An aero-engine control scheme determination method and system based on an agent

Abstract

The application discloses an aero-engine control scheme determination method and system based on an intelligent agent, establishes a basic tool model library for control scheme determination; provides task information of control design to a design intelligent agent, calls part of basic tools in the basic tool model library; acquires a basic tool calling request generated by the design intelligent agent; and performs analysis and verification; if the verification is passed, the basic tools are called to generate a control scheme and generate an execution receipt; the control scheme and the execution receipt are verified, and the control scheme passed by the verification is taken as a final control scheme; and the execution receipt and the verification result of the control scheme not passed by the verification are provided to the design intelligent agent as existing iteration records. Each structure of the control scheme is established as a basic tool model library, part of tools in the basic tool model library is called by the design intelligent agent, part of tools in the control scheme determination is fixed and not modified by the design intelligent agent, and the risk of incorrect modification of the design intelligent agent is reduced.

Description

Technical Field

[0001] This invention relates to aero-engine control design, and more particularly to a method and system for determining aero-engine control schemes based on intelligent agents. Background Technology

[0002] The design of aero-engine control schemes requires simultaneous consideration of control quality and safety boundaries. The control scheme must not only meet performance requirements such as steady-state accuracy, response speed, and overshoot, but also avoid safety issues such as over-revving, over-temperature, and surge, while ensuring that fuel flow and actuator change rates do not exceed actual capabilities. These performance requirements and safety constraints must be met simultaneously under different flight altitudes, speeds, and degradation conditions, but the dynamic characteristics of an engine vary under different conditions. Therefore, control schemes typically require multiple iterative processes, including control structure selection, parameter tuning, multi-condition simulation, performance evaluation, and result correction. In actual design, designers often need to determine the source of problems based on simulation curves, evaluation indicators, and engineering experience, and then adjust control parameters, optimize objectives, or control structures, resulting in numerous trial calculations, long design cycles, and difficulties in reusing experience.

[0003] In recent years, large-scale model agents have acquired the capabilities of task decomposition, tool invocation, code generation, and result processing, which can assist in tasks such as program invocation, parameter modification, and simulation execution. However, the design of aero-engine control schemes is not a typical automated program execution task. Such tasks rely on defined engine mathematical models, control quality calculation methods, safety constraint verification methods, and multi-condition testing procedures. If a general-purpose large-scale model agent is directly allowed to modify local programs, generate simulation scripts, or judge simulation results independently, although it may produce runnable code or procedures, the calculations in different rounds may not be based on the same set of models, indicators, and constraints. This could lead to problems such as erroneous modification of evaluation rules, erroneous modification of safety constraints, and referencing results not executed by real tools. For safety-critical objects like aero-engine control, the qualification of a control scheme should not be determined solely by the large-scale model itself, but should be evaluated based on pre-defined and validated engineering calculation and constraint verification procedures.

[0004] Existing intelligent agent tool invocation technologies focus more on converting natural language tasks into program calls or code execution, which is not yet suitable for directly adapting to the controlled modifications and traceable verification requirements in aero-engine control scheme design. Furthermore, if each round of simulation results is only fed back to the design agent in the form of natural language or ordinary files, problems such as unclear result sources, unclear execution status, and unclear input-output correspondences can easily arise. This makes it difficult to determine whether the result corresponds to the preset tool execution record, and also hinders subsequent traceability and verification. Summary of the Invention

[0005] Purpose of the invention: To address the above-mentioned shortcomings, this invention provides an agent-based method for determining aero-engine control schemes to reduce the risk of erroneous modification of large models. This invention also provides an agent-based system for determining aero-engine control schemes.

[0006] Technical solution: The method for determining an aero-engine control scheme based on an intelligent agent, as described in this invention, includes the following steps:

[0007] Establish a basic tool model library to complete the determination of the control scheme; the basic tool model library includes several basic tools;

[0008] The system provides the design agent with task information for control design, including control design objectives, a list of available tools, constraints, and existing iteration records; the list of available tools includes some basic tools from the basic tool model library.

[0009] Obtain the basic tool call request generated by the design agent based on task information; and parse and verify the basic tool call request;

[0010] Based on the verified basic tool call request, the basic tool is invoked, and in conjunction with other basic tools in the basic tool model library, control schemes and execution receipts for the control schemes are generated.

