Data-driven AUV (Autonomous Underwater Vehicle) multidisciplinary design optimization method considering time consumption difference of simulation unit

Abstract

The invention belongs to the technical field of multidisciplinary design, and discloses a data-driven AUV (Autonomous Underwater Vehicle) multidisciplinary design optimization method considering time consumption difference of simulation units. The method comprises the steps of generating an initial sample for an optimization object and segmenting the initial sample into design samples of each subject sequence; analyzing the high-fidelity evaluation average time consumption of the design samples of each subject sequence, calculating the amplification multiplying power of the design samples of other non-time-consuming subject sequences on the basis of a total time consumption consistency principle by taking the design sample of the time-consuming subject sequence as a benchmark to construct the design samples, and performing high-fidelity evaluation on the constructed design samples to construct a subject sequence library; establishing an agent model of each subject sequence based on the subject sequence library; and segmenting the multidisciplinary system of the optimization object into each disciplinary sequence, and carrying out rapid evaluation based on the corresponding agent model. According to the method, the problem of time isomerism evaluation in the multidisciplinary design optimization process is solved, and the optimization efficiency is remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of AUV multidisciplinary design, and particularly relates to a data-driven AUV multidisciplinary design optimization method considering time consumption differences of simulation units. BACKGROUND

[0002] In the modern engineering field, the development of high-performance systems such as aerospace, automotive engineering and energy systems is increasingly complex, and the overall performance requirements of design schemes are increasingly stringent. In the traditional design process, different disciplines are often treated in isolation, and this separated design method is extremely likely to lead to a non-optimal overall design, and even conflicts between disciplines, which cannot meet the urgent needs of complex systems for efficient and reasonable design. Multidisciplinary design optimization (MDO) as a systematic engineering design method has emerged as the times require. Through a unified optimization framework, the models and constraints of multiple disciplines are integrated, and the coupling relationship between structure, aerodynamics, thermal control, control and other disciplines is coordinated, aiming to realize the collaborative optimization of the overall performance of complex systems, and is widely used in the development of the above high-performance systems.

[0003] The core of MDO is to solve the problem of multidisciplinary collaborative optimization through reasonable architecture design, advanced optimization algorithms and efficient computing resources. In the aspect of architecture design, various strategies have been developed, such as the multidisciplinary feasible (MDF) strategy, which performs a complete multidisciplinary analysis in each iteration to ensure the feasibility of each step, but the computational cost is high; the individual discipline feasible (IDF) strategy introduces auxiliary variables to reduce the dimension of the design space, but sacrifices the autonomy of disciplines; the simultaneous optimization (AAO) strategy decouples sub-disciplines by introducing auxiliary variables, has high parallelism and discipline autonomy, but increases the dimension of the design space and the consistency convergence condition. In the field of optimization algorithms, traditional gradient methods such as Newton's method and steepest descent method have slow convergence speed and are prone to fall into local optimum when dealing with large-scale problems. In order to overcome these problems, various modern optimization algorithms have been introduced, such as genetic algorithm (GA), particle swarm optimization (PSO), ant colony algorithm (ACO), simulated annealing (SA) and the like. These algorithms have the advantages of strong global search ability and independence from gradient information, and are suitable for complex MDO problems. In recent years, data-driven optimization methods have been widely used in MDO, especially in dealing with high-dimensional, nonlinear and multi-objective complex problems. It constructs proxy models (such as Kriging model, Gaussian process regression, neural network, etc.) to approximate complex physical models, effectively reduces the computational cost and improves the optimization efficiency.

[0004] With the increasing complexity and scale of engineering systems, distributed MDO has become a research hotspot. Distributed MDO improves computational efficiency and scalability by distributing optimization tasks to multiple computing nodes for parallel computing. Typical distributed MDO methods include concurrent subspace optimization (CSSO), collaborative optimization (CO), enhanced collaborative optimization (ECO), and two-level integrated algorithm (BLISS). For example, the CSSO method decomposes the system optimization problem into several sub-optimization problems with non-intersecting design variables. The optimization objective function of each subsystem is the original system optimization problem objective function, and each subsystem only optimizes local design variables. The required coupling input variables are replaced by an approximate model. The CO method hierarchically decomposes the system, and each subsystem only exchanges data with the parent system. The system layer objective function is the original problem objective function, and each subsystem minimizes the difference with the target scheme allocated by the system. The system layer coordinates each sub-discipline for parallel optimization through consistency constraints. The ECO method reverses the system layer and subsystem layer objective functions of the CO algorithm, enhancing the "influence" of each sub-discipline on the objective function. The BLISS method divides design variables into global design variables and discipline local design variables, constructs an approximate expression of the original problem objective function using Taylor expansion, and alternately optimizes global design variables and discipline local design variables until the system optimal solution is obtained. In addition, distributed MDO also uses high-performance computing technologies such as grid computing and cloud computing to improve the solution capability of large-scale MDO problems using distributed computing resources. Grid computing achieves the sharing and collaboration of computing resources through resource management, information services, security services, data migration and management mechanisms; cloud computing provides flexible computing resources and storage capacity to support large-scale data processing and analysis.

