Methods, devices and media for noise reduction of arc voltage signals in high-frequency power supply welding

Abstract

This invention discloses a method, apparatus, and medium for denoising high-frequency power welding arc voltage signals. The method includes: acquiring the original arc voltage signal f(t); performing modal decomposition on the original arc voltage signal f(t) according to optimal parameters, and conducting global search and iteration to obtain the optimal solution (k, α) with the optimal fitness value; decomposing the original arc voltage signal f(t) using the optimal solution (k, α) to obtain multiple modal components; calculating the approximate entropy of each modal component; determining an approximate entropy threshold based on the approximate entropy of each modal component, and discarding noise components according to the approximate entropy threshold to filter the modal components; and reconstructing the filtered modal components to obtain the denoised arc voltage signal. This invention can achieve reliable filtering and efficient denoising of arc voltage signals, improving the accuracy and stability of arc length measurement.

Description

Technical Field

[0001] This invention belongs to the field of welding arc voltage signal denoising technology, and more specifically, relates to a method, device and medium for denoising high-frequency power supply welding arc voltage signals. Background Technology

[0002] Welding, as an important joining and repair technology, has a wide range of applications in the manufacturing field. Achieving efficient welding and improving energy utilization are the current development goals and research hotspots of welding technology.

[0003] In arc voltage sensing-based arc length control technology, acquiring the arc voltage signal is a prerequisite for arc length control. However, welding is a nonlinear and time-varying process, accompanied by various random disturbances and high-frequency interferences, causing significant fluctuations in the acquired arc voltage data and substantially impacting subsequent arc length control. Therefore, filtering algorithms are needed to denoise the acquired arc voltage values ​​to obtain accurate and stable arc voltage values. Achieving intelligent welding and enabling it to self-improve based on production data is of great significance for reducing energy consumption, improving welding efficiency, and realizing welding automation.

[0004] The welding process is a complex, nonlinear, and time-varying process, such as... Figure 1 As shown, arc ignition generates high-frequency interference, shielding gas delivery generates unstable interference, and changes in wire feeding speed and welding speed generate random interference. Therefore, the arc voltage signal data obtained by the signal acquisition module has large fluctuations, which has a significant impact on the subsequent arc length control process. Summary of the Invention

[0005] This invention addresses the aforementioned problems in the prior art. Therefore, a method, apparatus, and medium for denoising high-frequency power supply welding arc voltage signals are needed.

[0006] According to a first aspect of the present invention, a method for denoising high-frequency power supply welding arc voltage signals is provided, the method comprising:

[0007] Obtain the original arc voltage signal f(t);

[0008] The original arc voltage signal f(t) is modally decomposed according to the optimal parameters, and a global search and iteration are performed to obtain the optimal solution (k, α) with the optimal fitness value.

[0009] The original arc voltage signal f(t) is decomposed using the optimal solution (k,α) to obtain multiple modal components;

[0010] Calculate the approximate entropy of each modal component;

[0011] The approximate entropy threshold is determined based on the approximate entropy of each modal component, and noise components are discarded according to the approximate entropy threshold to filter the modal components;

[0012] The selected modal components are reconstructed to obtain the denoised arc voltage signal.

[0013] Further, the original arc voltage signal f(t) is modally decomposed according to the optimal parameters, and a global search and iteration are performed to obtain the optimal solution (k, α) with the optimal fitness value, including:

[0014] Initialize relevant parameters, including setting the gray wolf population size, the maximum number of iterations, the data range of the number of modes k, and the data range of the quadratic penalty factor α;

[0015] The gray wolf population is initialized using a chaotic sequence logistic mapping, the expression of which is:

[0016] x n =(m n -l n )y n +l n (1)

[0017] In the formula l n With m n The independent variable x n Minimum and maximum values ​​of the domain; y n It is a chaotic variable;

[0018] The original arc voltage signal f(t) is decomposed using VMD according to the assigned decomposition parameters to obtain the VMD decomposed signal.

