Hierarchical design-based multi-objective optimization design method for marine vehicle

Abstract

The invention belongs to the technical field of ocean engineering, and particularly discloses a multi-objective optimization design method for an ocean vehicle based on hierarchical design, which comprises the following steps of: S1, defining design variables, optimization objectives and constraint conditions, and constructing a parameterized model of the vehicle; s2, sensitivity analysis is conducted on the design variables, and the design variables are divided into high-sensitivity design variables and low-sensitivity design variables according to sensitivity ranking; s3, processing the high-sensitivity design variables by a multi-objective optimization method based on adaptive sequential sampling to obtain an optimal combination of the high-sensitivity design variables; s4, optimizing the low-sensitivity design variables by utilizing a fuzzy reasoning Taguchi method to obtain an optimal combination of the low-sensitivity design variables; and completing optimization of the aircraft according to the optimal combination of the high-sensitivity design variables and the optimal combination of the low-sensitivity design variables. According to the method, through technical innovation such as variable layering, adaptive sampling and staged optimization, the bottleneck of a traditional modeling and optimization method on high-dimensional complex problems is effectively broken through, and the method has clear technical advantages and wide engineering adaptability; and reliable and efficient technical support is provided for subsequent parameter sensitivity analysis, uncertainty quantification and rapid design iteration.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ocean engineering, and in particular to a multi-objective optimization design method for ocean vehicles based on hierarchical design. BACKGROUND

[0002] With the continuous progress of vehicle technology and ocean exploration means, multi-vehicle state ocean unmanned vehicles that can freely switch between water surface and underwater are gradually becoming the forefront of research. Such vehicles need to consider multiple performance indicators under different working conditions, such as resistance control when sailing on the water surface, propulsion efficiency when gliding underwater, and layout rationality and system integration in limited structural space. Therefore, the design process often faces highly complex optimization problems with multiple objectives and multiple constraints, and reasonable trade-offs between different performances become key technical problems to be solved.

[0003] Target dimensionality reduction technology, as an important direction to alleviate the problem of high-dimensional multi-objective optimization, has gradually attracted attention. The core idea is to identify and remove redundant or highly correlated optimization objectives, reduce the effective dimension of the objective space, and thus simplify the structure of the optimization problem, improve the computational efficiency of the algorithm, and help to more clearly present the Pareto frontier. Although this theory has shown potential in finance, automobiles and other fields, the application research for complex engineering systems such as ships and multi-vehicle state ocean vehicles is relatively scarce. Especially in the case of limited simulation resources, how to achieve efficient and accurate target correlation analysis and dimensionality reduction processing based on a small amount of sample data, there is still a lack of mature engineering solutions.

[0004] In addition, the existing multi-objective optimization process usually fails to combine effective design variable screening and hierarchical modeling strategies, resulting in insufficient accuracy and generalization ability of the surrogate model, further affecting the reliability of the optimization and the global performance of the design scheme.

[0005] In summary, there is a need for a new optimization method that has efficient dimensionality reduction capability, can make full use of limited simulation data, and improves the efficiency of obtaining global optimal solutions, to further improve the design level of such vehicles. SUMMARY

[0006] The present application provides a multi-objective optimization design method for ocean vehicles based on hierarchical design, which solves the problems of large computational load, slow convergence speed, lack of effective variable screening mechanism, insufficient accuracy of surrogate model under limited sample conditions, and inability to balance multiple performance indicators for traditional optimization methods under high-dimensional design variables.

[0007] To achieve the purpose of solving the above technical problems, the present application adopts the following technical solutions:

[0008] The marine vehicle multi-objective optimization design method based on the hierarchical design comprises the following steps:

[0009] Step S1, defining design variables, optimization objectives and constraint conditions, and constructing a parameterized model of the vehicle;

[0010] Step S2, sensitivity analysis is performed on the design variables, and the design variables are divided into highly sensitive design variables (HSDVs) and low-sensitivity design variables (LSDVs) according to the sensitivity ranking;

[0011] Step S3, the highly sensitive design variables are processed based on the adaptive sequential sampling multi-objective optimization method to obtain the optimal combination of the highly sensitive design variables;

[0012] Step S4, the low-sensitivity design variables are optimized by using the fuzzy inference Taguchi method (FIMT) to obtain the optimal combination of the low-sensitivity design variables; and the optimization of the vehicle is completed according to the optimal combination of the highly sensitive design variables and the optimal combination of the low-sensitivity design variables.

[0013] In some embodiments of the present application, the step S2 specifically comprises:

[0014] The optimization objective values of each parameterized model are obtained by coarse grid numerical simulation;

[0015] The Pearson correlation between the design variables and the optimization objective values is calculated;

[0016] The comprehensive sensitivity coefficients of each design variable and the optimization objective value are calculated according to the Pearson correlation;

[0017] The comprehensive sensitivity coefficients of each design variable are sorted from high to low, and the design variables located in the front half of the sequence are defined as highly sensitive design variables, and the design variables located in the back half of the sequence are defined as low-sensitivity design variables.

[0018] In some embodiments of the present application, the step S3 specifically comprises:

[0019] S31, constructing a design space based on the highly sensitive design variables, and generating initial samples and verification samples by using the optimized Latin hypercube sampling method (OLHS);

[0020] S32, establishing a Kriging surrogate model according to the initial samples and the verification samples, and running the Kriging surrogate model to output a Pareto front;

[0021] S33, the multi-objective optimization method based on adaptive sequential sampling increases sample points, and the increased sample points are added to the training set to update the surrogate model;

[0022] S34, the updated surrogate model is continuously run to output the current model Pareto front;

[0023] S35, whether the Pareto front in step S34 converges is judged, if yes, the high sensitivity design variable combination corresponding to the current model is output as the optimal combination of high sensitivity design variables, otherwise, the step S33 is returned to continue to cyclically optimize the surrogate model.

