Component collaborative optimization method in multi-target conflict scene

Abstract

The invention belongs to the field of multi-target optimization and component collaboration, and particularly relates to a component collaboration optimization method in a multi-target conflict scene, which aims at realizing global optimization among system components and comprises the following steps of: receiving input parameters of a multi-target optimization problem, establishing a multi-target conflict quantification model based on the input parameters, and establishing a multi-target conflict quantification model based on the multi-target conflict quantification model; outputting a quantitative index representing the conflict degree between the targets; the output of the quantitative index is obtained, processing is carried out through a weight dynamic adjustment algorithm, and a dynamic weight vector is output; receiving the dynamic weight vector, performing multi-target double-layer optimization in combination with a Pareto optimal principle and a deep reinforcement learning algorithm, and outputting task-specific weight distribution and a Pareto approximate solution set; and an improved sparrow search algorithm is adopted to carry out collaborative optimization among the components, and a global optimization result is output. According to the method, a multi-target conflict quantification model and a multi-target double-layer optimization algorithm are adopted, so that the problems of non-uniform quantification among targets, poor robustness of a single-target optimization problem and the like during multi-target combination optimization are solved.

Description

Technical Field

[0001] This invention belongs to the field of multi-objective optimization and component collaboration, specifically a component collaboration optimization method in multi-objective conflict scenarios. Background Technology

[0002] With the deepening of Industry 4.0 and intelligent manufacturing, the integration of information technology (IT) and operational technology (OT) in process industries is continuously deepening, leading to increasingly complex production systems. Multi-component collaborative optimization has become crucial for improving overall efficiency. Component-based technology, with its advantages of modularity, reusability, and ease of integration, has become an important vehicle for achieving full-process production optimization. However, in practical applications, different components often have their own independent optimization objectives, which may be fundamentally conflicting (such as efficiency versus energy consumption, cost versus quality). Traditional single-objective optimization methods struggle to effectively coordinate these conflicts, resulting in the system's overall performance failing to reach its optimal level.

[0003] The Multi-Objective Optimization Problem (MOOP) aims to find a set of Pareto optimal solutions that balance multiple conflicting objectives. Existing methods mainly include those based on weighted summation, evolutionary algorithms, metaheuristic algorithms, and reinforcement learning-based methods that have emerged in recent years. Among these, linear weighted methods are simple and efficient, but sensitive to weights and difficult to handle non-convex Pareto fronts; evolutionary algorithms (such as NSGA-II and MOEA / D) have good global search capabilities for large-scale problems, but have high computational costs and slow convergence speeds; deep reinforcement learning (DRL) methods can handle dynamic environments, but are highly dependent on the design of reward functions and lack training stability. In addition, existing methods still have shortcomings in multi-objective conflict quantification, dynamic weight adjustment, and inter-component collaboration mechanisms, making it difficult to adapt to the real-time and variable optimization requirements of complex industrial scenarios.

[0004] Therefore, there is an urgent need for an optimization method that can effectively quantify target conflicts, dynamically adjust optimization weights, and achieve efficient collaboration among components, in order to support the process industry in moving from single-point optimization to global optimization and promote the upgrading and development of intelligent manufacturing systems. Summary of the Invention

[0005] The purpose of this invention is to provide a component collaborative optimization method in multi-objective conflict scenarios, which effectively solves the problems of inconsistent quantification among objectives and poor robustness of single-objective optimization problems when optimizing multi-objective combinations, and provides a feasible solution for solving multi-objective combination optimization problems.

[0006] The technical solution adopted by this invention to achieve the above objectives is: a component collaborative optimization method in a multi-objective conflict scenario, which aims to achieve global optimization among system components and includes the following steps:

[0007] S1: Receives input parameters for a multi-objective optimization problem, establishes a multi-objective conflict quantification model based on the input parameters, and outputs a quantitative index characterizing the degree of conflict between objectives;

[0008] S2: Obtain the output of the quantification index, process it through the weight dynamic adjustment algorithm, and output a dynamic weight vector;

[0009] S3: Receive the dynamic weight vector, combine the Pareto optimality principle with the deep reinforcement learning algorithm to perform multi-objective two-layer optimization, and output the task-specific weight distribution and Pareto approximate solution set;

[0010] S4: Obtain the weight distribution and Pareto approximate solution set, perform inter-component collaborative optimization using the improved sparrow search algorithm, and output the global optimization result.