[0019] Different weighting coefficients are used to balance the relationship between the signal-to-noise ratio and the autocorrelation coefficient. The maximum value of the matching result is taken as the required fitness value of the gray wolf, expressed as:

[0020] fitness = w α *SNR+w β *AC (6)

[0021] Among them, w α and w β The adaptive weighting coefficients for the signal-to-noise ratio (SNR) and autocorrelation coefficient (AC) are respectively calculated using the following equation (7):

[0022]

[0023] Among them, w αi and w αf For w α Initial and final values, w βiand w βf For w β Initial value and final value, g is the current iteration number, G is the maximum iteration number;

[0024] By calculating the objective function value of the gray wolf optimization algorithm, the optimal solution, the good solution, and the second-best solution are obtained as α wolf, β wolf, and δ wolf. The positions of the other gray wolves ω move towards α wolf, β wolf, and δ wolf to complete the encirclement of the prey. The position of α wolf, β wolf, and δ wolf is determined by the positions of α wolf, β wolf, and δ wolf. The moving formulas are shown in Equations (8) to (10):

[0025]

[0026]

[0027]

[0028] Where C1, C2, C3 and A1, A2, A3 are coefficient vectors used to correct the prey's position and adjust the direction of the gray wolf's movement, calculated using equations (11) and (12) respectively, X α X β X δ The new position is generated based on the positions of α, β, and δ wolves. Equation (10) is the formula for updating the new position of individual ω wolf. X(t+1) represents the estimated position of the next generation of prey. X1, X2, and X3 represent the estimated prey positions of α wolf, β wolf, and δ wolf, respectively.

[0029] C = 2r² (11)

[0030] A=2φr1-a (12)

[0031] In equations (11) and (12), r1 and r2 are random values ​​between (0, 1), and φ is a control parameter calculated by equation (13):

[0032]

[0033] Where a takes values ​​in the range [0, 2], and decreases linearly with the number of iterations;

[0034] The new search formula for gray wolf populations is as follows:

[0035]

[0036] Among them, X best Let q be the best gray wolf position in the current iteration, and let q be the step size control factor, calculated by equation (15); Levy(λ) is the random step size following a Levy distribution, calculated by equation (16):

[0037]

[0038]

[0039] In equation (16), This indicates that μ follows a mathematical expectation of 0 and a variance of . The expression for v follows a normal distribution, v ~ N(0,1), indicating that v follows a standard normal distribution with a mathematical expectation of 0 and a variance of 1. Determined by equation (17):

[0040]

[0041] Where χ is usually taken as 1.5;

[0042] Given that the current iteration count has reached the maximum iteration count, output the optimal solution (k, α) for the best gray wolf individual, i.e., the optimal fitness value.

[0043] Further, the original arc voltage signal f(t) is decomposed using VMD according to the assigned decomposition parameters to obtain the VMD decomposed signal, including:

[0044] For the variational problem of arc voltage signal, the signal decomposition process, i.e., the corresponding constraint variational expression, is as follows:

[0045]

[0046] In the formula, f represents the original arc voltage signal, k is the number of modes to be decomposed, and {u k},{ω k} represents the modal components and center frequencies for all modes, t represents time, δ(t) is the Dirichlet function, δ(t) represents the unit impulse function, and * represents the convolution operation. Let j be the signal gradient, and j be the imaginary unit. Represents the L2 norm;

[0047] Solving equation (2), by introducing the penalty factor α and the Lagrange operator λ, the constrained variational problem is transformed into an unconstrained variational problem, yielding the augmented Lagrange expression:

[0048]

[0049] During the update process, the modal components, center frequencies, and Lagrange operators of the components are continuously updated. Once the accuracy criterion is met, the VMD decomposition signal is output.

[0050] Furthermore, the calculation process of the signal-to-noise ratio (SNR) is shown in equation (4):

[0051]

[0052] Among them, F(n) is the denoised signal, and f(n) is the original signal.

[0053] Furthermore, the calculation process of the autocorrelation coefficient AC is shown in Equation (5):

[0054]

[0055] Among them, s i is the original signal, y i is the reconstructed signal, i is the index, n is the signal length, and are the average values of s i and y i respectively.

[0056] Furthermore, calculate the approximate entropy of each modal component, including:

[0057] Each modal component forms an m-dimensional vector X(i) in sequence, where:

[0058] X(i) = [u(i), u(i + 1), …, u(i + m - 1)] (18)

[0059] In the above formula, the value range of i is [1, N - m + 1], N is the data length of the modal component, and μ(i) represents each modal component after decomposition;

[0060] Calculate the distance between X(i) and X(j):

[0061]

[0062] Among them, μ(i + k) represents the (i + k)-th modal component, μ(j + k) represents the (j + k)-th modal component, and k is the index;

[0063] Set a threshold r (r > 0), is the ratio of d[X(i), X(j)] to (N - m + 1) when d[X(i), X(j)] < r:

[0064]

[0065] For take the logarithm, and then find its average value for all i, denoted as o m (r), that is:

[0066]

[0067] Increase the dimension m by 1, and obtain and σ m+1 (r);

[0068] The final approximate entropy is expressed as:

[0069]

[0070] Furthermore, the value of the optimal solution (k, α) is (10, 2045).

[0071] Furthermore, the approximate entropy threshold is 0.3.