[0005] In the overall design of underwater vehicles, multi-disciplinary coupling problems are particularly prominent, involving complex interactions between disciplines such as hydrodynamics, structure, energy, and control. Due to the significant differences in the computational time of each discipline simulation unit, for example, high-precision flow field analysis (CFD) requires several hours, structural strength simulation requires minutes to hours, and control response simulation only requires seconds. This time heterogeneity poses a serious challenge to existing distributed multi-disciplinary design optimization methods. Current mainstream architectures (such as CSSO, CO, and BLISS) support parallel computing, but their static task allocation mechanism cannot dynamically adapt to the time differences between disciplines, causing fast disciplines (such as control simulation) to idle due to waiting for slow disciplines (such as CFD). Meanwhile, data transfer congestion further restricts the optimization efficiency of new configurations such as wing-body blends, severely limiting engineering practicality. SUMMARY

[0006] In view of the evaluation time heterogeneity problem faced in the existing AUV multidisciplinary design optimization (MDO) process, that is, the significant difference in the time required by different disciplines in the evaluation link, which further adversely affects the overall optimization efficiency and rational allocation of computing resources, the present application proposes a data-driven AUV multidisciplinary design optimization method considering the time consumption difference of simulation units. The method and system aim to break through the key restricting factors of MDO in the aspects of evaluation efficiency improvement and computing resource management optimization from the system level, and build an optimization system with high efficiency, good scalability and high engineering practicality.

[0007] In order to achieve the above technical purposes, the following technical solutions are specifically adopted in the present application: In one aspect of the present application, a data-driven AUV multidisciplinary design optimization method considering the time consumption difference of simulation units is provided, comprising the following steps: S1: generating initial samples for the optimization object and dividing them into design samples of each discipline sequence; the discipline sequence is composed of disciplines and data interaction between disciplines; S2: analyzing the high-fidelity evaluation average time consumption of the design samples of each discipline sequence, and calculating the expansion multiple of the design samples of other non-most time-consuming discipline sequences based on the total time consumption consistency principle and taking the design samples of the most time-consuming discipline sequence as the benchmark; S3: generating expanded samples according to the corresponding expansion multiple for the design samples of each non-most time-consuming discipline sequence and replacing the nearest neighbor samples; S4: performing high-fidelity evaluation on the design samples of each discipline sequence to build a discipline sequence library; S5: establishing Kriging surrogate models for each discipline sequence based on the discipline sequence library; S6: dividing the multidisciplinary system of the optimization object into each discipline sequence and performing rapid evaluation based on the corresponding Kriging surrogate model.

[0008] In one embodiment, step S1 includes: according to the Latin hypercube sampling criterion, generating a plurality of initial samples for the entire design space of the optimization object, and dividing the initial samples into design samples corresponding to different discipline sequences.

[0009] In one embodiment, analyzing the average time consumption of each discipline sequence in step S2 includes: generating the same number of Latin hypercube sampling samples for the design samples of each discipline sequence, performing real physical simulation, recording the evaluation time of the design samples of each discipline sequence and calculating the arithmetic mean.

[0010] In one embodiment, step S3 comprises: generating augmented samples for each non-most time-consuming discipline sequence design sample according to corresponding amplification ratio; deleting the closest equivalent size design sample from the discipline sequence design samples split from the initial sample from the remaining augmented samples; and using the remaining augmented samples and the remaining discipline sequence design samples split from the initial sample to build the surrogate model.

[0011] In one embodiment, the AUV multidisciplinary design optimization method further comprises step S7: generating a child population according to a given optimization algorithm according to the current population state; and performing fast evaluation on the child population using the Kriging surrogate model; periodically decomposing the frontier sample to each discipline sequence to trigger high-fidelity evaluation and surrogate model updating.

[0012] In one embodiment, periodically decomposing the frontier sample to each discipline sequence to trigger high-fidelity evaluation and surrogate model updating comprises: when the optimization iteration reaches a preset period, decomposing the frontier sample design variables of the current population into discipline sequence design samples; analyzing the high-fidelity evaluation average time consumption of each discipline sequence design sample, generating augmented samples for each non-most time-consuming discipline sequence design sample according to corresponding amplification ratio and replacing the nearest neighbor sample; triggering high-fidelity evaluation and surrogate model updating using the reconstructed discipline sequence design samples.

[0013] In another aspect of the present application, a heterogeneous distributed computing system for AUV multidisciplinary design optimization is provided, comprising: a computing node configured to perform high-fidelity evaluation on each discipline sequence in parallel and build a discipline sequence library; and analyze the high-fidelity evaluation average time consumption of each discipline sequence, calculate the sampling size of other non-most time-consuming discipline sequences based on the total time consumption consistency principle and taking the most time-consuming discipline sequence as a benchmark; a training node configured to build a Kriging surrogate model for each discipline sequence based on the discipline sequence library to form a surrogate model library; a data node configured to store the discipline sequence library, the surrogate model library and a computing sequence; the computing sequence is used as a storage buffer to cache design samples that have been allocated to the computing node but have not been evaluated; an optimization node comprising an optimizer and a multidisciplinary evaluation module; the optimizer is configured to generate a child population according to an optimization algorithm; and the multidisciplinary evaluation module is configured to call parameters in the surrogate model library to perform fast performance prediction on the child population.

