A frequency-domain filter-based active absorption method and device for a pusher wave maker

Abstract

The application relates to a kind of active absorption method and equipment of frequency domain filter-based push plate wave generator, and belongs to the field of ocean engineering.The specific process of the method is as follows: a two-dimensional model of active absorption of push plate wave generator is established, and the expression of wave height in front of push plate wave generator is derived.The absorption transfer function of the push plate wave generator of active absorption is derived from the expression of wave height in front of push plate wave generator.The weight distribution combination of all is determined by the target wave to be generated.The error with theoretical displacement under the weight distribution of each combination is calculated by iterative weighted least squares.The optimal solution suitable for the current working condition is determined through error thermal distribution diagram.The application takes the displacement calculated by fitting filter and the theoretical displacement of push plate wave generator as the basis for measuring the effect of active absorption based on weighted iterative least squares, and the optimal filter suitable for target working condition can be automatically fitted through free combination of different frequency bands and weight distribution, without human intervention.

Description

TECHNICAL FIELD

[0001] The application relates to a frequency-domain filter-based active absorption method and device for a push plate wave generator and belongs to the field of ocean engineering. BACKGROUND

[0002] In the research of coastal engineering and water conservancy engineering, physical model wave experiment is an important means for studying problems. However, due to the limitation of the size of the laboratory, the wave will form multiple reflections between the pool boundary, the structure and the wave plate, that is, secondary reflection waves will be generated, so that the whole wave field is chaotic. Therefore, a technology for absorbing secondary reflection waves is needed.

[0003] In view of the existing problems, many scholars have carried out a lot of research on this, and have achieved certain results. The active absorption theory was first proposed by Milgram in 1970. Bullock and Murton further developed the theory in 1989 and proposed a non-reflection system for a two-dimensional wedge-shaped wave generator. The system measures the wave height in front of the wave plate and converts it into a motion control signal for the wave plate by using a filter. Hirakuchi et al. proposed an improved active absorption control method in 1990. The method directly installs a wave height sensor on the wave plate, effectively overcoming the phase lag problem of the feedback wave height in front of the plate caused by the position deviation of the sensor. The method significantly simplifies the structure of the control algorithm and shows good absorption effect, and is still widely used around the world. Schaeffer et al. developed an active absorption control system at the Danish Hydraulic Institute (DHI) in 1994. The system is based on the feedback of the wave height sensor installed in front of the wave plate and fully considers the influence of the non-propagating mode. In the time-domain implementation of the algorithm, an infinite impulse response (IIR) digital filter is designed by using the best approximation method, which can effectively approximate the theoretical transfer function in a wide frequency range. In the same year, Christensen and Frigaard (1994) proposed another implementation approach, which separates the incident wave and the reflected wave by measuring the wave height signals at two different positions in front of the wave plate, and uses the separated incident wave component as the control input to design a finite impulse response (FIR) active absorption filter, which also realizes effective active absorption control. Yang Hongqi et al. proposed a recursive reweighted least squares algorithm based on Gauss-Newton. However, the least squares algorithm is prone to uniformization error and lacks pertinence. There are still defects in the research on filters with good absorption effect in a specific frequency band, and such research needs to be improved. SUMMARY

[0004] In view of the problems existing in the prior art, the application provides a frequency-domain filter-based active absorption method for a wave generator.

[0005] The technical scheme of the application comprises the following steps:

[0006] A kind of active absorption method of frequency domain filter-based push plate wave generator, comprising the following steps:

[0007] S1, the two-dimensional model of push plate wave generator active absorption is established, and the expression of push plate wave generator wave height before board is deduced.

[0008] S2, the expression of push plate wave generator wave height before board in S1 is used to deduce the absorption transfer function F of push plate wave generator active absorption.

[0009] S3, the total weight distribution combination is determined by the target wave to be generated.

[0010] S4, the error with the theoretical displacement under the weight distribution of each combination is calculated using iterative weighted least squares.

[0011] S5, the optimal solution suitable for the current working condition is determined by the error thermal distribution map.

