Data-driven AUV multidisciplinary optimization method based on state matching

Abstract

The invention discloses a data-driven AUV (Autonomous Underwater Vehicle) multidisciplinary optimization method based on state matching, which comprises the following steps of: determining disciplines related to an AUV design optimization problem and design variables related to each discipline; setting an optimization target and a constraint condition of the design optimization problem; constructing an analysis environment of the design optimization problem, wherein the analysis environment comprises a shape subject analysis environment, a fluid subject analysis environment, a layout subject analysis environment and a structure subject analysis environment; the genetic algorithm is initialized; in the design space range of the AUV, initial design samples are generated, each design sample is composed of a set of design variables, each design sample serves as an individual, and an initial population is constructed; evaluating a population state and determining an optimization strategy; and aiming at the determined optimization strategy, solving by adopting a genetic algorithm based on an agent model, screening out a non-dominated solution set meeting constraint conditions, and outputting a design vector corresponding to an individual contained in the non-dominated solution set as an AUV optimization design scheme.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of multidisciplinary design optimization, and particularly relates to a data-driven AUV (Autonomous Underwater Vehicle) multidisciplinary optimization method based on state matching. BACKGROUND

[0002] In the development process of pursuing high precision and high performance, the design scale of underwater equipment is also increasing. From the initial simple mechanical and electrical system to the current large-scale complex system, in order to make the equipment obtain better performance, the influence factors of multiple disciplines need to be considered, and the product needs to be comprehensively analyzed and optimized. There are a large number of coupling relationships between disciplines, and the analysis of each discipline itself has high professional nature, which puts forward very high requirements on the professional knowledge level and system analysis ability of researchers. Multidisciplinary systems exist widely in industry and manufacturing. In order to systematically study the optimization method of such multidisciplinary systems, multidisciplinary design optimization (MDO) has attracted widespread attention from practitioners around the world in recent decades.

[0003] The main task of MDO is to analyze the coupling effect between various disciplines in a complex system and the influence of these disciplines on the overall system performance, so that the system has better performance through the analysis of system state and iteration of design variables. In the MDO process, the designer sorts out the system topology according to his own discipline knowledge, simplifies the complex multidisciplinary system into a series of simple sub-disciplines, so as to improve the efficiency of analysis and optimization. MDO theory was first proposed by Schmit and Haftka, and was used to solve the MDO problem of aircraft wing design. With the increasing demand for overall performance optimization in various fields, MDO has been widely used in the design of aircraft, buildings, wind turbines and underwater vehicles.

[0004] In the early development of MDO, only single system performance was often concerned, but as the design scale expands and the number of disciplines increases, the system usually has multiple performance indicators to be optimized, and these performance indicators are generally contradictory and mutually restrictive. Single-objective optimization algorithms gradually cannot meet the requirements of MDO, and multi-objective optimization (MOO) based optimizers emerge as the times require. In addition, MDO problems have characteristics such as multiple constraints and difficult performance evaluation, so the development of optimizers for multi-objective multi-disciplinary optimization architecture often focuses on the development of expensive constrained multi-objective optimization algorithms (ECMOAs).

[0005] Unlike general multi-objective optimization algorithms, the constraint multi-objective algorithm for time-consuming and expensive problems needs to consider the expensive evaluation cost of the objective and constraint functions, and to achieve performance improvement within a limited number of evaluations, especially for the multi-disciplinary design optimization of underwater equipment.

[0006] In the current field of underwater equipment design, with the increase of system complexity, the traditional multi-disciplinary design optimization method faces many challenges. The existing multi-objective optimization algorithm cannot approximate the modeling of such complex systems due to the single modeling method and insufficient information extraction capability of the objective space, and often fails to achieve the ideal optimization effect within a limited number of evaluations. SUMMARY

[0007] The purpose of the present application is to provide a data-driven autonomous underwater vehicle (AUV) multi-disciplinary optimization method based on state matching, which can improve the overall performance of underwater equipment design under limited computing resources and solve the problems of high computational cost and low optimization efficiency in the prior art.