[0014] The present application has the following beneficial effects: The application proposes an AUV multidisciplinary design optimization method based on a discipline sequence and a heterogeneous distributed agent model, aiming to improve optimization efficiency and solve the problems of computational resource allocation and efficiency bottleneck faced by traditional methods in engineering applications. In view of the problems of high computational overhead and low resource utilization in the existing multidisciplinary design optimization process, the application introduces an agent model to accelerate system performance evaluation, and combines a heterogeneous distributed computing architecture to realize more efficient task scheduling and resource management. In the embodiments, the application is superior to the traditional method in achieving the effect of each optimization objective under the same number of iterations; under the same working time condition, the constructed architecture can effectively avoid idle of the computing nodes, and improve the overall parallel computing efficiency. In addition, the heterogeneous distributed architecture used has good adaptability and scalability, and significantly improves the operability and practicality of the method in actual engineering. Therefore, the application realizes innovative improvement on the multidisciplinary design optimization method, has clear technical progress, and can bring significant performance improvement and application value to related engineering fields. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is a schematic diagram of the heterogeneous distributed multidisciplinary design optimization process in the application; Figure 2 is a parameterized definition of the outer dimension of the BWBUG in the embodiment of the application; Figure 3 is a parameterized definition of the structure size of the BWBUG in the embodiment of the application; Figure 4 is a population state diagram of the final result output of the method of the embodiment of the application and the traditional MDF method; the result is better as it goes to the lower left corner in the diagram, the red dot is the final optimization result of the method of the application, and the blue dot is the final optimization result of the traditional method; Figure 5 is a comparison diagram of the method of the embodiment of the application and the traditional multidisciplinary feasible method; wherein, the vertical axis IGD represents the reverse generation distance of the optimization result front population, and the smaller the better, the horizontal axis represents the number of real evaluations, according to the content of the application, the real evaluation is performed once based on the agent model every 100 times of discipline evaluation, the number in the bracket represents the number of real evaluations, the blue data is the traditional multidisciplinary feasible method, and the red data is the method of the embodiment of the application. DETAILED DESCRIPTION

[0016] The technical solutions of the present application will be described clearly and completely in conjunction with specific embodiments. However, those skilled in the art will understand that the following described embodiments are part of the embodiments of the present application, not all the embodiments, and are only used to illustrate the present application, and should not be regarded as limiting the scope of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0017] In one specific embodiment of the present application, a multi-disciplinary design optimization method based on discipline sequence and heterogeneous distributed agent model is provided, which specifically comprises the following steps: S1: Construct a heterogeneous distributed computing architecture, including data nodes, computing nodes, training nodes and optimization nodes.

[0018] The present application proposes the concept of discipline sequence, establishes the input-output mapping relationship of numerical variables, and the discipline sequence meets: composed of data interaction between disciplines and disciplines; the variable data of the input and output discipline sequence is in pure numerical form; cannot be further decomposed into smaller discipline sequences.

[0019] The data node is deployed on a storage type node with large storage space and high data throughput, including a discipline sequence library, an agent model library and a computing sequence.

[0020] The discipline sequence library is used to centrally store the time-consuming evaluation results generated by the discipline sequence in the evaluation process. These results include the performance data of each discipline sequence under different design variable samples, providing an information base for subsequent analysis and optimization. The agent model library is responsible for managing the parameter configuration of the agent model corresponding to each discipline sequence. The computing sequence acts as a temporary storage buffer, caching design samples that have been allocated to computing nodes but have not been evaluated. These samples are stored in the computing sequence while waiting for evaluation, so that the computing nodes can obtain and execute evaluation tasks in a timely manner when there are idle resources.

[0021] The computing node is used to execute specific discipline evaluation tasks, and is equipped with corresponding hardware devices according to the computing requirements of the discipline evaluation tasks it is responsible for. There are specific computing nodes for different discipline sequences, and these nodes work in parallel and are independent of each other. The computing node and the data node interact with each other, specifically: the computing node obtains design samples from the computing sequence of the data node through the network. These samples are transmitted in a specific data format, usually containing the design parameters required by each discipline sequence. After completing the discipline evaluation of the design sample, the computing node sends the evaluation results back to the discipline sequence library of the data node through the network.

[0022] The training node is used for training and updating the proxy model and for data interaction with the data node. Specifically, the training node retrieves high-fidelity data of each subject sequence design sample from the subject sequence library of the data node for proxy model training; the training node imports the proxy model parameters of the training results into the proxy model library of the data node; when subject sequence expansion is required, subject sequence design samples with given requirements are generated based on the Latin hypercube sampling criterion and input into the calculation sequence of the data node.