[0012] Further, in S1, the two-dimensional model of push plate wave generator active absorption is established: the water depth is h, one side of the two-dimensional model is push plate wave generator, and the movement amplitude of push plate wave generator is , and the wave height instrument is fixed on the push plate wave generator and moves with it. The other side of the two-dimensional model is a straight wall, which is used to simulate the full reflection of wave. The length of the whole two-dimensional model is C. First, the push plate wave generator movement generates a traveling wave , and propagates in the direction of travel, when the traveling wave meets the straight wall, a reflected wave is generated, and then propagates in the opposite direction of the travel direction. Then, the reflected wave meets the push plate wave generator and generates a secondary reflected wave , and the secondary reflected wave is superimposed with the traveling wave to form a new traveling wave.

[0013] In order to eliminate the secondary reflected wave, the push plate wave generator needs to generate a compensation wave with equal amplitude and 180° phase difference with the secondary reflected wave. When the push plate wave generator works in absorption mode, the movement of the push plate wave generator is divided into two parts: the movement for generating traveling wave and the movement for absorbing secondary reflected wave, and the sum of the two is the total movement of push plate wave generator active absorption. At this time, the wave height instrument before the push plate wave generator obtains the push plate wave generator wave height before board , which is the superposition of the elevation of each wave component before the push plate wave generator board, and is written as follows:

[0014] ​​(1.1)

[0015] In the above formula, is the traveling wave corresponding to the non-propagating mode; is the motion of the secondary reflection wave absorbed by the wave maker the compensation wave generated, with the same wave height as the secondary reflection wave wave front and a phase difference of 180°. is the compensation wave corresponding to the non-propagating mode. If the reflection coefficient of the wave-making plate of the wave maker is 1, then . Transforming the above formula (1.1) into the frequency domain of the corresponding signal to obtain:

[0016] (1.2)

[0017] In the above formula, are the complex amplitudes of the Fourier transform of the time domain , , , , , , corresponding to the signal respectively.

[0018] Further, in S2, the expression of the absorption transfer function F of the wave maker is derived from formula (1.2):

[0019] (1.3)

[0020] wherein, represents the motion of the secondary reflection wave absorbed the complex amplitude of the Fourier transform, represents the complex amplitude of the Fourier transform of the total motion actively absorbed by the wave maker, is the absorption transfer function, and the expression is:

[0021] (1.4)

[0022] In the above formula, , is the hydrodynamic transfer function, represents the imaginary unit, and n represents the number of non-propagating modes.

[0023] Further, in S3, the spectral peak circular frequency of the target wave is , and the following weight distribution is defined:

[0024] (1.5)

[0025] In the above formula, is the initial weight distribution to be iterated in S4, is the left boundary, is the right boundary, is the specific weight given in the interval Let , and be freely combined to determine all weight distribution combinations.

[0026] Further, in S4, the specific steps are as follows:

[0027] Step 4.1, define the error :

[0028] When there is only an incident wave and no reflected wave in the actual situation, in the ideal state, there should be in the active absorption mode, and the movement of the wave maker is only , at this time the generated wave conforms to the first-order linear wave theory. According to formula (1.3), when the frequency domain input is , the frequency domain output is .

[0029] When the time domain input is , the time domain output obtained is , since at this time, so at this time. According to the first-order linear wave theory, the input and the input in the time domain can be calculated from the theoretical value. Therefore, if the designed filter approximates the absorption transfer function F, when the input in the filter is , the output of the filter should approximate . Comparing and can determine the active absorption effect of the filter. Therefore, first define the error between and :

[0030] (1.7)

[0031] In the formula, represents the total length of the sequence input or output in the time domain, represents the th data of , and represents the th data of ; when the calculated error The smaller, the better the active absorption effect.

[0032] Step 4.2, weighted least square calculation is carried out.

[0033] The absorption transfer function is formula (1.4), and the corresponding amplitude response and phase response at different circular frequencies are known. To realize the time-domain active absorption function, a system function F is designed through a filter , which is as equal to F as possible. Since different circular frequencies are calculated , , and F is different, in order to unify the description, F is replaced by .

