Aero-engine design method based on multi-agent conflict negotiation mechanism

Abstract

The invention relates to the technical field of complex equipment collaborative design, and discloses an aero-engine design method based on a multi-agent conflict negotiation mechanism, and the method comprises the steps: constructing a collaborative architecture of a global collaborative platform, an overall agent (a first agent), and a distributed component agent (a second agent of a corresponding component); the global collaboration platform serves as an information interaction and coordination center and is responsible for receiving design schemes of all agents and organizing overall performance evaluation; the overall agent gives initial interface parameters and performance indexes according to a design task book and historical cases to guide the design direction; and the distributed component intelligent agent focuses on the optimization design of each component, and continuously adjusts and improves the design scheme according to the feedback of the global collaborative platform and the guidance of the overall intelligent agent. The collaborative architecture not only improves the design efficiency, but also enhances the robustness and optimization effect of the design.

Description

Technical Field

[0001] This invention relates to the field of collaborative design technology for complex equipment, and discloses an aero-engine design method based on a multi-agent conflict negotiation mechanism. Background Technology

[0002] In the collaborative design of complex systems such as aero-engines, there are problems such as strong coupling of design objectives among components (e.g., compressor, combustor, turbine), numerous internal geometric parameters, and difficulties in interface matching. Existing technologies often struggle to efficiently resolve parameter conflicts and iterative convergence stagnation caused by multidisciplinary coupling while simultaneously considering the design parameters and feasibility (design difficulty, R&D cycle) of all components. In particular, traditional numerical optimization methods are prone to getting trapped in local optima and often ignore the engineering implementation costs of excessively high design specifications, resulting in calculated solutions that are "theoretically good but difficult to implement in engineering" or "have uncontrollable R&D cycles." Summary of the Invention

[0003] The purpose of this invention is to provide an aero-engine design method based on a multi-agent conflict negotiation mechanism, which effectively solves the problem of non-convergence in multi-disciplinary coupled design iteration in the collaborative design of complex equipment, and realizes the intelligent accumulation of design experience.

[0004] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is as follows:

[0005] An aero-engine design method based on a multi-agent conflict negotiation mechanism includes: Using the overall engine performance requirements and historical design cases from the aero-engine design task book as input, a first intelligent agent is used to provide initial values ​​for the interface parameters and performance indicators of each component; the overall engine performance requirements include the lower limit of thrust and the upper limit of fuel consumption rate, and the performance indicators include compressor pressure ratio, flow rate, efficiency, and combustion chamber total pressure recovery coefficient. Based on the current interface parameters and performance indicators, the second intelligent agent of each component submits the design scheme of the corresponding component to the global collaborative platform, including interface parameters, performance parameters, and geometric parameters; The first intelligent agent conducts an overall performance evaluation of the engine based on the design schemes provided by the second intelligent agents of each component, and obtains the global performance including thrust prediction and fuel consumption rate prediction. If the residuals of the interface parameters of each component are less than the preset residual threshold, and the margin between the global performance and the design task requirements is greater than the preset margin threshold, then the design is determined to be converged and the design result is output; otherwise, the first agent regenerates new interface parameters and performance indicators, and the second agent resubmits the iterative design scheme based on the new interface parameters and performance indicators until the design converges and the design result is output.

[0006] Furthermore, the methods for the first intelligent agent to regenerate new interface parameters and performance metrics include: Based on the initial values ​​of the performance indicators and the cumulative sample set formed by the performance indicators in the iterative process, the SHAP method is used to calculate the marginal contribution of each performance indicator in the sample set to the conflicting whole machine indicators. The conflicting whole machine indicators are two whole machine indicators that are negatively correlated. Multiple performance indicators whose contribution to each conflicting overall system indicator exceeds a corresponding preset threshold are selected as key variables to construct a low-dimensional subspace; Using all the key variables in the low-dimensional subspace as input, a multi-objective optimization model for design cost and constraint violation is constructed. An evolutionary algorithm is used to generate a set of non-dominated solutions with the minimum design cost and the minimum constraint violation, forming candidate solutions for interface parameters and performance indicators. The first agent selects new interface parameters and performance indicators from the candidate solutions and sends them to the corresponding second agent.