[0023] The optimization node includes an optimizer and a multidisciplinary evaluation module. The optimizer is an independent optimization algorithm with modular characteristics. It outputs a progeny population with better performance based on the input current optimized population. The optimizer exchanges data with the multidisciplinary evaluation module and the calculation sequence of the data node. The optimizer inputs the excellent progeny population into the multidisciplinary evaluation module for proxy model evaluation, predicting the response of a given discipline sequence. When the optimizer's iteration count reaches a given cycle, the frontier solution of the current population is decomposed into design samples for each discipline sequence and input into the calculation sequence of the data node. The multidisciplinary evaluation module is derived from the multidisciplinary system of the optimization object and reconstructs the multidisciplinary system of the optimization object into a discipline sequence network according to the definition of the discipline sequence. A proxy model is established for each output variable of each discipline sequence. The multidisciplinary evaluation module exchanges data with the optimizer in the optimization node and the proxy model library in the data node. When performing a rapid evaluation of the discipline sequence, the multidisciplinary evaluation module calls the parameters of each proxy model in the proxy model library in the data node, achieving rapid evaluation from input to output and rapid prediction of system performance. The multidisciplinary evaluation module feeds back the prediction result of the given sample to the optimizer in the optimization node for the next round of optimization iteration.

[0024] Heterogeneous distributed multidisciplinary design optimization process reference Figure 1 The specific implementation steps are as follows: S2: Based on the discipline sequence definition, the multidisciplinary system being optimized is decomposed into independent discipline sequences. Each discipline sequence represents a specific aspect or discipline area within the system, such as structural engineering or fluid dynamics. The average time required for high-fidelity evaluation of each discipline sequence is analyzed. Based on the principle of consistency in total time, the amplification factor of the remaining discipline sequences is calculated using the most time-consuming discipline sequence as a benchmark.

[0025] Among them, the analysis of the average time consumption for high-fidelity evaluation of each subject sequence is as follows: the same number of Latin hypercube samples are generated for each subject sequence, real physical simulation is performed, the evaluation time of each subject sequence is recorded and the arithmetic mean is calculated as the average time consumption of the subject sequence.

[0026] The expansion ratio of other non-most time-consuming discipline sequences is calculated based on the total time consistency principle and taking the most time-consuming discipline sequence as the benchmark. The specific expansion ratio can be obtained by dividing the average time consumption of the most time-consuming discipline sequence by the average time consumption of other non-most time-consuming discipline sequences.

[0027] S3: initialization, including initial population sampling, discipline sequence division, sequence performance evaluation and proxy model establishment.

[0028] S31: the optimization node generates 100 initial samples based on the Latin hypercube criterion for the entire design space, and divides the 100 initial samples into design samples of different discipline sequences according to the decomposition principle of the discipline sequence, and inputs them into the calculation sequence of the data node.

[0029] S32: identify the most time-consuming discipline sequence, and the training node generates expanded samples for the design samples of each non-most time-consuming discipline sequence according to the corresponding expansion ratio, and deletes the design sample closest to the expanded sample from the design sample of the discipline sequence divided from the initial sample, and inputs the remaining expanded sample into the calculation sequence of each discipline sequence in the data node.

[0030] S33: the calculation node corresponding to each discipline sequence calls the discipline sequence samples in the corresponding calculation sequence to perform high-fidelity discipline sequence evaluation, and inputs the results into the discipline sequence library of the data node.

[0031] S34: the training node calls the sample data in the discipline sequence library, establishes the Kriging proxy model of the output data of each discipline sequence, and then saves the parameter data file of each proxy model in the proxy model library of the data node.

[0032] S4: the optimization node performs the iteration process of the optimizer and the multidisciplinary evaluation module, and performs high-fidelity evaluation when the preset evaluation period is reached.

[0033] S41: the optimizer generates a child population according to the current population state and the optimizer consistent with the type of the object to be optimized, and inputs it into the multidisciplinary evaluation module. For example, for single-objective optimization, SQP, GA and other algorithms can be selected, and for multi-objective optimization algorithm, NSGAII and other algorithms can be selected. The algorithm selection only needs to match the optimization problem, and the selection is independent.

[0034] S42: the multidisciplinary evaluation module is composed of proxy models of each discipline sequence, which performs rapid evaluation on the design samples given by the optimizer by calling the proxy model parameters of the proxy model library in the data node, and feeds back the results to the optimizer.

[0035] S43: When the optimizer reaches a preset iteration period, the frontier samples of the current population are decomposed into subject sequence design samples and input into the calculation sequence of the data node. The training node generates a subject sequence design sample of a specified size for each subject sequence design sample based on the principle of consistency in total time consumption, deletes the subject sequence design sample of the same size that is closest to the initial sample in the subject sequence generated, and inputs the remaining subject sequence design samples into the calculation sequence of the data node.

[0036] In another embodiment of the present application, a heterogeneous distributed computing system for AUV multidisciplinary design optimization is provided, comprising: The computing nodes are configured to execute high-fidelity evaluation of each subject sequence in parallel and construct a subject sequence library; analyze the average time consumption of the high-fidelity evaluation of each subject sequence, and calculate the sampling scale of other subject sequences that are not the most time-consuming based on the principle of consistency of total time consumption and taking the most time-consuming subject sequence as the benchmark; A training node establishes a Kriging proxy model of each subject sequence based on the subject sequence library to form the proxy model library; Data nodes are used to store the subject sequence library, the proxy model library, and the calculation sequence; the calculation sequence is used as a storage buffer to cache design samples that have been assigned to the calculation node but have not yet been evaluated; The optimization node includes an optimizer and a multidisciplinary evaluation module; the optimizer generates a progeny population according to the optimization algorithm; the multidisciplinary evaluation module calls the parameters in the agent model library to perform rapid performance prediction on the progeny population.

