Multi-agent collaborative design simulation and closed-loop optimization method for optoelectronic payload

Abstract

The present disclosure provides a multi-agent collaborative design simulation and closed-loop optimization method for optoelectronic payloads, relating to the field of optoelectronic payloads. Through overall design agent intelligent evaluation of overall design work, system-level requirements are analyzed and subject tasks are decomposed; each subject agent autonomously generates a candidate scheme, calls a simulation tool for verification and completes design baseline updating; based on a cross-disciplinary dependency graph, the coupling influence propagation analysis of design changes is carried out, and the scope of scheme adjustment and simulation verification is dynamically determined; finally, the system-level index unified evaluation and iterative decision-making are realized through multi-disciplinary joint optimization discrimination. The present disclosure constructs a demand intelligent analysis, subject collaborative design, rapid simulation verification, change influence dynamic propagation and multi-disciplinary optimization integrated closed-loop mechanism, realizes the autonomous collaboration and design state dynamic evolution among the multi-agents of optomechanical thermal electricity, reduces the cross-disciplinary design conflicts and repeated verification, and improves the automation degree, collaborative optimization efficiency and system-level performance of the multi-disciplinary design of complex optoelectronic payloads.

Description

Technical Field

[0001] This disclosure relates to the field of optoelectronic payloads, and in particular to a multi-agent collaborative design simulation and closed-loop optimization method for optoelectronic payloads. Background Technology

[0002] Optoelectronic payloads are crucial equipment in fields such as space remote sensing, astronomical observation, aerial reconnaissance, optoelectronic detection, and laser communication. Typical examples include space remote sensing cameras, ground-based telescopes, airborne optoelectronic pods, and other complex optoelectronic systems. As optoelectronic payloads evolve towards higher resolution, larger apertures, higher stability, lighter weight, miniaturization, and higher integration, their design process is no longer a task that can be completed independently by single optical, mechanical, thermal, or electronic designs. Instead, it becomes a systems engineering problem involving multiple disciplines—optics, mechanics, thermodynamics, and electronics—constrained by mutual coupling and iterative iteration. Therefore, there is a significant interdisciplinary coupling between optomechanics, thermodynamics, and electromechanics, necessitating joint simulation and collaborative optimization to obtain design solutions that meet system-level performance indicators.

[0003] In existing technologies, a scheme combining "optical-mechanical-thermal joint analysis workflow with multidisciplinary optimization platform" is adopted. This scheme typically involves engineers first establishing optical models, mechanical structure models, thermal analysis models, and some electronic boundary conditions. Then, temperature fields, displacement fields, surface errors, power consumption boundaries, and performance indicators are transferred between different disciplines' software through file interfaces, script interfaces, or software platforms. Finally, design variables are adjusted through manual judgment or optimization algorithms to complete multiple rounds of joint simulation and parameter optimization. While existing schemes can support the optical-mechanical-thermal-electric joint design and multidisciplinary simulation of optoelectronic loads to a certain extent, the intelligence level and efficiency of the optical-mechanical-thermal-electric joint simulation and closed-loop optimization design for optoelectronic loads are not high, specifically including: The requirements of optoelectronic payload systems are difficult to automatically analyze and reasonably break down; The design and simulation processes of various disciplines such as optomechanics, thermoelectricity, and optics are fragmented, and it is difficult to efficiently coordinate and utilize specialized software. The lack of a unified design baseline and the difficulty in tracing data status during the joint design of optomechanics, thermoelectricity, and optoelectronics; It is difficult to accurately determine the cross-disciplinary impact of design changes in a single discipline.

[0004] Therefore, there is an urgent need to design a method for optoelectronic payload design. Summary of the Invention

[0005] Therefore, it is necessary to provide a multi-agent collaborative design and closed-loop optimization method for optoelectronic loads to address the problem of low intelligence level and efficiency in the design, simulation and optimization process.

[0006] To solve the above problems, the present disclosure adopts the following technical solution: This disclosure provides a multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic payloads, including the following steps: Step 1: Obtain and structure the system-level design requirements for the optoelectronic payload; Step 2: The overall design intelligent agent performs requirement analysis and automatic indicator decomposition based on the structured optoelectronic payload system-level design requirements to construct a system-level indicator tree including optical disciplines, mechanical disciplines, thermal disciplines, and electronic disciplines. Each discipline branch includes discipline-level tasks. Step 3: Each subject-specific intelligent agent generates candidate solutions based on its own subject-level task, and calls simulation software to verify the obtained candidate solutions through simulation. Step 4: Update the design baseline library based on the candidate solutions obtained from all single-discipline intelligent agents and the simulation verification results; Step 5: The analyzer performs an impact analysis on the design changes of single-discipline agents based on the design baseline library and interdisciplinary dependency graph to obtain the impact level; Step 6: The analyzer determines whether the single-discipline intelligent agent corresponding to the impact level should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the generation range of candidate solutions, and the verification range of simulation verification to obtain the determination result; when the determination result includes determining that the single-discipline intelligent agent should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the single-discipline intelligent agent works according to the determination result; Step 7: When the impact level is no impact, the discriminator reads the current design baseline in the design baseline library and performs a unified evaluation of the subject-level indicators and system-level indicators. Based on the evaluation results, it determines whether the current scheme meets the preset requirements. Step 8: If the preset requirements are met, output the solution; otherwise, return to step 2 and adjust the indicator allocation or design variable boundaries, or return to step 4 and adjust the candidate solution generation strategy.

[0007] In a preferred embodiment, the optoelectronic payload system-level design requirements include the optoelectronic payload task specification, requirements specification, interface files, environmental conditions, and historical design cases; step 1 specifically includes: obtaining the optoelectronic payload system-level design requirements and converting the text, tables, parameters, and historical design cases therein into structured requirement objects; the structured requirement objects include system-level design objectives, optical performance constraints, mechanical performance constraints, thermal performance constraints, electronic performance constraints, interface constraints, and overall resource constraints.

[0008] In a preferred embodiment, the process of performing requirement analysis and automatic indicator decomposition to construct a system-level indicator tree including optical, mechanical, thermal, and electronic disciplines includes: performing requirement analysis and automatic indicator decomposition to obtain multiple candidate decomposition schemes; scoring the candidate decomposition schemes based on requirement coverage, historical similar scheme matching degree, cross-disciplinary constraint consistency, system-level performance prediction, design risk, and simulation calculation cost; determining the discipline task allocation scheme based on the scoring results; and constructing the system-level indicator tree based on the discipline task allocation scheme.

