CN-121980842-A - Underwater sound transducer design method based on weighted Bayesian optimization and finite element simulation

CN121980842ACN 121980842 ACN121980842 ACN 121980842ACN-121980842-A

Abstract

The invention discloses a design method of an underwater sound transducer based on weighted Bayesian optimization and finite element simulation, which relates to the technical field of electroacoustic transducers, wherein the design method of the underwater sound transducer based on weighted Bayesian optimization and finite element simulation comprises the steps of establishing a finite element simulation model, establishing constraint conditions, normalizing parameters, constructing an objective function, constructing a performance evaluation model, iterating parameters and selecting an optimal solution. Aiming at the underwater acoustic transducer which needs to consider two or more output characteristics (emission current response, resonant frequency, working bandwidth and the like), the invention utilizes the multi-objective optimization idea to dynamically adjust the optimization weight of each shell structure parameter according to different effects of different shell structure parameters on the output characteristics, optimally designs the shell structure parameters of the underwater acoustic transducer, obtains an underwater acoustic transducer model meeting the output characteristics, and shortens the design research and development period of the underwater acoustic transducer.

Inventors

NING QIAN
ZHANG BINGXIN
Shi Tengyang
LI CHUNCHEN
GAO XUAN
Luan Shixu
GAO BING
Zhao Nengtong
WU WENHUA
HE ZHIXING
XU QIANMING
She Yingsen

Assignees

湖南大学

Dates

Publication Date: 20260505
Application Date: 20251218

Claims (10)

