.

[0039] Figure 5 shows a recursive characteristic parameter for different material degradation levels provided by the present invention. Detailed Implementation

[0040] The present invention will be further described below with reference to the embodiments thereof. The embodiments described below are descriptive and not limiting, and should not be construed as limiting the scope of protection of the present invention.

[0041] Figure 1 is a flowchart of a method for optimizing the threshold and feature parameter extraction of a recursive graph based on the particle swarm optimization algorithm according to the present invention. The specific implementation is as follows:

[0042] (1) Using an ultrasonic signal acquisition system consisting of an ultrasonic testing device, an oscilloscope and a digital signal acquisition card, ultrasonic backscatter signals were acquired on samples of different materials that had deteriorated. Backscatter signals between 21μs and 32μs were extracted and used as the time series for subsequent research. The backscatter signals are shown in Figure 2.

[0043] (2) The points that tend to be stable and the positions corresponding to the first local minimum are calculated by the spurious neighbor method and the average mutual information method. These positions are used as the final values ​​of the embedding dimension and the delay time.

[0044] (3) Set the initial parameters of the particle swarm, including the number of particles in the particle swarm is 30, the position of the particles is randomly set, the number of threshold iterations is 50, and the threshold range is 0 to 10*σ (σ is the standard deviation of the signal). After iterative calculation of the optimal threshold, the fitness value reaches the minimum value. The iterative curve of the particle swarm optimization algorithm is shown in Figure 3.

[0045] (4) Using the optimized threshold, the ultrasonic backscatter signal of samples with different material degradation levels was plotted as a recursive graph. The result of the recursive graph of the first-level material degradation of the high-temperature pressure hot wall pipe is shown in Figure 4. The horizontal and vertical axes are the number of sampling points, and the image is symmetrically distributed about the main diagonal.

[0046] (5) Use recursive quantitative analysis to quantitatively calculate the optimized threshold recursive graph and extract characteristic parameters, namely recursion rate, determination rate, recursion entropy and layering rate.

[0047] (6) Repeat steps (1) to (5) to draw recursion diagrams for samples of high-temperature pressure-bearing hot-walled pipe material degradation at four different levels, and complete the extraction of recursion parameters. The normalized results of the extracted recursion feature parameters are shown in Figure 5. As can be seen from the figure, the recursion feature parameters gradually increase with the increase of the material degradation level, exhibiting a good linear relationship. Therefore, the optimized recursion threshold optimization method and material degradation evaluation method proposed in this invention are feasible.

[0048] The above description is only a preferred embodiment of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimizing threshold and feature parameter extraction of recursive graphs based on particle swarm optimization algorithm, characterized in that: The process includes the following steps: Step 1: Acquire ultrasonic backscattered signals using an ultrasonic signal acquisition system. These signals will serve as the time series of signals for subsequent research. Step 2: Determine the embedding dimension and delay time using the spurious neighbor method and the average mutual information method. Reconstruct the time series from Step 1 using phase space reconstruction technology and calculate the distance between any two vectors in the phase space. Step 3: Set particle swarm optimization parameters and convergence conditions, and optimize the recursion graph threshold using a particle swarm optimization algorithm. Step 4: Draw a recursion graph of the ultrasonic backscattered signals using the optimized threshold. Step 5: Quantitatively calculate the recursion graph using a recursive quantitative analysis method, extract feature parameters, and finally construct a mapping relationship between feature parameters and material degradation damage levels. In Step 2, the points that tend to stabilize and the positions corresponding to the first local minimum are calculated using the spurious neighbor method and the average mutual information method. These positions are used as the final values ​​of the embedding dimension m and the delay time τ. The spurious neighbor method reconstructs a time series to obtain an m-dimensional phase space. Let Xη(n) be the nearest neighbor of a phase point Xn. The distance between a phase point and its nearest neighbor is calculated as follows: After the phase space becomes m+l dimensional, the distance between phase points is expressed as: ； Calculate the ratio B of the distance between phase points in the (m+1)-dimensional phase space and the distance between phase points in the m-dimensional phase space. Define a threshold Br. If B > Br, then the point is a false neighbor. m is the optimal embedding dimension. The average mutual information method selects τ corresponding to the first local minimum of the mutual information function. The mutual information function expression is: Using the embedding dimension m and delay time τ obtained above, the ultrasonic backscatter signal is reconstructed according to the phase space reconstruction technique of Takens embedding theorem, as shown in equation (4): In the formula, Xi is the i-th phase point, m is the embedding dimension of the ultrasonic backscatter signal, and τ is the delay time of the ultrasonic backscatter signal; the number of reconstructed phase space vectors M = n - (m - 1), where n is the sampling point of the ultrasonic backscatter signal, the rows in the matrix are vectors reconstructed by adding the embedding dimension m, and the columns in the matrix are vectors reconstructed by adding the delay time τ; the ultrasonic backscatter signal vector Xi obtained by phase space reconstruction, each vector represents a point in the phase space, the distance between any two ultrasonic backscatter signal vectors in the phase space is calculated, and the recursive matrix is ​​shown in the following formulas (5) and (6): ； In the formula, ε is the recursion graph threshold. It's the Heaviside function. ， It is the reconstructed ultrasonic backscattered signal vector; in step three, the initial parameters for the recursion graph threshold particle swarm optimization of the high-temperature pressure hot wall pipe are set, including the number of particles in the particle swarm, the particle swarm calculation speed, the number of recursion graph threshold iterations, and the recursion graph threshold range; based on the above parameter settings, the global optimum and local optimum in the population are calculated according to the fitness function, and the velocity and position of the particles are calculated, as shown in formulas (7) and (8): ； Calculate the fitness function values ​​of all particles at their new positions, and compare and update the global optimum. Select the local optimum in the particle swarm based on the comparison between the current fitness function value and the previous fitness function value. When the threshold iteration number is satisfied, the objective function reaches its minimum value, thus completing the optimization selection of the recursive graph threshold.

2. The method for optimizing the threshold and feature parameter extraction of a recursive graph based on particle swarm optimization algorithm according to claim 1, characterized in that, Initialize the parameters in the particle swarm optimization algorithm, define the convergence condition, update the movement speed and position of each particle, and minimize the convergence condition.

3. The method for optimizing the threshold and feature parameter extraction of a recursive graph based on particle swarm optimization algorithm according to claim 1, characterized in that, The recurrence plot of the ultrasonic backscatter signal was drawn using the optimized threshold.

4. The method for optimizing the threshold and feature parameter extraction of a recursive graph based on particle swarm optimization algorithm according to claim 1, characterized in that, A recursive quantitative analysis method is used to extract feature parameters from the recursive graph and construct a comprehensive mapping model between recursive feature parameters and material degradation damage level.

Images