[0009] In a preferred embodiment, step 3 specifically involves: an optical-related single-discipline intelligent agent generating optical structure parameters, optical component parameters, and imaging quality analysis to generate optical task candidate schemes based on the optical task; a mechanical-related single-discipline intelligent agent generating support structures, mounting interfaces, lightweight structures, and structural analysis to generate mechanical task candidate schemes based on the mechanical task; a thermal-related single-discipline intelligent agent generating heat source distribution, heat dissipation paths, thermal control strategies, and temperature field analysis to generate thermal task candidate schemes based on the thermal task; and an electronics-related single-discipline intelligent agent generating electronic architecture, device layout, power consumption distribution, printed circuit board constraints, signal integrity, power integrity, and electromagnetic compatibility analysis to generate electronics task candidate schemes based on the electronics task.

[0010] In a preferred embodiment, step 4 includes: submitting the subject-level tasks, key design variables, interface variables, simulation tasks for simulation verification, simulation verification results, constraint satisfaction status, version information, and change records corresponding to each single-discipline intelligent agent to the design baseline library.

[0011] In a preferred embodiment, the candidate solution generation process of the single-discipline intelligent agent is expressed as follows: in, Indicates the subject number; Indicate subject Initial candidate design variables; This refers to the disciplinary tasks assigned by the overall design intelligent agent; Indicate subject Historical design cases; Indicate subject Knowledge base and design constraints; Indicate subject The corresponding candidate solution generation function for single-discipline intelligent agents.

[0012] In a preferred embodiment, step 5 specifically includes: the analyzer identifies effective change variables based on the difference between the current design baseline and the historical design baseline, and uses an interdisciplinary dependency graph to analyze the scope of influence of the effective change variables to obtain the influence level.

[0013] In a preferred embodiment, the impact level includes Level 1 impact, Level 2 impact, Level 3 impact, and no impact; the determination result corresponding to Level 1 impact is that the single-discipline intelligent agent corresponding to this impact level regenerates all candidate solutions and re-performs all simulation verifications; the determination result corresponding to Level 2 impact is that the single-discipline intelligent agent corresponding to this impact level regenerates some candidate solutions and re-performs some simulation verifications; the determination result corresponding to Level 3 impact is that the single-discipline intelligent agent corresponding to this impact level does not regenerate candidate solutions or re-perform simulation verifications.

[0014] In a preferred embodiment, the step of uniformly evaluating the subject-level and system-level indicators and determining whether the current solution meets the preset requirements based on the evaluation results specifically involves: determining whether the current solution meets the preset requirements based on the comprehensive objective function and convergence conditions; meeting the preset requirements means meeting the convergence conditions and the comprehensive objective function value reaching the preset value.

[0015] In a preferred embodiment, the comprehensive objective function includes optical objective terms, mechanical objective terms, thermal objective terms, electronic objective terms, system-level objective terms, and constraint violation penalty terms; the convergence condition is: simultaneously satisfying optical, mechanical, thermal, electronic, and system-level constraints, and the improvement of the comprehensive objective function is less than a set threshold in several consecutive iterations, at which point the analyzer no longer triggers first-level or second-level effects.

[0016] The aforementioned multi-agent collaborative design simulation and closed-loop optimization method for optoelectronic payloads intelligently generates discipline-level tasks in optics, mechanics, thermodynamics, and electronics based on the system-level design requirements of the optoelectronic payload. Each discipline automatically generates candidate solutions through its own single-discipline agent and calls its own simulation software to complete simulation verification. A unified design baseline library is used to manage the candidate solutions and simulation verification results of each discipline. An analyzer performs design change impact analysis on single-discipline agents based on the design baseline library and cross-disciplinary dependency graph to obtain the impact level. Based on the impact level, it determines whether to regenerate candidate solutions and perform simulation verification, the generation range of candidate solutions, and the verification range of simulation verification, and performs the corresponding operations. A discriminator reads the current design baseline in the design baseline library and performs a unified evaluation of each discipline-level indicator and system-level indicator, and determines whether to output or redefine the solution based on the evaluation results. This disclosure automatically analyzes and decomposes the system-level design requirements of optoelectronic payloads. Each discipline has an intelligent agent capable of generating candidate solutions and calling corresponding simulation software for simulation verification. The optomechanical-thermoelectric joint design process has a design baseline library, which can be used to trace the data status, accurately determine the cross-disciplinary impact range after a single discipline design change, and achieve closed-loop optimization. This disclosure realizes system-level, intelligent optomechanical-thermoelectric joint simulation and closed-loop optimization design, improving design efficiency and intelligence level. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method in one embodiment of the present disclosure. Detailed Implementation

[0018] The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0019] See Figure 1 This embodiment provides a multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic payloads, including: Step 1: Obtain and structure the system-level design requirements for the optoelectronic payload; Step 2: The overall design intelligent agent performs requirement analysis and automatic indicator decomposition based on the structured optoelectronic payload system-level design requirements to construct a system-level indicator tree including optical disciplines, mechanical disciplines, thermal disciplines, and electronic disciplines. Each discipline branch includes discipline-level tasks. Step 3: Each subject-specific intelligent agent generates candidate solutions based on its own subject-level task, and calls simulation software to verify the obtained candidate solutions through simulation. Step 4: Update the design baseline library based on the candidate solutions obtained from all single-discipline intelligent agents and the simulation verification results; Step 5: The analyzer performs an impact analysis on the design changes of single-discipline agents based on the design baseline library and interdisciplinary dependency graph to obtain the impact level; Step 6: The analyzer determines whether the single-discipline intelligent agent corresponding to the impact level should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the generation range of candidate solutions, and the verification range of simulation verification to obtain the determination result; when the determination result includes determining that the single-discipline intelligent agent should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the single-discipline intelligent agent works according to the determination result; Step 7: When the impact level is no impact, the discriminator reads the current design baseline in the design baseline library and performs a unified evaluation of the subject-level indicators and system-level indicators. Based on the evaluation results, it determines whether the current scheme meets the preset requirements. Step 8: If the preset requirements are met, output the solution; otherwise, return to step 2 and adjust the indicator allocation or design variable boundaries, or return to step 4 and adjust the candidate solution generation strategy.

[0020] The method is described in detail below, including: Step 1: Obtain and structure the system-level design requirements for the optoelectronic payload; Specifically, the system-level design requirements for the optoelectronic payload include the optoelectronic payload task specification, requirements statement, interface documents, environmental conditions, and historical design cases. The process involves obtaining these system-level design requirements and converting the text, tables, parameters, and historical design cases into a structured requirement object. This structured requirement object includes at least system-level design objectives, optical performance constraints, mechanical performance constraints, thermal performance constraints, electronic performance constraints, interface constraints, and overall resource constraints. It is understood that "system-level system" refers to the optoelectronic payload system itself.