[0008] In order to achieve the above-mentioned task, the present application adopts the following technical solutions: The data-driven AUV multi-disciplinary optimization method based on state matching comprises: Determine the disciplines involved in the AUV design optimization problem and the design variables involved in each discipline; set the optimization objectives and constraints of the design optimization problem; Construct the analysis environment of the design optimization problem, including the shape discipline analysis environment, the fluid discipline analysis environment, the layout discipline analysis environment and the structure discipline analysis environment; wherein: In the shape discipline analysis environment, the parameter-controlled airfoil section is lofted along the leading and trailing edge contour lines to construct the shape geometric solid model of the AUV; in the fluid discipline analysis environment, numerical simulation is performed for different working conditions of the AUV to determine the lift-drag ratio of the AUV; in the layout discipline analysis environment, the overall volume maximization design of the internal pressure cabin of the AUV is performed based on the shape geometric solid model of the AUV, and the cabin interference volume is determined; in the structure discipline analysis environment, the stress of the AUV skeleton under the self-weight load is evaluated to obtain the skeleton mass and the maximum stress in the skeleton; Initialize the genetic algorithm; within the design space of the AUV, generate initial design samples, each of which consists of a set of design variables, and construct the initial population by treating each design sample as an individual; Assess the population state and determine the optimization strategy; the population state includes infeasible population state, equilibrium state and feasible population state; in the infeasible population state, perform single-objective optimization based on the degree of constraint violation; in the equilibrium state, retain the optimization objective and incorporate the sum of the degree of constraint violation as an additional objective into the multi-objective optimization process; in the feasible population state, directly perform unconstrained optimization with multiple optimization objectives. For the determined optimization strategy, a genetic algorithm based on the surrogate model is used to solve the problem, and the non-dominated solution set that meets the constraints is selected. The design vectors corresponding to the individuals contained in the set are output as the AUV optimization design scheme.

[0009] Furthermore, the optimization problem is a design optimization problem for a blended wing-body underwater glider (BWBUG) that includes four disciplines: shape, fluid dynamics, layout, and structure. The design optimization problem includes multiple design variables, namely shape parameters and contour parameters in the shape analysis environment, pressure tank size parameters in the layout analysis environment, and structural control parameters in the structure analysis environment. The design optimization problem includes three optimization objectives: maximizing the lift-to-drag ratio obtained under the fluid dynamics analysis environment, maximizing the total volume of the pressure tank obtained under the layout analysis environment, and minimizing the skeleton mass obtained under the structural analysis environment. The constraints of the design problem are: the maximum stress of the skeleton is not greater than a preset ratio of the material yield strength; and the cabin interference volume in the layout analysis environment is required to reach a preset volume.

[0010] Furthermore, in the context of morphological analysis: The airfoil section of BWBUG is defined using CST and is controlled by multiple shape parameters; The leading and trailing edge contours start from the center of the head end of the BWBUG and are each composed of a third-order Bézier curve and a straight line segment; the leading and trailing edge contours are controlled by multiple contour parameters. Based on the NACA airfoil, the normalized airfoil shape value points are generated from the shape parameters according to the CST rule, and then the leading and trailing edge contours are generated based on the contour parameters. Multiple points are uniformly taken along the wingspan direction on the shape value points of the leading and trailing edge contour lines, and the lengths are calculated one by one to determine the chord length of the airfoil in the current wingspan direction; then the shape value points of the normalized airfoil are called up and enlarged proportionally to form the corresponding airfoil section shape value points. Based on the profile value points of each airfoil section, a corresponding solid surface is generated; then the profile value points of the leading and trailing edge contours are read, and each solid surface is lofted along the leading and trailing edge contours to form the external geometric solid model of BWBUG. Furthermore, in the fluid dynamics analysis environment: Using the symmetry plane of BWBUG as a reference plane, fluid science analysis and watershed setting are performed using half of the geometric solid model of BWBUG. A boundary layer is established at the interface between BWBUG and the watershed. The number of boundary layers and the thickness of each layer are defined independently. At the same time, the watershed grid is divided according to the requirement of fine to coarse from near to far. Fluid materials and turbulence models are set. In the boundary conditions, the reference plane is set as symmetric, the opposite side is set as zero shear, the contact surface between BWBUG and the flow domain is set as no slip, the front and bottom of the flow domain are defined as velocity inlets, and the back and top of the flow domain are defined as pressure outlets. The lift and drag are defined by calculating the integral on the BWBUG surface, and the lift and drag in the solution process are derived; after adding the angle of attack condition, the ratio is taken to calculate the lift-to-drag ratio.

[0011] Furthermore, in the context of setting up a subject-specific analytics environment: A buoyancy adjustment chamber is deployed in the middle of BWBUG, and a battery chamber is deployed on each side. Both the buoyancy adjustment chamber and the battery chamber adopt a capsule-shaped pressure tank. The objective variable is the total volume of the pressure tank, which needs to be maximized and optimized. The constraint variable is the interference volume between the pressure tank and the shape of BWBUG. The dimensions of the pressure tank were selected as design variables to generate the buoyancy regulating tank and the battery tank and determine their relative distances to the BWBUG wall, and to evaluate the volume of the pressure tank. Read the external geometric solid model of BWBUG; find the union of the external geometric solid model of BWBUG and the geometric model of the pressure tank body, subtract the external geometric solid model of BWBUG, and calculate the volume of the remaining part, which is the interference volume of the body.