[0007] Furthermore, the design cost is defined as a measure of the design difficulty that each component agent needs to undertake to achieve the interface parameters and performance indicators. Specifically, it is calculated as a weighted sum of the component performance difficulty value and the interface parameter difficulty value. The component performance difficulty value is calculated based on the Euclidean distance of the component's performance parameter value relative to a given optimal design performance limit. The interface parameter difficulty value is calculated based on the distance of the interface parameter relative to the optimal design interface parameter limit under given fluid parameters. The fluid parameters include the interface flow rate and the interface parameters include the flow area. The constraint violation degree is defined as the degree of violation of the design scheme in terms of interface parameter consistency and global performance indicators. Specifically, it is calculated as the weighted sum of interface consistency residual and global performance violation amount. The interface consistency residual is the Euclidean distance between the actual interface parameters output by each component and the target interface parameters, and the global performance violation amount is the degree to which the predicted value of the whole machine performance deviates from the threshold set in the design task.

[0008] Furthermore, the second agent is an agent with a preset penalty function, and the first agent selects new interface parameters and performance indicators from the candidate solutions based on the Nash negotiation decision process, specifically including: Define utility functions in the first intelligent agent and the second intelligent agent for each component design. The utility function is used to quantify the degree to which the current design scheme meets the performance indicators of each party. The utility function of the first intelligent agent is the utility function of the whole machine indicators including thrust and fuel consumption rate, and the utility function of each second intelligent agent is the function of the corresponding component efficiency and pressure ratio performance indicators. Define the utility gain of each agent. ,in For the first The agent in the th... The design scheme of the second negotiation is based on the utility value calculated according to the predefined utility function; For the first intelligent agent, the minimum performance target required to meet the design specifications is the minimum performance target that must be achieved; for each second intelligent agent, the maximum design difficulty boundary or performance limit that can be met by physical feasibility. Update the weighting of each agent. , For the first The agent in the th... The importance weight of each negotiation, the initial weight , For the first The agent in the th... The penalty function value after the second negotiation, and the initial penalty parameters. For preset input values, For the first The agent in the th... The residual value of the second negotiation, for the first agent, This represents the margin deviation between the global performance metrics and the design specifications; for each second agent, The interface parameter residual is the Euclidean distance between the actual interface parameters submitted by the component and the target interface parameters. With updated importance weight As the exponential weight of the Nash product, a weighted Nash set calculation model based on utility gain is constructed to find the solution that maximizes the weighted Nash product among the candidate solutions; The solution that maximizes the weighted Nash product This serves as a second intelligent agent, distributing new interface parameters and performance metrics to each component.

[0009] Furthermore, the weighted Nash product calculation model based on utility gain is as follows: .

[0010] Furthermore, it also includes using a large language model, combined with the contribution of key variables obtained from SHAP analysis and the weight allocation logic determined by Nash negotiation, to generate a design log in natural language form; the design log explains the reasons for selecting specific interface parameters and performance indicators, including why specific key variables were selected and why a balance of interests was reached at that point.

[0011] Furthermore, the first agent is equipped with an iterative memory to store the state, the strategy adopted, and the consequences of the historical iteration steps. Before the first agent regenerates new interface parameters and performance indicators, it uses a large language model to retrieve historical iteration records in the memory where the cosine similarity is greater than the similarity threshold. If it is found that the strategy to be adopted at present has caused parameter oscillation or non-convergence in the past, it calls on the historical successful experience to correct the strategy and avoid repeating the same mistakes.

[0012] Furthermore, after receiving the interface parameters and performance indicators, each component agent performs the following target tracking optimization process: Under the premise of satisfying its own physical constraints, the internal geometric parameters are optimized with the goal of minimizing the Euclidean distance between the interface parameters of the current design scheme and the interface parameters and performance indicators issued by the first intelligent agent.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: by constructing a collaborative architecture of "global collaborative platform + overall intelligent agent (first intelligent agent) + distributed component intelligent agent (second intelligent agent of corresponding component)," the present invention effectively solves the problem of non-convergence of multi-disciplinary coupled design iteration in the collaborative design of complex equipment, realizes the intelligent accumulation of design experience, not only improves design efficiency, but also enhances the robustness and optimization effect of the design. Attached Figure Description