[0024] In some embodiments of the present application, the step S33 of the multi-objective optimization method based on adaptive sequential sampling includes:

[0025] The sample points are determined as seed points to determine a Monte Carlo random point set;

[0026] The value function of the sample points and the Monte Carlo random point set is calculated to determine a candidate set of new sample points;

[0027] The expected hyper-volume gain brought by adding each candidate point in the candidate set to the current Pareto front is estimated;

[0028] Whether the candidate set converges is judged according to the expected hyper-volume gain, if not, the candidate points are optimized to obtain new sample points.

[0029] In some embodiments of the present application, the step of optimizing the candidate points to obtain new sample points includes:

[0030] The candidate points are taken as a double-objective optimization problem according to [value function, expected hyper-volume gain];

[0031] NSGA-II is adopted for non-dominated sorting and congestion calculation;

[0032] The multiple candidate points with the highest Pareto front level are selected for parametric model calculation, the output optimization target and the corresponding candidate points are added to the training set to update the surrogate model, and a closed-loop optimization is formed.

[0033] In some embodiments of the present application, the step of selecting the optimal combination of low sensitivity design variables in step S4 includes:

[0034] Orthogonal experimental design is performed on the low sensitivity design variables to obtain N low sensitivity design variable combinations;

[0035] For each low sensitivity design variable combination, the optimization target value corresponding to each group of low sensitivity design variables is obtained through a simulation platform, and the signal-to-noise ratio of the low sensitivity design variable and the optimization target is calculated;

[0036] The signal-to-noise ratio is normalized and input into the fuzzy inference system, and the level of the multi-objective performance characteristic index (MPCI) is output;

[0037] The level output by the fuzzy inference system is defuzzified by the barycentric method, and the accurate multi-objective performance characteristic index is calculated;

[0038] According to the accurate multi-objective performance characteristic index, the low-sensitivity design variable combination corresponding to the maximum value of the multi-objective performance characteristic index is selected as the optimal combination of low-sensitivity design variables.

[0039] In some embodiments of the present application, the fuzzy inference process of the fuzzy inference system comprises:

[0040] The signal-to-noise ratio of each optimization target is fuzzified into three levels by using a triangular membership function, wherein the three levels include small, medium and large;

[0041] According to the multi-objective weight distribution, a plurality of fuzzy rules are defined;

[0042] The multi-objective performance characteristic index is divided into seven levels by using the fuzzy rules.

[0043] In some embodiments of the present application, a multi-objective optimization design system for marine vehicles based on hierarchical design is provided, comprising:

[0044] A parameterized model construction module is configured to define design variables, optimization targets and constraint conditions, and to construct a parameterized model of the vehicle;

[0045] A design variable screening module is configured to divide the design variables into high-sensitivity design variables and low-sensitivity design variables according to the sensitivity analysis results;

[0046] A design variable optimization module is configured to optimize the high-sensitivity design variables and the low-sensitivity design variables by using different methods, respectively, to obtain the optimized design variables of the vehicle;

[0047] A communication module is configured to communicate with external devices.

[0048] In some embodiments of the present application, an electronic device is provided, comprising:

[0049] A processor, a memory and a transceiver connected in communication with the processor;

[0050] The memory stores computer execution instructions; and the transceiver is configured to receive and transmit data;

[0051] The processor executes computer execution instructions stored in the memory to implement the above-mentioned multi-objective optimization design method of the marine vehicle.

[0052] In some embodiments of the present application, a computer readable storage medium is provided, characterized in that,

[0053] The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the above-mentioned multi-objective optimization design method of the marine vehicle.

[0054] The technical scheme of the present application has the following technical effects compared with the prior art:

[0055] The present application effectively reduces the dimension complexity of the proxy model by introducing a variable layering mechanism, and significantly improves the modeling efficiency. By using the adaptive importance sampling algorithm, the prediction accuracy of the constructed proxy model is greatly improved compared with the static sampling method under the same sample quantity condition, and the fitting ability and generalization performance of the model are significantly enhanced. By introducing a phased optimization mechanism, the key local variables are finely controlled on the basis of maintaining the global search ability, so that the optimization result is more practical in engineering, and the overall quality of the multi-objective optimization is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical scheme of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0057] Figure 1 The multi-objective optimization design method flowchart shown in the embodiments of the present application.

[0058] Figure 2 The Myring curve diagram shown in the embodiments of the present application.

[0059] Figure 3 The vehicle profile diagram shown in the embodiments of the present application.

[0060] Figure 4 The construction flowchart of the parametric model shown in the embodiments of the present application.

[0061] Figure 5 The Pearson correlation coefficient diagram of the design variables and the optimization objectives shown in the embodiments of the present application.

[0062] Figure 6 The membership function diagram of the input water resistance in the fuzzy inference system shown in the embodiments of the present application.

[0063] Figure 7 A membership function graph of the input variable of underwater resistance in the fuzzy inference system shown in the embodiment of the present application.

[0064] Figure 8 A membership function graph of the input variable of ship volume in the fuzzy inference system shown in the embodiment of the present application.

[0065] Figure 9 A membership function graph of the output variable of MPCI in the fuzzy inference system shown in the embodiment of the present application.

[0066] Figure 10 A graph of average MPCI response curves of low sensitivity design variables at different levels shown in the embodiment of the present application.

[0067] Figure 11 A simulation prediction graph of the surface model shown in the embodiment of the present application.

[0068] Figure 12 A simulation prediction graph of the underwater model shown in the embodiment of the present application.

