CN-122020855-A - Underwater boat body optimal design method based on continuous accompanying method

CN122020855ACN 122020855 ACN122020855 ACN 122020855ACN-122020855-A

Abstract

The invention discloses an underwater hull optimal design method based on a continuous accompanying method, which relates to the technical field of underwater hull optimal design and comprises the following steps of establishing a parameterized geometric model of an underwater hull, defining an optimal objective function containing cavitation inhibition performance indexes, executing multiphase flow forward flow field solution on the current geometric model to obtain flow field state data containing phase volume fraction distribution, executing cavitation singularity diagnosis and management flow, dynamically generating regularization calculation instructions based on physical phase change characteristics and numerical gradient characteristics in the flow field state data and according to convergence states of cavitation singular areas, introducing local mathematical smoothing terms into discrete equations in the cavitation singular areas to obtain sensitivity gradients subjected to stabilization treatment, and circularly executing steps S2 to S5 until the preset optimal convergence criteria are met.

Inventors

GAO XULIANG
ZHU XIAO

Assignees

上海桓领信息科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The underwater hull optimal design method based on the continuous accompanying method is characterized by comprising the following steps of: s1, establishing a parameterized geometric model of an underwater boat body, and defining an optimized objective function containing cavitation inhibition performance indexes; S2, carrying out multiphase flow forward flow field solution on the current geometric model to obtain flow field state data containing phase volume fraction distribution; S3, executing cavitation singular diagnosis and control flow, identifying cavitation singular areas in a flow field based on physical phase change characteristics and numerical gradient characteristics in the flow field state data, and dynamically generating regularization calculation instructions according to convergence states of the cavitation singular areas; S4, responding to the regularization calculation instruction by an accompanying solver, reconstructing and solving an accompanying equation by adopting a region self-adaptive regularization strategy based on physical feature perception, and introducing a local mathematical smoothing term into a discrete equation in the cavitation singular region so as to obtain a sensitivity gradient subjected to stabilization treatment; And S5, updating design variables of the parameterized geometric model based on the sensitivity gradient, and circularly executing the steps S2 to S5 until a preset optimization convergence criterion is met.
2. The method for optimizing design of an underwater hull based on the continuous accompanying method according to claim 1, wherein in step S3, the identifying cavitation singular areas in the flow field specifically includes: calculating the phase volume fraction gradient modulus of the full-field grid unit (∇ phi), and marking the unit with the modulus larger than the physical characteristic threshold as a physical phase change interface unit; Estimating a local norm I S adj I of a right-end term source item S adj of the accompanying equation, and marking a unit with a norm value larger than a numerical characteristic threshold as a potential numerical singular point unit; and judging the grid cell set which simultaneously meets the two marking conditions as the cavitation singular region, and endowing the cavitation singular region with an internal state identifier of the high-risk singular region.
3. The method for optimizing design of an underwater hull based on a continuous accompanying method according to claim 2, wherein the dynamically generating regularized calculation instruction according to the convergence state of the cavitation singular region comprises: monitoring a pressure residual error convergence curve of grid cells in the high-risk singular region in the last flow field iterations; analyzing the convergence curve, and extracting quantization indexes including average convergence rate eta and amplitude sigma of residual oscillation; Inputting the quantization index (eta, sigma) into a preset decision function F, and mapping and outputting a local regularized intensity coefficient lambda (x) matched with the current pathological degree of the region, wherein the function F is configured to monotonically increase the lambda value along with eta decrease or sigma increase; And generating the regularization calculation instruction based on the grid identification of the high-risk singular region and the corresponding lambda (x) coefficient.
4. The method for optimizing design of an underwater hull based on a continuous accompanying method according to claim 3, wherein the decision function F is implemented by a strategy based on a two-dimensional lookup table and bilinear interpolation, and specifically comprises: a discrete empirical value table with (eta, sigma) as the coordinate axis is pre-established, the nodes in the table store the reference lambda value, and in operation, grids are positioned in the table according to the calculated (eta, sigma) and interpolation is carried out so as to quickly obtain the adaptive lambda (x).