[0012] Furthermore, in the context of structural science analysis: Set structural control parameters; The BWBUG's airfoil section is divided into two parts: the fuselage and the wing surface by the line connecting the intersection of the leading and trailing edge profiles of the Bezier curve and the straight line segment. The wing and fuselage sections each have multiple ribs along the length of the fuselage from nose to tail and along the wingspan, respectively. First, the shape parameters of BWBUG are read in to determine the position and size of the fuselage end face, the fuselage-wing interface, and the wing tip face. Then, the structural control parameters are read in to determine the position of the ribs and generate a solid geometric model. Finally, the shape geometric model of BWBUG is combined to generate the skeleton geometric model of BWBUG, and the skeleton mass is exported simultaneously.

[0013] Furthermore, there are four structural control parameters, namely ρ1, ρ2, ρ3 and t; The fuselage has three ribs along the x-direction (length from nose to tail) and four ribs along the z-direction (wingspan). Of the three ribs along the x-direction, the positions of the two outer ribs are fixed, and their thickness is half of t. The position of the middle rib is controlled by parameter ρ3, which represents the ratio of the middle rib's position to the total width of the fuselage in the z-direction, starting from the left side. Of the four ribs along the z-direction, the two middle ribs, L1 and L2, connect to the ends of the two ribs along the z-direction on the wing surface. The positions of these two ribs, L1 and L2, are controlled by parameters ρ1 and ρ2. ρ1 represents the ratio of the upper rib L1's position to the fuselage's x-direction length, starting from the tail of the fuselage; ρ2 represents the ratio of the lower rib L2's position to the fuselage's x-direction length, starting from the nose of the fuselage.

[0014] Furthermore, for all individuals in the current population, check in turn whether they meet all constraints; assign individuals that do not meet any constraints to the "infeasible population"; assign individuals that meet all constraints to the "feasible population"; calculate the ratio of the number of individuals in the "feasible population" to the number of individuals in the current population to obtain a feasibility ratio, which is used to measure the proportion of individuals in the current population that meet all constraints.

[0015] Furthermore, the optimization strategy is specifically as follows: Set ratio thresholds lc and lo, where lc is less than lo; When the feasible proportion is less than lc, the current population belongs to the "infeasible population" state. At this time, the original multi-objective optimization objective will be ignored, and all the constraint violation degrees will be summed up as a single optimization objective. A single-objective optimization specifically targeting the constraint violation degree will be performed. The constraint violation degree is the sum of the values ​​of the individual's deviation from the preset range of various constraints. When the feasible ratio is between lc and lo, the current population is considered to be in an "equilibrium" state. At this time, multiple optimization objectives are retained, and on this basis, the sum of the degree of constraint violation is included as an additional objective in the multi-objective optimization process. When the feasible proportion is greater than or equal to lo, the current population is considered to be in the "feasible population" state; at this time, the constraints are ignored and multi-objective unconstrained optimization is performed directly.

[0016] A terminal device includes a processor, a memory, and a computer program stored in the memory; when the processor executes the computer program, it implements the state-matching-based data-driven AUV multidisciplinary optimization method.

[0017] A computer-readable storage medium storing a computer program; when executed by a processor, the computer program implements the state-matching-based data-driven AUV multidisciplinary optimization method.

[0018] Compared with the prior art, the present invention has the following technical features: This invention addresses the efficiency and accuracy bottlenecks inherent in the multidisciplinary design optimization of blended wing-body underwater gliders, including difficulties in multi-objective optimization, constraint handling, and high computational costs. Applying this method to the multidisciplinary design optimization of blended wing-body underwater gliders significantly improves the glider's hydrodynamic performance in terms of lift-to-drag ratio. Furthermore, it provides a richer and more diverse range of high lift-to-drag ratio design samples, fully meeting design requirements and validating the feasibility and superiority of this method in practical engineering. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the genetic algorithm part of this invention; Figure 2 This is a schematic flowchart of the method of the present invention; Figure 3 This is a schematic diagram of the NACA airfoil in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the generation of front and rear edge contour lines based on rear contour line parameters in an embodiment of the present invention. Figure 5 This is a schematic diagram of the cross-section of the skeleton model of BWBUG in an embodiment of the present invention; Figure 6 This is a simulation diagram of fluid science analysis in an embodiment of the present invention; Figure 7 This is a schematic diagram of the pressure-resistant chamber portion in a disciplinary analysis environment according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the geometric model of the pressure tank constructed in an embodiment of the present invention; Figure 9 This is a schematic diagram of the skeleton model adjustment process in an embodiment of the present invention; Figure 10 This diagram illustrates the relationship between the skeleton mass, size ratio, and total volume of the pressure chamber before and after optimization in this embodiment of the invention. Detailed Implementation

[0020] This invention aims to provide a data-driven multidisciplinary optimization method for AUVs based on state matching, which can improve the overall performance of underwater equipment design under limited computational resources. By introducing data-driven surrogate model technology, this invention constructs an approximate model to replace the actual objective and constraint functions for optimization evaluation. This method continuously samples advantageous sample points and updates the optimization population and surrogate model in real time, achieving the optimization objective with fewer actual evaluations, thereby effectively solving the problems of high computational cost and low optimization efficiency in existing technologies.