[0034] First, the least square method is used to define the least square error :

[0035] (1.8)

[0036] In the formula, The expression is as follows:

[0037] (1.9)

[0038] In the formula, is the numerator coefficient, is the denominator coefficient, M is the numerator order, and N is the denominator order; for the convenience of subsequent calculation, here , .

[0039] The problem is further described as:

[0040] (1.10)

[0041] In order to solve formula (1.10), formula (1.8) is first simplified:

[0042] (1.11)

[0043] For IIR filter , the term is selected to be ignored, and the above formula is linearized. For a circular frequency , there is:

[0044] (1.12)

[0045] In the formula, respectively represent the corresponding at a circular frequency .

[0046] The above equation can be written in matrix form as:

[0047] (1.13)

[0048] where K denotes the total number of circular frequencies, and substituting equation (1.13) into equation (1.10) is a standard least squares problem.

[0049] Equation (1.13) can be further simplified as:

[0050] (1.14)

[0051] where correspond to the first term on the left side of equation (1.13), and the first, second, and third terms on the right side of equation (1.13), respectively.

[0052] Therefore, the parameter vector is obtained by solving:

[0053] (1.15)

[0054] When the weight vector is assigned to the given frequency band, equation (1.15) is transformed as:

[0055] (1.16)

[0056] where is a simplified form of .

[0057] The initial weight distribution given in S3 is used for iteration. First, the parameter vector is initialized, and the weight distribution is given, with the weight of other frequency bands being 1. The denominator in equation (1.11) is ignored, converting the non-linear equation into a linear equation, which can be directly solved using the least squares method to obtain the initial parameter vector . The system function corresponding to the initial parameter vector is further solved , and the residual vector corresponding to the current system function and the theoretical transfer function under the weight distribution at the pth iteration is calculated as :

[0058] (1.17)

[0059] where is the system function at the kth circular frequency at the pth iteration, is the weight at the kth circular frequency;

[0060] The Jacobian matrix corresponding to the residual is further calculated as :

[0061] (1.18)

[0062] where, is the first n columns of the Jacobian matrix, is the last m columns of the Jacobian matrix. and together form the Jacobian matrix . The sensitivity of the residual to the parameter variation can be described to determine the search direction for the next iteration .

[0063] (1.19)

[0064] By searching the value of the step size factor , such that:

[0065] (1.20)

[0066] where, is the parameter vector at the pth iteration; Q is the objective function, which measures the error between the system function and the theoretical transfer function. represents the system function calculated when the parameter vector is ; represents the system function calculated when the parameter vector is ;

[0067] This ensures that the objective function is reduced at each iteration, ensuring global convergence of the algorithm. Starting from , if the objective function does not decrease, the step size is reduced according to until the condition is met.

[0068] Finally, update the parameter vector:

[0069] (1.21)

[0070] Update the parameters in the descending direction, expecting to approximate the optimal solution.

[0071] The iteration termination condition is:

[0072] (1.22)

[0073] Take , when the convergence condition is met, output the optimal parameter vector .

[0074] (1.23)

[0075] Meanwhile, the denominator polynomial needs to be stabilized to ensure that the system is stable (all poles are within the unit circle). Solve all roots of the denominator . That is, convert it to domain ( ). Then , solve all roots of the equation . Reflect each unstable root , and let

[0076] (1.24)

[0077] Determine the new denominator polynomial from the updated roots:

[0078] (1.25)

[0079] where is the Nth denominator coefficient solved from the updated roots; is the denominator expression after updating the roots.

[0080] Further, the parameter vector is updated as:

[0081] (1.26)

[0082] This parameter vector is the final solution. Calculate the final solution corresponding to each weight distribution respectively. Use the final solution to solve the filter displacement, and further solve the filter displacement error and the theoretical displacement error.

[0083] Further, in S5, through the calculation in S4 above, an error distribution heat map is obtained, and the weight distribution corresponding to the point with the smallest error in the error distribution heat map is the optimal weight distribution suitable for the target working condition.

[0084] A push plate wave generator active absorption device based on a frequency domain filter, comprising: at least one processor, and a memory in communication connection with the at least one processor;

[0085] Wherein the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned push plate wave generator active absorption method based on a frequency domain filter.