[0011] The establishment of a multi-objective conflict quantification model based on input parameters is specifically as follows:

[0012] Define a multi-objective conflict quantification model, expressed as:

[0013] minF(x)=(f1(x),f2(x),...f m (x))

[0014] x = (x1, x2, ... x) n )∈R n

[0015] Gl(x)={gl(x1),gl(x2),...gl(x p )}

[0016] Htf(x)={htf(x1),htf(x2),...htf(x q )}

[0017] Where x = (x1, x2, ... x n )∈R n It is a vector of decision variables, F(X) is the vector of conflict objectives, f m (x) is the m-th conflict target sub-vector, x n Let R be the i-th decision variable. n Let the space of values ​​for the decision variable vector x be an n-dimensional real space, and Gl(x) denote the set of inequality constraints of the problem, gl(x) i Let Htf(x) be the i-th inequality constraint function, and let Htf(x) be the set of equality constraints in the problem. i Let ) represent the i-th equality constraint function;

[0018] This model achieves multi-objective trade-offs by quantifying the degree of conflict between objectives.

[0019] The conflict intensity and reconcilability indicators are quantified in any of the following ways:

[0020] Using the correlation coefficient method, the correlation coefficient ρ between targets is calculated, where ρ∈[-1,1]. The larger the absolute value, the more obvious the conflict.

[0021] Using the improvement space method, we define the improvement space S for objective i. i S i The larger the value, the greater the optimization potential of the objective.

[0022] Using the maximum and minimum operation minmax i [λ i |f i (x)-Z i |] Quantify the trade-offs between objectives;

[0023] Where, λ i Z is the weight coefficient for the i-th sub-objective, used to represent the importance or priority of the objective in weighted summation or trade-offs; i Let be the expected value of the i-th sub-objective.

[0024] Step S2 specifically includes:

[0025] S2-1: Define a weight vector λ that satisfies:

[0026]

[0027] Where, λ i Let K be the dynamic weight of the i-th sub-objective in the i-th iteration, and K be the total number of sub-objectives that need to be optimized.

[0028] S2-2: Dynamically calculate the dynamic weight vector through gradient response adjustment and load balancing strategies.

[0029] Gradient response adjustment is expressed as:

[0030]

[0031] Where η is the adaptive learning rate. For the i-th sub-objective function f in the t-th iteration i The gradient vector and gradient magnitude of the decision variable x The larger the value, the lower the weight of the target.

[0032] Step S2-2 specifically includes:

[0033] a. The gradient response adjustment is expressed as:

[0034]

[0035] Where η is the adaptive learning rate. For the i-th sub-objective function f in the t-th iteration i The gradient vector and gradient magnitude of the decision variable x The larger the value, the lower the weight of the target.

[0036] b. The aforementioned dynamic weight adjustment algorithm also employs a load balancing strategy, which is expressed as follows:

[0037]

[0038] in, Let U1 and U2 be the load balancing factors for the weight of the i-th sub-objective at the t-th iteration, and let t be the system utility or performance index. response The system response time is denoted by α, β, and γ, which are equilibrium coefficients determined using the analytic hierarchy process (AHP).

[0039] In step S3, the two-layer optimization includes:

[0040] S3-1: Predict task-specific weight distributions using high-level weight generation algorithms;

[0041] S3-2: Calculate the Pareto optimal solution set using a low-level policy optimization algorithm.

[0042] The high-level weight generation algorithm is expressed as follows:

[0043] h t =LSTM(s t ,h t-1 ;θh)

[0044] α t =softmax(W α tanh(Whh t +bh))

[0045]

[0046] Among them, h t To be in a hidden state, s t Given the state input at time t, θ h Here are the LSTM parameters, η is the adaptive learning rate, and α is the α value. t W is the attention weight vector. α W h b h Let be the trainable parameter matrix and vector, and softmax and tanh be the activation functions.

[0047] The low-level strategy optimization algorithm is expressed as follows:

[0048]

[0049] Where ρ is the equilibrium parameter, z * and z nad These are the ideal point and the worst point, respectively, and minmax[.] is the weighted Chebyshev distance.