[0021] The constraints on optical performance include focal length, field of view, aperture, modulation transfer function, wavefront error, and line-of-sight stability; mechanical performance constraints include structural mass, envelope size, stiffness, strength, modal frequency, and mounting interface; thermal performance constraints include operating temperature range, temperature gradient, thermal control power consumption, and thermal settling time; and electronic performance constraints include device layout, total power consumption, signal integrity, power integrity, electromagnetic compatibility (EMC), and printed circuit board constraints. Here, signal integrity refers to the quality of high-speed signal transmission, power integrity refers to the stability of the power supply network, and EMC refers to the ability of electronic equipment to operate normally in an electromagnetic environment without generating unacceptable interference.

[0022] Step 2: Utilize the overall design intelligent agent to perform requirement analysis and automatic indicator decomposition based on the structured optoelectronic payload system-level design requirements to construct a system-level indicator tree.

[0023] The overall design agent generates subject-level tasks based on the structured optoelectronic payload system-level design requirements; the overall design agent is an agent used for the overall design of the optoelectronic payload system.

[0024] The overall design agent is used for requirement analysis and automatic indicator decomposition: Based on the structured optoelectronic payload design requirements, this agent constructs a system-level indicator tree. This tree includes multiple disciplines, specifically optical, mechanical, thermal, and electronic branches. Each discipline includes discipline-level tasks, which specifically include executable design objectives, design variables, constraints, and evaluation indicators. Discipline-level tasks refer to optical, mechanical, thermal, and electronic tasks. Generating discipline-level tasks includes generating optical, mechanical, thermal, and electronic tasks.

[0025] The overall design agent reads structured requirement objects, constructs a system-level indicator tree, and decomposes the system-level indicators into executable design goals, design variables, constraints, and evaluation indicators for each discipline: optics, mechanics, thermodynamics, and electronics. The overall design agent can be implemented using one or more combinations of large-scale language models, knowledge graphs, rule engines, expert systems, and case-based reasoning algorithms. This invention does not limit the specific algorithmic form; its key lies in automatically generating discipline-level tasks based on the system-level requirements of the optoelectronic payload.

[0026] Step 2 also includes the evaluation and selection of candidate decomposition schemes. Based on requirements coverage, matching degree of historical similar schemes, consistency of cross-disciplinary constraints, system-level performance prediction, design risk, and simulation calculation cost, multiple candidate decomposition schemes are scored (which can be a comprehensive score), and the subject task allocation scheme that meets the requirements is selected. The index decomposition in this step is not a simple rule assignment, but has an evaluation and selection mechanism.

[0027] The process of performing requirement analysis and automatic indicator decomposition to construct a system-level indicator tree encompassing optical, mechanical, thermal, and electronic disciplines includes: performing requirement analysis and automatic indicator decomposition to obtain multiple candidate decomposition schemes; scoring the candidate decomposition schemes based on requirement coverage, historical similar scheme matching degree, cross-disciplinary constraint consistency, system-level performance prediction, design risk, and simulation calculation cost; determining the disciplinary task allocation scheme based on the scoring results; and constructing the system-level indicator tree based on the disciplinary task allocation scheme.

[0028] Demand coverage refers to the degree to which the disciplinary indicators included in the candidate decomposition scheme cover the functional requirements, performance requirements, and constraint requirements in the structured requirement object; historical similarity matching degree refers to the degree of similarity between the candidate decomposition scheme and historical design cases in terms of system function, indicator configuration, disciplinary task division, and constraint relationship; cross-disciplinary constraint consistency refers to the degree to which optical performance constraints, mechanical performance constraints, thermal performance constraints, and electronic performance constraints in the candidate decomposition scheme satisfy mutual coupling relationships and do not have constraint conflicts; system-level performance prediction refers to the evaluation result obtained by predicting the overall system performance that the candidate decomposition scheme can achieve based on the disciplinary indicators, interdisciplinary coupling relationships, and system-level performance mapping relationships; design risk refers to the degree of risk that the overall system performance of the candidate decomposition scheme may fail to meet the design requirements due to insufficient indicator margin, cross-disciplinary constraint conflicts, insufficient technology maturity, and design parameter uncertainty; simulation computation cost refers to the comprehensive measure of the computational resources, computation time, and simulation task complexity required in the process of single-disciplinary and multi-disciplinary joint simulation verification of the candidate decomposition scheme in optics, mechanics, thermal, and electronics.

[0029] Unlike traditional solutions that rely on human experience for indicator decomposition and task assignment, step 2 of this embodiment protects the requirements analysis and automatic indicator decomposition of the overall design agent. It comprehensively analyzes the task book, requirements specification, interface files, environmental conditions and historical design cases to form a system-level indicator tree, and automatically decomposes it into optical, mechanical, thermal and electronic discipline tasks.

[0030] Step 3: Utilize the single-discipline intelligent agents corresponding to each discipline to generate candidate solutions based on their respective discipline-level tasks, and then call simulation software to verify the obtained candidate solutions through simulation. That is, the single-discipline intelligent agents corresponding to each discipline generate candidate solutions based on their respective discipline-level tasks, and then call simulation software to verify the obtained candidate solutions through simulation.

[0031] The single-discipline intelligent agent is a single agent that can only be used for generating candidate solutions based on subject-level tasks within a single discipline and calling simulation software to verify the obtained candidate solutions. This embodiment does not limit the specific implementation form of the single-discipline intelligent agent; for example, it can be implemented by one or more combinations of large-scale language models, knowledge graphs, rule engines, etc. This invention does not limit its specific algorithmic form.

[0032] Specifically, the corresponding intelligent agents (referred to as single-discipline intelligent agents) for each of the optomechanical and thermoelectric disciplines generate candidate solutions and call specialized software to complete simulation verification. The optical intelligent agent (the single-discipline intelligent agent corresponding to optics, i.e., the intelligent agent corresponding to the optics discipline) generates optical structure parameters, optical component parameters, and imaging quality analysis based on the optical task, generating candidate solutions for its discipline-level task (optical task); the mechanical intelligent agent generates support structures, installation interfaces, lightweight structures, and structural analysis based on the mechanical task, generating candidate solutions for its discipline-level task (mechanical task); the thermal intelligent agent generates heat source distribution, heat dissipation paths, thermal control strategies, and temperature field analysis based on the thermal task, generating candidate solutions for its discipline-level task (thermal task); and the electronic intelligent agent generates electronic architecture, device layout, power consumption distribution, printed circuit board constraints, signal integrity, power integrity, and electromagnetic compatibility analysis based on the electronic task, generating candidate solutions for its discipline-level task (electronic task).