[0086] The beneficial effects of the present application are: based on the weighted iterative least square method, the displacement calculated by the fitting filter and the theoretical displacement of the push plate wave generator are taken as the basis for measuring the good and bad of the active absorption effect, and through the free combination of different frequency bands and weight distributions, the optimal filter suitable for the target working condition can be automatically fitted without human intervention. ​

[0087] Compared to traditional pusher-plate wave generator active absorption filters, the pusher-plate wave generator active absorption filter designed in this invention is more targeted. Different weight distributions are designed for different target operating conditions to achieve the optimal absorption effect for those conditions. Attached Figure Description

[0088] Figure 1 Calculation flowchart.

[0089] Figure 2 A schematic diagram of a two-dimensional model of an active absorption mechanism for a pusher-plate wave generator.

[0090] Figure 3 Error distribution heatmap.

[0091] Figure 4 A comparison of the amplitudes of theoretical and fitted values ​​(Tp=1.0s).

[0092] Figure 5 A comparison of the amplitudes of theoretical and fitted values ​​(Tp=1.5s).

[0093] Figure 6 A comparison of the magnitudes of theoretical and fitted values ​​(Tp=2.0s).

[0094] Figure 7 Comparison of theoretical displacement and filtered displacement (Tp=1.0s).

[0095] Figure 8 Comparison of theoretical displacement and filtered displacement (Tp=1.5s).

[0096] Figure 9 Comparison of theoretical displacement and filtered displacement (Tp=2.0s). Detailed Implementation

[0097] An active absorption method for wave generators based on frequency domain filters includes the following steps, such as... Figure 1 As shown:

[0098] S1. Establish a two-dimensional model of the active absorption of the pusher plate wave generator, and derive the wave height in front of the pusher plate wave generator. The expression.

[0099] S2, Wave height in front of the wave generator plate from S1 The expression is used to derive the absorption transfer function F of the active absorption pusher wave generator.

[0100] S3. Determine all weight distribution combinations based on the target wave to be generated.

[0101] S4. Calculate the error between the weight distribution and the theoretical displacement for each combination using iterative weighted least squares.

[0102] S5, determining the optimal solution suitable for the current working condition through the error thermodynamic distribution map.

[0103] Further, in S1, a two-dimensional model of active absorption of the push plate wave maker is established, as shown in the figure: Figure 2 The water depth is h, one side of the two-dimensional model is the push plate wave maker, and the movement amplitude of the push plate wave maker is The wave height instrument is fixed on the push plate wave maker and moves with it. The other side of the two-dimensional model is a straight wall for simulating total reflection of the wave. The length of the entire two-dimensional model is C. First, the push plate wave maker movement generates a traveling wave , which propagates in the direction of travel, and when the traveling wave meets the straight wall, a reflected wave is generated, which then propagates in the opposite direction of the travel direction. Then, the reflected wave meets the push plate wave maker and generates a secondary reflected wave , and the secondary reflected wave is superimposed with the traveling wave to form a new traveling wave.

[0104] In order to eliminate the secondary reflected wave, the push plate wave maker needs to generate a compensation wave with the same amplitude as the secondary reflected wave but with a phase difference of 180°. When the push plate wave maker works in the absorption mode, the movement of the push plate wave maker is divided into two parts: the movement that generates the traveling wave and the movement that is used to absorb the secondary reflected wave, and the sum of the two is the total movement of the push plate wave maker in active absorption . At this time, the wave height in front of the push plate wave maker obtained by the wave height instrument in front of the push plate wave maker is , which is the superposition of the elevations of each wave component in front of the push plate wave maker, and can be written as follows:

[0105] (1.1)

[0106] In the above formula, is the corresponding non-propagating mode of the traveling wave . is the compensation wave generated by the movement of the push plate wave maker to absorb the secondary reflected wave, which has the same wave height as the wave surface of the secondary reflected wave and a phase difference of 180°. is the corresponding non-propagating mode of the compensation wave . If the reflection coefficient of the wave plate of the push plate wave maker is 1, then . Transforming the above formula (1.1) into the frequency domain of the corresponding signal obtains:

[0107] (1.2)​

[0108] In the above formula, are the complex amplitudes of Fourier transform of the time domain 、 、 、 、 、 、 corresponding to the signal respectively.