[0011] The control scheme and execution receipts are verified. The control scheme that passes the verification is taken as the final control scheme. The execution receipts and verification results of the control schemes that fail the verification are provided to the design agent as existing iteration records.

[0012] Furthermore, the basic tools include an engine mathematical model interface, a control method library, a control scheme generation interface, a parameter optimization tool, a simulation evaluation tool, and execution receipt generation and verification rules. The engine mathematical model interface is used to call established aero-engine mathematical models; the control method library provides optional control laws or control algorithm frameworks; the control scheme generation interface generates configuration files that can be recognized by the simulation program, including control loops, parameter lists, and variable mapping relationships; the parameter optimization tool performs numerical optimization under given parameter ranges and objective function conditions; the simulation evaluation tool calculates control quality indicators and safety constraint indicators under multi-condition simulations; and the execution receipt generation and verification rules specify the receipt format and verification rules for the execution results when the basic tools are actually called.

[0013] Furthermore, the basic tool call request is represented as: tool name and input parameters; the input parameters include one or more of the following: control method, variable to be optimized, parameter range, objective function, constraints, simulation case set, and existing execution receipt reference.

[0014] Furthermore, when verifying the basic tool call request, the verification content includes whether the tool name exists, whether the input parameters are complete, whether the parameter values ​​exceed the allowed range, and whether it involves modifying fixed content; the fixed content includes the engine mathematical model, the evaluation index calculation method, the safety constraint verification logic, and the execution receipt verification rules.

[0015] Furthermore, the specific steps for generating the control scheme include:

[0016] Call the control method library and select a control method template;

[0017] Call the control scheme generation interface to generate the control loop configuration file of the control scheme based on the control method template;

[0018] Call the engine mathematical model interface to obtain the aero-engine mathematical model;

[0019] The parameter optimization tool is invoked to drive the mathematical model of the aero-engine for iterative simulation based on the control loop configuration file, and numerical optimization is achieved. During the numerical optimization process, the simulation evaluation tool is invoked to calculate the control quality index and safety constraint index of the current parameter combination.

[0020] Furthermore, the execution receipt includes one or more of the following: execution number, tool name, input parameter summary, output structure summary, execution time, execution status, and tool signature or verification code.

[0021] Furthermore, when verifying the execution receipt, the verification content includes: whether the execution number exists, whether the input parameter digest matches, whether the output result digest matches, and whether the tool signature or verification code is valid.

[0022] Furthermore, when verifying the control scheme, the verification content includes whether the control quality indicators and safety constraint indicators generated by the simulation evaluation tool meet the requirements; the control quality indicators include steady-state error, response time, and overshoot, and the safety constraint indicators include speed, temperature, surge margin, fuel flow rate, actuator amplitude, or actuator change rate.

[0023] Furthermore, when the control scheme fails the verification, an evaluation report with a valid execution receipt is fed back to the design agent. The design agent adjusts the control parameters, parameter search range, optimizes target settings, simulation condition configuration or control method selection based on the evaluation report, and generates the next round of basic tool call request.

[0024] The present invention discloses an agent-based aero-engine control scheme determination system, comprising:

[0025] The model library construction module is used to establish a basic tool model library for determining the control scheme; the basic tool model library includes several basic tools.

[0026] The agent invocation request generation module is used to provide the design agent with task information for control design. The task information includes control design objectives, a list of callable tools, constraints, and existing iteration records. The list of callable tools includes some basic tools in the basic tool model library. The module also obtains basic tool invocation requests generated by the design agent based on the task information.

[0027] The local tool execution module is used to parse and verify the basic tool call request, call the basic tool based on the verified basic tool call request, and combine other basic tools in the basic tool model library to generate control schemes and control scheme execution receipts.

[0028] The result verification module is used to verify the control scheme and execution receipts. The control scheme that passes the verification is taken as the final control scheme. The execution receipts and verification results of the control schemes that fail the verification are provided to the design agent as existing iteration records.

[0029] To address the problem that existing general-purpose large-model intelligent agents cannot directly meet the requirements of controlled modification, fixed evaluation, and traceable verification in the design of aero-engine control schemes, this invention provides an agent-based method for determining aero-engine control schemes. This method designs the engine mathematical model, control method library, parameter optimization program, simulation evaluation program, and execution feedback verification rules into a basic tool model that can be structurally invoked by the design intelligent agent. This enables the design intelligent agent to complete control method selection, parameter adjustment, simulation condition setting, and iterative correction within a defined scope of objects and parameters, reducing the risks of freely modifying the model and verification rules and referencing non-realistic execution results, and minimizing repeated parameter tuning and verification work.