[0033] Step 3 protects the collaborative simulation process of optical intelligent agents (single-discipline intelligent agents corresponding to optics), mechanical intelligent agents (single-discipline intelligent agents corresponding to mechanics), thermal intelligent agents (single-discipline intelligent agents corresponding to thermal), and electronic intelligent agents (single-discipline intelligent agents corresponding to electronics). Each discipline-specific intelligent agent automatically generates candidate solutions based on the discipline's task and calls upon its own discipline's specialized software to complete model updates, simulation solutions, result extraction, and local optimization.

[0034] Unlike schemes that only treat electronics as a power consumption boundary input, in this embodiment, the processing scope of the single-discipline intelligent agent corresponding to electronics includes device layout, power consumption distribution, printed circuit board constraints, signal integrity, power integrity and electromagnetic compatibility, and allows the electronic results to affect thermal, mechanical and optical design through heat source distribution, interface constraints and power consumption constraints.

[0035] Step 4: Establish or update the design baseline library based on the candidate solutions obtained by all single-discipline agents. It can be understood that the first execution of Step 4 is to establish the design baseline library based on the candidate solutions obtained by all single-discipline agents, and subsequent executions of Step 4 are to update the design baseline library based on the candidate solutions obtained by all single-discipline agents.

[0036] The design baseline library can be understood as a database, specifically a design database for multi-agent collaborative design, comprising several design baselines. A baseline is understood to be the foundation upon which subsequent work is based. The database includes information for each discipline, such as discipline-level tasks, key design variables, interface variables, simulation tasks for simulation verification, simulation verification results, and constraint satisfaction status.

[0037] Specifically, after each subject-specific agent completes simulation verification and local optimization, the final candidate solutions obtained by each agent are submitted to the design baseline library. The subject-level tasks, key design variables, interface variables, simulation verification tasks, simulation results, and constraint satisfaction status of each agent are also submitted to the design baseline library, forming a recordable, queryable, and traceable design baseline for the current round. Furthermore, version information and change records (version information and change records of candidate solutions) are also submitted to the design baseline library.

[0038] Single-discipline intelligent agents can intelligently analyze key design variables and critical variables based on subject-level tasks (using the capabilities of large models). For example, the design variables identified by an optical discipline intelligent agent may include optical lens aperture, focal length, field of view, number of lens groups, detection distance, etc., among which lens aperture and focal length are key design variables.

[0039] Unlike solutions that involve scattered storage of data across disciplines and difficulty in uniformly managing version status, step 4 involves the construction and updating of a unified design baseline library. This embodiment uniformly writes original design requirements, disciplinary tasks, design variables, interface variables, simulation results, constraint satisfaction status, version information, and change records into the design baseline to support subsequent change impact analysis, local iteration, joint optimization judgment, and solution tracing.

[0040] Step 5: Using the analyzer, based on the design baseline library and interdisciplinary dependency graph, perform an impact analysis on the design changes of single-discipline intelligent agents to obtain the impact level.

[0041] Specifically, the step is as follows: when any subject submits new content, the design change impact analyzer reads the current round baseline and the previous round baseline, and compares the changes in design variables and interface variables.

[0042] The Design Change Impact Analyzer is an analyzer used to perform design change impact analysis on single-discipline agents based on a design baseline library and an interdisciplinary dependency graph to obtain the impact level. It is also used to determine, based on the impact level analysis, whether the single-discipline agent corresponding to the impact level should regenerate candidate solutions and perform simulation verification according to the simulation verification results, the generation range of candidate solutions, and the verification range of simulation verification to obtain the determination results.

[0043] The design change impact analyzer reads the current round baseline and the previous round baseline. Based on the differences between the current design baseline and historical design baselines, it identifies valid change variables and uses an interdisciplinary dependency graph to analyze the scope of their impact. The interdisciplinary dependency graph is a structured analysis tool that integrates multidisciplinary knowledge systems and visually presents the knowledge connections and dependencies between different disciplines. It is a core auxiliary means for interdisciplinary research and knowledge system organization. It is understood that in some embodiments, a difference refers to a variable that reaches a preset difference threshold as a valid change variable.

[0044] Step 5 involves identifying valid change variables based on the differences between the current baseline and historical baselines. This embodiment uses the amount of change and relative rate of change to judge numerical variables, and uses state differences to judge non-numerical variables such as material type, device model, structural topology, and boundary condition type, thereby identifying whether they constitute valid changes requiring propagation analysis.

[0045] Step 5 involves change impact analysis based on a cross-disciplinary dependency graph. This embodiment uses design variables, interface variables, simulation models, simulation results, and performance indicators as nodes, and the influence relationships between nodes as directed edges. The affected disciplines, variables, models, and indicators are determined based on the intensity of the impact. This approach differs from relying on manual experience to determine the scope of change impact.

[0046] Step 6: Using the design change impact analyzer, determine whether the single-discipline intelligent agent corresponding to the impact level should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the generation range of candidate solutions, and the verification range of simulation verification to obtain the determination result; if the determination result includes determining that the single-discipline intelligent agent should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the single-discipline intelligent agent shall work according to the determination result; if the impact level is no impact, proceed to step 7; The design change impact analyzer categorizes impact levels into Level 1, Level 2, Level 3, and no impact based on the intensity and importance of the impact indicators. Level 1 impact directly affects core system-level indicators or hard constraints, requiring redesign and resimulation across relevant disciplines. The result is that the single-discipline agent corresponding to this impact level regenerates all candidate solutions and performs full simulation verification again; that is, the scope of candidate solution generation is all, and the scope of simulation verification is all simulation verification (simulating all content that can be verified by the single-discipline agent). Level 2 impact affects important discipline-level indicators or key interface variables, requiring local candidate solutions and simulation verification. The result is that the single-discipline agent corresponding to this impact level regenerates local candidate solutions and performs local simulation verification again (simulating all content that can be verified by the single-discipline agent). (Simulation verification is performed on a portion of the entire content). Only the affected disciplines, affected models, and affected indicators are triggered for local recalculation or local optimization to avoid indiscriminate duplication of calculations across all disciplines. Understandably, the verification content of local simulation verification corresponds to the newly generated content of generating local candidate solutions. Level 3 impact indicates that it only affects auxiliary variables or recorded variables, which can be written into the design baseline library but does not trigger re-simulation. The result is that the single-discipline agent corresponding to this impact level does not regenerate candidate solutions or re-perform simulation verification. No impact indicates that the effective change of variables has no impact on the system-level core indicators or hard constraints, no impact on the discipline-level important indicators or key interface variables, and no impact on auxiliary variables or recorded variables. When there is no impact, no iteration is performed, and step 7 is directly performed from step 6.