[0037] Example In an embodiment, this application is applied to the multidisciplinary design optimization of a blended-wing-body underwater glider (BWBUG). The BWBUG's unique wing panel design and the seamless transition between the fuselage and wings provide it with excellent lift-to-drag ratio and endurance performance. Like traditional rotary autonomous underwater vehicles, the BWBUG is a complex multidisciplinary system involving multiple coupled disciplines, including shape, fluid flow, and structure. Therefore, its effective implementation in the BWBUG's multidisciplinary design optimization problem demonstrates the effectiveness of this application in the design optimization of multidisciplinary AUV systems.

[0038] A BWBUG multidisciplinary design optimization problem including three disciplines of shape, fluid and structure is proposed in this embodiment. The optimization problem includes 18 design variables, which are 14 parameters of the shape parametric model and 4 parameters of the structure parametric model of the BWBUG. The optimization problem includes 3 objective variables, which are the shape volume evaluated by the shape discipline, the lift-drag ratio evaluated by the fluid discipline and the skeleton mass evaluated by the structure discipline. The optimization problem includes one constraint, which requires that the maximum skeleton stress is not greater than 10% of the material yield strength. Specifically, the optimization problem can be described as follows:

[0039] wherein, V represents the volume of the BWBUG, C ld represents the lift-drag ratio of the BWBUG, M represents the skeleton mass of the BWBUG, x represents the design variable, represents the maximum skeleton stress, represents the material yield strength.

[0040] In this embodiment, the length of the BWBUG is specified as 1 meter, and the wingspan is 1.75 meters. The design variable ranges are as follows:

[0041] wherein, the first to the fourteenth variables belong to the shape discipline, and the fifteenth to the eighteenth variables belong to the structure discipline. The following is the specific application process of this embodiment.

[0042] 1) Complete the hardware environment construction of the heterogeneous distributed computing platform 11) Select different threads of a personal computer with a central processing unit (CPU) of Intel i7-10700 and a memory of 32 GB as optimization nodes and training nodes.

[0043] 12) Select a commercial network attached storage (NAS) device with a storage capacity of 8 TB as a data node; 13) Select a workstation with a CPU of dual-channel EPYC7502 and a memory of 256 GB as a computing node.

[0044] 14) Use six types of network cables to connect the personal computer and the workstation respectively with the data node as the center, so as to form a local area network.

[0045] 2) Complete the software environment construction of the heterogeneous distributed computing platform based on the folder sharing function of the local area network of the Windows operating system 21) In the data node, create a folder called ComputingSequence to store the computational sequence design variables to be solved, a folder called DisciplinarySequenceLibrary to store the high-fidelity evaluation results of each discipline sequence, and a folder called SurrogateModelLibrary to store or update the computational results of the surrogate model.

[0046] 22) Create a folder DistributedMDOPlatform_NodeA in the optimization node to store the optimization program and the training program of the proxy model.

[0047] 23) Create a folder DistributedMDOPlatform_NodeB in the computing node to store the high-fidelity evaluation programs of each subject sequence.

[0048] 24) Map the folders in the data node to the optimization node and computing node respectively to ensure that both the optimization node and computing node can normally write, read, and delete files to each folder of the data node.

[0049] 3) Complete the construction of a multidisciplinary analysis environment for the wing-body blended underwater glider (BWBUG) on the computing node 31) Complete the construction of the analysis environment for the external appearance discipline 311) In this embodiment, the shape generation logic of BWBUG is: lofting the parameter-controlled airfoil section along the parameter-controlled leading and trailing edge contour lines.

[0050] 312) The airfoil section is defined using CST and is controlled by five parameters, which only determine the shape but not the size.

[0051] 313) The leading edge contour line starts from the center of the vehicle's head and is composed of a third-order Bezier curve and a straight line segment. The trailing edge contour line is similar. The leading and trailing edge contour lines are controlled by nine parameters.

[0052] 314) Taking NACA airfoil as the benchmark and CST rule as the parameterization requirement, the type value points of the normalized airfoil section are generated by five parameters and saved as a data file.

[0053] 315) Then, the leading and trailing edge contours are generated based on the last nine parameters. Figure 2 As shown, the nine variable parameters are D2, D3, D4, D5, Z1, Z2, Z3, Z4, Z5 in order. The control point of the leading edge Bezier curve is P 11 and P 12 , where P 11 The z coordinate is Z5, the x coordinate is a constant value 0, and P 12The z coordinate of P is Z4, and the vertical coordinate is the intersection of 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 of the trailing edge Bezier curve are P 21 and P 22 , wherein the z coordinate of P 21 is Z1, and the x coordinate is a constant D1, and the z coordinate of P 22 is Z2, and the vertical coordinate is the intersection of 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). The program outputs the coordinate type points of the leading and trailing edges according to the input of nine parameters.

[0054] 316) On the coordinate type points of the leading and trailing edges, 51 points are evenly taken along the z direction respectively, and the length is determined as the chord length of the airfoil at the current z coordinate. Then the normalized airfoil type points generated in step 314) are called by the program to be scaled up to form 51 airfoil cross-section type points, which are stored as data files respectively.