[0069] Figure 13 A simulation prediction graph of the volume model shown in the embodiment of the present application.

[0070] Figure 14 A comparison graph of ship hull profiles before and after optimization shown in the embodiment of the present application.

[0071] Figure 15 A comparison graph of ship surface resistance before and after optimization shown in the embodiment of the present application.

[0072] Figure 16 A comparison graph of ship underwater resistance before and after optimization shown in the embodiment of the present application.

[0073] Figure 17 A comparison graph of bow waves of the surface model before and after optimization shown in the embodiment of the present application.

[0074] Figure 18 A comparison graph of fluid pressure distribution of the underwater model before and after optimization shown in the embodiment of the present application.

[0075] Figure 19 A structural schematic diagram of the multi-objective optimization design system of the marine vehicle based on layered design.

[0076] Figure 20 A structural schematic diagram of the electronic device. DETAILED DESCRIPTION

[0077] With reference to the drawings and the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of the present application.

[0078] In the description of the present application, it should be noted that unless specifically defined and limited, the terms "mounting", "connection", "connecting" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral connection. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances. In the description of the above embodiments, the specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0079] Embodiment 1: Reference Figure 1 As shown in the figure, the embodiment provides a multi-objective optimization design method of marine vehicle based on hierarchical design, including the following steps:

[0080] Step S1, defining design variables, optimization objectives and constraint conditions, and constructing a parameterized model of the vehicle; specifically, referring to Figure 4 As shown in the figure.

[0081] Step S11, obtaining sampling points: defining the design variables of the USGV hull parameterized model, for example, the parameterized model includes 14 key design variables (as shown in Table 1), including hull section parameters (such as d s , d l , d u ) and curve control parameters (such as d s n, d s θ, d s a, etc.), and setting the variable feasible region. 140 initial sample points (10 times the variable dimension) are generated by OLHS to ensure uniform coverage of the design space.

[0082] Based on the Morris-Mitchell criterion, the sampling distribution is optimized, which minimizes the evaluation function Φ q to improve the spatial uniformity of the sample distribution. The criterion is defined based on the distribution of the pairwise distance between sample points:

[0083]

[0084] Where p i represents the distance between a pair of sample points, and J i represents the distance equal to p iwhere q is the order of the norm, and the larger the value of q, the more the optimization process focuses on maximizing the minimum distance between the sampling points.

[0085] Table 1:

[0086]

[0087]

[0088] S12, generating a three-dimensional model of the vehicle using CAESES software:

[0089] By calling the CAESES software batch mode to map the sampling points to the parameterized model, as shown in Figure 2 The profile curve of the multi-hull marine vehicle is controlled by the Myring curve, but the vehicle is not a solid of revolution, but is controlled by three different parameter Myring curves, which control the top, side and bottom profiles, respectively, and automatically output a three-dimensional ship body STEP model, as shown in Figure 3 The fixed parallel mid-length is used to simplify the optimization dimension. The Myring curve control formula is as follows:

[0090]

[0091] where the bow feature of the ship body is a modified semi-elliptical radius distribution, and the stern is defined by a cubic function. In the formula, x represents the axial distance from the bow, a, b and c represent the lengths of the bow, midship and stern, respectively; d is the diameter of the midship body; n is the bow shape index; and θ is the stern half-cone angle. By adjusting the above six parameters, different shaped ship bodies can be generated.

[0092] S13, calculating the volume of the vehicle and the waterline using PythonOCC:

[0093] The total volume f V of the ship body is calculated by calling the PythonOCC library, and the waterline position of each scheme is determined to ensure the same displacement; specifically, the three-dimensional ship body STEP model is divided horizontally by calling the PythonOCC library, and the volume after division is equal to the displacement volume, at which time the division position is the waterline position.

[0094] S14, automatic numerical calculation:

[0095] By driving STAR-CCM+ with Java script, the whole process is automatically executed. The generated three-dimensional ship body model is imported into STAR-CCM+ software for simulation calculation, involving mesh division, solution setting and result extraction.

[0096] Mesh: half-hull symmetry model, 1.36 million grids for surface condition (VOF multiphase flow model, local encryption of free surface), 1.06 million grids for underwater condition. 12 layers of boundary layer, Y+≈30, expansion ratio 1.1.

[0097] Solution settings: RANS equation, k-ε turbulence model, velocity inlet 1.5 times the length of the bow (Lpp=2.0m), pressure outlet 4 times the length of the stern. The surface navigation speed is 1.54m / s, and the waterline position is set as the waterline position obtained in step S13, so as to ensure that the displacement of the simulation of different models is the same; the underwater navigation speed is 0.5m / s.

[0098] Result extraction: the resistance convergence criterion is that the fluctuation amplitude is less than 1%, and the mean value of the last three oscillation periods is taken. Finally, the performance prediction results (surface resistance, underwater resistance and volume) obtained are combined with the design variable values of the corresponding model to form a data set.

[0099] Step S2, sensitivity analysis of design variables, according to the sensitivity ranking, the design variables are divided into high sensitivity design variables and low sensitivity design variables;

[0100] Step S21, obtaining the optimization target value in each parameterized model by coarse grid numerical simulation;

[0101] Specifically, using coarse grid numerical simulation can reduce the calculation time to 1 / 4 of high-precision simulation.

[0102] For example, the optimization target includes surface resistance f Rs , underwater resistance f Ru , volume f V ; the target response (surface resistance f Rs , underwater resistance f Ru , volume f V ) of each sample point obtained and the corresponding design variable form a primary data set.

