Multi-disciplinary design optimization method and device
By initializing and normalizing the decision variable samples, combining vectors and perturbation operations of the probability iteration function, the problem of numerical differences caused by different dimensions of decision variables in multidisciplinary design optimization is solved, and rapid convergence and efficient search are achieved to obtain optimized solutions that meet design requirements.
Patent Information
- Application Number
- CN202510489726.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-18
- Publication Date
- 2025-08-01
AI Technical Summary
In the lightweighting of aircraft engines, automobile structures and multidisciplinary design optimization of spacecraft, existing optimization devices are difficult to effectively solve the problems of discontinuous, undirected, high-dimensional, highly nonlinear, and multimodality. The different dimensions of decision variables lead to huge numerical differences, affecting the convergence speed and result quality.
A multidisciplinary design optimization method is adopted, and the decision variable samples are initialized and normalized, and a probability iteration function is used to perform vector operations and perturbation operations to ensure the stability and effectiveness of decision variables during each step of iteration, and finally the optimal solution is obtained.
Convergence of optimization problems is achieved in a short time, engineering solutions that meet the design requirements are obtained, search efficiency and convergence speed of decision variables are improved, and local optimization search ability is enhanced.
Smart Images
Figure CN120408980A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of multidisciplinary design optimization, and in particular relates to a multidisciplinary design optimization method and device. Background Art
[0002] Most existing optimization devices are based on simulations of biological evolutionary behavior or inspiration from natural phenomena, iteratively updating solutions to problems to search for optimal or near-optimal solutions. These algorithms generally lack comprehensive mathematical theories; most are inspired by biological evolution, group behavior, or certain physical laws. While a small number of optimization devices incorporate mathematical theories, these are often logically incomplete and merely employ corresponding mathematical formulas, significantly deviating from the original mathematical theories.
[0003] Existing optimization methods for multidisciplinary design optimization of aircraft engines, lightweight automotive structures, and spacecraft, which are discontinuous, non-differentiable, high-dimensional, highly nonlinear, sparse, multimodal, require long single-point computation times, and have vastly different dimensions and magnitudes of decision variables, often suffer from unacceptable convergence speeds or difficulty obtaining ideal results. Furthermore, existing optimization methods often overlook a long-overlooked issue in multidisciplinary design optimization: the varying dimensions of decision variables, which lead to vastly different numerical values, sometimes even differing by several orders of magnitude. Summary of the Invention
[0004] In response to the above problems, the present invention provides a multidisciplinary design optimization method and device, which adopts the following technical solutions:
[0005] A multidisciplinary design optimization method comprises the following steps:
[0006] Initialize the decision variable samples, randomly generate the decision variable data in the decision variable samples into vectors within a set value range, and obtain the decision variable sample space matrix;
[0007] Determine the real sample space matrix based on the decision variable sample space matrix and the lower limit vector and upper limit vector of the decision variable of the actual optimization problem;
[0008] According to the function mapping relationship of the actual optimization problem and the real sample space matrix, the output sample matrix is determined, and the best sample individual corresponding to the optimal value of the output sample matrix is obtained;
[0009] Based on the probability iteration function, vector operations and perturbation operations are performed on each decision variable in the decision variable sample space matrix to obtain the decision variable for the next iteration step, and the following conditions are met: the samples at the same sequence position in the decision variable sample space matrix at each iteration step are no worse than those in the previous step;
[0010] After running the maximum number of iterative steps, according to the obtained optimal decision variables, determine the best sample individual corresponding to the optimal value of the output sample matrix, and use it as the optimal solution.
[0011] Further, perform an initialization process on the decision variable samples, randomly generate the decision variable data in the decision variable samples into vectors within the set value range, and obtain the decision variable sample space matrix, including the following steps:
[0012] Normalize and centralize the decision variable samples so that all decision variables in the decision variable samples are within the interval [a, b];
[0013] Randomly generate m vectors with a value range of [a, b] for the d-dimensional decision variables to form the decision variable sample space matrix X d×m , where m is the decision variable sample size.