[0050] In step S4, the improved sparrow search algorithm uses the weight distribution and Pareto approximate solution set output in step S3 as the initial solution space, and specifically includes the following steps:

[0051] S4-1: Generate the initial population based on the Pareto approximate solution set, and use chaotic mapping to enhance the diversity of population distribution;

[0052] S4-2: During the iteration process, the discoverer individual guides the update of solutions with low global conflict degree according to the weight distribution; the follower individual adjusts its own solution vector with reference to the current optimal weight vector; and the vigilant individual avoids getting trapped in local optima by perturbing solutions with high conflict degree.

[0053] S4-3: Through the above multi-role collaborative update, a candidate solution set that can take into account both the target conflict degree and the dynamic nature of the weights is obtained.

[0054] In step S4, the improved sparrow search algorithm further introduces an adaptive weight adjustment mechanism during the iteration process, namely:

[0055] An adaptive inertia weighting factor is introduced, defined as follows:

[0056]

[0057] Where, ω max and ω min These are the initial maximum value and the final minimum value of the inertial weight, respectively; t is the current iteration number; and T is the maximum iteration number.

[0058] After each iteration, the weights are dynamically adjusted and the step size is updated based on the distribution entropy of the solution set, thereby improving the convergence accuracy.

[0059] Meanwhile, the non-dominated solutions are screened by combining congestion degree calculation to ensure that the candidate solutions are evenly distributed in the target space; finally, the optimized Pareto optimal solution set is output, which provides highly adaptable decision support for component collaboration in multi-objective conflict scenarios.

[0060] The present invention has the following beneficial effects and advantages:

[0061] 1. This invention achieves global optimization and fusion through local information interaction by improving the sparrow search algorithm, thus realizing a leap from single-point optimization to global optimization.

[0062] 2. This invention employs a multi-objective conflict quantization model and a multi-objective bi-layer optimization algorithm to effectively solve the problems of inconsistent quantization among objectives in multi-objective combined optimization and poor robustness of single-objective optimization problems.

[0063] 3. This invention achieves effective quantification of multi-objective conflicts, dynamic weight adjustment, two-level optimization solution, and inter-component collaboration, significantly improving the robustness, adaptability, and global optimization capability of the optimization system, and is applicable to multi-objective collaborative optimization problems in complex industrial environments. Attached Figure Description

[0064] Figure 1 This is a flowchart of the component collaborative optimization method of the present invention. Detailed Implementation

[0065] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0066] like Figure 1 The diagram shown is a flowchart of the component collaborative optimization method of the present invention. The present invention provides a component collaborative optimization method in a multi-objective conflict scenario. The method aims to achieve global optimization among system components and includes the following steps:

[0067] S1: Receives input parameters for a multi-objective optimization problem, establishes a multi-objective conflict quantification model based on the input parameters, and outputs a quantitative index characterizing the degree of conflict between objectives;

[0068] A multi-objective conflict quantification model is established based on the input parameters, specifically as follows:

[0069] Define a multi-objective conflict quantification model, expressed as:

[0070] minF(x)=(f1(x),f2(x),...f m (x))

[0071] x = (x1, x2, ... x) n )∈R n

[0072] Gl(x)={gl(x1),gl(x2),...gl(x p )}

[0073] Htf(x)={htf(x1),htf(x2),...htf(x q )}

[0074] Where x = (x1, x2, ... x n )∈R n It is a vector of decision variables, F(X) is the vector of conflict objectives, f m (x) is the m-th conflict target sub-vector, xn Let R be the i-th decision variable. n Let the space of values ​​for the decision variable vector x be an n-dimensional real space, and Gl(x) denote the set of inequality constraints of the problem, gl(x) i Let Htf(x) be the i-th inequality constraint function, and let Htf(x) be the set of equality constraints in the problem. i Let ) represent the i-th equality constraint function;

[0075] This model achieves multi-objective trade-offs by quantifying the degree of conflict between objectives.

[0076] The intensity and reconcilability of conflict can be quantified in any of the following ways:

[0077] Using the correlation coefficient method, the correlation coefficient ρ between targets is calculated, where ρ∈[-1,1]. The larger the absolute value, the more obvious the conflict.