[0103] Step S22, calculating the Pearson correlation coefficient of the design variable and the optimization target value;

[0104] Specifically, the Pearson correlation coefficient method is used to quantify the linear correlation between the design variable and the optimization target value:

[0105]

[0106] Where: N p is the total number of samples, x i is the design variable value, and y i is the optimization target value. The calculation result is as follows: Figure 5As shown, the abscissa is the design variable, the ordinate is the Pearson correlation coefficient, and each design variable corresponds to the Pearson correlation coefficient of the three optimization objectives to be optimized. A positive number indicates a positive correlation, and a negative number indicates a negative correlation.

[0107] Step S23, according to the Pearson correlation coefficient, the comprehensive sensitivity coefficient of each design variable and the optimization target value is calculated;

[0108] For multi-objective weight allocation (surface resistance priority weight ω S = 0.5, underwater resistance priority weight ω U = 0.3, volume priority weight ω V = 0.2), the comprehensive sensitivity coefficient of each design variable is calculated using the following formula:

[0109]

[0110] wherein, and respectively represent the sensitivity value of the zth design variable to the objective functions f Rs , f Ru and f V , that is, the Pearson correlation coefficient in Figure 5 . The weights ω S , ω U and ω V correspond to the relative importance of f Rs , f Ru and f V , and are allocated according to the specific design requirements, and satisfy ω S + ω U + ω V = 1.

[0111] Step S24, the comprehensive sensitivity coefficients of each design variable are sorted from high to low, and the design variables located in the front half of the sequence are defined as high sensitivity design variables, and the design variables located in the back half of the sequence are defined as low sensitivity design variables.

[0112] In some other embodiments, step S25 is further included for verifying hierarchical reliability, capturing nonlinear monotonic relationship and enhancing robustness, and simultaneously calculating the Spearman's rank correlation coefficient:

[0113]

[0114] wherein, rg X and rg Y are the ranks of the design variable X and the optimization target Y, cov(rg X , rgY ) is the covariance of the rank variable; σrg X and σrg Y is the standard deviation of the rank variable.

[0115] Both the ranking results of the comprehensive sensitivity coefficient and the Spearman coefficient identify the same high-sensitive variables, and finally confirm that the 7 high-sensitive design variables constitute the core optimization space, and the remaining 7 low-sensitive design variables enter the subsequent simplified optimization process.

[0116] In the simulation process, the two ranking results are basically consistent. Among them, the comprehensive sensitivity coefficient mainly reflects the main relationship between the design variables and the optimization objective, and the Spearman coefficient refers to more nonlinear relationships between the design variables and the optimization objective; when the two ranking results are inconsistent, the ranking result of the comprehensive sensitivity coefficient can be selected as the main one or the ranking result of the Spearman coefficient can be selected as the main one according to the demand.

[0117] Step S3, processing the high-sensitive design variables by using the adaptive sampling algorithm to obtain the optimal combination of the high-sensitive design variables; after completing the screening of the high-sensitive design variables, the adaptive sequential sampling algorithm is used for fine optimization. The specific steps are as follows:

[0118] S31, constructing a design space based on the high-sensitive design variables, and generating initial samples and verification samples by the OLHS method;

[0119] For example, based on the 7 HSDVs, the design space is constructed, and the initial samples and verification samples are generated by OLHS. Specifically, the training data set and the verification data set are formed by calculating the target response of the sample point and the corresponding design variable.

[0120] S32, establishing a Kriging surrogate model according to the initial samples and the verification samples, which can effectively reduce the calculation amount, and running the Kriging surrogate model to output the Pareto front;

[0121] S321, the establishment process of the Kriging surrogate model:

[0122] The Kriging model interpolates through the spatial correlation between known data points, and establishes a spatial correlation model based on the concept of covariance function. The establishment process is as follows:

[0123] The unknown objective function y(x) is composed of a deterministic trend function and a random process with spatial correlation:

[0124] y(x) = μ(x) + Z(x);

[0125] where μ(x) is a trend function (usually a polynomial, such as constant, linear, quadratic terms); Z(x) is a zero-mean Gaussian process used to model local perturbations or bias terms with a covariance structure (spatial correlation), i.e.

[0126]

[0127] where: σ 2 : process variance; R(x,x') : covariance function (or correlation function) reflecting the correlation between two points. Commonly used correlation functions are Gaussian kernel function and exponential kernel function,

[0128]

[0129] The joint distribution of training samples is established, and the sample training and covariance matrix construction are constructed; given n sample points: [{x (1) ,y (1)},…,{x (n) ,y (n)}], a correlation matrix and a response vector y can be constructed according to the sample points,

[0130] The correlation matrix element is:

[0131] R ij = R(x (i) ,x (j) );

[0132] The response vector is:

[0133]

[0134] The expected prediction value of the Kriging model for a new input point x * is:

[0135]

[0136] where r = [R(x * ,x (1) ),...,R(x * ,x (n) )] T , C is the covariance matrix between data points x, y is the target value of data points x, and R is an n-dimensional vector, The maximum likelihood estimate is:

[0137]

[0138] and its related mean square error:

[0139]

[0140] where, is the process variance, whose estimation formula is:

[0141]

[0142] S322, run the Kriging surrogate model, which can output the Pareto front corresponding to the current input; wherein the multi-objective optimization principle based on adaptive sequential sampling is as follows:

[0143] The optimization problem can be expressed as the following multi-objective mathematical model:

[0144] Minimize F(x) = [f1(x), f2(x), -f3(x)]

[0145] Subject x e Ω

[0146] Where x is the decision variable vector, here represents the high sensitivity design variables of the ship body, Ω represents the high sensitivity design variable space; f1, f2, f3 are the objective functions to be optimized, wherein f1, f2 are the water surface resistance and underwater resistance objective functions respectively, and f3 is the volume objective function. The negative sign in front indicates that this objective function takes the maximum value in the optimization process, rather than the minimum value.