[0047] Step 6 establishes a local iteration mechanism for affected disciplines. Based on the impact level of the change, only the affected disciplines and affected models are triggered for local optimization, avoiding duplicate calculations for optics, mechanics, thermodynamics, and electronics due to a single local change, while also reducing the risk of missing implicit cross-disciplinary impacts.

[0048] Step 7: The multidisciplinary joint optimization discriminator reads the current design baseline from the design baseline library and performs a unified evaluation of the discipline-level indicators and system-level indicators. Based on the evaluation results, it determines whether the current solution meets the preset requirements.

[0049] The multidisciplinary joint optimization discriminator is a discriminator used to read the current design baseline from the design baseline library and perform a unified evaluation of various discipline-level and system-level indicators. Based on the evaluation results, it determines whether the current solution meets the preset requirements. Unified evaluation means that the evaluation results include evaluations of both discipline-level and system-level indicators.

[0050] Each subject-level indicator, that is, each subject-level task, includes not only the subject-level task itself but also the subject-level indicators. The subject-level indicators include: Optical discipline-level indicators, which are selected from one or more of the following: effective aperture, focal length, relative aperture, field of view, working band, modulation transfer function, wavefront error, distortion, optical transmittance, spot size, energy concentration and stray light suppression ratio. Mechanical engineering-level indicators are selected from one or more of the following: structural mass, envelope size, center of mass position, moment of inertia, installation interface position, structural stiffness, first-order natural frequency, structural strength, safety factor, structural deformation, vibration response, and impact response. The thermal science-level indicators are selected from one or more of the following: highest temperature, lowest temperature, average temperature, temperature gradient, temperature uniformity, temperature stability, temperature control accuracy, heat load, heat dissipation capacity, thermal control power consumption, thermally induced deformation, and thermally induced coke surface drift. Electronics-level indicators are selected from one or more of the following: power supply voltage, power consumption, sampling frequency, imaging frame rate, analog-to-digital conversion bit depth, readout noise, signal-to-noise ratio, dynamic range, data transmission rate, interface bandwidth, signal processing delay, and electromagnetic compatibility.

[0051] System-level metrics include one or more of the following: Mission performance metrics include target detection range, target recognition range, target identification range, target detection probability, target identification probability, and mission coverage area; The comprehensive imaging and detection performance indicators include the system's operating band, system field of view, spatial resolution, angular resolution, ground sampling distance, system modulation transfer function, system signal-to-noise ratio, noise equivalent temperature difference, noise equivalent irradiance, radiometric resolution, spectral resolution, and system dynamic range. Real-time performance and data processing metrics include system imaging frame rate, system response time, end-to-end processing latency, data throughput, and data compression ratio. System resource constraints include overall machine weight, overall machine envelope size, overall machine power consumption, peak power consumption, heat dissipation requirements, and storage capacity; Reliability and environmental adaptability indicators include system reliability, mission life, operating temperature range, and adaptability to vibration, shock, and radiation environments.

[0052] The multidisciplinary joint optimization judgment is as follows: after the simulation results and necessary iteration results of each discipline in the current round are completed, the discriminator reads the optical, mechanical, thermal, electronic and system-level indicators of the current design baseline in the design baseline library, and judges whether the current scheme meets the preset requirements based on the comprehensive objective function and convergence conditions.

[0053] The unified evaluation of various subject-level and system-level indicators, and the determination of whether the current solution meets the preset requirements based on the evaluation results, specifically involves: determining whether the current solution meets the preset requirements based on the comprehensive objective function and convergence conditions. The comprehensive objective function includes optical, mechanical, thermal, electronic, system-level objective items, and constraint violation penalty items. The convergence condition is: simultaneously satisfying optical, mechanical, thermal, electronic, and system-level constraints; and in several consecutive iterations, the improvement in the comprehensive objective function (the difference between two comprehensive objective functions, where the comprehensive objective function value in the later iteration is better than the comprehensive objective function value in the earlier iteration) is less than a set threshold, and the design change impact analyzer no longer triggers first-level or second-level effects. Meeting the preset requirements means that the comprehensive objective function value reaches a preset value and the convergence condition is met.

[0054] The simultaneous satisfaction of optical, mechanical, thermal, electronic, and system-level constraints means simultaneously satisfying optical performance constraints, mechanical performance constraints, thermal performance constraints, electronic performance constraints, and system-level constraints. System-level constraints refer to the constraints of the entire payload system, i.e., the constraints of the overall optoelectronic payload system. In one embodiment, system-level constraints are system-level indicators. In another embodiment, system-level constraints are system functional objectives of the optoelectronic payload system. In yet another embodiment, system-level constraints are system-level target items. In a practical application, as an example, consider designing an optoelectronic carrier system with a target range ≥20km. This can be broken down into four disciplines: optics, mechanics, thermodynamics, and electronics. Optics requires an aperture ≥200 m, focal length ≥1000 mm, MTF ≥0.3, and transmittance ≥80%. Electronics requires a NETD (Noise Equivalent Temperature Difference) ≤30 mK, detector resolution ≥1920×1080, and signal-to-noise ratio ≥ a certain threshold. Mechanics requires line-of-sight jitter ≤15 μrad and optical axis shift due to structural deformation ≤10 μrad. Thermodynamics requires thermal defocus ≤ a specified value and optical axis thermal drift ≤10 μrad. Ultimately, these specifications collectively ensure target detection within 20 km, constituting a system-level constraint.

[0055] Step 8: If the preset requirements are met, output the solution; otherwise, readjust the indicator allocation, design variable boundaries and / or candidate solution generation strategy.

[0056] Closed-loop feedback or final solution output. When the current solution meets the preset requirements, the optomechanical-thermoelectric joint optimization design solution is output. The output includes optical design results, mechanical design results, thermal design results, electronic design results, system-level indicator satisfaction status, simulation report, design baseline record, and change traceability record. When the current solution does not meet the preset requirements, the unmet indicators, impact paths, and suggested adjustment variables are returned to the overall design agent. The indicator allocation (corresponding to the indicator allocation in step 2 for building the system-level indicator tree), design variable boundaries (corresponding to the generation of discipline-level tasks in step 2), or candidate solution generation strategy (corresponding to step 4) are readjusted, and the next round of closed-loop optimization is entered. The agent is driven again to perform candidate solution generation, simulation verification, change impact analysis, and joint optimization discrimination, until the system-level convergence condition is met.