[0030] Beneficial effects: Compared with the prior art, the significant advantages of this invention are:

[0031] (1) It helps reduce the risk of accidental modification by the design agent. This method establishes the various structures of the control scheme into a basic tool model library. By having the design agent call some tools in the basic tool model library, some tools in the control scheme determination process are fixed and cannot be modified by the design agent, thus reducing the risk of accidental modification by the design agent. For example, the engine mathematical model, evaluation index calculation method, safety constraint verification logic, and execution acknowledgment verification rules are set as fixed content. The design agent can only modify some content such as control method, parameters, objective function, and simulation condition parameter settings within a preset range, thus reducing the risk of accidental modification of the engine mathematical model, evaluation index calculation method, safety constraint verification logic, and execution acknowledgment verification rules by the design agent.

[0032] (2) It helps to improve the traceability of the design process and the credibility of the results. This method requires that each execution of the local tool generates a bound execution receipt, and records the input and output parameters, execution status and verification information in a closed loop, providing an immutable basis for subsequent scheme iterations; at the same time, since the engine model, evaluation program and safety constraints remain fixed, it ensures that multiple schemes can be compared horizontally under a unified benchmark, and maintains the comparability of simulation results; finally, the fixed tool and authentication rules replace the large model to give the control quality and safety evaluation, eliminating the risk of fabricating results, and helping to improve the applicability in the field of safety-critical aero-engine engineering.

[0033] (3) It helps reduce repeated parameter tuning and result processing. The design agent generates the next round of structured tool call requests based on the verified evaluation results. The local tool execution module automatically completes the scheme generation, parameter optimization and simulation evaluation, thus forming a closed-loop iterative process. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the method for determining the control scheme of the present invention.

[0035] Figure 2 This is a flowchart illustrating the process of establishing and executing return receipt verification rules for the basic tool model library in this invention.

[0036] Figure 3 This is a flowchart illustrating the process of designing a structured tool call request for intelligent agents in this invention, which is then verified by the system.

[0037] Figure 4 This is a flowchart illustrating how the local tool execution module in this invention completes the generation of control schemes, parameter optimization, simulation evaluation, and generation of execution receipts.

[0038] Figure 5 This is a flowchart of the result verification module in this invention, which performs verification and feedback iterative correction based on execution receipts and simulation evaluation results.

[0039] Figure 6 This is the structure of a single-shaft turbojet engine speed control system in an embodiment of the present invention.

[0040] Figure 7 The following are simulation results of the single-shaft turbojet engine speed control in an embodiment of the present invention. Detailed Implementation

[0041] like Figure 1 As shown in this embodiment, a method for determining an aero-engine control scheme based on an intelligent agent includes the following steps:

[0042] Establish a basic tool model library to complete the determination of the control scheme; the basic tool model library includes several basic tools;

[0043] The system provides the design agent with task information for control design, including control design objectives, a list of available tools, constraints, and existing iteration records; the list of available tools includes some basic tools from the basic tool model library.

[0044] Obtain the basic tool call request generated by the design agent based on task information; and parse and verify the basic tool call request;

[0045] Based on the verified basic tool call request, the basic tool is invoked, and in conjunction with other basic tools in the basic tool model library, control schemes and execution receipts for the control schemes are generated.

[0046] The control scheme and execution receipts are verified. The control scheme that passes the verification is taken as the final control scheme. The execution receipts and verification results of the control schemes that fail the verification are provided to the design agent as existing iteration records.

[0047] This embodiment uses the design of a speed control scheme for a single-shaft turbojet engine as an example to illustrate the specific implementation process of the method of the present invention. The core control objective of this turbojet engine is to regulate the fuel flow rate W. f To stabilize or change engine speed N, while limiting turbine inlet temperature T t4 and compressor surge margin S m .

[0048] Step 1: Establish a basic tool model library to complete the determination of the control scheme.

[0049] like Figure 2 As shown, firstly, a basic tool model library is established based on control design requirements. The physical models and calculation rules within the basic tool model library remain fixed and are not directly modified by the design agent. The basic tools in the basic tool model library include an engine mathematical model interface, a control method library, a control scheme generation interface, parameter optimization tools, simulation evaluation tools, and execution feedback generation and verification rules.