[0055] 317) In COMSOL software, a script is written based on Java language to read 51 airfoil cross-section type points, generate 51 entity surfaces, and then read the type points of the leading and trailing edge contour lines to loft the 51 entity surfaces along the contour lines to form a geometric entity. The volume of the geometric entity is evaluated as an output parameter using the built-in evaluation tool of the software, and finally the geometric entity is saved as a standard geometry file.

[0056] 32) After completing the analysis environment setup of the outer shape discipline, the analysis environment setup of the fluid dynamics discipline is performed 321) The BWBUG is simulated numerically under the condition of 2 knots and 2 degrees of attack angle underwater, which is divided into two parts of grid drawing and numerical simulation.

[0057] 322) The grid drawing is completed by automatic script in ICEM. Considering the symmetry of the evaluation results, a geometric model is established for half of the BWBUG and its flow domain. A rectangle with a length of ten meters and a width of five meters is drawn with the symmetry plane of the BWBUG as a reference plane, and a cuboid is stretched five meters on one side of the BWBUG model. The remaining geometric entity is defined as the flow domain by subtracting the entity geometry of the BWBUG using Boolean operation.

[0058] 323) The boundary layer is established on the contact surface of the BWBUG and the flow domain, and the number of layers and the thickness of each layer of the boundary layer are specified. At the same time, the flow domain grid is divided according to the requirement from near to far and from fine to coarse, and the result is output as a grid file.

[0059] 324) Fluent software calls script, automatically reads ICEM exported grid file, sets fluid material as water, turbulence model as k-Ω model, and solution format as second order SIMPLEC format.

[0060] 325) In the boundary condition, the reference surface defined in step 322) is set as symmetry condition, the opposite side is zero shear condition, the contact surface of BWBUG and flow field is no slip condition, the front and lower surface of flow field are defined as velocity inlet, the input components are 2cos2 and 2sin2 respectively, and the back and upper surface of flow field are defined as pressure outlet.

[0061] 326) The integral on the surface of BWBUG is calculated to define the lift and drag, the lift and drag in the solution process are exported as data file, and the upper limit of iteration number is set as 500 times; 327) After the end of CFD evaluation, the computer thread continues to execute the main program, reads the lift and drag calculated in step 327), and calculates the lift-drag ratio after adding the angle of attack condition.

[0062] 33) After the analysis environment of the two disciplines of shape and fluid dynamics is completed, the analysis environment of the structure discipline is continued to be built 331) The analysis of the structure discipline is mainly to evaluate the stress of the designed BWBUG skeleton under the self-weight load, the skeleton section is specified as a rectangle, and the definition is as shown in Figure 3 .

[0063] 332) In this embodiment, there are four input parameters for the structure discipline, which are , , and t. In step 315), the composition of the front and rear edge contour lines of BWBUG is specified as a Bezier curve and a straight line segment respectively, and the connecting line of the intersection points A and B of the two lines divides BWBUG into two parts of fuselage and wing. In the wing part, five ribs are uniformly distributed along the x direction, and two ribs are uniformly distributed along the wing direction; while in the fuselage part, three ribs are distributed along the x direction, and four ribs are distributed along the z direction.

[0064] 333) The x direction distance from the head end to point B is defined as d1, the z direction distance from the tail end to point A is defined as d2, and the z direction distance from the symmetric end surface of BWBUG to line segment AB is defined as d3. The x direction ribs on both sides of the fuselage are fixed, and are half of the thickness t, the position of the middle rib is controlled by the parameter , which represents the ratio of the position of the middle rib to the total width of the fuselage, and the actual distance is . The right side of the z direction rib in the center of the fuselage is connected with the two ribs of the wing, and the z direction ribs on both sides of the fuselage are respectively defined as and Control, wherein the z-direction distance of the rib proximate the head end to the head end is and the z-direction distance of the rib proximate the tail end to the tail end is .

[0065] 334) The skeleton model of BWBUG is automatically drawn by COMSOL. First, the shape parameters are read in to determine the location and size of the fuselage end surface, the fuselage-wing interface and the wing end surface, and then the structure parameters are read in , and The location of the two end points of each rib is determined, then the structure parameter t is read in, the rectangle and parallelogram are drawn, and the solid geometric model is formed by stretching along the y direction. Finally, the generated geometric model is read in (317), and the intersection Boolean operation is performed with the original geometric model to form the skeleton geometric model, and the skeleton mass is exported synchronously.

[0066] 335) The skeleton geometric model drawn in step 334) is automatically meshed using COMSOL.

[0067] 336) Using the structural mechanics module in COMSOL, set the boundary condition as the fixed constraint of the fuselage section, and the load condition as the volume force equal to the buoyancy applied to the whole skeleton, automatically carry out finite element solving, and evaluate the stress distribution in the skeleton.

[0068] 337) The maximum stress in the skeleton is automatically output after the calculation is completed.

[0069] 4) According to prior knowledge, after completing the multidisciplinary automatic analysis environment setup in step 3), write three main analysis programs to solve the shape discipline sequence according to step 31), solve the shape-fluid discipline sequence according to steps 31) and 32), and solve the shape-structure discipline sequence according to steps 31) and 33).