[0014] Figure 1 This is a flowchart of the aero-engine design method based on a multi-agent conflict negotiation mechanism in Example 1; Figure 2 This is a flowchart of the aero-engine design method based on a multi-agent conflict negotiation mechanism in Example 2. Detailed Implementation

[0015] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0016] Example 1 See Figure 1 An aero-engine design method based on a multi-agent conflict negotiation mechanism includes: Using the overall engine performance requirements and historical design cases from the aero-engine design task book as input, a first intelligent agent is used to provide initial values ​​for the interface parameters and performance indicators of each component; the overall engine performance requirements include the lower limit of thrust and the upper limit of fuel consumption rate, and the performance indicators include compressor pressure ratio, flow rate, efficiency, and combustion chamber total pressure recovery coefficient. Based on the current interface parameters and performance indicators, the second intelligent agent of each component submits the design scheme of the corresponding component to the global collaborative platform, including interface parameters, performance parameters, and geometric parameters; The first intelligent agent conducts an overall performance evaluation of the engine based on the design schemes provided by the second intelligent agents of each component, and obtains the global performance including thrust prediction and fuel consumption rate prediction. If the residuals of the interface parameters of each component are less than the preset residual threshold, and the margin between the global performance and the design task requirements is greater than the preset margin threshold, then the design is determined to be converged and the design result is output; otherwise, the first agent regenerates new interface parameters and performance indicators, and the second agent resubmits the iterative design scheme based on the new interface parameters and performance indicators until the design converges and the design result is output.

[0017] In this embodiment, a collaborative architecture is constructed consisting of a "global collaboration platform + overall intelligent agent (first intelligent agent) + distributed component intelligent agents (second intelligent agents corresponding to the components)". The global collaboration platform serves as the center for information interaction and coordination, responsible for receiving design schemes from each intelligent agent and organizing overall performance evaluation. The overall intelligent agent, based on the design task and historical cases, provides initial interface parameters and performance indicators to guide the design direction. The distributed component intelligent agents focus on the optimization design of their respective components, continuously adjusting and improving the design schemes based on feedback from the global collaboration platform and guidance from the overall intelligent agent. This collaborative architecture not only improves design efficiency but also enhances the robustness and optimization effect of the design.

[0018] Example 2 See Figure 2 This embodiment takes the collaborative design of the compressor and combustion chamber of an aero-engine as an example to describe in detail the aero-engine design method based on a multi-agent conflict negotiation mechanism of the present invention. The design process is as follows: Step 1: Using the overall engine performance requirements and historical design cases from the aero-engine design task book as input, the first intelligent agent provides the initial values ​​of the interface parameters and performance indicators of each component; the overall engine performance requirements include the lower limit of thrust and the upper limit of fuel consumption rate, and the performance indicators include the compressor pressure ratio, flow rate, efficiency, and the total pressure recovery coefficient of the combustion chamber. In this embodiment, based on historical excellent design cases or overall model calculations, initial values ​​of interface parameters and initial values ​​of performance indicators of each component are given.

[0019] Step 2: Based on the current interface parameters and performance indicators, the second intelligent agent of each component submits the design scheme of the corresponding component to the global collaboration platform, including interface parameters, performance parameters, and geometric parameters. In this embodiment, the second intelligent agent of each component submits a proposal to the global collaboration platform based on the current internal design state, the initial values ​​of interface parameters, and the initial performance target values ​​of each component: Parallel design: The second agent corresponding to the compressor, combustion chamber and other components performs preliminary parallel design based on the interface parameters and initial values ​​of performance indicators transmitted by the first agent, and submits message interface parameters (such as calculated outlet flow field parameters), performance indicators or performance characteristic diagrams (such as efficiency and margin corresponding to the current design point).

[0020] Step 3: Conduct an overall performance evaluation of the engine based on the design scheme provided by the second intelligent agent of each component, and obtain the global performance including thrust prediction value and fuel consumption rate prediction value.

[0021] Step 4: State Assessment and Convergence Determination 4.1 Calculate the original residuals of the interface parameters of each component. And calculate the global performance metrics and the performance margin required by the design specifications. ,in This is a global performance metric vector. This is a vector of performance indicators required by the design task specification.