[0050] (1) Engine Mathematical Model Interface: Used to call the established aero-engine mathematical model and use the engine's component-level physical model for simulation. This model receives the fuel flow rate W. f As control inputs, calculate and output engine speed N and turbine inlet temperature T. t4 and compressor surge margin S m .

[0051] (2) Control method library: used to provide optional control laws or control algorithm frameworks, including standard aero-engine control algorithm frameworks or auxiliary control links such as traditional PID control, PID control with feedforward compensation, gain scheduling control, acceleration and deceleration fuel limiting.

[0052] (3) Control scheme generation interface: This interface is used to generate configuration files that can be recognized by the simulation program, including control loops, parameter lists, and variable mapping relationships. It specifies the parameter expression format for candidate schemes, including the controller proportional coefficient K. p Integral coefficient K i Differential coefficient K d The value boundaries and configuration channels also include the mapping relationship between control input variables, fuel flow command output interface and simulation variables, and finally the parameters are formed into a configuration file.

[0053] (4) Parameter optimization tool: used to complete numerical optimization under given control quality objectives, parameter boundaries and safety constraints.

[0054] (5) Simulation evaluation tool: used to provide control quality evaluation indicators and safety constraint verification logic during engine dynamic adjustment. Control quality indicators are set as follows: steady-state speed error not exceeding 0.25%, step response time not exceeding 3 seconds, and overshoot not exceeding 2%. Safety constraints are set as follows: maximum physical speed N... max Not exceeding 103%, maximum turbine inlet temperature T t4,max Minimum compressor surge margin S not higher than 1150 K m,min No less than 10%.

[0055] (6) Execution Receipt Generation and Verification Rules: This section specifies the receipt format and verification rules for the execution results when the basic tool is actually invoked. As a specific implementation, the SHA-256 digest algorithm and asymmetric signature logic can be configured. The receipt fields include "execution number, input parameter digest, output result digest, execution time, execution status, tool signature or verification code". Input parameters and output results are converted into input parameter digests and output result digests using the SHA-256 digest algorithm to improve data security. The above digest algorithm and signature method are only implementation examples; other verification methods that can bind input, output, and execution status can also be used.

[0056] In this basic tool model library, the mathematical model of the aero-engine, the calculation methods of evaluation indicators in the simulation evaluation tools, the safety constraint verification logic, and the execution receipt verification rules remain fixed. The adjustable content that the system opens to the design agent includes: control method selection, initial values ​​of controller parameters, search range of the variable to be optimized, weights of the optimization objective function, and simulation condition configuration.

[0057] Step 2: Design the intelligent agent to generate structured tool call requests and have them verified by the system.

[0058] like Figure 3 As shown, during the initial design round or each iterative design round, the system provides the design agent with the current control design task information. The task information includes the control design objective, a list of available tools, constraints, and existing iteration records. For example, the current task is: "Design a speed controller for a speed range of 70% to 100% under static takeoff conditions on the ground."

[0059] When the design agent proposes candidate control schemes, if it generates illegal instructions that relax safety constraints, such as setting the maximum turbine inlet temperature T... t4,max When the size is changed from 1150 K to 1200 K, the local tool execution module parses and verifies the received request. It identifies that the request involves modifying fixed content such as security constraint verification logic, and determines that the verification fails. The system returns the error reason and requires the designed agent to regenerate the request. The local tool execution module is a local agent on the local computer that executes tool calls.

[0060] The design agent corrects the design scheme based on the returned error reasons and generates a structured tool call request that conforms to the specifications. This request is expressed in the form of "tool name + input parameters," specifically calling a parameter optimization tool and a simulation evaluation tool. The input control engineering parameters include: selecting gain-scheduled PI and feedforward control methods, using COBYLA as the optimization algorithm, and K as the variable to be optimized. p With K i The initial values ​​are set to [1.0, 0.1]. The weights of tracking error and overshoot in the objective function are set to [0.8, 0.2], respectively. An optimization algorithm is selected, and simulation evaluation is conducted under fixed safety constraints. The local tool execution module verifies that the parameters in the request are complete and do not exceed the allowed boundaries, and allows entry into the execution process.

[0061] Step 3: The local tool execution module completes the scheme generation, optimization, and simulation, and generates an execution receipt.

[0062] like Figure 4 As shown, after receiving a verified tool call request, the local tool execution module calls the corresponding tool in the basic tool model library to complete the actual control simulation and optimization calculation.

