LEA-VMD-based nonlinear ultrasonic signal processing method
Through the nonlinear ultrasonic signal processing method based on LEA-VMD, the amplitudes of the one-frequency and two-frequency are extracted using variational modal decomposition and Hilbert transform, solving the problem that the nonlinear ultrasonic signal is weakly susceptible to noise, and achieving more accurate material damage detection.
Patent Information
- Application Number
- CN202411832104.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-12
- Publication Date
- 2025-03-18
AI Technical Summary
The nonlinear ultrasonic signals are very weak and are easily affected by noise. The prior art is difficult to effectively extract nonlinear features, resulting in difficulty in detecting early performance degradation of materials.
The nonlinear ultrasonic signal processing method based on LEA-VMD is used to extract the amplitude of the one-fold and two-fold frequency through variational modal decomposition and Hilbert transform to characterize the degree of material damage.
This method can significantly reduce noise interference, improve the quality of nonlinear ultrasonic signals, more accurately characterize the degree of material damage, and support early performance detection of materials.
Smart Images
Figure CN119643724A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of nonlinear ultrasonic detection signal processing, and in particular to a nonlinear ultrasonic signal processing method based on LEA-VMD. Background Art
[0002] With the development of industrial level, high-end equipment usually works in complex and harsh environments such as high temperature, high pressure, fatigue, and creep. Long-term service will cause components in the equipment to face potential risks such as creep damage and fatigue damage, which not only affects the service life of the components, but also affects the operation quality of high-end equipment.
[0003] To avoid these potential risks, technical personnel must conduct timely testing of equipment status. Research has shown that early material degradation accounts for a significant portion of a structure's lifespan, making early material performance testing particularly important. Nonlinear ultrasonic testing, among other methods, can accurately detect early microcracks in materials and is the most effective nondestructive testing method for evaluating material performance.
[0004] However, nonlinear ultrasonic signals are very weak and easily affected by noise. To better assess material properties using nonlinear ultrasonic testing, preprocessing of the nonlinear ultrasonic signals is necessary to facilitate the extraction of nonlinear features. Therefore, those skilled in the art need a processing method that can produce clearer nonlinear ultrasonic testing signals. Summary of the Invention
[0005] The purpose of the present invention is to solve the above problems and design a nonlinear ultrasonic signal processing method based on LEA-VMD.
[0006] The technical solution of the present invention to achieve the above-mentioned object is a nonlinear ultrasonic signal processing method based on LEA-VMD, which comprises:
[0007] Acquire nonlinear ultrasonic detection signals, set parameters of the LEA algorithm, input the nonlinear ultrasonic detection signals into the LEA algorithm, and obtain the optimal parameter penalty factor and number of decompositions for variational mode decomposition;
[0008] The nonlinear ultrasonic detection signal is decomposed into modal components using the optimal parameter penalty factor and decomposition number, and then the first and second harmonic frequencies of the excitation signal are selected as the reconstructed signals.
[0009] The reconstructed signal is Hilbert transformed to obtain the analytical signal, and the analytical signal is Fourier transformed to obtain the amplitudes of the first and second frequencies. The degree of material damage is characterized according to the amplitudes of the first and second frequencies, thereby completing nonlinear ultrasonic testing.
[0010] The nonlinear ultrasonic detection signal x(t) is obtained by detecting a component specimen with microcrack damage using a nonlinear ultrasonic phased array probe and an ultrasonic signal acquisition instrument. The parameter setting process of the LEA algorithm is as follows: the penalty factor and the number of decompositions in the variational modal decomposition are used as optimization targets, and the envelope entropy is used as the fitness function. The optimal parameter penalty factor and the number of decompositions of the variational modal decomposition are obtained by inputting the nonlinear ultrasonic detection signal x(t) into the LEA algorithm and using the minimum envelope entropy as the target value of the fitness function.