[0022] 4.2 Determine whether the residuals of the interface parameters of each component are less than the preset residual threshold, and whether the margin of global performance and design task requirements is greater than the preset margin threshold. If yes, the design is considered converged and the design result is output. If no, the first agent regenerates new interface parameters and performance indicators, and the second agent resubmits the iterative design scheme based on the new interface parameters and performance indicators until the design converges and the design result is output.

[0023] In this embodiment, the method for the first intelligent agent to regenerate new interface parameters and performance indicators includes: 4.2.1 Based on the initial values ​​of the performance indicators and the cumulative sample set formed by the performance indicators in the iterative process, the SHAP method is used to calculate the marginal contribution of each performance indicator in the sample set to the conflicting overall system indicators. The conflicting overall system indicators are two overall system indicators that are negatively correlated, such as the conflict between high thrust and low fuel consumption, or the conflict between high thrust and long life. 4.2.2 Select multiple performance indicators whose contribution to each conflicting overall system indicator exceeds the corresponding preset threshold as key variables to construct a low-dimensional subspace; 4.2.3 Using all the key variables in the low-dimensional subspace as input, a multi-objective optimization model for design cost and constraint violation is constructed. An evolutionary algorithm is used to generate a set of non-dominated solutions with the minimum design cost and the minimum constraint violation, forming candidate solutions for interface parameters and performance indicators. In this embodiment, the design cost is defined as a measure of the design difficulty that each component agent needs to undertake to achieve the interface parameters and performance indicators. Specifically, it is calculated as a weighted sum of the component performance difficulty value and the interface parameter difficulty value. The component performance difficulty value is calculated based on the Euclidean distance of the component's efficiency, pressure ratio, and other performance values ​​relative to the optimal design performance limit. The interface parameter difficulty value is calculated based on the distance of the interface parameter (e.g., flow area) relative to the optimal design interface parameter limit under a given fluid parameter (e.g., flow rate). The constraint violation degree is defined as the degree of violation of the design scheme in terms of interface parameter consistency and global performance indicators. Specifically, it is calculated as the weighted sum of interface consistency residual and global performance violation amount. The interface consistency residual is the Euclidean distance between the actual interface parameters output by each component and the target interface parameters. The global performance violation amount is the degree to which the predicted value of the overall machine performance (such as thrust and fuel consumption rate) deviates from the threshold set in the design task.

[0024] 4.2.4 The first intelligent agent selects new interface parameters and performance indicators from the candidate solutions and sends them to the corresponding second intelligent agent.

[0025] In some other embodiments, the second agent is an agent with a preset penalty function, and the first agent selects new interface parameters and performance indicators from candidate solutions based on a Nash negotiation decision-making process, specifically including: 1) Define utility functions in the first agent and the second agent for each component design. The utility function is used to quantify the degree to which the current design scheme meets the performance indicators of each party. The utility function of the first intelligent agent is a function of performance indicators such as thrust and fuel consumption rate, and the utility function of each second intelligent agent is a function of the component's efficiency and pressure ratio performance indicators. 2) Define the utility gain for each agent. ,in For the first The agent in the th... The design scheme of the second negotiation is based on the utility value calculated according to the predefined utility function; For the first intelligent agent, the minimum performance target required to meet the design specifications is the minimum performance target that must be achieved; for the second intelligent agent, the maximum design difficulty boundary or performance limit that can be tolerated by physical feasibility. 3) Update the weighting of each component. , For the first The agent in the th... The importance weight of each negotiation, the initial weight , For the first The agent in the th... The penalty function value after the second negotiation, and the initial penalty parameters. For preset input values, For the first The agent in the th... The residual value of the second negotiation, for the first agent, This represents the margin deviation between the global performance metrics and the design specifications; for each second agent, The interface parameter residual is the Euclidean distance between the actual interface parameters submitted by the component and the target interface parameters. 4) Using the updated importance weights As the exponential weight of the Nash product, a weighted Nash set calculation model based on utility gain is constructed to find the solution that maximizes the weighted Nash product among the candidate solutions; 5) Maximize the solution of the weighted Nash product This serves as a second intelligent agent, distributing new interface parameters and performance metrics to each component.