[0072] According to a second technical solution of the present invention, a noise reduction device for high-frequency power welding arc voltage signals is provided, the device comprising:

[0073] The signal acquisition module is configured to acquire the raw arc voltage signal f(t);

[0074] The search iteration module is configured to perform mode decomposition of the original arc voltage signal f(t) according to the optimal parameters, and perform global search and iteration to obtain the optimal solution (k, α) with the optimal fitness value;

[0075] The signal decomposition module is configured to decompose the original arc voltage signal f(t) using the optimal solution (k,α) to obtain multiple modal components;

[0076] The entropy calculation module is configured to calculate the approximate entropy of each modal component;

[0077] The noise removal module is configured to determine an approximate entropy threshold based on the approximate entropy of each modal component, and discard noise components according to the approximate entropy threshold to filter the modal components;

[0078] The signal reconstruction module is configured to reconstruct the selected modal components to obtain the denoised arc voltage signal.

[0079] According to a third technical solution of the present invention, a readable storage medium is provided, wherein the readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the method described above.

[0080] The present invention has at least the following beneficial effects:

[0081] This invention utilizes an efficient optimization algorithm to improve variational mode decomposition, obtains the optimal mode components of the signal, calculates the approximate entropy to discard the noise mode components, and then reconstructs the signal, thereby achieving reliable filtering and efficient noise reduction of the arc voltage signal and improving the accuracy and stability of arc length measurement. Attached Figure Description

[0082] Figure 1 A schematic diagram of an arc pressure sensing arc length control welding test system according to an embodiment of the present invention is shown.

[0083] Figure 2A flowchart illustrating an overall process for denoising a high-frequency power supply welding arc voltage signal according to an embodiment of the present invention is shown.

[0084] Figure 3 A schematic diagram of the original arc voltage signal f(t) according to an embodiment of the present invention is shown;

[0085] Figure 4 A schematic diagram of the modal components under the optimal solution (k, α) according to an embodiment of the present invention is shown;

[0086] Figure 5 A schematic diagram of the reconstructed arc voltage signal according to an embodiment of the present invention is shown;

[0087] Figure 6 A flowchart illustrating the optimal solution determination process according to an embodiment of the present invention is shown.

[0088] Figure 7 A structural diagram of a high-frequency power supply welding arc voltage signal noise reduction device according to an embodiment of the present invention is shown. Detailed Implementation

[0089] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and specific examples, but this is not intended to limit the present invention. If there is no necessary sequential relationship between the various steps described herein, the order in which they are described as examples should not be considered a limitation. Those skilled in the art should understand that the order can be adjusted, as long as it does not disrupt the logical consistency between them and render the entire process impossible.

[0090] Figure 1 A schematic diagram of an arc voltage sensing arc length control welding test system according to an embodiment of the present invention is shown. This system includes a welding torch, a shielding gas cylinder, a wire feeding mechanism mounted on a robotic arm, a welding power source, a controller, a Hall sensor, and a DSP embedded data acquisition unit. The shielding gas cylinder is connected to the welding torch to provide shielding gas for welding. The wire feeding mechanism mounted on the robotic arm provides welding wire to the welding torch. The welding power source provides an arc voltage signal to the welding torch, which is detected by the Hall sensor and fed to the controller via the DSP embedded data acquisition unit. The controller can also control the wire feeding mechanism and the welding power source. The welding torch generates a welding arc, which includes a keyhole, a tail flame, and a molten pool for welding the base material.

[0091] Figure 2 A flowchart illustrating an overall process for denoising high-frequency power supply welding arc voltage signals according to an embodiment of the present invention is shown, as follows: Figure 2As shown in the figure, this embodiment of the invention provides a method for denoising high-frequency power supply welding arc voltage signals. This method is used to denoise signals such as... Figure 1 The system shown performs noise reduction processing on the arc voltage signal collected by the system. This method can also be implemented by the controller.

[0092] It should be noted that this method can be applied to, for example... Figure 1 The system shown performs noise reduction on the arc voltage signal acquired by the system. This method can also be applied to perform noise reduction on arc voltage signals acquired by other related systems. This embodiment is merely an example and is not intended to limit the invention.