[0109] Further, in S2, the expression of the absorption transfer function F of the wave maker is derived from formula (1.2):

[0110] (1.3)

[0111] wherein, represents the motion of absorbing the secondary reflected wave the complex amplitude of Fourier transform, represents the complex amplitude of Fourier transform of the total motion of the wave maker actively absorbed, F is the absorption transfer function, and the expression is:

[0112] (1.4)

[0113] In the above formula, , is the hydrodynamic transfer function, represents an imaginary unit, and n represents the number of non-propagation modes.

[0114] Further, in S3, it is assumed that the spectral peak circular frequency of the target wave is Since the period T commonly used in laboratory wave making is in the range of (0.5s~5s), and , the commonly used circular frequency range is: (1.3rad / s~12.6rad / s). Therefore, the weight distribution is defined as follows:

[0115] (1.5)

[0116] In the above formula, is the initial weight distribution to be iteratively calculated in S4, is the left boundary, is the right boundary, is a specific weight given on the interval Let 、 and be freely combined to determine all the weight distribution combinations.

[0117] Further, in S4, the specific steps are as follows:

[0118] Step 4.1, define error :

[0119] When there is only incident wave but no reflected wave in the actual situation, in the ideal state, there should be in the active absorption mode, and the movement of the push plate wave maker is only At this time, the generated wave conforms to the first-order linear wave theory. According to equation (1.3), when the frequency domain input is , the frequency domain output is .

[0120] When the time domain input is , the obtained time domain output is , and since at this time, so at this time. According to the first-order linear wave theory, the input and the input in the time domain can be calculated from the theoretical value. Therefore, if the designed filter approximates the absorption transfer function F, when the input in the filter is , the output of the filter should approximate . Comparing and can determine the active absorption effect of the filter. Therefore, first define the error between and :

[0121] (1.7)

[0122] In the formula, represents the total length of the input or output sequence in the time domain, represents the th data of , and represents the th data of ; the smaller the calculated error , the better the active absorption effect.

[0123] Step 4.2, perform weighted least squares calculation.

[0124] The absorption transfer function is equation (1.4), and the corresponding amplitude response and phase response at different circular frequencies are known. To achieve the time domain active absorption function, a system function needs to be designed through the filter, so that it is equal to F as much as possible. Since different circular frequencies are calculated , Different, and then F different, in order to describe the following with Instead of F.

[0125] First of all, the least squares method to specify the least squares error :

[0126] (1.8)

[0127] In the formula, Expression as follows:

[0128] (1.9)

[0129] In the formula, The numerator coefficient, The denominator coefficient, M is the numerator order, N is the denominator order; in order to simplify the calculation, here let , .

[0130] The problem is further described as:

[0131] (1.10)

[0132] In order to solve equation (1.10) first equation (1.8) is simplified:

[0133] (1.11)

[0134] For IIR filter , choose to ignore The term, linearization of the above formula. For a certain circular frequency There are:

[0135] (1.12)

[0136] In the formula, Respectively, a certain circular frequency Corresponding to .

[0137] The above formula is written in matrix form, expressed as:

[0138] (1.13)

[0139] In the formula, K represents the K circular frequency, equation (1.13) into equation (1.10) is the standard least squares problem.

[0140] Equation (1.13) is further simplified as:

[0141] (1.14)

[0142] where correspond to the first term on the left side of the equation, the first, second and third terms on the right side of the equation in formula (1.13) respectively.

[0143] Therefore, the parameter vector is solved as:

[0144] (1.15)

[0145] When the weight vector is given in a certain frequency band, formula (1.15) is transformed as:

[0146] (1.16)

[0147] where, is the short form of .