[0011] The process of solving the optimal parameter penalty factor and the number of decompositions of the variational mode decomposition includes:
[0012] Initialize population parameters: set population size N, number of iterations T, penalty factor α and number of decompositions k;
[0013] Position X where the dragonfly pollinates i The signal is decomposed by VMD as the penalty factor α and the number of decompositions k, and the envelope entropy of the signal is calculated as the fitness function value, and the optimal individual X is recorded in the process. i ; The expression of envelope entropy is:
[0014]
[0015] Where, E i represents the envelope entropy value of the i-th modal component, N represents the number of signals in each layer, a(j) represents the envelope signal obtained after Hilbert demodulation of the i-th modal component, p j is the normalized form of a(j);
[0016] Dragonfly pollination location X i The update of is affected by five factors, which are calculated separately as follows:
[0017]
[0018] Among them, X t i represents the position of the current individual in evolutionary iteration t, X t j represents the position of the jth individual in the neighborhood at evolutionary iteration t, and N is the number of individuals in the neighborhood;
[0019] Then the alignment is calculated as follows:
[0020]
[0021] The aggregation is calculated as follows:
[0022]
[0023] The attractiveness towards a food source is calculated as follows:
[0024]
[0025] Among them, X t i represents the position of the current individual in evolutionary iteration t, X t + is the location of the food source, i.e., the best solution found in the current evolutionary iteration;
[0026] The repulsive force on the enemy is calculated as follows:
[0027]
[0028] Among them, X t i represents the position of the current individual in evolutionary iteration t, X t - is the enemy's position, i.e., the worst solution found in the current evolutionary iteration;
[0029] The step vector is used to show the direction of the dragonfly's movement, which is defined as follows:
[0030]
[0031] Where s is the separation coefficient, S t i is the degree of separation of the i-th individual in evolutionary iteration i, a is the alignment coefficient, A t i is the degree of alignment of the i-th individual, c is the cohesion coefficient, C t i is the cohesion degree of the i-th individual, f is the food factor, F t i is the food source of the i-th individual, e is the enemy factor, E t i is the enemy of the i-th individual, w is the inertia weight, and t is the iteration counter of the algorithm;
[0032] After calculating the step vector, the dragonfly pollination position X i The position vector is calculated as follows:
[0033]
[0034] Where t is the current iteration counter and d is the dimension of the position vector; then Levy is calculated as follows:
[0035]
[0036] Where r1 and r2 are two random numbers between 0 and 1, β is a constant, and σ is calculated as follows:
[0037]
[0038] Using partial pollination (self-fertilization) as the development stage of the LEA algorithm, we have:
[0039]
[0040] Among them, X t+1 is the position of the pollen in iteration t+1, g * is the best pollen position found in all evolutionary iterations, and R is the growth region, which shrinks as the algorithm iterations proceed;
[0041]
[0042] Among them, t is the algorithm of the current evolutionary iteration, which iterates until the maximum number of cycles L is reached and the cycle ends, outputting the optimal individual position, i.e., the penalty factor α and the number of decompositions k.
[0043] The reconstructed signal is set to y(t), and the Hilport transform of the reconstructed signal is:
[0044]
[0045] The constructed analytical signal is:
[0046] z(t)=y(t)+iY(t)
[0047] The Fourier transform of z(t) is:
[0048]
[0049] The above formula gives the amplitude of the first harmonic frequency A1 and the amplitude of the second harmonic frequency A2. The degree of material damage can be characterized by the two amplitudes as follows:
[0050]
[0051] γ is the nonlinear coefficient of the material, and γ is positively correlated with the damage of the material.
[0052] Compared with the prior art, the present invention has the following advantages:
[0053] This method uses the LEA-VMD algorithm to process and analyze nonlinear ultrasonic signals. Compared with existing methods, it can greatly reduce noise interference and increase the amplitude of the first and second harmonic frequencies in nonlinear ultrasonic signals, thereby obtaining high-quality nonlinear ultrasonic signals and better characterizing material damage. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 is a flow chart of a nonlinear ultrasonic signal processing method based on LEA-VMD according to the present invention;
[0055] Figure 2 It is a frequency domain comparison diagram of the signal before and after the nonlinear ultrasonic signal processing described in the present invention. DETAILED DESCRIPTION
[0056] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 As shown;
[0057] The technical solution of the present application is to detect an aluminum specimen with microcrack damage by using a nonlinear ultrasonic phased array probe and an ultrasonic signal acquisition instrument to obtain a nonlinear ultrasonic detection signal x(t), and then set the parameters of the LEA algorithm (lotus effect algorithm), and use the penalty factor α and the number of decompositions k in the variational modal decomposition as optimization targets, and use the envelope entropy as the fitness function. The nonlinear ultrasonic detection signal x(t) is input into the LEA algorithm, and the minimum value of the envelope entropy is used as the target value of the fitness function to obtain the optimal parameters of the variational modal decomposition, namely the penalty factor α and the number of decompositions k.
[0058] Among them, the main innovation of the present invention is: using LEA (Lotus Effect Algorithm) to optimize the penalty factor α and the number of decompositions k in the variational modal decomposition, and integrating the Hilbert transform to construct an analytical signal, and performing Fourier transform on the analytical signal to obtain the amplitude of the first frequency and the amplitude of the second frequency. The two amplitudes can be used to characterize the degree of material damage.