[0014] Further, according to the decision variable sample space matrix, as well as the lower limit vector and upper limit vector of the decision variables of the actual optimization problem, determine the true sample space matrix, including the following steps:
[0015] Use the lower limit vector of the actual optimization problem as the minimum value of the decision variable sample, the upper limit vector as the maximum value of the decision variable sample, and determine the mapping relationship formula between the initialized decision variable data in the decision variable samples;
[0016] Based on the mapping relationship formula, according to the lower limit vector of the decision variables of the actual optimization problem, the decision variable sample space matrix, and the upper limit vector of the decision variables of the actual optimization problem, determine the true sample space matrix.
[0017] Further, based on the probability iteration function, perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix to obtain the decision variables of the next iteration step, including the following steps:
[0018] For the j-th sample individual in the decision variable sample space matrix at the i-th iteration step Randomly select two sample individuals in the current decision variable sample space matrix and According to the fitness value of and sort them to obtain the worst-position sample individual after sorting the middle-position sample individual and the best-position sample individual
[0019] If the sample individual is the worst-position sample individual Based on the first operation relation, the position vector at the current iteration step advances the iteration by using the worst position sample to obtain the position vector of the same sequence at the next iteration step;
[0020] If the sample individual is the middle position sample individual then based on the first operation relation and the second operation relation, the position vector at the current iteration step advances the iteration by using the middle position sample to obtain the position vector of the same sequence at the next iteration step;
[0021] If the sample individual is the best position sample individual then based on the second operation relation, the position vector at the current iteration step advances the iteration by using the best position sample to obtain the position vector of the same sequence at the next iteration step.
[0022] Further, based on the probability iteration function, vector operations and perturbation operations are performed on each decision variable in the decision variable sample space matrix to obtain the decision variables at the next iteration step, and the following steps are further included:
[0023] Based on the first operation relation and the second operation relation, the position vector of the same sequence at the next iteration step is determined according to the best position sample individual at the current iteration step and the average value of the decision variable sample matrix at the current iteration step;
[0024] Based on the third operation relation, the position vector of the same sequence at the next iteration step is determined according to the best sample at the current iteration step, the sample corresponding to the current sequence, the perturbation amount, and the average value of the decision variable sample matrix at the current iteration step;
[0025] Based on the third operation relation, the position vector of the same sequence at the next iteration step is determined according to the best sample at the current iteration step, the sample corresponding to the current sequence, the perturbation amount, the average value of the decision variable sample matrix at the current iteration step, and the probability iteration function.
[0026] Further, the samples at the same sequence positions in the decision variable sample space matrix at each iteration step are not worse than those in the previous step after each iteration, including:
[0027] If the output value corresponding to the new position vector of the same sequence is less than the output value corresponding to the position vector of the previous iteration step, then the new position vector is used as the position vector of the same sequence at the next iteration step; otherwise, the position vector of the previous iteration step is used.
[0028] Further, the first operation relation and the second operation relation characterize the quantitative relationship between the new position, the worst position, the middle position, and the best position.
[0029] Further, the third operation relation represents the quantitative relationship among the perturbed position, the perturbation position, and the perturbation amount.
[0030] Further, after running the maximum number of iteration steps, according to the obtained optimal decision variables, determine the best sample individual corresponding to the optimal value of the output sample matrix, and use it as the optimal solution, including the following steps:
[0031] Map the optimal decision variables to the decision variable space of the actual optimization problem based on the upper bound vector of the decision variables of the actual optimization problem to obtain the optimal true sample;
[0032] Determine the optimal value of the output sample matrix according to the optimal true sample, and use the best sample individual corresponding to the optimal value of the output sample matrix as the optimal solution.
[0033] Further, it further includes the following steps:
[0034] Determine the perturbation amount according to the maximum number of iteration steps and the exploration-exploitation balance rate, and determine the probability iteration function according to the perturbation amount and the exploration-exploitation balance rate.