[0021] This invention provides a data-driven multidisciplinary optimization method for AUV based on state matching, comprising: Step 1: Determine the disciplines involved in the AUV design optimization problem, as well as the design variables involved in each discipline; set the optimization objective and constraints for the design optimization problem.

[0022] (11) This embodiment proposes a multidisciplinary design optimization problem for a blended-wing-body underwater glider (BWBUG), encompassing four disciplines: shape, fluid dynamics, layout, and structure. The blended-wing-body underwater glider combines gliding characteristics with propulsion capabilities, belonging to the category of autonomous underwater vehicles. Its shape design, hydrodynamic performance, internal layout, and structural safety are highly compatible with the multidisciplinary problems in general AUV design. The integrated geometry of the hull and wing surfaces of the blended-wing-body glider requires simultaneous consideration of shape optimization and flow field response in the coupled analysis of shape and fluid dynamics disciplines, thus verifying the capabilities of this method in shape parameterization and automated CFD analysis.

[0023] (12) The design optimization problem includes multiple design variables, namely the shape parameters and contour parameters in the shape analysis environment, the pressure tank size parameters in the layout analysis environment, and the structural control parameters in the structure analysis environment.

[0024] In one embodiment of the present invention, there are a total of 14 design variable parameters, the range of which is as follows:

[0025] (13) The design optimization problem includes three optimization objectives: maximizing the lift-to-drag ratio obtained under the fluid science analysis environment, maximizing the total volume of the pressure tank obtained under the layout science analysis environment, and minimizing the skeleton mass obtained under the structural science analysis environment; among which the lift-to-drag ratio and skeleton mass are optimization variables.

[0026] (14) The design optimization problem includes two constraints: the maximum stress of the skeleton is required to be no greater than 10% of the material yield strength, and the cabin interference volume in the layout discipline analysis environment is required to be 0.

[0027] Step 2, construct the analysis environment for the design optimization problem, including an analysis environment for external morphology, fluid dynamics, layout, and structure; wherein: In the shape analysis environment, the airfoil section controlled by parameters is lofted along the leading and trailing edge contours to construct the AUV's external geometric solid model; in the fluid analysis environment, numerical simulations are performed for different operating conditions of the AUV to determine its lift-to-drag ratio; in the layout analysis environment, based on the AUV's external geometric solid model, the overall volume of the AUV's internal pressure chamber is maximized, while the chamber interference volume is determined; in the structural analysis environment, the stress of the AUV's frame under its own weight load is evaluated to obtain the frame mass and the maximum stress in the frame.

[0028] (21) External morphology analysis environment.

[0029] In this embodiment, the shape generation logic of BWBUG is as follows: the parameter-controlled airfoil section is lofted along the parameter-controlled leading and trailing edge contours to construct the geometric solid model of the AUV.

[0030] (212) The airfoil section is defined using CST and is controlled by multiple shape parameters (not shown in the figure). It only determines the shape and not the size.

[0031] (213) The leading edge contour line starts from the center of the head end of the BWBUG and consists of a third-order Bézier curve and a straight line segment. The trailing edge contour line is similar. The leading and trailing edge contour lines are controlled by multiple contour line parameters.

[0032] (214) First, using the NACA airfoil as the reference and the CST rule as the parameterization requirement, the normalized airfoil shape value points are generated from the shape parameters and saved as a data file.

[0033] (215) Next, generate the leading and trailing edge contours based on the contour parameters; such as... Figure 3 As shown, there are nine contour line parameters in this embodiment. The nine contour line parameters are D2, D3, D4, D5, Z1, Z2, Z3, Z4, and Z5 in sequence, which represent the length between the starting points of the leading and trailing edge straight segments, the length of the wingtip, the length from the starting point of the leading edge straight segment to the tip of the AUV, the length from the end point of the leading edge straight segment to the tip of the BWBUG, the spanwise length of P21 of the trailing edge Bézier curve, the spanwise length of P22 of the trailing edge Bézier curve, the spanwise length of the starting point of the leading edge straight segment, the spanwise length of P22 of the leading edge Bézier curve, and the spanwise length of P21 of the leading edge Bézier curve.

[0034] The control points for the leading edge Bézier curve are P11 and P12, where P11 has a z-coordinate of Z5 and an x-coordinate of a constant 0, while P12 has a z-coordinate of Z4 and its vertical coordinate is the intersection of its z-coordinate with the extension of the leading edge straight line segment. The coordinates of the leading edge straight line segment are (Z3, D4) and (L, D5). The control points for the trailing edge Bézier curve are P21 and P22, where P21 has a z-coordinate of Z1 and an x-coordinate of a constant D1, while P22 has a z-coordinate of Z2 and its vertical coordinate is the intersection of its z-coordinate with the extension of the trailing edge straight line segment. The coordinates of the trailing edge straight line segment are (Z3, D2+D4) and (L, D5+D3), respectively. The program outputs the coordinate values ​​of the leading and trailing edges based on the nine input contour parameters; L is the half-wingspan, and D1 is the BWBUG coordinate (from beginning to end). Figure 3 Body length (middle and lower end is the head end).