[0057] Unlike single-discipline local optimization, ordinary parameter search, or multi-software serial simulation schemes lacking feedback adjustment mechanisms, this method protects the multi-disciplinary joint optimization discrimination and closed-loop feedback mechanism in steps 7 and 8. It integrates optical performance, mechanical performance, thermal performance, electronic performance, and system-level comprehensive indicators into the comprehensive objective function, constraint violation penalty term, and convergence conditions for discrimination. When the preset requirements are not met, the unmet indicators, impact paths, and suggested adjustment variables are returned to the overall design agent, and the disciplinary agents are re-driven to enter the next round of joint simulation and optimization.

[0058] In step 2 of this invention, during the indicator decomposition process, the overall design agent generates multiple candidate decomposition schemes and evaluates these schemes. Specifically, the process of performing requirement analysis and automatic indicator decomposition to construct a system-level indicator tree includes: performing requirement analysis and automatic indicator decomposition to obtain multiple candidate decomposition schemes; scoring the candidate decomposition schemes based on requirement coverage, historical similarity scheme matching degree, cross-disciplinary constraint consistency, system-level performance prediction, design risk, and simulation computation cost; determining the disciplinary task allocation scheme based on the scoring results; and constructing the system-level indicator tree based on the disciplinary task allocation scheme. The comprehensive score of the candidate disassembly schemes can be expressed as: in, Indicates the first A comprehensive score for each candidate dismantling scheme; Indicates demand coverage; Indicates the matching degree of historically similar schemes; Indicates consistency across disciplines; This indicates a system-level performance prediction score; Indicates the design risk score; This indicates the estimated cost of simulation calculations; to This indicates the weight of each evaluation item. The overall design agent selects candidate decomposition schemes that meet the comprehensive score requirements, forming optical, mechanical, thermal, and electronic tasks, and distributes them to the corresponding subject-specific agents.

[0059] In step 3 of this invention, each discipline-specific intelligent agent invokes its corresponding specialized software through a specialized software scheduling process to complete parameter writing, model updating, solution execution, result extraction, and structured result feedback. The candidate solution generation process of a single-discipline intelligent agent can be represented as follows: in, This indicates the subject number, i.e., the subject type, which can be optics, mechanics, thermodynamics, or electronics; Indicate subject Initial candidate design variables; This refers to the disciplinary tasks assigned by the intelligent agent used for the overall design of the optoelectronic payload system; Indicate subject Historical design cases; This represents the subject-specific knowledge base and design constraints; This indicates the single-discipline intelligent agent (discipline) This generates candidate solutions for the corresponding single-discipline intelligent agents. In this way, instead of blindly traversing the entire design space, each discipline's intelligent agent generates verifiable candidate solutions based on task constraints, historical cases, and discipline knowledge.

[0060] In step 4 of this invention, each round of design baselines is used to record the complete state of the current optomechanical-thermoelectric co-design. The wheel design baseline can be represented as: in, Indicates the first Wheel design baseline; Indicate the original design requirements; Indicates the first A set of subject-specific tasks; This represents the set of design variables for each discipline; Represents a set of interdisciplinary interface variables; Represents the set of simulation results; Represents the set of states where constraints are satisfied; It indicates the version and change history. By using a unified design baseline library, the source, variable changes, interface relationships, simulation results, and indicator satisfaction status of each round of solutions can be clearly identified, providing a data foundation for subsequent change impact analysis and solution traceability.

[0061] In step 5 of the present invention, (a) For numerical variables, their change It can be represented as: The relative rate of change of a variable can be expressed as: A variable is considered a valid change variable when the following condition is met: ≥ in, Indicates the first In the first round of baselines The values ​​of the variables; Indicates the first round of baselines in the previous round. The values ​​of the variables; This indicates an extremely small positive number that prevents the denominator from being zero; Indicates the first The change thresholds corresponding to each variable; Indicates the first The relative rate of change of each variable. For non-numerical variables such as material type, device model, structural topology, and boundary condition type, the validity of the change is determined by the state difference.

[0062] (b) Interdisciplinary dependency diagrams are used to describe the influence relationships between optomechanical and thermoelectric variables, models, results, and indicators. Specifically, the scope of influence of an effective change variable analyzed using an interdisciplinary dependency diagram includes the influence relationships between optomechanical and thermoelectric variables, models, results, and indicators, which can be represented as: in, Represents interdisciplinary dependency graphs; This represents a set of nodes, which includes design variable nodes, interface variable nodes, simulation model nodes, simulation result nodes, and performance indicator nodes. This represents a set of directed edges, used to indicate the direction of influence between nodes; This represents the set of directed edge weights, used to indicate the intensity of the influence. For example, changes in electronic power consumption affect the heat source distribution, which in turn affects the temperature field. The temperature field affects the thermal deformation of the structure, which in turn affects the pose and surface shape of optical components, ultimately impacting image quality.

[0063] (c) For a given valid change node Its target node The intensity of the influence can be expressed as: in, Represents the target node Subject to change node The intensity of the impact; Indicates the change node The normalized change range; to Indicates from the change node To the target node The weights of each edge along a propagation path; This indicates that when multiple propagation paths exist, the maximum value among the influence strengths of each path is taken. If If the impact threshold is greater than or equal to that of the target node, the node is determined to be affected, and the affected disciplines, variables, models, and indicators are further identified.

[0064] In step 6 of this invention, the local iterative task can be represented as: in, Indicates the first Round-trip local iterative task; This represents the set of subject-specific intelligent agents that have been triggered. Represents the set of affected variables; This represents the set of models that need to be updated or re-called. This indicates the set of indicators that need to be recalculated. Through this mechanism, the system only performs local iterations on the affected disciplines and models, avoiding the indiscriminate recalculation of all disciplines due to a change in one place.

[0065] In step 7 of the present invention, (a) Clause The wheel system-level evaluation vector can be represented as: in, Represents a set of optical performance indicators; Represents a set of mechanical performance indicators; Represents a set of thermal performance indicators; Represents a set of electronic performance indicators; This represents a set of comprehensive system-level indicators.

[0066] (b) A multidisciplinary joint optimization discriminator constructs a comprehensive objective function, which can preferably be expressed as: in, Indicates the first Round-comprehension objective function; Indicates optical target items; Indicates the mechanical target item; Represents the thermal objective term; Represents the electronic target term; Indicates system-level target items; , , , and Each represents the weight of the objective item; Indicates the constraint penalty coefficient; Indicates the first Wheel constraint violation penalty term. Through this objective function, factors such as imaging quality, structural stability, temperature stability, electronic performance, mass, power consumption, and envelope can be uniformly incorporated into the system-level optimization judgment.