[0059] In the process of obtaining the optimal parameter penalty factor α and decomposition number k, the technical solution of the present application first initializes the population parameters: sets the population size N, the number of iterations T, the penalty factor α and the decomposition number k.
[0060] Position X where the dragonfly pollinates i As the penalty factor α and the number of decompositions k, the signal is decomposed by VMD (variational mode decomposition), the envelope entropy of the signal is calculated as the fitness function value, and the optimal individual X is recorded in the process. i ; The expression of envelope entropy is:
[0061]
[0062] Where, E i represents the envelope entropy value of the i-th modal component, N represents the number of signals in each layer, a(j) represents the envelope signal obtained after Hilbert demodulation of the i-th modal component, p j is the normalized form of a(j)
[0063] Dragonfly pollination location Xi The update of will be affected by five factors, including separation coefficient, alignment coefficient, cohesion coefficient, food factor and enemy factor; among them,
[0064] The separation is calculated as follows:
[0065]
[0066] where X t i represents the position of the current individual in evolutionary iteration t, X t j represents the position of the jth individual in the neighborhood at evolutionary iteration t, and N is the number of individuals in the neighborhood. The alignment is calculated as follows:
[0067]
[0068] The aggregation is calculated as follows:
[0069]
[0070] The attractiveness towards a food source is calculated as follows:
[0071]
[0072] where X t i represents the position of the current individual in evolutionary iteration t, X t + is the location of the food source, i.e., the best solution found in the current evolutionary iteration. The repulsive force on the enemy is calculated as follows:
[0073]
[0074] where X t i represents the position of the current individual in evolutionary iteration t, X t - is the enemy's position, i.e., the worst solution found in the current evolutionary iteration.
[0075] The dragonfly's behavior is a combination of these five hypothetical patterns. To update the artificial dragonfly's position in the search space and simulate its movement, two vectors are considered: step size and position. The step or rate vector shows the direction of the dragonfly's movement and is defined as follows:
[0076]
[0077] Where s is the separation coefficient, S t i is the degree of separation of the i-th individual in evolutionary iteration i, a is the alignment coefficient, A ti is the degree of alignment of the i-th individual, c is the cohesion coefficient, C t i is the cohesion degree of the i-th individual, f is the food factor, F t i is the food source of the i-th individual, e is the enemy factor, E t i is the enemy of the i-th individual, w is the inertia weight, and finally, t is the iteration counter of the algorithm. After calculating the step vector, the position vector is calculated as follows:
[0078]
[0079] Where t is the current iteration counter and d is the dimension of the position vector. Levy is calculated as follows:
[0080]
[0081] Where r1 and r2 are two random numbers between 0 and 1, and β is a constant.
[0082] The calculation relationship of σ is as follows:
[0083]
[0084] Partial pollination (self-fertilization) is the development phase of the algorithm. In this pollination, a coefficient specifies the size of the area of growth of each flower around the best found flower. The best solution is the basis of movement, and other solutions will move towards it.
[0085]
[0086] where X t+1 is the position of the pollen in iteration t+1, g * is the best pollen position found in all evolutionary iterations. R is the growth region, which shrinks as the algorithm iterations proceed.
[0087]
[0088] Where t is the algorithm of the current evolutionary iteration, and the iterative loop is performed until the maximum number of loops L is reached and the loop ends, and the optimal individual position, i.e., the penalty factor α and the number of decompositions k, are output.
[0089] The nonlinear ultrasonic detection signal is decomposed using the optimal parameter penalty factor α and decomposition number k of variational modal decomposition to obtain several modal components. The first and second harmonics of the excitation signal are selected as the reconstructed signal y(t).
[0090] The reconstructed signal y(t) is Hilbert transformed to obtain an analytical signal, and the analytical signal is Fourier transformed to obtain the amplitudes of the first and second frequencies. The degree of material damage is characterized according to the amplitudes of the first and second frequencies, thereby completing nonlinear ultrasonic testing.
[0091] Hilport transform:
[0092]
[0093] The analytical signal is constructed from this:
[0094] z(t)=y(t)+iY(t)
[0095] Perform the Fourier transform on z(t):
[0096]
[0097] The amplitude of the first harmonic frequency A1 and the amplitude of the second harmonic frequency A2 are obtained. The degree of material damage can be characterized according to the two amplitudes:
[0098]
[0099] γ is the nonlinear coefficient of the material, which is positively correlated with the damage of the material and can be used to characterize material damage.