[0070] 5) Optimization framework initialization. This step mainly carries out the initialization process of the wing-body blended underwater glider multidisciplinary optimization, including initial population sampling, discipline sequence division, sequence performance evaluation and proxy model establishment, which includes the following steps: 51) The optimization node generates initial samples based on the Latin hypercube sampling criterion, which is divided into shape discipline sequence samples, shape-fluid discipline sequence samples and shape-structure discipline sequence samples.

[0071] 52) In the design variable space of the shape-fluid discipline sequence and the shape-structure discipline sequence, generate additional discipline sequence samples four times the size of the population of the leading sample according to the Latin hypercube sampling criterion, and input the above three sets of discipline sequence samples into the calculation sequence; 53) The computing nodes perform the automated evaluation process of each discipline sequence in step 4 in parallel, and store the results in the discipline sequence library of the corresponding data node.

[0072] 54) The training node retrieves the shape discipline sequence sample, shape-fluid discipline sequence sample and shape-structure discipline sequence sample data in the library, establishes the Kriging surrogate model of the BWBUG shape volume, lift-drag ratio, skeleton mass and maximum skeleton stress with respect to the design variable space of the corresponding discipline sequence, and then saves the parameter data file of each surrogate model in the surrogate model library of the data node.

[0073] 6) Optimization node optimization. The RVEA in the optimization node generates the offspring population according to the current population state.

[0074] 7) Surrogate model assisted BWBUG multidisciplinary system performance evaluation. The surrogate model established in the optimization node reads the offspring population generated by the RVEA to quickly predict the system performance.

[0075] 8) Discipline sequence evaluation. This step is continuously performed independently after step 5 is completed. The computing nodes monitor if there are discipline sequence samples to be evaluated in the computing sequence, and start the parallel execution of the automated evaluation process of each discipline sequence, and store the results in the discipline sequence library of the corresponding data node.

[0076] 9) Surrogate model update. This step is continuously performed independently after step 5 is completed. When the training node monitors that a certain amount of new discipline sequence samples are generated in the discipline sequence library, it retrieves the shape discipline sequence sample, shape-fluid discipline sequence sample and shape-structure discipline sequence sample data in the library, establishes the Kriging surrogate model of the BWBUG shape volume, lift-drag ratio, skeleton mass and maximum skeleton stress with respect to the design variable space of the corresponding discipline sequence, and then saves the parameter data file of each surrogate model in the surrogate model library of the data node.

[0077] 10) whether the number of function evaluations in the surrogate-based multidisciplinary analysis process in the optimization module has reached the upper limit of 500, if so, the optimization process stops, and the current optimization population of the frontier sample population is output, which is decomposed into the shape discipline sequence sample, shape-fluid discipline sequence sample and shape-structure discipline sequence sample, and input into the calculation sequence; if not, whether the number of evaluations has reached the result output period of 100, if so, the optimization continues, and the current optimization population of the frontier sample population is output, which is decomposed into the shape discipline sequence sample, shape-fluid discipline sequence sample and shape-structure discipline sequence sample, and at the same time, four times the size of the frontier sample population of discipline sequence samples is generated in the design variable space of the shape-fluid discipline sequence and the shape-structure discipline sequence according to the Latin hypercube sampling criterion, and finally the above three sets of samples are input into the calculation sequence, otherwise, the iterative optimization continues.

[0078] The embodiment method of the present application is compared with the traditional single-discipline feasible (IDF) method. In the IDF method, each sample is decomposed into a shape discipline sequence, a shape-fluid discipline sequence and a shape-structure discipline sequence, which are input into each calculation node for high-fidelity evaluation, without sample size expansion considering the time consumption of disciplines. The embodiment method is denoted as the experimental group (Experimental MHDCF), and the IDF method is denoted as the control group (Control MHDCF).

[0079] After 50,000 optimizer iterations, the final frontiers of the experimental and control MHDCF are shown in Figure 4 . Obviously, the MHDCF considering the difference in time consumption performs better. The evolutionary process generates six non-dominated solution sets, and the convergence is quantified by the inverse generational distance (IGD) in Figure 5 . Obviously, the IGD of the MHDCF considering the difference in time consumption decreases faster and converges to a lower value compared with the control MHDCF. In addition, the numbers in the brackets in each point in Figure 5 represent the number of shape-fluid discipline sequence evaluations performed to obtain the non-dominated set. It can be seen that the experimental MHDCF achieves better results with less expensive evaluation time consumption.

[0080] Only the IGD performance of the MHDCF considering the difference in time consumption is concerned, and the decline of this indicator proves the effectiveness of this framework. In addition, by comparing the evaluation difference of the experimental MHDCF and the control MHDCF in the experiment, the former achieves about 25% improvement, while the cost is only 198 times the most expensive evaluation time consumption, while the latter needs 307 times the most expensive evaluation time consumption. This shows that the MHDCF considering the difference in time consumption has a significant advantage in the strategy of supplementing and completing expensive evaluations through the evaluation gap of each node.