[0093] like Figure 2 As shown, the high-frequency power supply welding arc voltage signal denoising method includes steps 1-6, which are described in detail below:

[0094] Step 1: Construct an arc voltage sensing arc length control welding test system to acquire the voltage signal containing welding arc length information under high-frequency power supply. Utilize a Hall voltage sensor and related data acquisition equipment to obtain the original arc voltage signal f(t), such as... Figure 3 As shown;

[0095] Step 2: Using the original arc voltage signal f(t) as the input signal, perform optimization and decomposition processing using the improved iterative VMD-ICGWO (variable mode decomposition algorithm and improved chaotic gray wolf optimization algorithm) algorithm: Perform mode decomposition on the original arc voltage signal f(t) according to the optimal parameters of VMD, and use ICGWO for global search and iteration to obtain the optimal solution (k, α) with the optimal fitness value. In this invention, the optimal decomposition parameter (k, α) obtained by the VMD-ICGWO algorithm iteration is (10, 2045).

[0096] Step 3: Substitute the optimal solution (k, α) into the VMD decomposition of the original arc voltage signal f(t) to obtain k modal components IMF1, IMF2, ..., IMFk, as follows: Figure 4 As shown;

[0097] Step 4: Calculate the approximate entropy of each modal component. The results are shown in Table 1.

[0098] Step 5: From Table 1, we can see that the approximate entropy value of IMF1 is 0.037, while the approximate entropy value of IMF2 increases to 0.346. Furthermore, the approximate entropy values ​​of subsequent modal components differ significantly from those of IMF1. Therefore, the approximate entropy threshold is determined to be 0.3, and noise components are discarded according to the approximate entropy threshold.

[0099] Step 6: Reconstruct the selected modal components to obtain the denoised arc voltage signal, such as... Figure 5As shown, this step effectively removes the base noise from the original signal, resulting in a reconstructed signal with a high signal-to-noise ratio. It also effectively extracts and restores the arc voltage signal that was submerged in noise.

[0100] Table 1 Approximate entropy for each mode

[0101]

[0102] Figure 6 This is a schematic diagram of the optimization and decomposition process of the improved VMD-ICGWO algorithm in step 2 above, which specifically includes the following steps:

[0103] Step 2.1: Initialize the relevant parameters of the VMD-ICGWO algorithm, including setting the gray wolf population size N=100, the maximum number of iterations G=100, the data range of the number of modes k to [2,15], and the data range of the secondary penalty factor α to [100,5000];

[0104] Step 2.2: In the GWO algorithm, the diversity of the population during initialization significantly impacts subsequent iterations. Therefore, a chaotic sequence Logistic mapping is used to initialize the gray wolf population, enriching its diversity and generating a broad and diverse initial solution space. This helps the algorithm explore a wider range of potential solutions. Simultaneously, the dynamic and uncertain characteristics of chaotic systems can help the optimization algorithm escape local optima, accelerating the global search capability during the search process. The Logistic model expression is:

[0105] x n =(m n -l n )y n +l n (1)

[0106] In the formula l n With m n The independent variable x n Minimum and maximum values ​​of the domain; y n It is a chaotic variable.

[0107] Step 2.3: After initialization, perform VMD decomposition on the original arc voltage signal f(t) according to the assigned decomposition parameters (k, α) to obtain the VMD decomposed signal; the details are as follows:

[0108] Variational mode decomposition (VMD) obtains the center frequency and bandwidth of each mode component by iteratively seeking the optimal solution of the variational model, thereby completing the full-band decomposition of the signal. For the arc voltage signal, a variational problem is established, and the signal decomposition process, i.e., the corresponding constrained variational expression, is as follows:

[0109]

[0110] In the formula, f represents the original arc voltage signal, k is the number of modes to be decomposed, and {u k},{ω k} represents the modal components and center frequencies for all modes, t represents time, δ(t) is the Dirichlet function, δ(t) represents the unit impulse function, and * represents the convolution operation. Let j be the signal gradient, and j be the imaginary unit. This represents the L2 norm.

[0111] Solving equation (2), we introduce a penalty factor α and a Lagrange operator λ, transforming the constrained variational problem into an unconstrained variational problem. Introducing these two factors benefits both from the good convergence of the penalty factor under finite weights and from the strict enforcement of constraints by the Lagrange multipliers. The augmented Lagrange expression is then obtained as:

[0112]

[0113] During the update process, the modal components, center frequencies, and Lagrange operators of the components are continuously updated. Once the accuracy criterion is met, the waveforms of each decomposed mode are output.

[0114] Step 2.4: When optimizing parameters using the ICGWO algorithm, a fitness function needs to be selected. This invention uses the signal-to-noise ratio (SNR) and autocorrelation coefficient (AC) of the signal to noise as evaluation criteria for denoising performance. The higher the quality of the denoising operation, the higher its SNR will be, defined as follows:

[0115]

[0116] The calculation process of the autocorrelation coefficient is shown in equation (5):

[0117]

[0118] Among them, s i The original signal, y i The AC value represents the reconstructed signal. A larger AC value indicates a smaller deviation between the reconstructed signal and the original signal, and a higher degree of fit.