[0035] The present invention also provides a multidisciplinary design optimization device, including:
[0036] A data processing module, configured to perform initialization processing on the decision variable samples, randomly generate the decision variable data in the decision variable samples into a vector within a set value range, and obtain a decision variable sample space matrix;
[0037] A first data calculation module, configured to determine a true sample space matrix according to the decision variable sample space matrix, and the lower bound vector and the upper bound vector of the decision variables of the actual optimization problem;
[0038] A second data calculation module, configured to determine an output sample matrix according to the function mapping relationship of the actual optimization problem and the true sample space matrix, and obtain the best sample individual corresponding to it according to the optimal value of the output sample matrix;
[0039] A third data calculation module, configured to perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix based on the probability iteration function to obtain the decision variables for the next iteration step, and satisfy: the samples at the same sequence position in the decision variable sample space matrix in each iteration step are not worse than the previous step after each iteration;
[0040] A fourth data calculation module, configured to, after running the maximum number of iteration steps, determine the best sample individual corresponding to the optimal value of the output sample matrix according to the obtained optimal decision variables, and use it as the optimal solution.
[0041] Further, the data processing module is specifically configured to:
[0042] Normalize and centralize the decision variable samples so that all decision variables in the decision variable samples are within the interval [a, b].
[0043] Randomly generate m vectors with a value range of [a, b] for d-dimensional decision variables to form the decision variable sample space matrix X d×m , where m is the decision variable sample size.
[0044] Advantages of the present invention:
[0045] 1. The present invention initializes the decision variable samples, assuming that all decision variables in the search and optimization process are vectors within the set value range, avoiding large differences in the values of decision variables and reducing the computational amount.
[0046] 2. The present invention performs an inverse operation on the true decision variables through the decision variable sample space matrix to obtain the true output solution, forming a closed loop with the initialization process of the decision variable samples.
[0047] 3. In the exploration stage of the present invention, vector operations are performed on the best and worst points of the decision variables, ensuring that each step is iteratively advanced based on the sequence sample j, and accelerating the convergence speed of each sequence j.
[0048] 4. In the development stage of the present invention, perturbation operations are performed on the globally optimal decision variable points, which can enhance the local optimization ability.
[0049] Other features and advantages of the present invention will be described in the subsequent specification, and some of them will be obvious from the specification or understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structures pointed out in the specification and the drawings. Brief Description of the Drawings
[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0051] Figure 1 Shows a flowchart of a multidisciplinary design optimization method according to an embodiment of the present invention;
[0052] [[ID=X]] Figure 2 Shows the schematic of obtaining a new coordinate position through vector operations according to an embodiment of the present invention Figure 1 ;
[0053] Figure 3Shows a schematic diagram of obtaining a new coordinate position through vector operations according to an embodiment of the present invention Figure 2 ;
[0054] Figure 4 Shows a structural schematic diagram of a multidisciplinary design optimization device according to an embodiment of the present invention Detailed implementation manners
[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention
[0056] It should be noted that the terms "first", "second", etc. in this application are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances for the embodiments of the present application described herein
[0057] The present invention provides a multidisciplinary design optimization method and device, which can both converge in a short time for the optimization problem of multidisciplinary design and obtain an engineering solution that meets the requirements of designers
[0058] As Figure 1 shown, a multidisciplinary design optimization method includes the following steps
[0059] S1. Determine the decision variable sample size, the maximum number of iterations, and the exploration-exploitation balance rate
[0060] S2. Perform initialization processing on the decision variable sample, randomly generate the decision variable data in the decision variable sample as vectors within a set value range, and obtain a decision variable sample space matrix
[0061] S21. The initialization processing of the decision variable sample includes: normalizing and centralizing the decision variable sample so that all decision variables in the decision variable sample are within the interval [a, b]
[0062] S22. Randomly generate m vectors with a value range of [a, b] for d-dimensional decision variables. For example, a can be -1 and b can be +1 to form a decision variable sample space matrix X d×m , where m is the decision variable sample size, and the acquisition of any single decision variable is given by the following formula
[0063]
[0064] where rand() is a random number in [0, 1], so the decision variable sample space matrix X d×m can be expressed as:
[0065] X d×m = [x 1 … x j … x m
[0066] where each value in the j-th decision variable sample position belongs to a number in [-1, +1], and j = 1, 2,..., m is a column vector.