[0148] The initial weight distribution given in S3 is used for iteration. First, the parameter vector is initialized, the weight distribution is given, and the weight of other frequency bands is 1. And the influence of the denominator in formula (1.11) is ignored, which is transformed from a nonlinear equation to a linear equation, so that the least square method can be directly used to solve, and the initial parameter vector is obtained. Further, the system function corresponding to the initial parameter vector is solved, and then the residual vector corresponding to the current system function and the theoretical transfer function under this weight distribution at the pth iteration is calculated as :

[0149] (1.17)

[0150] where, is the system function at the kth circular frequency at the pth iteration, is the weight at the kth circular frequency;

[0151] Further, the Jacobian matrix corresponding to the residual is calculated as :

[0152] (1.18)

[0153] where, is the first n columns of the Jacobian matrix, is the last m columns of the Jacobian matrix; and together constitute the Jacobian matrix . The Jacobian matrix can describe the sensitivity of the residual to the change of the parameter, and is used to determine the search direction of the next iteration.

[0154] (1.19)

[0155] By searching the value of step factor , so that:

[0156] (1.20)

[0157] Where, is the parameter vector at the pth iteration; Q is the objective function, which measures the error between the system function and the theoretical transfer function. represents the system function calculated when the parameter vector is ; represents the system function calculated when the parameter vector is ;

[0158] This ensures that each iteration reduces the objective function, ensuring global convergence of the algorithm. Starting from , if the objective function does not decrease, reduce the step size according to , until the condition is met.

[0159] Finally, update the parameter vector:

[0160] (1.21)

[0161] Update the parameters in the descending direction, expecting to approximate the optimal solution.

[0162] The iteration termination condition is:

[0163] (1.22)

[0164] Take , when the convergence condition is met, output the optimal parameter vector .

[0165] (1.23)

[0166] Meanwhile, the denominator polynomial needs to be stabilized to ensure that the system is stable (all poles fall within the unit circle). Solve all roots of the denominator . That is, convert it to the domain ( ). Then , solve all roots of the equation. Reflect each unstable root , and let:

[0167] (1.24)

[0168] The new denominator polynomial is determined by the updated root:

[0169] (1.25)

[0170] wherein, is the Nth denominator coefficient solved by the updated root; is the denominator after the updated root. is the domain expression.

[0171] The parameter vector is further updated as:

[0172] (1.26)

[0173] The parameter vector is the final solution. The final solution corresponding to each weight distribution is calculated respectively. The filter displacement is solved by using the final solution, and the error between the filter displacement and the theoretical displacement is further solved.

[0174] Further, in S5, the error distribution heat map is obtained by the calculation in S4. The weight distribution corresponding to the point with the minimum error in the error distribution heat map is the optimal weight distribution suitable for the target working condition. For example, the value corresponding to (4.8, 12.6) in the heat map is 0.00022. This represents that when the weight is 10, 60, 110…, 960, the corresponding is calculated respectively in the interval of circular frequency 4.8 rad / s~12.6 rad / s. The minimum is 0.00022.

[0175] Further, in order to better illustrate the advantages of the present application, Figure 4 , Figure 5 and Figure 6 respectively give the frequency response comparison diagram of the theoretical value and the fitted value of the active absorption transfer function of the push plate wave generator under the working condition of the spectral peak period T p =1.0s, 1.5s, 2.0s, and the effective wave height 0.04m. The shaded part in the figure is the range covered by 0.2 on the left and right of the spectral peak frequency. From the figure, it can be seen that no matter what the spectral peak period is, the theoretical value and the fitted value in the shaded part are in good agreement, which shows that the present method has high absorption effect in a specific frequency band. In addition, Figure 7 , Figure 8 and Figure 9 also give the comparison diagram of the theoretical displacement and the filter displacement under the above three working conditions. From the figure, it can be easily seen that both of them are in good agreement under the three working conditions, which further shows that the active absorption effect of the present application is good.

Claims

1. An active absorption method for a pusher-plate wave generator based on a frequency domain filter, characterized in that, Includes the following steps: S1. Establish a two-dimensional model of the active absorption of the pusher plate wave generator, and derive the wave height in front of the pusher plate wave generator. The expression; S2, Wave height in front of the wave generator plate from S1 The expression is used to derive the absorption transfer function F of the active absorption pusher wave generator; S3. Determine all weight distribution combinations based on the target wave to be generated; In S3, let the spectral peak angular frequency of the target wave be... The following weight distribution is defined: (1.5) ； In the above formula, This is the initial weight distribution for S4 to be iteratively calculated. The left boundary The right boundary In the interval The specific weights given above; let , and Free combination, determining all possible combinations of weight distributions; S4. Calculate the filter displacement and theoretical displacement error under the weight distribution of each combination using iterative weighted least squares. S5. Determine the optimal solution applicable to the current working conditions using the error heat map.