[0026] In some other embodiments, the first agent is provided with an iterative memory to store the state, the strategy adopted, and the consequences of the historical iteration steps. Before the first agent generates new interface parameters and performance indicators through a negotiation mechanism, the large language model is used to retrieve historical iteration records in the memory where the cosine similarity is greater than the similarity threshold. If it is found that the strategy to be adopted at present has caused parameter oscillation or non-convergence in the past, the successful experience of the past is called to correct the strategy and avoid repeating the same mistakes.

[0027] Furthermore, a design log in natural language form can be generated by leveraging a large language model (LLM) and combining the contribution of key variables obtained from SHAP analysis with the weight allocation logic determined by Nash negotiation. This design log explains the reasons for selecting specific interface parameters and performance metrics, including why specific key variables were selected and why a balance of interests was reached at that point. For example, if a large language model (LLM) retrieves historical memory and confirms that the strategy has not triggered oscillations, a log is generated suggesting adjustments to the interface parameters and performance metrics. Reason: Although some compressor efficiency was sacrificed, according to current... Based on the weighting, this scheme represents the only 'Nash equilibrium point' that satisfies the combustion chamber emission threshold. Historical experience indicates that this path converges the fastest.

[0028] In some other embodiments, after receiving new interface parameters and performance metrics, each component agent performs the following target tracking optimization process: Under the premise of satisfying its own physical constraints, the internal geometric parameters are optimized with the goal of minimizing the Euclidean distance between the interface parameters of the current design scheme and the interface parameters and performance indicators issued by the first intelligent agent:

[0029] in For components The current interface parameters, The new target issued to the first intelligent agent.

[0030] This embodiment proposes a conflict resolution method that integrates Explainable Artificial Intelligence (XAI) dimensionality reduction, multi-objective agent optimization, and Nash game decision-making. This mechanism introduces an Episodic Memory module during the negotiation process, utilizes XAI technology to screen key variables in the high-dimensional design space, and directly maps the adaptive constraint weights generated during the convergence determination phase to exponential weights in the weighted Nash negotiation model. Finally, by solving the Nash product maximization problem, it automatically searches for the globally optimal solution that balances physical performance matching and the lowest engineering implementation difficulty, and generates an interpretable design decision log. This effectively solves the problem of non-convergence in multi-disciplinary coupled design iterations in the collaborative design of complex equipment, and realizes the intelligent accumulation of design experience.

[0031] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An aero-engine design method based on a multi-agent conflict negotiation mechanism, characterized in that, include: Using the overall engine performance requirements and historical design cases from the aero-engine design task book as input, a first intelligent agent is used to provide initial values ​​for the interface parameters and performance indicators of each component; the overall engine performance requirements include the lower limit of thrust and the upper limit of fuel consumption rate, and the performance indicators include compressor pressure ratio, flow rate, efficiency, and combustion chamber total pressure recovery coefficient. Based on the current interface parameters and performance indicators, the second intelligent agent of each component submits the design scheme of the corresponding component to the global collaborative platform, including interface parameters, performance parameters, and geometric parameters; The first intelligent agent conducts an overall performance evaluation of the engine based on the design schemes provided by the second intelligent agents of each component, and obtains the global performance including thrust prediction and fuel consumption rate prediction. If the residuals of the interface parameters of each component are less than the preset residual threshold, and the margin of global performance and design task requirements is greater than the preset margin threshold, then the design is determined to be converged and the design result is output. Otherwise, the first agent regenerates new interface parameters and performance indicators, and the second agent resubmits the iterative design scheme based on the new interface parameters and performance indicators until the design converges and the design result is output.

2. The aero-engine design method according to claim 1, characterized in that, Methods for the first intelligent agent to regenerate new interface parameters and performance metrics include: Based on the initial values ​​of the performance indicators and the cumulative sample set formed by the performance indicators in the iterative process, the SHAP method is used to calculate the marginal contribution of each performance indicator in the sample set to the conflicting whole machine indicators. The conflicting whole machine indicators are two whole machine indicators that are negatively correlated. Multiple performance indicators whose contribution to each conflicting overall system indicator exceeds a corresponding preset threshold are selected as key variables to construct a low-dimensional subspace; Using all the key variables in the low-dimensional subspace as input, a multi-objective optimization model for design cost and constraint violation is constructed. An evolutionary algorithm is used to generate a set of non-dominated solutions with the minimum design cost and the minimum constraint violation, forming candidate solutions for interface parameters and performance indicators. The first agent selects new interface parameters and performance indicators from the candidate solutions and sends them to the corresponding second agent.