[0119] By using different weighting coefficients to balance the relationship between the signal-to-noise ratio and the autocorrelation coefficient, the maximum value of the matching result is taken as the required fitness value of the gray wolf, which can be specifically expressed as:

[0120] fitness = w α *SNR+w β *AC (6)

[0121] Among them, w α and w βThe adaptive weighting coefficients for SNR and AC are respectively, and their calculation is as follows:

[0122]

[0123] w αi and w αf For w α Initial and final values, w βi and w βf For w β Initial and final values, g is the current iteration number, and G is the maximum iteration number.

[0124] Step 2.5: In GWO (Grey Wolf Algorithm), the optimal solution, the good solution and the second best solution are obtained by calculating the objective function value of the Grey Wolf optimization algorithm. Let α wolf, β wolf and δ wolf be the optimal solution and δ wolf respectively. The positions of other Grey wolves ω move towards α wolf, β wolf and δ wolf to complete the encirclement of the prey. The position of α wolf, β wolf and δ wolf is determined by the position of α wolf, β wolf and δ wolf. The moving formula is shown in Equation (8) to Equation (10).

[0125]

[0126]

[0127]

[0128] Where C1, C2, and C3 are calculated using equation (11), and A1, A2, and A3 are calculated using equation (12), X α X β X δ The new position is generated based on the positions of α, β, and δ wolves. Equation (10) is the formula for updating the new position of individual ω wolf.

[0129] C = 2r² (11)

[0130] A=2φr1-a (12)

[0131] In equations (11) and (12), r1 and r2 are random values ​​between (0, 1), and φ is a control parameter calculated by the following formula:

[0132]

[0133] Here, a takes values ​​in the range [0, 2] and decreases linearly with the number of iterations.

[0134] To effectively avoid local optima, accelerate convergence, and improve global search capabilities, the improved search formula for the gray wolf population is as follows:

[0135]

[0136] Among them, Xbest is the best position of the gray wolf in the current iteration process, q is the step size control factor, calculated by Equation (15); Levy(λ) is a random step size obeying the Levy distribution, calculated by Equation (16):

[0137]

[0138]

[0139] In Equation (16), v ∼ N(0, 1), determined by Equation (17):

[0140]

[0141] where χ usually takes the value of 1.5.

[0142] Through the above update process, the improved gray wolf optimization algorithm can find the global optimal solution faster.

[0143] Step 2.6: Judge whether the current iteration number reaches the maximum iteration number. If it reaches, output the optimal gray wolf individual, that is, the optimal (k, α) parameters; otherwise, return to Step 2.3 and add 1 to the iteration number.

[0144] The optimal parameters (k, α) are obtained through the improved VMD - ICGWO, and the original arc voltage signal is decomposed into different modal components. The concept of approximate entropy is introduced. By calculating the approximate entropy of each modal component and discarding the noise modal components according to the approximate entropy threshold, the selected modal components are reconstructed, so as to realize modal division and achieve the best denoising effect. The specific calculation steps of approximate entropy are as follows:

[0145] Step 4.1: The modal components obtained by the above variational modal decomposition are sequentially formed into an m - dimensional vector X(i), that is, from X(1) to X(N - m + 1), where:

[0146] X(i) = [u(i), u(i + 1), …, u(i + m - 1)] (18)

[0147] In the above formula, the value range of i is [1, N - m + 1].

[0148] Step 4.2: The distance between X(i) and X(j) is determined by the maximum difference of the corresponding elements, and the calculation is as follows:

[0149]

[0150] Step 4.3: Set a threshold r (r > 0), which is the ratio of d[X(i), X(j)] to (N - m + 1) when d[X(i), X(j)] < r:

[0151]

[0152] Step 4.4: For Taking the logarithm, we get:

[0153]

[0154] Step 4.5: Increment the dimension m by 1, and repeat the first four steps to obtain... and σ m+1 (r)

[0155] Step 4.6: The final approximate entropy can be expressed as:

[0156]

[0157] This invention provides a noise reduction device for high-frequency power supply welding arc voltage signals, such as... Figure 7 As shown, the device 700 includes:

[0158] The signal acquisition module 701 is configured to acquire the original arc voltage signal f(t);

[0159] The search iteration module 702 is configured to perform mode decomposition on the original arc voltage signal f(t) according to the optimal parameters, and perform global search and iteration to obtain the optimal solution (k, α) with the optimal fitness value.