[0081] Under the same evaluation conditions, traditional multidisciplinary design optimization (MDO) methods, whether integrated or distributed, have a certain degree of resource waste due to not considering time differences. In contrast, the framework alleviates the impact of time differences to a certain extent with the help of surrogate models and heterogeneous deployment. The less time-consuming discipline sequence can obtain more effective samples. In addition to the advantages of MHDCF in computational efficiency, the implementation mode of the framework also has great research value. The framework decomposes and sets the distributed MDO workflow on different nodes, and realizes the interconnection between nodes through offline data rather than online instructions, thereby enhancing stability and ease of implementation. In addition, the importance of MHDCF is also increased based on the fact that it is based on an integrated architecture. It strictly separates the optimizer from the discipline analysis, allowing the optimizer and the framework to optimize independently. This function conforms to the design mode of complex equipment in practical applications and is suitable for practical use.

[0082] Although the embodiments of the present application are described above with reference to the drawings, the present application is not limited to the specific embodiments and application fields described above, and the specific embodiments described above are merely illustrative, instructive, and not restrictive. Those of ordinary skill in the art can make many forms under the inspiration of the present specification and without departing from the scope protected by the claims of the present application, which are all within the scope of protection of the present application.

Claims

1. A data-driven AUV multidisciplinary design optimization method considering the time consumption differences of simulation units, characterized by: include: S1: Generate initial samples for the optimization object and divide them into design samples for each subject sequence; The subject sequence is composed of data interactions between subjects; S2: Analyze the average time consumption for high-fidelity evaluation of each discipline sequence design sample, and calculate the amplification factor of other non-most time-consuming discipline sequence design samples based on the principle of consistency in total time consumption and taking the most time-consuming discipline sequence design sample as the benchmark; S3: Generate expansion samples for each non-most time-consuming subject sequence design sample according to the corresponding amplification factor and replace the nearest neighbor sample; S4: Conduct high-fidelity evaluation of each discipline sequence design sample to construct a discipline sequence library; S5: establishing a Kriging proxy model for each subject sequence based on the subject sequence library; S6: Divide the multidisciplinary system of the optimization object into discipline sequences and perform rapid evaluation based on the corresponding Kriging proxy model.

2. The AUV multidisciplinary design optimization method according to claim 1, characterized in that: Step S1 includes: generating a number of initial samples for the entire design space of the optimization object according to the Latin hypercube sampling criterion, and dividing the initial samples into design samples corresponding to different subject sequences.

3. The AUV multidisciplinary design optimization method according to claim 1, characterized in that: Analyzing the average time consumption of each subject sequence in step S2 includes: generating the same number of Latin hypercube sampling samples for the design samples of each subject sequence, performing real physical simulation, recording the evaluation time of the design samples of each subject sequence and calculating the arithmetic mean.

4. The AUV multidisciplinary design optimization method according to claim 1, characterized in that: Step S3 includes: generating expanded samples for the design samples of each non-most time-consuming subject sequence according to the corresponding amplification ratio, deleting the design sample of the same size that is closest to the corresponding expanded sample from the subject sequence design samples divided from the initial sample, and establishing a proxy model based on the remaining expanded samples and the remaining subject sequence design samples divided from the initial sample.

5. The AUV multidisciplinary design optimization method according to claim 1, characterized in that: The multidisciplinary design optimization method further comprises step S7: Generate a progeny population according to a given optimization algorithm based on the current population state, and quickly evaluate the progeny population using the Kriging proxy model; Periodically decompose cutting-edge samples into discipline sequences, triggering high-fidelity evaluation and agent model updates.

6. The AUV multidisciplinary design optimization method according to claim 5, characterized in that: Periodically decompose cutting-edge samples into discipline sequences, triggering high-fidelity evaluation and proxy model updates, including: When the optimization iteration reaches the preset cycle, the frontier sample design variables of the current population are decomposed into sequence design samples of each discipline; Analyze the average time consumption of high-fidelity evaluation of each discipline sequence design sample, generate expansion samples according to the corresponding amplification factor for each discipline sequence design sample that is not the most time-consuming, and replace the nearest neighbor sample; Utilize reconstructed subject sequence design samples to trigger high-fidelity evaluation and agent model updates.

7. A heterogeneous distributed computing system for data-driven AUV multidisciplinary design optimization considering the time consumption differences of simulation units, characterized by: include: Computational nodes, configured to execute high-fidelity evaluation of each subject sequence in parallel and build a subject sequence library; And analyze the average time consumption of high-fidelity assessment of each subject sequence, and calculate the sampling scale of other subject sequences that are not the most time-consuming based on the principle of consistency of total time consumption and taking the most time-consuming subject sequence as the benchmark; Training node, based on the subject sequence library, establishes the Kriging proxy model of each subject sequence to form a proxy model library; Data nodes are used to store the subject sequence library, the proxy model library, and the calculation sequence; the calculation sequence is used as a storage buffer to cache design samples that have been assigned to the calculation node but have not been evaluated; An optimization node, comprising an optimizer and a multidisciplinary evaluation module; the optimizer generates a progeny population according to an optimization algorithm; The multidisciplinary evaluation module calls the parameters in the agent model library to perform rapid performance prediction on the offspring population.

Images