[0067] S3. Determine the true sample space matrix according to the decision variable sample space matrix, and the lower limit vector and upper limit vector of the decision variable of the actual optimization problem, including the following steps:
[0068] S31. Take the lower limit vector of the actual optimization problem as the minimum value of the decision variable sample, the upper limit vector as the maximum value of the decision variable sample, and determine the mapping relationship formula between the initialized decision variable data in the decision variable sample, as follows:
[0069]
[0070] where X real is the true sample space matrix, X norm is the initialized decision variable data in the decision variable sample. Obviously, X norm ∈[-1, +1], X min represents the minimum value of the decision variable sample, and X max represents the maximum value of the decision variable sample.
[0071] S32. Based on the mapping relationship formula, determine the true sample space matrix according to the lower limit vector of the decision variable of the actual optimization problem, the decision variable sample space matrix, and the upper limit vector of the decision variable of the actual optimization problem, as follows:
[0072]
[0073] where L = [l1, l2,…, l d T is the lower limit vector of the decision variable of the actual optimization problem, and each of l1, l2,…, l d is a corresponding lower limit of the actual optimization problem, which is a specific number; U = [u1, u2,…, u d T is the upper limit vector of the decision variable of the actual optimization problem, and u1, u2,…, u d Each of them is the upper limit of a corresponding actual optimization problem and is a specific number.
[0074] S4. Determine the output sample matrix according to the function mapping relationship of the actual optimization problem and the real sample space matrix, and obtain the corresponding best sample individual according to the optimal value of the output sample matrix.
[0075] Y real = f(X real ) (4)
[0076] In the formula, f represents the function mapping relationship of the actual optimization problem, and X real represents the real sample space matrix.
[0077] S5. Determine the perturbation amount according to the maximum number of iteration steps and the exploration-exploitation balance rate, and determine the probability iteration function according to the perturbation amount and the exploration-exploitation balance rate.
[0078] For example, before the iteration starts, the exploration-exploitation balance rate er = 0.1, and the exploration-exploitation balance rate er is the only pre-adjustable parameter.
[0079] For example, determine the perturbation amount according to the maximum number of iteration steps and the exploration-exploitation balance rate, as follows:
[0080]
[0081] Among them, ε represents the perturbation amount, iter is the iteration step, and MaxIter is the maximum number of iteration steps.
[0082] For example, determine the probability iteration function according to the perturbation amount and the exploration-exploitation balance rate, as follows:
[0083]
[0084] R2 = 1 - R1 (7)
[0085] In the formula, R1 and R2 represent a determined value that changes with the iteration step.
[0086] S6. Based on the probability iteration function, perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix to obtain the decision variable of the next iteration step.
[0087] S601. Determine the first operation relationship and the second operation relationship adopted by the vector operation. Among them, the first operation relationship and the second operation relationship characterize the quantitative relationship between the new position, the worst position, the intermediate position, and the best position, as follows:
[0088] First operation relationship: x = x w + (x b - xw )+(x m -x w ) (8)
[0089] Second operation relation: x = x b +(x b -x w )+(x b -x m ) (9)
[0090] where, x w is the worst position, x m is the middle position, x b is the best position, and x is the new position (next). Figure 2 and Figure 3 show schematic diagrams of obtaining new coordinate positions through two vector operations.
[0091] S602. Determine the third operation relation used in the perturbation operation, where the third operation relation characterizes the quantitative relationship among the position after perturbation (new position), perturbation position, and perturbation amount, specifically as follows:
[0092] Third operation relation: x = x 0 + εx 1 + ε 2 x 2 +…(10)
[0093] where, x 0 , x 1 and x 2 are perturbation positions, x is the position after perturbation (new position), and ε is the perturbation amount.