[0035] (216) Take 51 points evenly along the z direction on the shape value points of the leading and trailing edge contour lines, and calculate the length of each point to determine the chord length of the airfoil under the current z coordinate. Then, the program calls the shape value points of the normalized airfoil section generated in (214), enlarges them proportionally to form 51 airfoil section shape value points, and saves them as data files.

[0036] (217) Write a script in the COMSOL software based on Java language to read 51 airfoil section value points and generate 51 solid surfaces. Then read the value points of the leading and trailing edge contour lines and make the 51 solid surfaces loft along the contour lines to form the BWBUG shape geometry solid model. Use the software's built-in evaluation tool to evaluate its volume as the output parameter. Finally, save the shape geometry solid model as a standard geometry file.

[0037] (22) Fluid science analysis environment.

[0038] (221) In the fluid analysis environment, the simulation will be carried out for BWBUG underwater under preset working conditions, such as a speed of 2 knots and an angle of attack of 2 degrees; the work is divided into two parts: mesh drawing and numerical simulation.

[0039] (222) Mesh drawing was completed using an automated script in ICEM; considering the symmetry of the evaluation results, half of the BWBUG's geometric solid model was used; taking the symmetry plane of BWBUG as a reference plane, a rectangle with a length of 10 meters and a width of 5 meters was drawn with this cross section as the center, and a cuboid was formed by stretching 5 meters on one side of the BWBUG model; the BWBUG's geometric solid model was subtracted using Boolean operations, and the final remaining geometric solid model was defined as the watershed.

[0040] (223) Establish a boundary layer on the interface between BWBUG and the watershed, define the number of boundary layers and the thickness of each layer, and divide the watershed grid according to the requirements of fine to coarse from near to far, and output the result as a grid file.

[0041] (224) The Fluent software calls the script to automatically read the mesh file exported by ICEM, sets the fluid material to water, solves the turbulence model as the k-Ω model, and solves the second-order SIMPLEC format.

[0042] (225) In the boundary conditions, the reference plane defined in (222) is set as a symmetric condition, the opposite side is a zero shear condition, the contact surface between BWBUG and the flow domain is a no-slip condition, the front and bottom of the flow domain are defined as velocity inlets, and the back and top of the flow domain are defined as pressure outlets.

[0043] (226) Calculate the integral on the BWBUG surface to define lift and drag, export the lift and drag in the solution process as a data file, and set the upper limit of the number of iterations to 500.

[0044] (227) After the CFD evaluation is completed, the computer thread continues to execute the main program, reads the lift and drag calculated in (226), adds the angle of attack condition, and calculates the lift-to-drag ratio by taking the ratio.

[0045] (23) Layout of the subject analysis environment.

[0046] The layout analysis of BWBUG mainly includes two parts: hull formation and parameter calculation. Based on the external geometric solid model of BWBUG, the overall volume of the internal pressure chamber of BWBUG is maximized. All of these tasks are completed in COMSOL combined with COMSOL with MATLAB controls.

[0047] (231) The layout of the discipline analysis environment mainly studies the design of the internal pressure chamber of the glider; a buoyancy adjustment chamber is deployed in the middle of the BWBUG and a battery chamber is deployed on each side. Both the buoyancy adjustment chamber and the battery chamber adopt the capsule-shaped pressure chamber; the shape of the pressure chamber is specified as capsule-shaped, that is, the two planes of the cylinder are each attached to a hemisphere. This design can improve the pressure resistance performance on the one hand, and increase the volume utilization rate on the other hand.

[0048] (232) In the layout discipline analysis environment, based on the geometric solid model of BWBUG, the objective variable is the total volume of the pressure tank, which needs to be maximized and optimized, and the constraint variable is the interference volume between the pressure tank and the shape of BWBUG.

[0049] (233) The layout discipline analysis environment includes six design variables, which are the dimensional parameters of the pressure tank, including the battery compartment length x. 20 Length of buoyancy adjustment chamber x 16 The distance x from the buoyancy adjustment chamber to the head end 17 The distance x between the bow ends of the battery compartment and the buoyancy regulating compartment 18, buoyancy adjustment chamber diameter x 19 and battery compartment diameter x 20 Detailed illustration is attached. Figure 4 As shown.

[0050] (234) Call MATLAB code to automatically generate the buoyancy regulating chamber and battery chamber in COMSOL Server and determine the relative distance with the BWBUG wall, evaluate the volume of the pressure chamber, and export the geometric model file of the pressure chamber.