5. The method for optimizing design of an underwater hull based on a continuous accompanying method according to claim 4, wherein in step S4, the specific way of reconstructing and solving the accompanying equation by using the region adaptive regularization strategy based on physical feature perception is as follows: Forming a correction objective function J by adding a regularization functional term with spatial weighting to a standard objective function J, and performing variation based on J to derive the reconstructed accompanying equation, wherein the expression of the correction objective function J is: ; Where Ω is the computational domain, λ (x) is the local regularized intensity coefficient, φ (x) is the phase volume fraction, ψ (x) is the companion variable vector corresponding to the flow field conservation variable vector, R (φ) is the interface activation function, |·| 2 represents the L2 norm square of the vector.
6. The method for optimizing design of an underwater hull based on a continuous concomitant method according to claim 5, wherein said interface activation function R (Φ) is a smooth function continuously conductive in the vicinity of the phase transition interface: The function value is approximately zero in the pure liquid phase and the pure vapor phase, and takes a positive value in the mixed phase region.
7. The method for optimizing design of an underwater hull based on the continuous accompanying method according to claim 5, wherein in the step S4, the reconstructed accompanying equation is solved by an iterative strategy including delay correction and mixing accuracy calculation, and specifically comprises the following steps: s4.1, constructing a companion linear system with a high diagonal dominance by using a first group of regularization parameters lambda 1 (x), and solving to obtain an initial companion solution field psi 0 ; S4.2, taking the psi 0 as an iteration initial value, carrying out a plurality of defect correction iterations, namely solving a correction equation by using a k-th regularization parameter lambda k (x) in a k-th step: [A+Λk]δΨ k =-R k-1 ; and updating the solution field: Ψ k =Ψ k-1 +δΨ k ; Until the norm of the correction δψ k is smaller than a preset tolerance or the iteration step reaches a preset maximum K max ; Wherein A is a standard jacobian matrix, lambada k is a diagonal matrix derived from lambada k , R k-1 is the residual error of the last step, and the value of any element in the first group of regularization parameters lambada 1 (x) is larger than the value of the corresponding position element in any subsequent group of regularization parameters lambada k (x).
8. The method for optimizing design of an underwater hull based on the continuous concomitant method according to claim 5, further comprising the following processing steps after obtaining the stabilized sensitivity gradient and before step S5: Calculating an included angle cosine cos theta between the resistance performance gradient g D and the cavitation suppression gradient g C , if cos theta is less than 0 and the absolute value of cos theta is greater than a preset conflict threshold value, entering an arbitration mode, projecting g D to a zero space of g C to obtain an arbitrated gradient g ́ D , and taking the weighted sum of g ́ D and g C as a final updating direction; And calculating the change rate of the principal curvature of the surface of the boat body based on the gradient prediction appearance to be updated, and if the change rate of any region exceeds the geometric tolerance allowed by the manufacturing process, performing smooth amplitude limiting processing on the gradient component of the region based on spline filtering.
9. The method for optimizing the design of the underwater hull based on the continuous accompanying method according to claim 1, wherein the method for optimizing the design of the underwater hull further comprises: If the accompanying solution is detected to be not converged or numerical value abnormality occurs, automatically backing to the check point, and executing the step from the step S3 again after the global reference value of the regularized intensity coefficient is increased by one level; Continuously counting grid units marked as high-risk singular areas, if the same physical area is continuously marked for N times, and the minimum value of jacobian of the grid is continuously reduced, interrupting the current optimization iteration, initiating local grid reconstruction of the area, and re-executing flow field solving and subsequent steps of the current iteration step based on a new grid after the reconstruction is completed.
10. The method for optimizing the design of the underwater hull based on the continuous accompanying method according to claim 1, wherein the method for optimizing the design of the underwater hull further comprises the steps of learning and predicting an optimizing process, and the method comprises the following specific steps: recording the spatial position distribution, the area and the average regularization intensity of the high-risk singular region in historical iteration to form a time sequence data set; Training a lightweight time sequence prediction model using the dataset; In the subsequent optimization, the position of the high-risk singular region of the next iteration step is predicted by using the model, and the prediction information is used as priori knowledge to be provided for the diagnosis flow in the step S3 for initializing the diagnosis parameters thereof.