[0094] S603. This step is the exploration stage, and the main task is to search in the entire space. For the jth sample individual in the decision variable sample space matrix at the ith iteration step arbitrarily select two sample individuals and
[0095] (j ≠ r1 ≠ r2) in the current decision variable sample space matrix, and sort according to the fitness values of and to obtain the sorted worst position sample individual middle position sample individual and best position sample individual
[0096] S604. If the sample individual is the worst position sample individual Based on the first operation relation, the position vector at the current iteration step advances the iteration by using the worst position sample to obtain the position vector of the same sequence at the next iteration step, as follows:
[0097] If j = worst, then
[0098] (In short, it advances the iteration with j)
[0099] S605. If the sample individual is the middle position sample individual then based on the first operation relation and the second operation relation, the position vector at the current iteration step advances the iteration by using the middle position sample to obtain the position vector of the same sequence at the next iteration step.
[0100] If j ≠ worst ≠ best, then
[0101]
[0102] S606. If the sample individual is the best position sample individual then based on the second operation relation, the position vector at the current iteration step advances the iteration by using the best position sample to obtain the position vector of the same sequence at the next iteration step, as follows:
[0103] If j = best, then
[0104]
[0105] S607. In order to also perform exploitation during the exploration stage and further improve the search efficiency, the following exploitation operations can be carried out during the exploration stage. Based on the first operation relation and the second operation relation, according to the average value of the best position sample individual at the current iteration step and the decision variable sample matrix at the current iteration step, determine the position vector of the same sequence at the next iteration step, as follows:
[0106]
[0107] where x0 = xbest, xm represents the average value of the decision variable sample matrix, x0 represents a solution equal to xbest, and xbest is the best position in the decision variable sample space matrix X d×m at the current iteration step.
[0108]
[0109]
[0110] Among them, x0 is the current optimal solution, xm = mean(X).
[0111] S609. In order to also conduct exploration during the development stage and prevent falling into local optima, the following search is carried out. Based on the third operation relation, according to the best sample of the current iteration step, the sample corresponding to the current sequence, the perturbation amount, the average value of the decision variable sample matrix of the current iteration step, and the probability iteration function, determine the position vector of the same sequence in the next iteration step.
[0112]
[0113] Among them, rd = rand(d, 1).
[0114]
[0115] S610. In the selection stage, the sequence-optimal selection method is adopted, which can ensure that the samples at the same sequence positions in the decision variable sample space matrix of each iteration step are not worse than the previous step after each iteration, including: if the output value corresponding to the new position vector of the same sequence is less than the output value corresponding to the position vector of the previous iteration step, then use the new position vector as the position vector of the same sequence in the next iteration step; otherwise, use the position vector of the previous iteration step. Specifically as follows:
[0116]
[0117] S7. After running the maximum number of iteration steps, according to the obtained optimal decision variable, determine the best sample individual corresponding to the optimal value of the output sample matrix, and use it as the optimal solution. Specifically as follows:
[0118] S71. Map the optimal decision variable to the decision variable space of the actual optimization problem based on the upper limit vector of the decision variable of the actual optimization problem to obtain the optimal true sample.
[0119] S72. Determine the optimal value of the output sample matrix according to the optimal true sample, and use the best sample individual corresponding to the optimal value of the output sample matrix as the optimal solution.
[0120] The process of the multidisciplinary design optimization method according to the embodiment of the present invention is as follows:
[0121] The first step is initialization: set the decision variable sample capacity m, the maximum number of iteration steps MaxIter, and the exploration-exploitation balance rate er, initialize the decision variable sample, use Equation 1 to obtain the decision variable sample space matrix, use Equation 3 to calculate the true sample space matrix, and use Equation 4 to obtain the best sample individual xbest.
[0122] Second iteration:
[0123]
[0124]
[0125] Step 3: Return the optimal solution.
[0126] Based on the above multidisciplinary design optimization methods, such as Figure 4 As shown, the present invention also provides a multidisciplinary design optimization system including a data processing module, a first data calculation module, a second data calculation module, a third data calculation module and a fourth data calculation module.