[0160] The signal decomposition module 703 is configured to decompose the original arc voltage signal f(t) using the optimal solution (k,α) to obtain multiple modal components;

[0161] Entropy calculation module 704 is configured to calculate the approximate entropy of each modal component;

[0162] The noise removal module 705 is configured to determine an approximate entropy threshold based on the approximate entropy of each modal component, and discard noise components according to the approximate entropy threshold to filter the modal components.

[0163] The signal reconstruction module 706 is configured to reconstruct the selected modal components to obtain the denoised arc voltage signal.

[0164] In some embodiments, the search iteration module is further configured to:

[0165] Initialize relevant parameters, including setting the gray wolf population size, the maximum number of iterations, the data range of the number of modes k, and the data range of the quadratic penalty factor α;

[0166] The gray wolf population is initialized using a chaotic sequence logistic mapping, the expression of which is:

[0167] x n =(m n -l n )y n +l n (1)

[0168] In the formula l n With m n The independent variable x n Minimum and maximum values ​​of the domain; y n It is a chaotic variable;

[0169] The original arc voltage signal f(t) is decomposed using VMD according to the assigned decomposition parameters to obtain the VMD decomposed signal.

[0170] Different weighting coefficients are used to balance the relationship between the signal-to-noise ratio and the autocorrelation coefficient. The maximum value of the matching result is taken as the required fitness value of the gray wolf, expressed as:

[0171] fitness = w α *SNR+w β *AC (6)

[0172] Among them, w α and w β The adaptive weighting coefficients for the signal-to-noise ratio (SNR) and autocorrelation coefficient (AC) are respectively calculated using the following equation (7):

[0173]

[0174] Among them, w αi and w αf For w α Initial and final values, w βi and w βf For w β Initial value and final value, g is the current iteration number, G is the maximum iteration number;

[0175] By calculating the objective function value of the gray wolf optimization algorithm, the optimal solution, the good solution, and the second-best solution are obtained as α wolf, β wolf, and δ wolf. The positions of the other gray wolves ω move towards α wolf, β wolf, and δ wolf to complete the encirclement of the prey. The position of α wolf, β wolf, and δ wolf is determined by the positions of α wolf, β wolf, and δ wolf. The moving formulas are shown in Equations (8) to (10):

[0176]

[0177]

[0178]

[0179] Where C1, C2, C3 and A1, A2, A3 are coefficient vectors used to correct the prey's position and adjust the direction of the gray wolf's movement, calculated using equations (11) and (12) respectively, X α X β X δ The new position is generated based on the positions of α, β, and δ wolves. Equation (10) is the formula for updating the new position of individual ω wolf. X(t+1) represents the estimated position of the next generation of prey. X1, X2, and X3 represent the estimated prey positions of α wolf, β wolf, and δ wolf, respectively.

[0180] C = 2r² (11)

[0181] A=2φr1-a (12)

[0182] In equations (11) and (12), r1 and r2 are random values ​​between (0, 1), and φ is a control parameter calculated by equation (13):

[0183]

[0184] Where a takes values ​​in the range [0, 2], and decreases linearly with the number of iterations;

[0185] The new search formula for gray wolf populations is as follows:

[0186]

[0187] Among them, X best Let q be the best gray wolf position in the current iteration, and let q be the step size control factor, calculated by equation (15); Levy(λ) is the random step size following a Levy distribution, calculated by equation (16):

[0188]

[0189]

[0190] In equation (16), This indicates that μ follows a mathematical expectation of 0 and a variance of . The expression for v follows a normal distribution, v ~ N(0,1), indicating that v follows a standard normal distribution with a mathematical expectation of 0 and a variance of 1. Determined by equation (17):

[0191]

[0192] Where χ is usually taken as 1.5;

[0193] Given that the current iteration count has reached the maximum iteration count, output the optimal solution (k, α) for the best gray wolf individual, i.e., the optimal fitness value.

[0194] In some embodiments, the search iteration module is further configured to:

[0195] For the variational problem of arc voltage signal, the signal decomposition process, i.e., the corresponding constraint variational expression, is as follows:

[0196]

[0197] In the formula, f represents the original arc voltage signal, k is the number of modes to be decomposed, and {u k},{ω k} represents the modal components and center frequencies for all modes, t represents time, δ(t) is the Dirichlet function, δ(t) represents the unit impulse function, and * represents the convolution operation. Let j be the signal gradient, and j be the imaginary unit. Represents the L2 norm;

[0198] Solving equation (2), by introducing the penalty factor α and the Lagrange operator λ, the constrained variational problem is transformed into an unconstrained variational problem, yielding the augmented Lagrange expression:

[0199]

[0200] During the update process, the modal components, center frequencies, and Lagrange operators of the components are continuously updated. Once the accuracy criterion is met, the VMD decomposition signal is output.