[0147] First, initialize the population, randomly generate the initial population P0 in the solution space, and the population size is N0=500. Evaluate the objective function for each individual in the population, and calculate its value on all objectives through the constructed surrogate model. Then, according to the Pareto dominance relationship, the population is non-dominated sorted, and the individuals are divided into different non-dominated levels F1, F2, … For two individuals x i and x j , if:

[0148] and

[0149] x i dominates x j .

[0150] For individuals in each non-dominated level, calculate the "crowding degree" index di to maintain the diversity of solutions. The index is calculated as follows:

[0151]

[0152] Where and represent the objective values of the two adjacent individuals of the ith individual on the kth objective function.

[0153] The tournament selection strategy based on non-dominated rank and crowding degree sorting is used to select individuals from the current population to form a mating pool. Then genetic operations are performed, which are divided into crossover operation and mutation operation. The crossover operation is to generate new offspring individuals using simulated binary crossover, and the crossover probability P c = 0.9; the mutation operation is to disturb part of the variables using polynomial mutation to increase the diversity of the population. Then the current population P t is merged with the offspring population Q t to form a joint population R t . The non-dominated sorting is performed on the joint population, and the top N0individuals are selected as the next generation population P t+1 . The maximum number of iterations of the population T0= 200. Through the above implementation, the non-dominated solution set with uniform distribution and approximate true Pareto front can be effectively obtained.

[0154] The result of running the Kriging model in this step S322 is an initial Pareto front, which is used to participate in the subsequent judgment of the convergence of the Kriging model.

[0155] S33, in order to improve the accuracy of the surrogate model constructed on the initial training sample set, an adaptive sampling loop is introduced, that is, the sample points are increased by using the adaptive sampling method, and the increased sample points are added to the training set to update the surrogate model; specifically including the following steps:

[0156] S331, the sample points are determined as seed points to determine their Monte Carlo random point set;

[0157] A large number of random points (10000 in this example) are generated by the Monte Carlo method to fill the design space, and the design space is divided into a plurality of units C i , each unit contains the nearest Monte Carlo random point set to the sample point x (i) .

[0158] S332, the candidate set of new sample points is determined by calculating the merit function of the sample points and the Monte Carlo random point set;

[0159] The local Lipschitz constant of each unit is calculated, where x (j) represents any random point in the Monte Carlo random point set; the constant represents the change gradient of the objective function in the region; the unit radius radius i is defined as the Euclidean distance from the farthest random point in the unit to the seed point, and then the merit function Merit = L i x radius i is calculated. For example, the top 20% of random points with Merit value are selected to form the candidate set.

[0160] S333, estimate the expected hypervolume gain brought by each candidate point in the candidate set after adding to the current Pareto front;

[0161] The expected hypervolume gain (EHVI) is calculated for the candidate set. The distribution is predicted by using the Kriging surrogate model, and the expected hypervolume gain of each candidate point x c may bring to the current Pareto front is estimated by the PlatEMO toolbox:

[0162]

[0163] where y is the predicted value of the new sampling point x in the objective space; p(y|x) is the posterior distribution obtained by Gaussian process prediction; is the current Pareto front solution set; HV(·) is the hypervolume calculation function.

[0164] S334, determine whether the candidate set converges according to the expected hypervolume gain, if not, optimize the candidate points to obtain new sample points.

[0165] If the maximum EHVI value in all candidate points is <0.01, it is determined that the first convergence, then return to step S331 to reselect the Monte Carlo random point set of sample points.

[0166] Otherwise, it is determined that the current candidate set does not converge, that is, the current candidate set can be optimized to obtain new sample points. The following steps are used:

[0167] I. The candidate points are optimized as a bi-objective optimization problem according to [value function Merit, expected hypervolume gain EHVI];

[0168] II. Non-dominated sorting and crowding degree calculation are performed by using NSGA-II, which is a multi-objective optimization algorithm based on evolutionary calculation, and has good convergence and solution distribution, which is suitable for solving complex optimization problems containing multiple conflicting objective functions.

[0169] III. Select the multiple candidate points with the highest Pareto front level for parameterized model calculation, and add the output optimization target and corresponding candidate points to the training set to update the surrogate model, forming a closed loop optimization.

[0170] S34, continue to run the updated surrogate model to output the current model Pareto front f new ;

[0171] S35, determine whether the Pareto front in step S34 converges, by calculating the relative change rate of the optimal value of each objective where f oldthe Pareto front output in the last round of the Kriging model;

[0172] If δ f <1%, the cycle is terminated; the high-sensitivity design variable combination corresponding to the current model is output as the optimal high-sensitivity design variable combination, otherwise the cycle of optimizing the surrogate model is continued.

[0173] Step S4, the low-sensitivity design variable optimal combination is obtained by optimizing the low-sensitivity design variable using the fuzzy reasoning Taguchi method; and the optimization of the aircraft is completed according to the high-sensitivity design variable optimal combination and the low-sensitivity design variable optimal combination. After the high-sensitivity design variable optimal solution set is obtained, the fuzzy reasoning Taguchi method is used for the optimization of the low-sensitivity design variable, and the multi-objective optimization problem is converted into a single-objective optimization problem, and the following steps are specifically used:

[0174] S41, orthogonal experimental design is performed on the low-sensitivity design variable to obtain a plurality of low-sensitivity design variable combinations; for example, orthogonal experimental design is performed on 7 low-sensitivity design variables. In order to avoid the calculation burden of full-factorial experiment, an L32(43 7 orthogonal table is selected, and only 32 numerical experiments are required. Each experiment calculates the water resistance f Rs , the underwater resistance f Ru and the ship volume f V by the parameterized model in step S1.