[0127] The data processing module is used to initialize the decision variable samples, randomly generate the decision variable data in the decision variable samples into vectors within a set value range, and obtain the decision variable sample space matrix;
[0128] A first data calculation module is used to determine a real sample space matrix based on the decision variable sample space matrix and the lower limit vector and upper limit vector of the decision variable of the actual optimization problem;
[0129] The second data calculation module is used to determine the output sample matrix according to the function mapping relationship of the actual optimization problem and the real sample space matrix, and obtain the best sample individual corresponding to the optimal value of the output sample matrix;
[0130] The third data calculation module is used to perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix based on the probability iteration function to obtain the decision variable for the next iteration step, and to meet the following requirements: the samples at the same sequence position in the decision variable sample space matrix at each iteration step are no worse than those in the previous step after each iteration;
[0131] The fourth data calculation module is used to determine the best sample individual corresponding to the optimal value of the output sample matrix according to the obtained optimal decision variable after running the maximum number of iteration steps, and use it as the optimal solution.
[0132] The multidisciplinary design optimization method and device of the present invention have been verified on 32 basic test cases and 10 engineering cases. The optimization method of the embodiment of the present invention can converge on the optimization problem in a short time and quickly obtain an engineering solution that meets the design requirements, which is superior to the most advanced optimization algorithm currently published.
[0133] Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A multidisciplinary design optimization method, characterized in that, Including the following steps: Perform initialization processing on the decision variable samples, randomly generate the decision variable data in the decision variable samples into vectors within the set value range, and obtain the decision variable sample space matrix; Determine the true sample space matrix according to the decision variable sample space matrix, as well as the lower limit vector and upper limit vector of the decision variables of the actual optimization problem; Determine the output sample matrix according to the function mapping relationship of the actual optimization problem and the true sample space matrix, and obtain the best sample individual corresponding to its optimal value according to the output sample matrix; Based on the probability iteration function, perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix to obtain the decision variables of the next iteration step, and satisfy: the samples at the same sequence position in the decision variable sample space matrix in each iteration step are not worse than the previous step after each iteration; After running the maximum number of iteration steps, determine the best sample individual corresponding to the optimal value of the output sample matrix according to the obtained optimal decision variables, and use it as the optimal solution.
2. The multidisciplinary design optimization method according to claim 1, characterized in that Perform initialization processing on the decision variable samples, randomly generate the decision variable data in the decision variable samples into vectors within the set value range, and obtain the decision variable sample space matrix, including the following steps: Normalize and centralize the decision variable samples so that all decision variables in the decision variable samples are within the interval [a, b]; Randomly generate m vectors with d - dimensional decision variables within the range [a, b] to form the decision variable sample space matrix X d×m , where m is the sample size of the decision variables.
3. The multidisciplinary design optimization method according to claim 1, characterized in that Determine the true sample space matrix according to the decision variable sample space matrix, as well as the lower limit vector and upper limit vector of the decision variables of the actual optimization problem, including the following steps: Use the lower limit vector of the actual optimization problem as the minimum value of the decision variable samples, the upper limit vector as the maximum value of the decision variable samples, and determine the mapping relationship formula between the initialized decision variable data in the decision variable samples; Based on the mapping relationship formula, determine the true sample space matrix according to the lower limit vector of the decision variables of the actual optimization problem, the decision variable sample space matrix, and the upper limit vector of the decision variables of the actual optimization problem.
4. The multidisciplinary design optimization method according to claim 1, characterized in that Based on the probability iteration function, perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix to obtain the decision variables of the next iteration step, including the following steps: For the j-th sample individual in the sample space matrix of decision variables at the i-th iteration step Arbitrarily select two sample individuals in the current sample space matrix of decision variables and According to the fitness value pair and Sort them to obtain the worst-position sample individual after sorting The middle-position sample individual and the best-position sample individual If the sample individual is the sample individual at the worst position then, based on the first operation relation, the position vector at the current iteration step is advanced by the sample at the worst position to obtain the position vector of the same sequence at the next iteration step; If the sample individual is the sample individual at the middle position then, based on the first operation relation and the second operation relation, the position vector at the current iteration step adopts the method of advancing the iteration with the sample at the middle position to obtain the position vector of the same sequence at the next iteration step; If the sample individual is the best-position sample individual then, based on the second operation relation, the position vector at the current iteration step is advanced by the best-position sample for iteration to obtain the position vector of the same sequence at the next iteration step.