2. The active absorption method for a pusher-plate wave generator based on a frequency domain filter according to claim 1, characterized in that, In S1, a two-dimensional model of the active absorption by the pusher wave generator is established: the water depth is h, one side of the two-dimensional model is the pusher wave generator, and the amplitude of the pusher wave generator's motion is... The wave height meter is fixed to the pusher wave generator and moves with it; the other side of the two-dimensional model is a straight wall to simulate total internal reflection of waves; the entire two-dimensional model has a length of C; first, the pusher wave generator moves to generate traveling waves. And propagates in the direction of travel, when the traveling wave A reflected wave is generated when it encounters a straight wall. Then it propagates in the opposite direction of travel; then, a reflected wave... When it encounters the pusher plate wave generator, it will generate a secondary reflected wave. Secondary reflected wave With traveling waves Superposition forms a new traveling wave; To eliminate the secondary reflected wave, the pusher wave generator needs to produce a compensating wave with the same amplitude but a 180° phase difference from the secondary reflected wave. When the pusher wave generator is operating in absorption mode, its motion is divided into two parts: the motion that generates the traveling wave. and motion used to absorb secondary reflected waves The sum of the two constitutes the total motion actively absorbed by the pusher wave generator. At this time, the wave height obtained by the wave height meter in front of the pusher wave generator plate is... The superposition of the elevations of each wave component in front of the pusher wave generator plate is expressed as follows: (1.1) ； In the above formula, It is a traveling wave The corresponding non-propagation mode; The motion of the pusher plate wave generator to absorb secondary reflected waves The resulting compensation wave, With secondary reflected wavefront The wave heights are equal, but the phase difference is 180°; To compensate for the wave The corresponding non-propagation mode; assuming the reflection coefficient of the wave-generating plate of the pusher wave generator is 1, then Transforming equation (1.1) into the frequency domain of the corresponding signal yields: (1.2) ； In the above formula, These correspond to the time domain of the signal. , , , , , , The complex amplitude value after Fourier transform.

3. The active absorption method for a pusher-plate wave generator based on a frequency domain filter according to claim 2, characterized in that, In S2, the expression for the absorption transfer function F of the pusher wave generator is derived from equation (1.2): (1.3) ； in, Indicates the motion of absorbing secondary reflected waves The complex amplitude of the Fourier transform, This indicates that the pusher wave generator actively absorbs the total motion. The complex amplitude of the Fourier transform, The absorption transfer function is expressed as follows: (1.4) ； In the above formula, , For hydrodynamic transfer function, The imaginary unit is represented by , and n represents the number of non-propagating modes.