Description

Underwater boat body optimal design method based on continuous accompanying method Technical Field The invention relates to the technical field of underwater hull optimal design, in particular to an underwater hull optimal design method based on a continuous accompanying method. Background In engineering design of underwater vehicles (such as high-speed submarines and torpedoes), shape optimization is a key link for improving tactical performance, and currently, a gradient optimization method based on continuous accompanying theory is widely applied to pneumatic and hydrodynamic optimization of complex three-dimensional shapes due to the fact that the calculation cost of the gradient optimization method is high-efficiency independent of the number of design variables. According to the method, the sensitivity gradient of the objective function on massive design variables can be obtained at extremely low calculation cost by introducing the accompanying equation, so that an automatic optimization flow is driven, and in practical application, in order to accurately predict the cavitation performance of an underwater vehicle, a forward flow field solution generally adopts a multiphase flow Computational Fluid Dynamics (CFD) model (such as a cavitation model or a VOF model) capable of describing water/gas phase change and density discontinuity. In order to accurately predict and suppress cavitation, flow field calculations must introduce multiphase flow models describing water/gas phase transitions and density discontinuities; However, the mathematical derivation of the continuous accompanying method strictly depends on the basic assumption that flow field variables are continuous and can be made tiny, when carrying out accompanying sensitivity analysis based on such flow field solutions containing physical discontinuities, in the vicinity of a phase transition interface, mathematical singularities (non-conductive points in the form of dirac-like functions) caused by step changes of physical parameters such as density, sound velocity and the like occur in the gradient calculation of a design variable by an objective function (such as resistance), and the mathematical singularities can cause the sensitivity gradient output by an accompanying solver to generate severe and non-physical numerical oscillations in a key area, thereby misleading an optimization algorithm, causing unreasonable geometric distortion of the appearance generated by iteration to occur locally, even causing divergence of flow field numerical calculation, and causing the optimization process to be invalid in practical engineering application. Disclosure of Invention In order to solve the defects in the prior art, the invention provides an underwater hull optimal design method based on a continuous accompanying method. In order to solve the technical problems, the invention provides the following technical scheme: the invention provides an underwater boat body optimal design method based on a continuous accompanying method, which comprises the following steps: s1, establishing a parameterized geometric model of an underwater boat body, and defining an optimized objective function containing cavitation inhibition performance indexes; S2, carrying out multiphase flow forward flow field solution on the current geometric model to obtain flow field state data containing phase volume fraction distribution; S3, executing cavitation singular diagnosis and control flow, identifying cavitation singular areas in a flow field based on physical phase change characteristics and numerical gradient characteristics in the flow field state data, and dynamically generating regularization calculation instructions according to convergence states of the cavitation singular areas; S4, responding to the regularization calculation instruction by an accompanying solver, reconstructing and solving an accompanying equation by adopting a region self-adaptive regularization strategy based on physical feature perception, namely introducing a local mathematical smoothing term into a discrete equation in the cavitation singular region, so as to obtain a sensitivity gradient subjected to stabilization treatment; And S5, updating design variables of the parameterized geometric model based on the sensitivity gradient, and circularly executing the steps S2 to S5 until a preset optimization convergence criterion is met. As a preferred technical solution of the present invention, in step S3, the identifying cavitation singular area in the flow field specifically includes: calculating the phase volume fraction gradient modulus of the full-field grid unit (∇ phi), and marking the unit with the modulus larger than the physical characteristic threshold as a physical phase change interface unit; Estimating a local norm I S adj I of a right-end term source item S adj of the accompanying equation, and marking a unit with a norm value larger than a numerical characteristic threshold as a pot