[0201] In some embodiments, the search iteration module is further configured to calculate the signal-to-noise ratio (SNR) of the signal against noise as shown in equation (4):

[0202]

[0203] Where F(n) is the denoised signal and f(n) is the original signal.

[0204] The calculation process of the autocorrelation coefficient AC is shown in equation (5):

[0205]

[0206] Among them, s i The original signal, y i To reconstruct the signal, i is the index and n is the signal length. and s i and y i The average value.

[0207] In some embodiments, the entropy calculation module is further configured to:

[0208] The modal components are arranged in order to form an m-dimensional vector X(i), where:

[0209] X(i) = [u(i), u(i + 1), …, u(i + m - 1)] (18)

[0210] In the above formula, the value range of i is [1, N - m + 1], N is the data length of the modal component, and μ(i) represents each modal component after decomposition;

[0211] Calculate the distance between X(i) and X(j):

[0212]

[0213] where, μ(i + k) represents the (i + k)-th modal component, μ(j + k) represents the (j + k)-th modal component, and k is the index;

[0214] Set a threshold r (r > 0), When d[X(i), X(j)] < r, it is the ratio of d[X(i), X(j)] to (N - m + 1):

[0215]

[0216] For Take the logarithm, and then find its average value for all i, denoted as o m (r), that is:

[0217]

[0218] Increase the dimension m by 1, and obtain according to formulas (18) to (21) and σ m+1 (r);

[0219] The final approximate entropy is expressed as:

[0220]

[0221] In some embodiments, the value of the optimal solution (k, α) is (10, 2045).

[0222] In some embodiments, the approximate entropy threshold is 0.3.

[0223] It should be noted that the device described in this embodiment belongs to the same technical concept as the method described above, and can achieve the same technical effects, which will not be elaborated here.

[0224] The embodiment of the present invention provides a readable storage medium, and the readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the methods described in the above various embodiments.

[0225] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. Other embodiments can be used by those skilled in the art when reading the above description. Furthermore, in the above detailed description, various features may be grouped together to simplify the invention. This should not be construed as an intention that a feature of an unclaimed invention is necessary for any claim. Rather, the subject matter of the invention may be less than all the features of a particular embodiment of the invention. Thus, the following claims are incorporated herein by reference as examples or embodiments, wherein each claim is an independent, separate embodiment, and these embodiments are contemplated to be combined with each other in various combinations or arrangements. The scope of the invention should be determined by reference to the appended claims and the full scope of their equivalents.

Claims

1. A method for denoising high-frequency power supply welding arc voltage signals, characterized in that, The method includes: Obtain the original arc voltage signal f(t); The original arc voltage signal f(t) is modally decomposed according to the optimal parameters, and a global search and iteration are performed to obtain the optimal solution (k, α) with the optimal fitness value. The original arc voltage signal f(t) is decomposed using the optimal solution (k,α) to obtain multiple modal components; Calculate the approximate entropy of each modal component; The approximate entropy threshold is determined based on the approximate entropy of each modal component, and noise components are discarded according to the approximate entropy threshold to filter the modal components; The filtered modal components are reconstructed to obtain the denoised arc voltage signal; The original arc voltage signal f(t) is modally decomposed according to the optimal parameters, and a global search and iteration are performed to obtain the optimal solution (k, α) with the optimal fitness value, including: Initialize relevant parameters, including setting the gray wolf population size, the maximum number of iterations, the data range for the number of modalities k, and the quadratic penalty factor. Data range; The gray wolf population is initialized using a chaotic sequence logistic mapping, the expression of which is: （1） In the formula and They are the independent variables. Minimum and maximum values ​​of the domain; It is a chaotic variable; The original arc voltage signal f(t) is decomposed using VMD according to the assigned decomposition parameters to obtain the VMD decomposed signal. Different weighting coefficients are used to balance the relationship between the signal-to-noise ratio and the autocorrelation coefficient. The maximum value of the matching result is taken as the required fitness value of the gray wolf, expressed as: （6） in, and The adaptive weighting coefficients for the signal-to-noise ratio (SNR) and autocorrelation coefficient (AC) are respectively calculated using the following equation (7): （7） in, and for Initial value and final value, and for Initial value and final value, g is the current iteration number, G is the maximum iteration number; By calculating the objective function value of the Grey Wolf optimization algorithm, the optimal solution, good solution, and suboptimal solution are obtained. Wolf, wolves and Wolves, other gray wolves Position towards Wolf, wolves and The wolf moves, completing the encirclement of its prey, by Wolf, wolves and The wolf's position is determined by the movement formulas shown in equations (8) to (10): （8） （9） （10） in, , , and , , These are all coefficient vectors used to correct the prey's position and adjust the direction of the gray wolf's movement, calculated using equations (11) and (12) respectively. , , It is based on The new position generated by the wolf's position is given by equation (10). Formula for updating the new position of a wolf individual X ( t +1 indicates the estimated location of the next generation of prey. X 1. X 2. X 3 respectively represent Wolf, wolves and Wolves can pinpoint the location of their prey. （11） （12） In equations (11) and (12) For random values ​​between (0, 1) The control parameter is calculated using equation (13): （13） in, It takes values ​​in the range [0, 2], and decreases linearly with the number of iterations. The new search formula for gray wolf populations is as follows: （14） in, Let q be the best position of the gray wolf in the current iteration, and let q be the step size control factor, calculated by equation (15); The random step size, which follows a Lévy distribution, is calculated using equation (16): （15） （16） In equation (16), ,express It follows a mathematical expectation of 0 and a variance of . The normal distribution ,express It follows a standard normal distribution with a mathematical expectation of 0 and a variance of 1. Determined by equation (17): （17） in, It is a constant, with a value of 1.5; Given that the current iteration count has reached the maximum iteration count, output the optimal solution (k, α) for the best gray wolf individual, i.e., the optimal fitness value.