4. The active absorption method for a pusher-plate wave generator based on a frequency domain filter according to claim 1, characterized in that, In S4, the specific steps are as follows: Step 4.1, Define the error : When there is only an incident wave and no reflected wave in reality, the active absorption mode exists. The movement of the pusher wave generator is only The waves generated at this time conform to the first-order linear wave theory; According to equation (1.3), when the frequency domain input is At that time, the frequency domain output is ; When the input domain is When, the obtained time-domain output is Because at this time Therefore, at this time According to the first-order linear wave theory, the input in the time domain... and input All of these can be calculated from theoretical values; therefore, if the designed filter approximates the absorption transfer function F, the input to the filter is... At that time, the filter output Approaching ;contrast and This allows us to determine the active absorption effect of the filter; therefore, we first define... and Error between: (1.7) ； In the formula, This represents the total length of the input or output sequence in the time domain. express The One data point, express The One data point; when the calculated error The smaller the value, the better the active absorption effect; Step 4.2: Perform weighted least squares calculation; The absorption transfer function is given by equation (1.4), and the amplitude and phase responses at different angular frequencies are known. To achieve active absorption in the time domain, a system function needs to be designed using a filter. Make it as equal to F as possible; due to different angular frequencies Calculated , The differences lead to different F values, which will be used later. Replace F; First, the least squares method is used to define the least squares error. : (1.8) ； In the formula, The expression is as follows: (1.9) ； In the formula, The numerator coefficient, Let M be the denominator coefficient, M be the numerator order, and N be the denominator order; let , ; The problem is further described as follows: (1.10) ； To solve equation (1.10), we first simplify equation (1.8): (1.11) ； For IIR filters Choose to ignore The term linearizes the above equation; for a certain angular frequency There are places where: (1.12) ； In the formula, Representing a certain circular frequency Corresponding to ; The above equation can be written in matrix form as follows: (1.13) ； In the formula, K represents a total of K angular frequencies. Substituting equation (1.13) into equation (1.10) yields the standard least squares problem. Equation (1.13) can be further simplified to: (1.14) ； In the formula These correspond to the first term matrix on the left side of the equation and the first, second, and third terms matrix on the right side of the equation (1.13), respectively. Therefore, the parameter vector The solution yields: (1.15) ； When a weight vector is assigned to a given frequency band Then, equation (1.15) transforms into: (1.16) ； In the formula, for The abbreviated form; Iteration is performed using the initial weight distribution given in S3; first, the parameter vector is initialized, given the weight distribution. The weights for other frequency bands are 1; and the influence of the denominator in equation (1.11) is ignored, transforming it from a nonlinear equation into a linear equation. This allows for direct solution using the least squares method to obtain the initial parameter vector. Further solve for the system function corresponding to the initial parameter vector. Then, at the p-th iteration, the residual vectors corresponding to the current system function and the theoretical transfer function under this weight distribution are calculated as follows: : (1.17) ； In the formula, Let k be the system function at the k-th circular frequency during the p-th iteration. The weight at the k-th circular frequency; Further calculate the Jacobian matrix corresponding to the residuals. : (1.18) ； In the formula, The first n columns of the Jacobian matrix, The last m columns of the Jacobian matrix; and Together they form the Jacobian matrix ; It can describe the sensitivity of residuals to parameter changes, and is used to determine the search direction for the next iteration. ; (1.19) ； By searching for step size factor The value of is such that: (1.20) ； in, Let be the parameter vector at the p-th iteration; Q is the objective function, used to measure the error between the system function and the theoretical transfer function. The parameter vector is The system function calculated at that time; The parameter vector is The system function calculated at that time; This ensures that the objective function decreases in each iteration, guaranteeing the global convergence of the algorithm; from Initially, if the objective function does not decrease, then follow... Reduce the step size until the condition is met; Finally, update the parameter vector: (1.21) ； Update the parameters along the descent direction in order to approximate the optimal solution; The iteration termination condition is: (1.22) ； Pick When the convergence condition is met, the optimal parameter vector is output. ; (1.23) ； Simultaneously, the denominator polynomial needs to be stabilized to ensure the system is stable; solving the denominator... All roots; that is, transform them to domain, ;but Solve for all roots of this equation. For each unstable root To perform reflection, let: (1.24) ； The updated roots determine the new denominator polynomial: (1.25) ； In the formula, The coefficient of the Nth denominator is obtained from the updated roots; To update the denominator after the root Domain expression; The parameter vector is then updated as follows: (1.26) ； This parameter vector is the final solution; calculate the final solution for each weight distribution; use the final solution to solve for the filter displacement, and further solve for the error between the filter displacement and the theoretical displacement.

5. The active absorption method for a pusher-plate wave generator based on a frequency domain filter according to claim 4, characterized in that, In S5, the error distribution heatmap is obtained through the calculation in S4 above. The weight distribution corresponding to the point with the smallest error in the error distribution heatmap is the optimal weight distribution suitable for the target working condition.

6. An active absorption device for a pusher-plate wave generator based on a frequency domain filter, characterized in that, include: At least one processor, and a memory communicatively connected to said at least one processor; The memory stores instructions executable by the processor, which are executed by the at least one processor to cause the at least one processor to perform the active absorption method of a pusher wave generator based on a frequency domain filter as described in any one of claims 1-5.

Images