1. A design method of an underwater acoustic transducer based on weighted Bayesian optimization and finite element simulation is characterized by comprising the following steps: Taking shell structure parameters of the underwater acoustic transducer as design variables, establishing a mechanical acoustic coupling finite element simulation model of the underwater acoustic transducer, completing the setting of a solid mechanical field and a sound field of the finite element simulation model, and simulating the output characteristics of the underwater acoustic transducer; Establishing constraint conditions, and determining the range of values of shell structure parameters to be optimized of the underwater acoustic transducer; normalizing the obtained shell structure parameter value range, and importing normalized data into a parameter space of a Bayesian algorithm; invoking a finite element simulation model as an objective function of a weighted Bayesian optimization algorithm, regarding the finite element simulation as a black box system capable of acquiring an input-output mapping relation only through a numerical experiment, and carrying out probability modeling on the objective function through a Gaussian process agent model; performing performance evaluation on the output characteristics of the simulated underwater acoustic transducer by a multi-target guide scoring method, and guiding the Bayesian algorithm to optimize the direction by an evaluation result; updating the probability model to obtain a shell structure parameter scheme with higher scores, and carrying the group of parameters back to the simulation model to calculate to obtain new output characteristics; And storing the shell structure parameter solutions with scores higher than the set value into a solution set, iterating repeatedly until the iteration number is set, stopping iteration after the iteration number is reached, and selecting the shell structure parameter solution with the optimal scores as an optimal solution to obtain the optimal shell structure parameters of the underwater sound transducer.
2. The method for designing the underwater acoustic transducer based on weighted Bayesian optimization and finite element simulation as claimed in claim 1, wherein the step of using the shell structure parameters of the underwater acoustic transducer as design variables, establishing a mechano-acoustic coupling finite element simulation model of the underwater acoustic transducer, completing the setting of the solid mechanical field and the sound field of the finite element simulation model, and obtaining the output characteristics of the underwater acoustic transducer by simulation comprises the following steps: inputting shell structure parameters of the underwater acoustic transducer as design variables into a finite element simulation model; According to the actual model and the operation principle of the transducer, an underwater acoustic transducer electromechanical acoustic coupling model is established; setting boundary conditions of a pressure acoustic field and a solid mechanical field of the underwater acoustic transducer; In the coupling boundary of the mechanical and acoustic structure, the acoustic transducer meets two acoustic boundary conditions of sound pressure continuity and speed continuity, and can be arranged as follows: ; Wherein u tt is a structural acceleration matrix, n is a normal unit vector of each node of the radiation surface, ρ c is the density of water, and p t is a sound pressure matrix; For a transducer with symmetry, a 1/8 symmetric model can be built, and all symmetric dividing planes are added to the symmetric boundary of the solid mechanics, wherein the normal vibration speed of the symmetric boundary is 0, namely ; Wherein u is the vibration velocity matrix of the boundary, and n is the unit matrix of the normal direction of the boundary.
3. The method for designing an underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 2, wherein the step of constructing a transducer electromechanical-acoustic coupling model according to an actual transducer model and an operation principle comprises the steps of: a water area is established for the underwater acoustic transducer in the range of 2 times of the maximum size of the underwater acoustic transducer on the outer side of the underwater acoustic transducer by taking the appearance of the underwater acoustic transducer as a reference, and a plurality of grids are drawn in the underwater acoustic transducer and the water area by grids which are not more than 1/6 of the acoustic wave with the minimum wavelength in simulation.
4. The method for designing the underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 2, wherein the setting of boundary conditions of the underwater acoustic transducer pressure acoustic field and the solid mechanical field comprises: In a solid mechanical field, the underwater acoustic transducer satisfies the motion control equation: ; wherein ρ is the density of the solid material, ω is the angular frequency, u is the displacement matrix, and T is the stress matrix; In the underwater acoustic transducer, each group of bars and the shell meet the following mechanical equation on the contact surface of the bars and the shell: ; Wherein, the The work done for the load on the boundary is, For virtual displacement, a is the boundary area and F tot is the total force exerted on the boundary surface, i.e., magnetostrictive force F λ , which satisfies the following equation: ; Wherein A 0 is the cross-sectional area of a single magnetostrictive rod, Is Young's modulus of magnetostrictive material under axial constant magnetic field intensity, The magnetic material is subjected to axial magnetostriction strain, N is the number of turns of a coil, i ac is the coil current, and l c is the length of the coil; in a pressure acoustic field, the underwater acoustic transducer satisfies the acoustic wave equation: ; Where ρ c is the density of water, k is the wave number, k=ω/c is the sound velocity in water, and p t is the sound pressure matrix.
5. The method for designing an underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 4, wherein the setting the boundary conditions of the underwater acoustic transducer pressure acoustic field and the solid mechanical field further comprises: the outer water area is set to be a perfect matching layer and is used for calculating far-field sound field distribution conditions, and a far-field sound field calculation formula is as follows: ; Wherein R is a spatial coordinate vector of an observation point in a far field, p ext (R) is sound pressure calculated at the far field observation point R, S is a coordinate vector of a certain point on an integration surface, p (R) is total sound pressure calculated at the certain point on the integration surface, n is an outward unit normal vector on the integration surface S, and G (R, R) is a free field green function, and the expression is as follows: 。
6. The method for designing an underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 1, wherein establishing constraint conditions to determine a range of values of a shell structure parameter to be optimized for the underwater acoustic transducer comprises: the range of the values of the structural parameters of the shell to be optimized of the underwater acoustic transducer is as follows: 。
7. The method for designing an underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 1, wherein the normalizing the obtained range of values of the shell structure parameters and importing the normalized data into a parameter space of a bayesian algorithm comprises: By passing through The formula normalizes the parameter space.
8. The method for designing an underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 1, wherein the step of calling the finite element simulation model as an objective function of the weighted bayesian optimization algorithm, treating the finite element simulation as a black box system capable of obtaining the input-output mapping relation only through numerical experiments, and the step of probability modeling the objective function through the gaussian process proxy model comprises the steps of: obtaining the output characteristic of the transducer by using the accessed finite element simulation model, wherein the calculation expression of the output characteristic is y=f (x); wherein x is a shell structure parameter set, and f (x) is a black box function represented by a finite element simulation model.
9. The method for designing the underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to claim 1, wherein the performance evaluation of the output characteristics of the underwater acoustic transducer obtained by simulation through the multi-objective guide scoring method to guide the bayesian algorithm optimization direction according to the evaluation result comprises: the scoring formula of the multi-objective guiding scoring method is as follows: ; ; ; ; Wherein P TCR 、P BD and P f are respectively an emission current response score, a bandwidth score and a resonant frequency score, D tar 、B tar and F tar are respectively a target maximum emission current response, a target bandwidth and a target resonant frequency, D max 、B cur and F cur are respectively a current maximum emission current response, a current bandwidth and a current resonant frequency, n 1 、n 2 and n 3 are respectively fractional weights of P TCR 、P BD and P f , and P all is a model output characteristic total score.
10. The method for designing an underwater acoustic transducer based on weighted bayesian optimization and finite element simulation according to any of claims 1 to 9, wherein a kernel function used in the weighted bayesian optimization algorithm is expressed as follows: ; Wherein X is with Two shell structure parameter data matrixes, i is a length scale, Is a structural parameter of the shell Is used for the calculation of the weight of the model (c), The larger is the corresponding shell structure parameter The greater the effect on the output characteristics of the transducer; weight w and gaussian process super-parameters (length scale l, noise variance ) By maximizing edge log likelihood joint optimization: ; Wherein, the Is the marginal likelihood in gaussian process regression, i.e. the probability of observing the output data y given the input data X, K AWK is the weighting kernel matrix calculated based on the weighted input.