[0051] (235) Read the external geometric solid model of BWBUG; (236) Find the union of the external geometric solid model of BWBUG and the geometric model of the pressure tank, subtract the external geometric solid model of BWBUG, and calculate the volume of the remaining part as the interference volume of the tank; under the constraint of the interference, maximize the optimization of the total volume of the pressure tank.

[0052] (24) Structural science analysis environment.

[0053] (241) The structural analysis environment mainly assesses the stress state of the designed BWBUG frame under its own weight load to determine the frame mass; the frame cross-section is specified as rectangular, defined as follows: Figure 5 As shown.

[0054] (242) In this embodiment, there are four structural control parameters in the structural science, namely ρ1, ρ2, ρ3 and t; (213) explains that the leading and trailing edge contours of the BWBUG airfoil section are composed of a Bezier curve and a straight line segment, respectively. The line connecting the intersection of the two lines divides the BWBUG into two parts: the fuselage and the wing surface; that is, the part enclosed by the Bezier curve is the fuselage part, and the part containing the remaining straight line segment is the wing surface.

[0055] On the wing surface: five ribs are evenly distributed along the x-direction (head-to-tail direction), and two ribs are arranged along the z-direction of the wingspan; In the fuselage section, three ribs are arranged along the x-direction and four ribs are arranged along the z-direction. Among the three ribs along the x-direction, the positions of the two outer ribs are fixed, and their thickness is half of t. The position of the middle rib is controlled by parameter ρ3, which represents the ratio of the position of the middle rib to the total width of the fuselage in the z-direction, starting from the left side. Among the four ribs in the z-direction of the fuselage, the two middle ribs L1 and L2 are connected to the ends of the two ribs along the z-direction of the wing. The positions of these two ribs L1 and L2 are controlled by parameters ρ1 and ρ2. ρ1 represents the ratio of the position of the upper rib L1, starting from the tail of the fuselage, to the length of the fuselage in the x-direction. ρ2 represents the ratio of the position of the lower rib L2, starting from the nose of the fuselage, to the length of the fuselage in the x-direction.

[0056] (244) The skeleton model of BWBUG is drawn by COMSOL. First, the shape parameters are read in to determine the position and size of the fuselage end face, the fuselage and wing interface and the wing end face. Then, the structural control parameters ρ1, ρ2 and ρ3 are read in to determine the position of the ribs. Next, the structural control parameter t is read in and a solid geometric model is generated. Finally, the shape geometric solid model of BWBUG generated by (21) and (23) is read in to generate the skeleton geometric model of BWBUG and the skeleton mass is exported simultaneously.

[0057] (245) Use COMSOL to automatically mesh the drawn skeleton geometry model.

[0058] (246) Using the structural mechanics module in COMSOL, set the boundary conditions to fixed constraints on the fuselage end face and the load conditions to the volume force applied to the frame as a whole, which is equal to the buoyancy force, and automatically perform finite element solution to evaluate the stress distribution in the frame.

[0059] (247) The maximum stress in the skeleton will be automatically output after the calculation is completed.

[0060] Step 3: Initialize the genetic algorithm. Within the design space of the AUV (the space formed by the range of each design variable parameter given in step (12)), automatically generate several initial design samples using specific sampling strategies such as Latin hypercube sampling. Each design sample consists of a set of design variables, namely the shape parameters and contour parameters in the shape analysis environment, the pressure tank size parameters in the layout analysis environment, and the structural control parameters in the structure analysis environment mentioned in (12). Treat each design sample as an individual to construct the initial population.

[0061] Step 4: Evaluate the population state and determine the optimization strategy; the population state includes infeasible population state, equilibrium state, and feasible population state; in the infeasible population state, perform single-objective optimization based on the degree of constraint violation; in the equilibrium state, retain the optimization objective and incorporate the sum of constraint violation degrees as an additional objective into the multi-objective optimization process; in the feasible population state, directly perform unconstrained optimization of multiple optimization objectives.

[0062] For all individuals in the current population, check in turn whether they meet the constraints in (14); assign individuals in the population that do not meet any of the constraints to the "infeasible population"; assign individuals that meet all the constraints to the "feasible population"; calculate the ratio of the number of individuals in the "feasible population" to the number of individuals in the current population to obtain a feasible ratio, which is used to measure the proportion of individuals in the current population that meet all the constraints.

[0063] To facilitate subsequent optimization strategy identification and branch processing, two pre-defined ratio thresholds, lc and lo, were designed, where lc is less than lo; the specific logic is as follows: (41) When the feasible proportion is less than lc, it means that the vast majority of individuals do not meet the constraints and the current population is in the state of "infeasible population". At this time, the original multi-objective optimization objective will be ignored, and all the degree of constraint violation (i.e. the sum of the values ​​of individuals deviating from the preset range of various constraints) will be uniformly summed as a single optimization objective. A single-objective optimization (unconstrained processing) specifically targeting the degree of constraint violation will be performed to reduce the total amount of constraint violation.