[0078] Using the improvement space method, we define the improvement space S for objective i. i S i The larger the value, the greater the optimization potential of the objective.

[0079] Using the maximum and minimum operation minmax i [λ i |f i (x)-Z i |] Quantify the trade-offs between objectives;

[0080] Where, λ i Z is the weight coefficient for the i-th sub-objective, used to represent the importance or priority of the objective in weighted summation or trade-offs; i Let be the expected value of the i-th sub-objective.

[0081] S2: Obtain the output of the quantification index, process it through the weight dynamic adjustment algorithm, and output a dynamic weight vector;

[0082] S2-1: Define a weight vector λ that satisfies:

[0083]

[0084] Where, λ i Let K be the dynamic weight of the i-th sub-objective in the i-th iteration, and K be the total number of sub-objectives that need to be optimized.

[0085] S2-2: Dynamically calculate the dynamic weight vector through gradient response adjustment and load balancing strategies.

[0086] Gradient response adjustment is expressed as:

[0087]

[0088] Where η is the adaptive learning rate. For the i-th sub-objective function f in the t-th iteration i The gradient vector and gradient magnitude of the decision variable x The larger the value, the lower the weight of the target.

[0089] Step S2-2 is as follows:

[0090] a. The gradient response adjustment is expressed as:

[0091]

[0092] Where η is the adaptive learning rate. For the i-th sub-objective function f in the t-th iteration i The gradient vector and gradient magnitude of the decision variable x The larger the value, the lower the weight of the target.

[0093] b. The dynamic weight adjustment algorithm also employs a load balancing strategy, which is expressed as follows:

[0094]

[0095] in, Let U1 and U2 be the load balancing factors for the weight of the i-th sub-objective at the t-th iteration, and let t be the system utility or performance index. response The system response time is denoted by α, β, and γ, which are equilibrium coefficients determined using the analytic hierarchy process (AHP).

[0096] S3: Receives dynamic weight vectors, combines Pareto optimality principle with deep reinforcement learning algorithm to perform multi-objective two-layer optimization, and outputs task-specific weight distribution and Pareto approximate solution set;

[0097] Two-layer optimization, including:

[0098] S3-1: Predict task-specific weight distributions using high-level weight generation algorithms;

[0099] S3-2: Calculate the Pareto optimal solution set using a low-level policy optimization algorithm.

[0100] The high-level weight generation algorithm is expressed as follows:

[0101] h t =LSTM(s t ,h t-1 ;θh)

[0102] α t =softmax(W α tanh(W h ht +b h ))

[0103]

[0104] Among them, h t To be in a hidden state, s t Given the state input at time t, θ h Here are the LSTM parameters, η is the adaptive learning rate, and α is the α value. t W is the attention weight vector. α W h b h Let be the trainable parameter matrix and vector, and softmax and tanh be the activation functions.

[0105] The low-level policy optimization algorithm is represented as:

[0106]

[0107] Where ρ is the equilibrium parameter, z * and z nad These are the ideal point and the worst point, respectively, and minmax[.] is the weighted Chebyshev distance.

[0108] S4: Obtain the weight distribution and Pareto approximate solution set, perform inter-component collaborative optimization using the improved sparrow search algorithm, and output the global optimization result.

[0109] In step S4, the improved sparrow search algorithm uses the weight distribution and Pareto approximate solution set output in step S3 as the initial solution space, and specifically includes the following steps:

[0110] S4-1: Generate the initial population based on the Pareto approximate solution set, and use chaotic mapping to enhance the diversity of population distribution;

[0111] S4-2: During the iteration process, the discoverer individual guides the update of solutions with low global conflict degree according to the weight distribution; the follower individual adjusts its own solution vector with reference to the current optimal weight vector; and the vigilant individual avoids getting trapped in local optima by perturbing solutions with high conflict degree.

[0112] S4-3: Through the above multi-role collaborative update, a candidate solution set that can take into account both the target conflict degree and the dynamic nature of the weights is obtained.

[0113] In step S4, the improved sparrow search algorithm further introduces an adaptive weight adjustment mechanism during the iteration process, namely:

[0114] An adaptive inertia weighting factor is introduced, defined as follows:

[0115]

[0116] Where, ω max and ω min These are the initial maximum value and the final minimum value of the inertial weight, respectively; t is the current iteration number; and T is the maximum iteration number.