Description

Underwater sound transducer design method based on weighted Bayesian optimization and finite element simulation Technical Field The invention relates to the technical field of underwater acoustic transducers, in particular to a design method of an underwater acoustic transducer based on weighted Bayesian optimization and finite element simulation. Background In the field of ocean effective information transmission, the traditional electromagnetic wave, laser, infrared and other signals are transmitted under water to decay too fast, and the effective transmission distance is only within hundreds of meters. Acoustic waves are the only information carrier currently available for long-range, stable transmission in seawater. The underwater acoustic transducer can convert electric energy into acoustic energy, and is key equipment for realizing ocean communication and ocean detection. The giant magnetostrictive underwater acoustic transducer relates to conversion among various energies such as electromagnetic machine sound and the like, and the shell structure of the giant magnetostrictive underwater acoustic transducer directly determines the output characteristic of the giant magnetostrictive underwater acoustic transducer. In recent years, with the expansion of the underwater communication demands, the demands for the output characteristics of the transducer are not limited to a single evaluation index such as the maximum sound source level and the resonance frequency, and a plurality of mutually restricted output characteristic indexes often need to be satisfied at the same time. For example, the IV-shaped curved Zhang Chao magnetostrictive material transducer adopts an elliptical shell configuration, which can remarkably increase the radiation surface area, enlarge the vibration amplitude and realize the design target of a high sound source level, but can also lead to the narrowing of the bandwidth, and the curved transducer has a complex structure, the influence of the shell structure parameters on the output characteristics of the curved transducer is larger, the influence of different shell structure parameters on the output characteristics of the curved transducer is different, the coupling exists between the shell structure parameters, and the problem of designing the curved transducer is solved by searching the optimal shell structure parameters to meet the performance requirements of the bandwidth, the resonant frequency, the sound source level and the like. When designing the structural parameters of the transducer shell with various output characteristic indexes, the structural parameters of the multi-target shell with various output characteristics of the electroacoustic transducer need to be optimally designed. In order to realize the optimal design of the shell structure parameters of the underwater acoustic transducer, firstly, the relation between the shell structure parameters and the output characteristic evaluation indexes is established, and a researcher introduces a BP neural network for constructing a prediction model in the design process, and a Genetic Algorithm (GA) for global optimization so as to establish a transducer optimization model. The design method has the advantages that the optimizing process is fast in convergence and high in searching capability, and the defects that a large number of sample sets are needed for establishing the prediction model, the training process is long in time, the universality of the BP model cannot be guaranteed, and the genetic algorithm can possibly search for an error result. Therefore, there is a need for an underwater acoustic transducer optimization design method that can efficiently handle multi-parameter coupling and multi-objective collisions, and can flexibly adapt to different design preferences, so as to shorten the design development period of the underwater acoustic transducer while ensuring the design accuracy. Disclosure of Invention The invention mainly aims to provide a design method of an underwater acoustic transducer based on weighted Bayesian optimization and finite element simulation, aiming at shortening the design research and development period of the underwater acoustic transducer. In order to achieve the above object, the underwater acoustic transducer design method based on weighted bayesian optimization and finite element simulation provided by the invention comprises the following steps: Taking shell structure parameters of the underwater acoustic transducer as design variables, establishing a mechanical acoustic coupling finite element simulation model of the underwater acoustic transducer, completing the setting of a solid mechanical field and a sound field of the finite element simulation model, and simulating the output characteristics of the underwater acoustic transducer; Establishing constraint conditions, and determining the range of values of shell structure parameters to be optimized of the underwater acousti