[0064] (42) When the feasible proportion is between lc and lo, it is considered that the current computational population is in a "balanced" state, with a considerable proportion of individuals satisfying the constraints and some individuals not satisfying them. At this time, the system will retain multiple optimization objectives in (13), and on this basis, the sum of the degree of constraint violation will be included as an additional objective in the multi-objective optimization process. For individuals that satisfy all constraints, the additional objective value is zero; for the remaining individuals, the additional objective value is equal to the sum of their degree of constraint violation.

[0065] (43) When the feasible proportion is greater than or equal to lo, it means that the vast majority of individuals in the current population have met all the constraints and are in a feasible population state. At this time, the original constraints in (14) can be temporarily ignored, and multi-objective unconstrained optimization can be carried out directly. That is, for the multi-objective optimization of (13), the constraints of (14) are not set, so as to find individuals with better objective values ​​in the feasible solution domain more efficiently.

[0066] Step 5: For the determined optimization strategy, a genetic algorithm based on the surrogate model is used to solve the problem, and the non-dominated solution set that meets the constraints is selected. The design vectors corresponding to the individuals contained in the set are output as the AUV optimization design scheme.

[0067] In the genetic algorithm solution process, the Kriging surrogate model corresponding to the optimization objective and constraints is constructed to screen offspring individuals, and the reference vector evolution algorithm is used to determine whether the offspring individuals are retained or replace the parent individuals; finally, the non-dominated solution set that simultaneously satisfies the constraints is selected and output as the final optimization result.

[0068] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A data-driven multidisciplinary optimization method for AUVs based on state matching, characterized in that, include: Identify the disciplines involved in the AUV design optimization problem, and the design variables involved in each discipline; Define the optimization objective and constraints for the design optimization problem. The analysis environment for the design optimization problem is constructed, including an analysis environment for external morphology, fluid dynamics, layout, and structure; wherein: In the shape analysis environment, the airfoil section controlled by parameters is lofted along the leading and trailing edge contours to construct the AUV's geometric solid model. In the fluid analysis environment, numerical simulations are performed for different operating conditions of the AUV to determine its lift-to-drag ratio. In the layout analysis environment, based on the AUV's geometric solid model, the overall volume of the AUV's internal pressure chamber is maximized, while the chamber interference volume is determined. In the structural analysis environment, the stress of the AUV's frame under its own weight load is evaluated to obtain the frame mass and the maximum stress in the frame. Initialize the genetic algorithm; within the design space of the AUV, generate initial design samples, each of which consists of a set of design variables, and construct the initial population by treating each design sample as an individual; Assess the population state and determine the optimization strategy; the population state includes infeasible population state, equilibrium state and feasible population state; in the infeasible population state, perform single-objective optimization based on the degree of constraint violation; in the equilibrium state, retain the optimization objective and incorporate the sum of the degree of constraint violation as an additional objective into the multi-objective optimization process; in the feasible population state, directly perform unconstrained optimization with multiple optimization objectives. For the determined optimization strategy, a genetic algorithm based on the surrogate model is used to solve the problem, and the non-dominated solution set that meets the constraints is selected. The design vectors corresponding to the individuals contained in the set are output as the AUV optimization design scheme.

2. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, The optimization problem is the design optimization problem of the blended wing-body underwater glider BWBUG, which includes four disciplines: shape, fluid, layout and structure. The design optimization problem includes multiple design variables, namely shape parameters and contour parameters in the shape analysis environment, pressure tank size parameters in the layout analysis environment, and structural control parameters in the structure analysis environment. The design optimization problem includes three optimization objectives: maximizing the lift-to-drag ratio obtained under the fluid dynamics analysis environment, maximizing the total volume of the pressure tank obtained under the layout analysis environment, and minimizing the skeleton mass obtained under the structural analysis environment. The constraints of the design problem are: the maximum stress of the skeleton is not greater than a preset ratio of the material yield strength; and the cabin interference volume in the layout analysis environment is required to reach a preset volume.

3. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, In the context of morphological analysis: The airfoil section of BWBUG is defined using CST and is controlled by multiple shape parameters; The leading and trailing edge contours start from the center of the head end of the BWBUG and are each composed of a third-order Bézier curve and a straight line segment; the leading and trailing edge contours are controlled by multiple contour parameters. Based on the NACA airfoil, the normalized airfoil shape value points are generated from the shape parameters according to the CST rule, and then the leading and trailing edge contours are generated based on the contour parameters. Multiple points are evenly taken along the wingspan direction at the shape value points of the leading and trailing edge contour lines, and the lengths are calculated one by one to determine the chord length of the airfoil in the current wingspan direction. Then, the normalized airfoil profile points are called and scaled up proportionally to form the corresponding airfoil section profile points; Based on the profile value points of each airfoil section, a corresponding solid surface is generated; then the profile value points of the leading and trailing edge contours are read, and each solid surface is lofted along the leading and trailing edge contours to form the external geometric solid model of BWBUG.

4. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, In the fluid analysis environment: Using the symmetry plane of BWBUG as a reference plane, fluid science analysis and watershed setting are performed using half of the geometric solid model of BWBUG. A boundary layer is established at the interface between BWBUG and the watershed. The number of boundary layers and the thickness of each layer are defined independently. At the same time, the watershed grid is divided according to the requirement of fine to coarse from near to far. Fluid materials and turbulence models are set. In the boundary conditions, the reference plane is set as symmetric, the opposite side is set as zero shear, the contact surface between BWBUG and the flow domain is set as no slip, the front and bottom of the flow domain are defined as velocity inlets, and the back and top of the flow domain are defined as pressure outlets. The lift and drag are defined by calculating the integral on the BWBUG surface, and the lift and drag in the solution process are derived; after adding the angle of attack condition, the ratio is taken to calculate the lift-to-drag ratio.

5. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, Layout of subject analysis environment: A buoyancy adjustment chamber is deployed in the middle of BWBUG, and a battery chamber is deployed on each side. Both the buoyancy adjustment chamber and the battery chamber adopt a capsule-shaped pressure tank. The objective variable is the total volume of the pressure tank, which needs to be maximized and optimized. The constraint variable is the interference volume between the pressure tank and the shape of BWBUG. The dimensions of the pressure tank were selected as design variables to generate the buoyancy regulating tank and the battery tank and determine their relative distances to the BWBUG wall, and to evaluate the volume of the pressure tank. Read the external geometric solid model of BWBUG; find the union of the external geometric solid model of BWBUG and the geometric model of the pressure tank body, subtract the external geometric solid model of BWBUG, and calculate the volume of the remaining part, which is the interference volume of the body.

6. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, In a structural science analysis environment: Set structural control parameters; The BWBUG's airfoil section is divided into two parts: the fuselage and the wing surface by the line connecting the intersection of the leading and trailing edge profiles of the Bezier curve and the straight line segment. The wing and fuselage sections each have multiple ribs along the length of the fuselage from nose to tail and along the wingspan, respectively. First, the shape parameters of BWBUG are read in to determine the position and size of the fuselage end face, the fuselage-wing interface, and the wing tip face. Then, the structural control parameters are read in to determine the position of the ribs and generate a solid geometric model. Finally, the shape geometric model of BWBUG is combined to generate the skeleton geometric model of BWBUG, and the skeleton mass is exported simultaneously.

7. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, There are four structural control parameters: ρ1, ρ2, ρ3, and t. The fuselage has three ribs along the x-direction (length from nose to tail) and four ribs along the z-direction (wingspan). Of the three ribs along the x-direction, the positions of the two outer ribs are fixed, and their thickness is half of t. The position of the middle rib is controlled by parameter ρ3, which represents the ratio of the middle rib's position to the total width of the fuselage in the z-direction, starting from the left side. Of the four ribs along the z-direction, the two middle ribs, L1 and L2, connect to the ends of the two ribs along the z-direction on the wing surface. The positions of these two ribs, L1 and L2, are controlled by parameters ρ1 and ρ2. ρ1 represents the ratio of the upper rib L1's position to the fuselage's x-direction length, starting from the tail of the fuselage; ρ2 represents the ratio of the lower rib L2's position to the fuselage's x-direction length, starting from the nose of the fuselage.

8. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, For all individuals in the current population, check in turn whether they meet all constraints; assign individuals that do not meet any of the constraints to the "infeasible population"; assign individuals that meet all constraints to the "feasible population". The ratio of the number of individuals in the "feasible population" to the number of individuals in the current population is used to obtain a feasibility ratio, which measures the proportion of individuals in the current population that meet all constraints.

9. The data-driven AUV multidisciplinary optimization method based on state matching according to claim 1, characterized in that, The optimization strategy is specifically as follows: Set ratio thresholds lc and lo, where lc is less than lo; When the feasible proportion is less than lc, the current population belongs to the "infeasible population" state. At this time, the original multi-objective optimization objective will be ignored, and all the constraint violation degrees will be summed up as a single optimization objective. A single-objective optimization specifically targeting the constraint violation degree will be performed. The constraint violation degree is the sum of the values ​​of the individual's deviation from the preset range of various constraints. When the feasible ratio is between lc and lo, the current population is considered to be in an "equilibrium" state. At this time, multiple optimization objectives are retained, and on this basis, the sum of the degree of constraint violation is included as an additional objective in the multi-objective optimization process. When the feasible proportion is greater than or equal to lo, the current population is considered to be in the "feasible population" state; at this time, the constraints are ignored and multi-objective unconstrained optimization is performed directly.

10. A terminal device, comprising a processor, a memory, and a computer program stored in the memory; characterized in that, When the processor executes a computer program, it implements the data-driven AUV multidisciplinary optimization method based on state matching as described in any one of claims 1-9.

Images