[0063] (1) Call the control method library and select the gain-scheduled PI control method template; call the control scheme generation interface and generate a control scheme including speed command N based on the gain-scheduled PI control method template. refThe configuration file for the control loop includes the speed feedback N, speed deviation input, proportional-integral control loop, fuel flow command output, parameter boundaries, and simulation interface variable mapping relationships. This configuration file is generated by a preset template and structured parameters, and the underlying control program is not freely written by the design agent.

[0064] (2) Call the engine mathematical model interface to obtain the aero-engine mathematical model; call the parameter optimization tool to drive the engine mathematical model to perform iterative simulation of speed step response within the specified parameter range based on the control loop configuration file, and the parameter optimization tool performs numerical optimization according to the objective function and fixed safety constraints.

[0065] (3) During the optimization process, the simulation evaluation tool is used to record the dynamic response curve and calculate the control quality index and safety constraint index under the current parameter combination. For example, after the first round of optimization, the steady-state error is 0.2%, the response time is 2.8 seconds, the overshoot is 1.5%, the highest turbine inlet temperature is 1130 K, and the highest physical speed is 101.2%. However, the lowest compressor surge margin S during acceleration is... m,min The result is 8.5%. This indicates that while meeting the speed response requirements, the fuel command rises too quickly, causing the compressor operating point to approach the surge boundary.

[0066] After the calculation is completed, a corresponding execution receipt is automatically generated. This receipt records the execution code, tool name, input parameter summary, output result summary, execution time, execution status, security constraint violation status, and tool signature or verification code in the result database. The execution status is then fed back to the design agent as a reference for the next iteration. The generation of this receipt is not interfered with by the design agent.

[0067] Step 4: Based on the execution receipts and evaluation results, complete the result verification and iterative correction.

[0068] like Figure 5 As shown, after the simulation calculation is completed, the execution receipt is retrieved first for validity verification. The digest of the input and output data is recalculated using the digest algorithm and the tool signature is verified to determine whether the simulation result corresponds to the specified input parameter digest and output result digest obtained by applying the preset basic tool model library.

[0069] After the acknowledgment verification is performed, the result verification module evaluates the control quality and safety constraints of the candidate control scheme based on the specific physical indicators generated by the simulation evaluation tool. This is because the compressor's minimum surge margin S in the first round of results... m,min The margin is 8.5%, which is lower than the 10% margin indicator. The result verification module determines that the solution does not meet the requirements and feeds back the evaluation with a valid receipt to the design agent.

[0070] The intelligent agent adjusts the control design strategy based on the feedback evaluation results. To mitigate the surge risk during acceleration, the agent proposes an improvement scheme after completing relevant literature searches. In the next iteration, without changing the engine model and safety constraints, a fuel rate limiting loop is added outside the original PI control loop, and the fuel flow rate change limit parameter is used as an optimizable variable for the next iteration. This new scheme proceeds to the simulation step three after passing the verification in step two.

[0071] The execution receipts and evaluation results generated in the second round show that the steady-state error is 0.22%, the step response time is 2.9 seconds, the overshoot is 0.9%, the highest turbine inlet temperature is 1125 K, the highest physical speed is 100.8%, and the lowest compressor surge margin S during the entire dynamic acceleration process is... M,min The value is 13.2%, which meets the 10% safety constraint requirement.

[0072] The result verification module performs a final verification of the new round of execution receipts and various control quality and safety indicators, determining that the scheme meets the preset requirements. Finally, the design agent generates a final scheme description based on the verified simulation data. After the result verification module performs a consistency check and confirms that there are no errors, the system outputs the final speed control scheme for the target engine, the optimized controller parameter configuration, the effective execution receipt chain, and a complete set of simulation evaluation records.