2. The method according to claim 1, characterized in that, The original arc voltage signal f(t) is decomposed using VMD according to the assigned decomposition parameters to obtain the VMD decomposed signal, including: For the variational problem of arc voltage signal, the signal decomposition process, i.e., the corresponding constraint variational expression, is as follows: （2） In the formula, f represents the original arc voltage signal, and k is the number of modes to be decomposed. These represent the modal components and center frequencies for all modes. Indicates time, For Dirichlet functions, The symbol represents the unit impulse function, and * represents the convolution operation. For the signal gradient, The imaginary unit, Represents the L2 norm; To solve equation (2), we introduce the penalty factor α and the Lagrange operator. Transforming the constrained variational problem into an unconstrained variational problem, we obtain the augmented Lagrange expression as follows: （3） During the update process, the modal components, center frequencies, and Lagrange operators of the components are continuously updated. Once the accuracy criterion is met, the VMD decomposition signal is output.

3. The method according to claim 1, characterized in that, The signal-to-noise ratio (SNR) of a signal to noise is calculated as shown in equation (4): （4） in, F ( n () represents the noise-reduced signal. f ( n () represents the original signal.

4. The method according to claim 1, characterized in that, The calculation process of the autocorrelation coefficient AC is shown in equation (5): （5） in, The original signal, To reconstruct the signal, i For indexing, n For signal length, and They are respectively and The average value.

5. The method according to claim 1, characterized in that, Calculate the approximate entropy of each modal component, including: The modal components are arranged in sequence to form an m-dimensional vector. ,in: （18） In the above formula The value of is [1, N-m+1], where N is the data length of the modal component. Represents the modal components after decomposition; calculate and Distance between: （19） in, Indicates the first One modal component, Indicates the first One modal component, For indexing; Set threshold , for hour and The ratio: （20） right Take the logarithm, then calculate its average over all i, denoted as ,Right now: （21） Increment the dimension m by 1, and obtain the result from formulas (18) to (21). and ; The final approximate entropy is expressed as: （22）。 6. The method according to claim 1, characterized in that, The optimal solution (k, α) has a value of (10, 2045).

7. The method according to claim 1, characterized in that, The approximate entropy threshold is 0.

3.

8. A method and apparatus for denoising high-frequency power supply welding arc voltage signals, used to implement the method as described in any one of claims 1 to 7, characterized in that, The device includes: The signal acquisition module is configured to acquire the raw arc voltage signal f(t); The search iteration module is configured to perform mode decomposition of the original arc voltage signal f(t) according to the optimal parameters, and to perform global search and iteration to obtain the optimal solution (k, α) with the optimal fitness value. The signal decomposition module is configured to decompose the original arc voltage signal f(t) using the optimal solution (k,α) to obtain multiple modal components; The entropy calculation module is configured to calculate the approximate entropy of each modal component; The noise removal module is configured to determine an approximate entropy threshold based on the approximate entropy of each modal component, and discard noise components according to the approximate entropy threshold to filter the modal components; The signal reconstruction module is configured to reconstruct the selected modal components to obtain the denoised arc voltage signal.

9. A readable storage medium, characterized in that, The readable storage medium stores one or more programs, which can be executed by one or more processors to implement the method as described in any one of claims 1 to 7.

Images