[0117] After each iteration, the weights are dynamically adjusted and the step size is updated based on the distribution entropy of the solution set, thereby improving the convergence accuracy.

[0118] Meanwhile, the non-dominated solutions are screened by combining congestion degree calculation to ensure that the candidate solutions are evenly distributed in the target space; finally, the optimized Pareto optimal solution set is output, which provides highly adaptable decision support for component collaboration in multi-objective conflict scenarios.

[0119] Example

[0120] Taking production scheduling in a smart manufacturing workshop as an example, the system needs to simultaneously optimize three conflicting objectives: production efficiency, energy consumption, and equipment wear. After applying the method of this invention, the system can dynamically adjust the weights of each objective, generate a high-quality Pareto solution set, and achieve collaborative optimization of each production component through an improved sparrow search algorithm. Ultimately, while ensuring production efficiency, it significantly reduces energy consumption and equipment wear, verifying the effectiveness and practicality of this invention.

[0121] In summary, based on the embodiments of this invention, this invention formally defines indicators such as conflict intensity and reconciliation through a multi-objective conflict quantification model and conflict quantification method. Combined with a dynamic weight adjustment algorithm and a multi-objective bi-layer optimization algorithm, it achieves optimization from single-point to global levels, significantly improving the quality and diversity of Pareto solutions and providing an efficient and robust solution for end-to-end intelligent optimization. The system can adapt to the dynamic changes of complex industrial scenarios, possessing broad applicability and significant engineering value. This invention fills a technological gap in the field of component collaborative optimization, providing core technological support for the digital transformation and intelligent manufacturing upgrade of the process industry.

[0122] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. This is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0123] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A component collaborative optimization method for multi-objective conflict scenarios, characterized in that, This method aims to achieve global optimization among system components and includes the following steps: S1: Receives input parameters for a multi-objective optimization problem, establishes a multi-objective conflict quantification model based on the input parameters, and outputs a quantitative index characterizing the degree of conflict between objectives; S2: Obtain the output of the quantification index, process it through the weight dynamic adjustment algorithm, and output a dynamic weight vector; S3: Receive the dynamic weight vector, combine the Pareto optimality principle with the deep reinforcement learning algorithm to perform multi-objective two-layer optimization, and output the task-specific weight distribution and Pareto approximate solution set; S4: Obtain the weight distribution and Pareto approximate solution set, perform inter-component collaborative optimization using the improved sparrow search algorithm, and output the global optimization result.

2. The component collaborative optimization method in a multi-objective conflict scenario according to claim 1, characterized in that, The establishment of a multi-objective conflict quantification model based on input parameters is specifically as follows: Define a multi-objective conflict quantification model, expressed as: minF(x)=(f1(x),f2(x),...f m (x)) x=(x1,x2,...x n )∈R n Gl(x)={gl(x1),gl(x2),...gl(x p )} Htf(x)={htf(x1),htf(x2),...htf(x q )} Where x = (x1, x2, ... x n )∈R n It is a vector of decision variables, F(X) is the vector of conflict objectives, f m (x) is the m-th conflict target sub-vector, x n Let R be the i-th decision variable. n Let the space of values ​​for the decision variable vector x be an n-dimensional real space, and Gl(x) denote the set of inequality constraints of the problem, gl(x) i Let Htf(x) be the i-th inequality constraint function, and let Htf(x) be the set of equality constraints in the problem. i Let ) represent the i-th equality constraint function; This model achieves multi-objective trade-offs by quantifying the degree of conflict between objectives.

3. The component collaborative optimization method in a multi-objective conflict scenario according to claim 1, characterized in that, The conflict intensity and reconcilability indicators are quantified in any of the following ways: Using the correlation coefficient method, the correlation coefficient ρ between targets is calculated, where ρ∈[-1,1]. The larger the absolute value, the more obvious the conflict. Using the improvement space method, we define the improvement space S for objective i. i S i The larger the value, the greater the optimization potential of the objective. Minmax is used as the maximum and minimum operation. i [λ i |f i (x)-Z i |] Quantify the trade-offs between objectives; Where, λ i Z is the weight coefficient for the i-th sub-objective, used to represent the importance or priority of the objective in weighted summation or trade-offs; i Let be the expected value of the i-th sub-objective.