[0067] Among them, the penalty items for violating the constraints It can be represented as: It can also be further expressed as: in, to These represent different inequality constraint functions, each corresponding to a design boundary in a different discipline; specifically, This represents the first inequality constraint function. This represents the second inequality constraint function. Indicates the first Inequality constraint functions, Let be the total number of inequality constraint functions. For example, the lower bound constraint of the modulation transfer function in optics, the upper bound constraint of mass in mechanics, the maximum temperature constraint in thermodynamics, and the power consumption constraint in electronics can be expressed as follows: in, Indicates the first The second iteration or the first The design variable vector corresponding to each candidate design scheme. This represents the lower bound constraint function for the modulation transfer function in optics. This represents the minimum permissible value of the modulation transfer function. express The corresponding actual modulation transfer function value, The normalized scaling factor represents the constraint of the modulation transfer function. This represents the upper limit constraint function for mass in mechanical engineering. express The corresponding actual system quality value, Indicates the maximum allowable mass of the system. The normalized scaling factor representing the quality constraint. This represents the highest temperature constraint function in the field of thermal science. express The corresponding highest system temperature, Indicates the highest temperature allowed by the system. The normalized scaling factor representing the temperature constraint. This represents the power consumption constraint function in the field of electronics. express The corresponding actual power consumption of the system Indicates the maximum allowable power consumption of the system. The normalized scaling factor represents the power consumption constraint.

[0068] When a constraint is satisfied, its corresponding penalty term is 0; when a constraint is not satisfied, the greater the degree of violation, the greater the penalty on the comprehensive objective function.

[0069] The convergence condition can be expressed as: in, Indicates the first Does the round-robin scheme meet the convergence condition? This indicates that the hard constraint satisfaction criterion is met, meaning that key optical, mechanical, thermal, electronic, and system-level constraints are all satisfied. This represents the convergence criterion for the objective function, i.e., the improvement of the comprehensive objective function is less than a set threshold in several consecutive iterations. This indicates the design baseline stability criterion, meaning the design change impact analyzer no longer triggers first-level or second-level effects. When When established, it is determined that the current solution meets the preset requirements; when If the condition is not met, the multidisciplinary joint optimization discriminator generates the unmet criteria, the impact path, suggested adjustment variables and iterative strategies, and returns the overall design agent.

[0070] This embodiment also provides a multi-agent cooperative design simulation and closed-loop optimization system for optoelectronic payloads, including: Acquire and structure modules to acquire and structure the system-level design requirements for optoelectronic payloads; The task generation module is used to utilize the overall design intelligent agent to perform requirement analysis and automatic indicator decomposition based on the structured optoelectronic payload system-level design requirements to construct a system-level indicator tree including optical disciplines, mechanical disciplines, thermal disciplines, and electronic disciplines. Each discipline branch includes discipline-level tasks. The scheme generation and simulation verification module is used to generate candidate schemes by utilizing the single-discipline intelligent agents corresponding to each discipline according to their own discipline-level tasks, and call the simulation software to perform simulation verification on the candidate schemes obtained. The update module is used to update the design baseline library based on the candidate solutions and simulation verification results obtained from all single-discipline agents; The impact analysis module is used to perform impact analysis on design changes of single-discipline agents based on the design baseline library and interdisciplinary dependency graph, and obtain the impact level. The determination module is used to determine, based on the influence level, whether the single-discipline intelligent agent corresponding to that influence level should regenerate candidate solutions and conduct simulation verification according to the simulation verification results, the generation range of candidate solutions, and the verification range of simulation verification to obtain the determination result; when the determination result includes determining that the single-discipline intelligent agent should regenerate candidate solutions and conduct simulation verification according to the simulation verification results, the module is used to start the single-discipline intelligent agent to work according to the determination result. The evaluation module is used to read the current design baseline in the design baseline library using a discriminator when the impact level is no impact, and to conduct a unified evaluation of various discipline-level indicators and system-level indicators. Based on the evaluation results, it determines whether the current scheme meets the preset requirements. The output module is used to output a solution when preset requirements are met. The task generation module is also used to adjust the indicator allocation or design variable boundaries when the preset requirements are not met, or the update module is also used to adjust the generation strategy of candidate solutions when the preset requirements are not met.

[0071] In specific implementation, the multi-agent collaborative design simulation and closed-loop optimization system for optoelectronic loads can refer to the implementation of the multi-agent collaborative design simulation and closed-loop optimization method for optoelectronic loads in any of the above embodiments. The specific implementation steps will not be repeated here.

[0072] An electronic device can be implemented according to the method of this disclosure, the electronic device comprising: a memory; one or more processors; one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for executing the multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic loads according to any of the above embodiments.

[0073] This disclosure also provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the steps of the multi-agent collaborative design simulation and closed-loop optimization method for optoelectronic loads described in any of the above embodiments.

[0074] This disclosed method for multi-agent collaborative design simulation and closed-loop optimization of optoelectronic payloads (OAP) systems generates subject-specific tasks for optics, mechanics, thermodynamics, and electronics based on the OAP system-level design requirements. It automatically analyzes and rationally decomposes these requirements. Each subject automatically generates candidate solutions through its own single-subject agent and calls its respective simulation software for simulation verification. Each subject has one agent capable of generating candidate solutions and calling the corresponding simulation software for verification, solving the problems of fragmented design simulation processes across optomechanics, thermodynamics, and electronics, and the difficulty in efficiently coordinating and calling specialized software. The method has a unified design baseline library, enabling the integration of various subjects. The system manages candidate solutions and simulation verification results in a unified manner, making the optomechanical-thermoelectric co-design process recordable, searchable, and traceable. Based on the design baseline library and interdisciplinary dependency graph, it performs impact analysis on design changes of single-disciplinary agents to obtain the impact level, accurately determining the interdisciplinary impact range after a single-disciplinary design change. Based on the impact level, it determines whether to regenerate candidate solutions and conduct simulation verification, the generation range of candidate solutions, and the verification range of simulation verification, and performs corresponding operations. The discriminator reads the current design baseline in the design baseline library and performs a unified evaluation of each discipline-level indicator and system-level indicator, determining whether to output or re-determine the solution based on the evaluation results. This disclosure intelligently realizes a system-level joint optimization discrimination closed loop, forming optomechanical-thermoelectric co-simulation and closed-loop optimization design, improving design efficiency and intelligence level.