[0175] S42, for each low-sensitivity design variable combination, the optimization target value corresponding to each low-sensitivity design variable combination is obtained through the simulation platform, and then the signal-to-noise ratio (S / N) of the low-sensitivity design variable and the optimization target is calculated.

[0176] The signal-to-noise ratio (S / N) is calculated based on the target characteristic classification:

[0177]

[0178] Wherein T represents the target for which the signal-to-noise ratio needs to be calculated, and in this patent, the optimization targets are the water resistance f Rs , the underwater resistance f Ru and the ship volume f V . When the optimization target has the "large characteristic", the negative sign is taken in the formula; when the optimization target has the "small characteristic", the positive sign is taken in the formula.

[0179] For example, when the optimization target is the ship volume f V , the signal-to-noise ratio calculation formula takes the negative sign; when the optimization target is the water resistance f Rs and the underwater resistance f Ru , the signal-to-noise ratio calculation formula takes the positive sign.

[0180] S43, the signal-to-noise ratio is normalized and input into the fuzzy inference system, and the grade of the multi-objective performance characteristic index is output;

[0181] The fuzzy inference process includes:

[0182] S431, the signal-to-noise ratio of each optimization target is fuzzed into three grades by using a triangular membership function, wherein the three grades include small (S), medium (M), and large (L); refer to Figures 6-8 , Figure 6 The membership function graph of the water surface resistance as the input 1 of the fuzzy inference system; Figure 7 The membership function graph of the outboard resistance as the input 2 of the fuzzy inference system; Figure 8 The membership function graph of the ship volume as the input 3 of the fuzzy inference system.

[0183] S432, according to the multi-objective weight distribution (the water surface resistance has the highest priority), the variable multi-objective performance characteristic index is divided into seven grades: ES (extremely small) to EL (extremely large), and the membership degree function of the MPCI is as shown in Figure 9 . 27 fuzzy rules are defined (an example rule: if the S / N of f Rs is "large", the S / N of f Ru is "medium", and the S / N of f V is "small", the output MPCI is "medium"), as shown in Table 2:

[0184] Table 2 Fuzzy rules

[0185]

[0186] S44, the grade output by the fuzzy inference system is de-fuzzed by the barycentric method, and the accurate multi-objective performance characteristic index is calculated:

[0187]

[0188] In the formula, k is the number of discrete output levels, y i is the single-point value of the output grade, and μC(y i ) is the membership degree thereof. The mean value analysis is performed on the MPCI results of 32 groups of experiments, and the average MPCI response curves of each variable at different levels are drawn, as shown in Figure 10 .

[0189] S45, according to the accurate multi-objective performance characteristic index, the low-sensitivity design variable combination corresponding to the maximum multi-objective performance characteristic index is selected as the optimal low-sensitivity design variable combination.

[0190] For example, refer to Figure 10As shown, the horizontal combination corresponding to the maximum MPCI is selected as the optimal solution of the LSDVs, which are A2, B3, C2, D2, E1, F2 and G3, respectively.

[0191] The optimal combination of the high-sensitivity design variables obtained in the output step S35 and the optimal combination of the low-sensitivity design variables obtained in the step S45 complete the optimization of the marine vehicle.

[0192] The performance verification and comparative analysis of the above optimization method will be described below.

[0193] Verification of the adaptive sampling algorithm: after the convergence of the Kriging surrogate model, the surrogate model obtained by the adaptive sampling algorithm of the present embodiment is used to regress and predict the optimization target water resistance f Rs , underwater resistance f Ru and ship volume f V , and the results are shown in Figures 11-13 From the figure, it can be seen that the Kriging surrogate model proposed in the present embodiment has superior effect, and the determination coefficients R 2 are 0.9931, 0.9962 and 0.9990, respectively, which are close to 1, indicating that the prediction accuracy of the model for the target response is extremely high, and the input-output relationship of the complex system can be almost perfectly captured and reproduced. Compared with the static sampling method and the method before improvement, the accuracy of the surrogate model obtained by the adaptive sampling algorithm of the present embodiment is improved by 10% and 4.21%, respectively, under the same number of samples.

[0194] Verification of the optimization effect:

[0195] High-precision calculation is performed on the optimized ship type to verify the drag reduction effect. The comparison of the ship type profiles before and after optimization is shown in Figure 14 The comparison of the ship body water resistance before and after optimization is shown in Figure 15 The comparison of the ship body underwater resistance before and after optimization is shown in Figure 16

[0196] Water resistance: compared with the initial ship type, it is reduced by an average of 14.9%, and the maximum reduction is 24.7% at a ship speed of 1.54 m / s.

[0197] Underwater resistance: compared with the initial ship type, it is reduced by an average of 33.3%, and the maximum reduction is 36.3% at a ship speed of 0.7 m / s.

[0198] Optimization comparison result analysis: by comparing the flow field in Figure 17 , it can be seen that the wave height distribution of the optimized ship type is more gentle, the wave peak of the bow is reduced, the range is reduced, and the wave resistance is significantly reduced. By comparing Figure 18 ​As can be seen from the flow field in the figure, the local pressure of the bow and stern of the optimized ship type is significantly reduced, and overall, the pressure distribution of the optimized ship type is more gentle, effectively reducing the water flow resistance.