5. The multidisciplinary design optimization method according to claim 4, wherein Based on the probability iteration function, perform vector operations and perturbation operations on each decision variable in the decision variable sample space matrix to obtain the decision variables of the next iteration step, and also include the following steps: Based on the first operation formula and the second operation formula, determine the position vector of the same sequence of the next iteration step according to the best position sample individual of the current iteration step and the average value of the decision variable sample matrix of the current iteration step; Based on the third operation formula, determine the position vector of the same sequence of the next iteration step according to the best sample of the current iteration step, the sample corresponding to the current sequence, the perturbation amount, and the average value of the decision variable sample matrix of the current iteration step; Based on the third operation formula, determine the position vector of the same sequence of the next iteration step according to the best sample of the current iteration step, the sample corresponding to the current sequence, the perturbation amount, the average value of the decision variable sample matrix of the current iteration step, and the probability iteration function.
6. The multidisciplinary design optimization method according to claim 1, wherein In each iteration step, the samples at the same sequence position in the decision variable sample space matrix are not worse than those in the previous step after each iteration, including: If the output value corresponding to the new position vector of the same sequence is less than the output value corresponding to the position vector of the previous iteration step, then the new position vector is used as the position vector of the same sequence in the next iteration step; otherwise, the position vector of the previous iteration step is adopted.
7. The multidisciplinary design optimization method according to claim 4, wherein The first operation relation and the second operation relation characterize the quantitative relationships among the new position, the worst position, the intermediate position, and the best position.
8. The multidisciplinary design optimization method according to claim 5, wherein The third operation relation characterizes the quantitative relationships among the perturbed position, the perturbation position, and the perturbation amount.
9. The multidisciplinary design optimization method according to claim 1, characterized in that After running the maximum number of iteration steps, based on the obtained optimal decision variable, determine the best sample individual corresponding to the optimal value of the output sample matrix and use it as the optimal solution, including the following steps: Map the optimal decision variable to the decision variable space of the actual optimization problem based on the upper limit vector of the decision variable of the actual optimization problem to obtain the optimal true sample; Determine the optimal value of the output sample matrix according to the optimal true sample, and use the best sample individual corresponding to the optimal value of the output sample matrix as the optimal solution.
10. The multidisciplinary design optimization method according to any one of claims 1-9, characterized in that, It also includes the following steps: Determine the perturbation amount according to the maximum number of iteration steps and the exploration-exploitation balance rate, and determine the probability iteration function according to the perturbation amount and the exploration-exploitation balance rate.
11. A multidisciplinary design optimization device, characterized in that Including: A data processing module for initializing the decision variable sample, randomly generating the decision variable data in the decision variable sample into a vector within the set value range to obtain the decision variable sample space matrix; A first data calculation module for determining the true sample space matrix according to the decision variable sample space matrix, as well as the lower limit vector and the upper limit vector of the decision variable of the actual optimization problem; A second data calculation module for determining the output sample matrix according to the function mapping relation of the actual optimization problem and the true sample space matrix, and obtaining the best sample individual corresponding to it according to the optimal value of the output sample matrix; A third data calculation module for performing vector operations and perturbation operations on each decision variable in the decision variable sample space matrix based on the probability iteration function to obtain the decision variable of the next iteration step, and satisfying that in each iteration step, the samples at the same sequence position in the decision variable sample space matrix are not worse than those in the previous step; A fourth data calculation module for, after running the maximum number of iteration steps, determining the best sample individual corresponding to the optimal value of the output sample matrix according to the obtained optimal decision variable and using it as the optimal solution.
12. The multidisciplinary design optimization device according to claim 11, characterized in that The data processing module is specifically used for: Normalizing and centralizing the decision variable sample so that all decision variables in the decision variable sample are within the interval [a, b]; Randomly generate m vectors with d-dimensional decision variables within the range [a, b] to form the decision variable sample space matrix X d×m , where m is the sample size of the decision variables.