[0075] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0076] The embodiments described above are merely illustrative of several implementations of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these all fall within the protection scope of this disclosure. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic loads, characterized in that, Includes the following steps: Step 1: Obtain and structure the system-level design requirements for the optoelectronic payload; Step 2: The overall design intelligent agent performs requirement analysis and automatic indicator decomposition based on the structured optoelectronic payload system-level design requirements to construct a system-level indicator tree including optical disciplines, mechanical disciplines, thermal disciplines, and electronic disciplines. Each discipline branch includes discipline-level tasks. Step 3: Each subject-specific intelligent agent generates candidate solutions based on its own subject-level task, and calls simulation software to verify the obtained candidate solutions through simulation. Step 4: Update the design baseline library based on the candidate solutions obtained from all single-discipline intelligent agents and the simulation verification results; Step 5: The analyzer performs an impact analysis on the design changes of single-discipline agents based on the design baseline library and interdisciplinary dependency graph to obtain the impact level; Step 6: The analyzer determines whether the single-discipline intelligent agent corresponding to the impact level should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the generation range of candidate solutions, and the verification range of simulation verification to obtain the determination result; when the determination result includes determining that the single-discipline intelligent agent should regenerate candidate solutions and conduct simulation verification based on the simulation verification results, the single-discipline intelligent agent works according to the determination result; Step 7: When the impact level is no impact, the discriminator reads the current design baseline in the design baseline library and performs a unified evaluation of the subject-level indicators and system-level indicators. Based on the evaluation results, it determines whether the current scheme meets the preset requirements. Step 8: If the preset requirements are met, output the solution; Otherwise, return to step 2 and adjust the indicator allocation or design variable boundaries, or return to step 4 and adjust the candidate solution generation strategy.

2. The multi-agent cooperative design simulation and closed-loop optimization method for photoelectric loads according to claim 1, characterized in that, The optoelectronic payload system-level design requirements include the optoelectronic payload task book, requirements specification, interface files, environmental conditions, and historical design cases; Step 1 specifically includes: obtaining the optoelectronic payload system-level design requirements and converting the text, tables, parameters, and historical design cases therein into structured requirement objects; the structured requirement objects include system-level design objectives, optical performance constraints, mechanical performance constraints, thermal performance constraints, electronic performance constraints, interface constraints, and overall resource constraints.

3. The multi-agent cooperative design simulation and closed-loop optimization method for photoelectric loads according to claim 1, characterized in that, The process of performing requirement analysis and automatic indicator decomposition to construct a system-level indicator tree encompassing optical, mechanical, thermal, and electronic disciplines includes: performing requirement analysis and automatic indicator decomposition to obtain multiple candidate decomposition schemes; scoring the candidate decomposition schemes based on requirement coverage, historical similar scheme matching degree, cross-disciplinary constraint consistency, system-level performance prediction, design risk, and simulation calculation cost; determining the disciplinary task allocation scheme based on the scoring results; and constructing the system-level indicator tree based on the disciplinary task allocation scheme.

4. The multi-agent cooperative design simulation and closed-loop optimization method for photoelectric loads according to claim 1, characterized in that, Step 3 specifically involves the following steps: For optics, a single-discipline intelligent agent generates optical structure parameters, optical component parameters, and imaging quality analysis based on the optical task to generate candidate solutions for the optical task; for mechanics, a single-discipline intelligent agent generates support structures, mounting interfaces, lightweight structures, and structural analysis based on the mechanical task to generate candidate solutions for the mechanical task; for thermodynamics, a single-discipline intelligent agent generates heat source distribution, heat dissipation paths, thermal control strategies, and temperature field analysis based on the thermodynamic task to generate candidate solutions for the thermodynamic task; and for electronics, a single-discipline intelligent agent generates electronic architecture, device layout, power consumption distribution, printed circuit board constraints, signal integrity, power integrity, and electromagnetic compatibility analysis based on the electronics task to generate candidate solutions for the electronics task.

5. The multi-agent cooperative design simulation and closed-loop optimization method for photoelectric loads according to claim 1, characterized in that, Step 4 includes submitting the subject-level tasks, key design variables, interface variables, simulation tasks for simulation verification, simulation verification results, constraint satisfaction status, version information, and change records corresponding to each single-discipline intelligent agent to the design baseline library.

6. The multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic loads according to claim 1, characterized in that, The candidate solution generation process for the single-discipline intelligent agent is represented as follows: in, Indicates the subject number; Indicate subject Initial candidate design variables; This refers to the subject-specific tasks assigned by the overall design intelligent agent; Indicate subject Historical design cases; Indicate subject Knowledge base and design constraints; Indicate subject The corresponding candidate solution generation function for single-discipline intelligent agents.

7. The multi-agent cooperative design simulation and closed-loop optimization method for photoelectric loads according to claim 1, characterized in that, Step 5 specifically includes: the analyzer identifies effective change variables based on the differences between the current design baseline and the historical design baseline, and uses an interdisciplinary dependency graph to analyze the scope of influence of the effective change variables to obtain the influence level.

8. The multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic loads according to claim 1, characterized in that, The impact levels include Level 1 impact, Level 2 impact, Level 3 impact, and no impact. The determination result corresponding to Level 1 impact is that the single-discipline intelligent agent corresponding to this impact level regenerates all candidate solutions and re-performs all simulation verifications. The determination result corresponding to Level 2 impact is that the single-discipline intelligent agent corresponding to this impact level regenerates some candidate solutions and re-performs some simulation verifications. The determination result corresponding to Level 3 impact is that the single-discipline intelligent agent corresponding to this impact level does not regenerate candidate solutions and does not re-perform simulation verifications.

9. The multi-agent cooperative design simulation and closed-loop optimization method for optoelectronic loads according to claim 8, characterized in that, The unified evaluation of various subject-level and system-level indicators, and the determination of whether the current solution meets the preset requirements based on the evaluation results, specifically involves: determining whether the current solution meets the preset requirements based on the comprehensive objective function and convergence conditions; meeting the preset requirements means meeting the convergence conditions and the comprehensive objective function value reaches the preset value.

10. The multi-agent cooperative design simulation and closed-loop optimization method for photoelectric loads according to claim 9, characterized in that, The comprehensive objective function includes optical objective terms, mechanical objective terms, thermal objective terms, electronic objective terms, system-level objective terms, and constraint violation penalty terms; the convergence condition is: simultaneously satisfying optical, mechanical, thermal, electronic, and system-level constraints, and if the improvement of the comprehensive objective function is less than a set threshold in several consecutive iterations, the analyzer will no longer trigger first-level or second-level effects.

Images