[0199] By the ship drag reduction optimization design method based on the hierarchical design and the multi-objective optimization method based on adaptive sequential sampling, the average resistance of the multi-navigation state vehicle is reduced by 14.9% and 33.3% respectively on the water surface and underwater, which is significantly better than the traditional proxy model, fully verifies the superiority of the method in fitting accuracy and generalization ability, provides a highly reliable model basis for subsequent optimization design, parameter analysis or uncertainty quantification engineering application, and can provide an efficient solution for multi-navigation state vehicle design.

[0200] The technical scheme of the present application has the following technical effects compared with the prior art:

[0201] By introducing the variable hierarchical mechanism, the dimension complexity of the proxy model is effectively reduced, and the modeling efficiency is significantly improved. In this example, the design variable space dimension can be reduced from 14 to 7, greatly shortening the design cycle and significantly reducing the number of samples and computing resource consumption required for optimization.

[0202] Adopting the adaptive sequential sampling algorithm, the prediction accuracy of the constructed proxy model is improved by 10% compared with the static sampling method, and by 4.21% compared with the improved method under the same sample number, significantly enhancing the fitting ability and generalization performance of the model.

[0203] Introducing the phased optimization mechanism, on the basis of maintaining the global search ability, the key local variables are finely controlled, so that the optimization result is more practical in engineering, further improving the overall quality of multi-objective optimization, effectively improving the model accuracy and design efficiency, and having positive significance for realizing the green design and low-carbon manufacturing goal in the field of marine vehicles.

[0204] When applied to the drag reduction optimization design of the multi-navigation state vehicle, the method of the present application significantly improves the key indicators such as the surface speed and underwater gliding efficiency of the vehicle. Specifically, the average resistance of the vehicle when sailing on the water surface is reduced by 14.9%, and the resistance when gliding underwater is reduced by 33.3%, which is much better than the traditional proxy modeling and optimization method.

[0205] Embodiment 2: This embodiment will be described according to Figure 19 and Figure 20 The multi-objective optimization design system 100 and the electronic device 200 based on hierarchical design of the marine vehicle are described.

[0206] For the multi-objective optimization design system 100 based on hierarchical design of the marine vehicle, referring to Figure 19 , it includes:

[0207] a parameterized model construction module 110, configured to define design variables, optimization objectives and constraint conditions, and construct a parameterized model of the marine vehicle;

[0208] a design variable screening module 120, configured to divide the design variables into high-sensitivity design variables and low-sensitivity design variables according to the sensitivity analysis result;

[0209] a design variable optimization module 130, configured to optimize the high-sensitivity design variables and the low-sensitivity design variables respectively by using different methods, to obtain optimized design variables of the marine vehicle;

[0210] a communication module 140, configured to communicate with an external device.

[0211] It should be understood that the marine vehicle multi-objective optimization design system 100 herein is embodied in the form of functional modules. The term "module" herein can refer to an application specific integrated circuit (ASIC), an electronic circuit, a processor (for example, a shared processor, a dedicated processor or a group processor, etc.) and a memory for executing one or more software or firmware programs, and other suitable components that integrate logic circuitry and / or other support described functions. In an optional example, those skilled in the art can understand that the computing system 100 can be embodied as the electronic device 200 in the above embodiment, or the functions of the electronic device 200 in the above embodiment can be integrated in the marine vehicle multi-objective optimization design system 100, and the marine vehicle multi-objective optimization design system 100 can be used to execute the respective processes and / or steps corresponding to the electronic device 200 in the above method embodiments. To avoid repetition, details are not described here.

[0212] The marine vehicle multi-objective optimization design system 100 described above has the functions of the respective steps executed by the electronic device 200 of the computing method in Embodiment 1. The above functions can be implemented by hardware, or the corresponding software can be executed by hardware. The hardware or software includes one or more modules corresponding to the above functions. For example, the above acquisition module can be a communication interface, such as a transceiver interface.

[0213] In the embodiments of the present application, Figure 19 The marine vehicle multi-objective optimization design system 100 in the above embodiment can also be a chip or a chip system, for example, a system on chip (SoC).

[0214] Referring to Figure 20 In the embodiment, an electronic device 200 is provided, which comprises:

[0215] The processor 210, and a memory 220 and a transceiver 230 connected in communication with the processor;

[0216] The memory 220 stores computer-executable instructions; and the transceiver 230 is configured to transceive data.

[0217] The processor 210 executes the computer-executable instructions stored in the memory 220 to implement the marine vehicle multi-objective optimization design method in Embodiment 1.

[0218] It should be understood that the electronic device 200 can be configured to perform the respective steps and / or procedures in the above-described method embodiments. Optionally, the memory 220 can include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory 220 can also include a non-volatile random access memory. For example, the memory 220 can also store device type information. The processor 210 can be configured to execute the instructions stored in the memory 220, and when the processor 210 executes the instructions, the processor 210 can perform the respective steps and / or procedures in the above-described method embodiments.

[0219] It should be understood that in the embodiments of the present application, the processor 210 can be a central processing unit (CPU), and the processor 210 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0220] In the implementation process, the steps of the above method can be completed by the integrated logic circuit of hardware in the processor 210 or the instructions in the form of software. The steps of the method disclosed in the embodiments of the present application can be directly embodied as hardware processor execution or executed by a combination of hardware and software modules in the processor 210. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor executes the instructions in the memory, and combines the hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0221] Embodiment 3: In this embodiment, a computer-readable storage medium is provided, and the computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the computer-executable instructions are configured to implement the marine vehicle multi-objective optimization design method in Embodiment 1.

[0222] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0223] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0224] In addition, the functional units in each embodiment of the present application can be integrated into a processing unit, or each unit can be physically present separately, or two or more units can be integrated into one unit.