3. The aero-engine design method according to claim 2, characterized in that, The design cost is defined as a measure of the design difficulty that each component agent needs to undertake to achieve the interface parameters and performance indicators. Specifically, it is calculated as a weighted sum of the component performance difficulty value and the interface parameter difficulty value. The component performance difficulty value is calculated based on the Euclidean distance of the component's performance parameter value relative to a given optimal design performance limit. The interface parameter difficulty value is calculated based on the distance of the interface parameter relative to the optimal design interface parameter limit under given fluid parameters. The fluid parameters include the interface flow rate and the interface parameters include the flow area. The constraint violation degree is defined as the degree of violation of the design scheme in terms of interface parameter consistency and global performance indicators. Specifically, it is calculated as the weighted sum of interface consistency residual and global performance violation amount. The interface consistency residual is the Euclidean distance between the actual interface parameters output by each component and the target interface parameters, and the global performance violation amount is the degree to which the predicted value of the whole machine performance deviates from the threshold set in the design task.

4. The aero-engine design method according to claim 2, characterized in that, The second agent is an agent with a preset penalty function. The first agent selects new interface parameters and performance indicators from the candidate solutions based on the Nash negotiation decision process, specifically including: Define utility functions in the first intelligent agent and the second intelligent agent for each component design. The utility function is used to quantify the degree to which the current design scheme meets the performance indicators of each party. The utility function of the first intelligent agent is the utility function of the whole machine indicators including thrust and fuel consumption rate, and the utility function of each second intelligent agent is the function of the corresponding component efficiency and pressure ratio performance indicators. Define the utility gain of each agent. ,in For the first The agent in the th... The design scheme of the second negotiation is based on the utility value calculated according to the predefined utility function; For the first intelligent agent, the minimum performance target required to meet the design specifications is the minimum performance target that must be achieved; for each second intelligent agent, the maximum design difficulty boundary or performance limit that can be met by physical feasibility. Update the weighting of each agent. , For the first The agent in the th... The importance weight of each negotiation, the initial weight , For the first The agent in the th... The penalty function value after the second negotiation, and the initial penalty parameters. For preset input values, For the first The agent in the th... The residual value of the second negotiation, for the first agent, This represents the margin deviation between the global performance metrics and the design specifications; for each second agent, The interface parameter residual is the Euclidean distance between the actual interface parameters submitted by the component and the target interface parameters. With updated importance weight As the exponential weight of the Nash product, a weighted Nash set calculation model based on utility gain is constructed to find the solution that maximizes the weighted Nash product among the candidate solutions; The solution that maximizes the weighted Nash product This serves as a second intelligent agent, distributing new interface parameters and performance metrics to each component.

5. The aero-engine design method according to claim 4, characterized in that, The weighted Nash product calculation model based on utility gain is as follows: .

6. The aero-engine design method according to claim 4, characterized in that, It also includes using a large language model, combined with the contribution of key variables obtained from SHAP analysis and the weight allocation logic determined by Nash negotiation, to generate a design log in natural language form; the design log explains the reasons for selecting specific interface parameters and performance indicators, including why specific key variables were selected and why a balance of interests was reached at that point.

7. The aero-engine design method according to claim 1, characterized in that, The first agent is equipped with an iterative memory to store the state, strategy adopted, and consequences of historical iterations. Before the first agent regenerates new interface parameters and performance metrics, it uses a large language model to search the iterative history records in the memory where the cosine similarity is greater than the similarity threshold. If the strategy to be adopted is found to have caused parameter oscillation or non-convergence in the past, it calls upon past successful experiences to correct the strategy and avoid repeating the same mistakes.

8. The aero-engine design method according to claim 1, characterized in that, After receiving the interface parameters and performance metrics, each component agent performs the following target tracking optimization process: Under the premise of satisfying the physical constraints of each component, the internal geometric parameters are optimized with the goal of minimizing the Euclidean distance between the interface parameters of the current design scheme and the interface parameters and performance indicators issued by the first intelligent agent.

Images