4. The component collaborative optimization method in a multi-objective conflict scenario according to claim 1, characterized in that, Step S2 specifically includes: S2-1: Define a weight vector λ that satisfies: Where, λ i Let K be the dynamic weight of the i-th sub-objective in the i-th iteration, and K be the total number of sub-objectives that need to be optimized. S2-2: Dynamically calculate the dynamic weight vector through gradient response adjustment and load balancing strategies. Gradient response adjustment is expressed as: Where η is the adaptive learning rate. For the i-th sub-objective function f in the t-th iteration i The gradient vector and gradient magnitude of the decision variable x The larger the value, the lower the weight of the target.

5. The component collaborative optimization method in a multi-objective conflict scenario according to claim 1, characterized in that, Step S2-2 specifically includes: a. The gradient response adjustment is expressed as: Where η is the adaptive learning rate. For the i-th sub-objective function f in the t-th iteration i The gradient vector and gradient magnitude of the decision variable x The larger the value, the lower the weight of the target. b. The aforementioned dynamic weight adjustment algorithm also employs a load balancing strategy, which is expressed as follows: in, Let U1 and U2 be the load balancing factors for the weight of the i-th sub-objective at the t-th iteration, and let t be the system utility or performance index. response The system response time is denoted by α, β, and γ, which are equilibrium coefficients determined using the analytic hierarchy process (AHP).

6. The component collaborative optimization method in a multi-objective conflict scenario according to claim 1, characterized in that, In step S3, the two-layer optimization includes: S3-1: Predict task-specific weight distributions using high-level weight generation algorithms; S3-2: Calculate the Pareto optimal solution set using a low-level policy optimization algorithm.

7. The component collaborative optimization method in a multi-objective conflict scenario according to claim 6, characterized in that, The high-level weight generation algorithm is expressed as follows: h t =LSTM(s t ,h t-1 ;θ h ) α t =softmax(W α tanh(W h h t +b h )) Among them, h t To be in a hidden state, s t Given the state input at time t, θ h Here are the LSTM parameters, η is the adaptive learning rate, and α is the α value. t W is the attention weight vector. α W h b h Let be the trainable parameter matrix and vector, and softmax and tanh be the activation functions.

8. The component collaborative optimization method in a multi-objective conflict scenario according to claim 6, characterized in that, The low-level strategy optimization algorithm is expressed as follows: Where ρ is the equilibrium parameter, z * and z nad These are the ideal point and the worst point, respectively, and minmax[.] is the weighted Chebyshev distance.

9. The component collaborative optimization method in a multi-objective conflict scenario according to claim 1, characterized in that, In step S4, the improved sparrow search algorithm uses the weight distribution and Pareto approximate solution set output in step S3 as the initial solution space, and specifically includes the following steps: S4-1: Generate the initial population based on the Pareto approximate solution set, and use chaotic mapping to enhance the diversity of population distribution; S4-2: During the iteration process, the discoverer individual guides the update of solutions with low global conflict degree according to the weight distribution; the follower individual adjusts its own solution vector with reference to the current optimal weight vector; and the vigilant individual avoids getting trapped in local optima by perturbing solutions with high conflict degree. S4-3: Through the above multi-role collaborative update, a candidate solution set that can take into account both the target conflict degree and the dynamic nature of the weights is obtained.

10. A component collaborative optimization method in a multi-objective conflict scenario according to claim 9, characterized in that, In step S4, the improved sparrow search algorithm further introduces an adaptive weight adjustment mechanism during the iteration process, namely: An adaptive inertia weighting factor is introduced, defined as follows: Where, ω max and ω min These are the initial maximum value and the final minimum value of the inertial weight, respectively; t is the current iteration number; and T is the maximum iteration number. After each iteration, the weights are dynamically adjusted and the step size is updated based on the distribution entropy of the solution set, thereby improving the convergence accuracy. Meanwhile, the non-dominated solutions are screened by combining congestion degree calculation to ensure that the candidate solutions are evenly distributed in the target space; finally, the optimized Pareto optimal solution set is output, which provides highly adaptable decision support for component collaboration in multi-objective conflict scenarios.

Images