[0100] Example 1;
[0101] like Figure 2 As shown in the frequency domain comparison diagram of the nonlinear ultrasonic signal measured experimentally and directly subjected to Fourier transform and the signal after using this method, before using this method, the amplitude of the first harmonic frequency (3MHz) is 1.281e-5, and the amplitude of the second harmonic frequency (6MHz) is 1.254e-6. After using this method to reconstruct the nonlinear ultrasonic signal, the amplitude of the first harmonic frequency is 2.561e-5, and the amplitude of the second harmonic frequency is 2.509e-6, both of which are increased by about twice.
[0102] The above technical solutions only reflect the preferred technical solutions of the technical solutions of the present invention. Any changes that may be made to certain parts thereof by those skilled in the art all reflect the principles of the present invention and fall within the scope of protection of the present invention.
Claims
1. A nonlinear ultrasonic signal processing method based on LEA-VMD, characterized in that: The method includes: Obtain nonlinear ultrasonic detection signals, set parameters of the LEA algorithm, input the nonlinear ultrasonic detection signals into the LEA algorithm, and obtain the optimal parameter penalty factor and decomposition number of variational mode decomposition; The nonlinear ultrasonic detection signal is decomposed by using the optimal parameter penalty factor and the number of decompositions to obtain the modal components and determine the reconstructed signal. The reconstructed signal is Hilbert transformed to obtain the analytical signal. The analytical signal is Fourier transformed to obtain the amplitudes of the first and second frequencies. The amplitudes of the first and second frequencies are used to characterize the degree of material damage.
2. The nonlinear ultrasonic signal processing method based on LEA-VMD according to claim 1, characterized in that: The nonlinear ultrasonic detection signal x(t) is obtained by detecting a component sample with microcrack damage through a nonlinear ultrasonic phased array probe and an ultrasonic signal acquisition instrument; the parameter setting process of the LEA algorithm is: the penalty factor and the number of decompositions in the variational mode decomposition are used as optimization targets, and the envelope entropy is used as the fitness function.
3. The nonlinear ultrasonic signal processing method based on LEA-VMD according to claim 2, characterized in that: The process of solving the optimal parameter penalty factor and the number of decompositions of the variational mode decomposition includes: Initialize population parameters: set population size N, number of iterations T, penalty factor α and number of decompositions k; Position the dragonfly pollinating X i The signal is decomposed by VMD as the penalty factor α and the number of decompositions k, and the envelope entropy of the signal is calculated as the fitness function value, and the optimal individual X is recorded in the process. i ; The expression of envelope entropy is: In the formula, E i represents the envelope entropy value of the i-th modal component, N represents the number of signals in each layer, a(j) represents the envelope signal obtained after Hilbert demodulation of the i-th modal component, p j is the normalized form of a(j); Dragonfly Pollination LocationX i The update of is represented by the step vector, and the movement direction of the dragonfly is defined as follows: Where s is the separation factor, S t i is the degree of separation of the ith individual in evolutionary iteration i, a is the alignment coefficient, A t i is the degree of alignment of the ith individual, c is the cohesion coefficient, C t i is the cohesion degree of the ith individual, f is the food factor, F t i is the food source of the ith individual, e is the enemy factor, E t i is the enemy of the ith individual, w is the inertia weight, and t is the iteration counter of the algorithm; After calculating the step vector, the dragonfly pollination position X i The position vector is calculated as follows: Where t is the current iteration counter and d is the dimension of the position vector; the Levy calculation is as follows: Among them, r1 and r2 are two random numbers between 0 and 1, β is a constant, and the calculation relationship of σ is as follows: Using local pollination as the development stage of the LEA algorithm, we have: Among them, X t+1 is the position of the pollen in iteration t+1, g * is the best pollen position found in all evolutionary iterations, and R is the growth region, which shrinks as the algorithm iterations proceed; Among them, t is the algorithm of the current evolutionary iteration, and the iteration cycle is performed until the maximum number of cycles L is reached and the cycle ends, and the optimal individual position, that is, the penalty factor α and the number of decompositions k, are output.
4. The nonlinear ultrasonic signal processing method based on LEA-VMD according to claim 1, characterized in that: The reconstructed signal is set to y(t), then the Hilport transform of the reconstructed signal is: The constructed analytical signal is: z(t)=y(t)+iY(t) The Fourier transform of z(t) is: The above formula gives the amplitude of the first harmonic frequency A1 and the amplitude of the second harmonic frequency A2. The degree of material damage can be characterized by the two amplitudes: γ is the nonlinear coefficient of the material, and γ is positively correlated with the damage of the material.