[0225] The functions, if implemented in the form of software functional units and sold or used as independent products, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that make contributions to the prior art, or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0226] In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0227] The above is merely specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed by the present application can be easily thought of by those skilled in the art, and should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-objective optimization design method for marine vehicles based on hierarchical design, characterized in that, Includes the following steps: Step S1: Define design variables, optimization objectives, and constraints to construct a parametric model of the aircraft; Step S2: Perform sensitivity analysis on the design variables and classify them into high-sensitivity design variables and low-sensitivity design variables according to their sensitivity ranking. Step S3: Using an adaptive sequential sampling-based multi-objective optimization method, the highly sensitive design variables are processed to obtain the optimal combination of the highly sensitive design variables; Step S4: Use fuzzy inference Taguchi method to optimize the low-sensitivity design variables to obtain the optimal combination of low-sensitivity design variables; The optimization of the aircraft is completed by combining the optimal combination of highly sensitive design variables and the optimal combination of low-sensitivity design variables.

2. The multi-objective optimization design method for marine vehicles according to claim 1, characterized in that, Step S2 specifically includes: The optimization target values ​​of each parameterized model are obtained through coarse-grid numerical simulation; Calculate the Pearson correlation coefficient between the design variables and the optimization objective value; Calculate the overall sensitivity coefficient between each design variable and the optimization objective value based on the Pearson correlation coefficient; The overall sensitivity coefficients of each design variable are sorted from high to low. The design variables in the first half of the sequence are defined as high-sensitivity design variables, and the design variables in the second half of the sequence are defined as low-sensitivity design variables.

3. The multi-objective optimization design method for marine vehicles according to claim 1, characterized in that, Step S3 specifically includes: S31. Construct a design space based on highly sensitive design variables, and generate initial and validation samples by optimizing the Latin hypercube sampling method; S32. Build a Kriging surrogate model based on the initial and validation samples, and run the Kriging surrogate model to output the Pareto front. S33. A multi-objective optimization method based on adaptive sequential sampling is used to increase sample points. The increased sample points are added to the training set and the surrogate model is then updated. S34. Continue running the updated proxy model and output the Pareto front of the current model; S35. Determine whether the Pareto front in step S34 has converged. If so, output the combination of high-sensitivity design variables corresponding to the current model as the optimal combination of high-sensitivity design variables. Otherwise, return to step S33 to continue iteratively optimizing the surrogate model.

4. The multi-objective optimization design method for marine vehicles according to claim 3, characterized in that, The multi-objective optimization method based on adaptive sequential sampling in step S33 includes the following steps in increasing sample points: Use the sample points as seed points to determine its Monte Carlo random point set; By calculating the value function of the sample points and the Monte Carlo random point set, a candidate set of new sample points is determined; Estimate the expected hypervolume gain of each candidate point in the candidate set after it is added to the current Pareto front; The convergence of the candidate set is determined based on the expected hypervolume gain. If the convergence is not achieved, the candidate points are optimized to obtain new sample points.

5. The multi-objective optimization design method for marine vehicles according to claim 4, characterized in that, The optimization of candidate points to obtain new sample points includes: Treat the candidate points as a bi-objective optimization problem based on [value function, expected hypervolume gain]; NSGA-II was used for non-dominated sorting and crowding calculation; Multiple candidate points with the highest Pareto front levels are selected for parameterized model calculation. The output optimization objective and the corresponding candidate points are added as new sample points to the training set to update the surrogate model, forming a closed-loop optimization.

6. The multi-objective optimization design method for marine vehicles according to claim 1, characterized in that, The selection steps for the optimal combination of low-sensitivity design variables in step S4 include: Orthogonal experimental design was performed on the low-sensitivity design variables to obtain N combinations of low-sensitivity design variables; For each combination of low-sensitivity design variables, the optimization target value corresponding to each group of low-sensitivity design variables is first obtained through the simulation platform, and then the signal-to-noise ratio between the low-sensitivity design variables and the optimization target is calculated. After normalizing the signal-to-noise ratio, the input is fed into the fuzzy inference system, and the output is the level of the multi-objective performance characteristic index; The level output of the fuzzy inference system is defuzzified by the centroid method, and the accurate multi-objective performance characteristic index is calculated. Based on the accurate multi-objective performance characteristic index, the combination of low-sensitivity design variables corresponding to the maximum value of the multi-objective performance characteristic index is selected as the optimal combination of low-sensitivity design variables.

7. The multi-objective optimization design method for marine vehicles according to claim 6, characterized in that, The fuzzy reasoning process of the fuzzy reasoning system includes: The signal-to-noise ratio of each optimization objective is fuzzified into three levels using triangular membership functions, namely small, medium, and large. Based on the multi-objective weight allocation, define multiple fuzzy rules; The multi-objective performance characteristic index is divided into 7 levels using fuzzy rules.

8. A multi-objective optimization design system for marine vehicles based on hierarchical design, characterized in that, The method for multi-objective optimization design of marine vehicles as described in any one of claims 1-7 includes: The parametric model building module is used to define design variables, optimization objectives and constraints, and to build a parametric model of the aircraft. The design variable screening module is used to classify design variables into high-sensitivity design variables and low-sensitivity design variables based on the sensitivity analysis results. The design variable optimization module is used to optimize high-sensitivity design variables and low-sensitivity design variables using different methods to obtain the optimized design variables of the aircraft. The communication module is used to communicate with external devices.

9. An electronic device, characterized in that, include: A processor, and a memory and a transceiver communicatively connected to the processor; The memory stores computer-executed instructions; the transceiver is used for sending and receiving data. The processor executes the computer execution instructions stored in the memory to implement the multi-objective optimization design method for marine vehicles as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the multi-objective optimization design method for marine vehicles as described in any one of claims 1-7.

Images