[0073] The final control scheme generated is as follows Figure 6 As shown, a speed control method combining feedforward compensation control and gain-scheduled PI controller is presented. The feedforward compensation control is based on the target speed. The pre-calculated control output is a coarse adjustment of an open-loop nature; target speed. With real-time feedback of rotational speed Obtain speed feedback error Gain-controlled PI controller based on speed feedback error The calculation and control output belongs to closed-loop regulation. The output of the feedforward compensation control and gain scheduling PI controller passes through parameter limiting protection to prevent rapid parameter changes from exceeding the actual fuel system supply capacity, ultimately outputting fuel flow. To the engine, to provide thrust to the engine. This achieves closed-loop speed regulation. Simulation results are as follows: Figure 7 As shown, Figure 7 (a) shows the change in rotational speed over time. Figure 7 (b) shows the change in fuel flow rate over time. Figure 7(c) shows the change of the gain-scheduled PI controller parameters over time. It can be seen that during the step change and ramp change of the target speed, the fuel flow changes rapidly, and the actual speed tracks it quickly. Due to the change in speed, the PI controller parameters kp and ki, which use speed as the scheduling parameter, also change accordingly, thereby ensuring a better speed tracking control effect.

[0074] As can be seen from the above embodiments, this invention combines the scheme generation capability of the design agent, the real execution capability of the local tool execution module, and the fixed verification capability of the result verification module, enabling the input, execution, output, and evaluation in the iterative design process of aero-engine control schemes to be recorded and verified. Those skilled in the art can make equivalent substitutions or adaptive adjustments to the control method library, parameter optimization algorithm, execution receipt field, or verification method without departing from the principles of this invention; all such adjustments should fall within the protection scope of this invention.

Claims

1. A method for determining an aero-engine control scheme based on an intelligent agent, characterized in that, Includes the following steps: Establish a basic tool model library to complete the determination of the control scheme; the basic tool model library includes several basic tools; The system provides the design agent with task information for control design, including control design objectives, a list of available tools, constraints, and existing iteration records; the list of available tools includes some basic tools from the basic tool model library. Obtain the basic tool call requests generated by the design agent based on task information; The basic tool call requests are then parsed and verified. Based on the verified basic tool call request, the basic tool is invoked, and in conjunction with other basic tools in the basic tool model library, control schemes and execution receipts for the control schemes are generated. The control scheme and execution receipts are verified. The control scheme that passes the verification is taken as the final control scheme. The execution receipts and verification results of the control schemes that fail the verification are provided to the design agent as existing iteration records.

2. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 1, characterized in that, The basic tools include an engine mathematical model interface, a control method library, a control scheme generation interface, a parameter optimization tool, a simulation evaluation tool, and execution receipt generation and verification rules. Specifically, the engine mathematical model interface is used to call established aero-engine mathematical models; the control method library provides optional control laws or control algorithm frameworks; the control scheme generation interface generates configuration files that can be recognized by the simulation program, including control loops, parameter lists, and variable mapping relationships; the parameter optimization tool performs numerical optimization under given parameter ranges and objective function conditions; the simulation evaluation tool calculates control quality indicators and safety constraint indicators under multi-condition simulations; and the execution receipt generation and verification rules specify the receipt format and verification rules for the execution results when the basic tools are actually called.

3. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 2, characterized in that, The basic tool call request is represented as: tool name and input parameters; the input parameters include one or more of the following: control method, variable to be optimized, parameter range, objective function, constraints, simulation case set, and existing execution receipt reference.

4. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 3, characterized in that, When verifying the basic tool call request, the verification content includes whether the tool name exists, whether the input parameters are complete, whether the parameter values ​​exceed the allowed range, and whether it involves modifying fixed content; the fixed content includes the engine mathematical model, the evaluation index calculation method, the safety constraint verification logic, and the execution receipt verification rules.

5. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 4, characterized in that, The specific steps for generating the control scheme include: Call the control method library and select a control method template; Call the control scheme generation interface to generate the control loop configuration file of the control scheme based on the control method template; Call the engine mathematical model interface to obtain the aero-engine mathematical model; The parameter optimization tool is invoked to drive the mathematical model of the aero-engine for iterative simulation based on the control loop configuration file, and numerical optimization is achieved. During the numerical optimization process, the simulation evaluation tool is invoked to calculate the control quality index and safety constraint index of the current parameter combination.

6. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 5, characterized in that, The execution receipt includes one or more of the following: execution number, tool name, input parameter summary, output result summary, execution time, execution status, and tool signature or verification code.

7. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 6, characterized in that, When verifying the execution receipt, the verification includes: whether the execution number exists, whether the input parameter digest matches, whether the output result digest matches, and whether the tool signature or verification code is valid.

8. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 7, characterized in that, When verifying the control scheme, the verification content includes whether the control quality indicators and safety constraint indicators generated by the simulation evaluation tool meet the requirements; the control quality indicators include steady-state error, response time, and overshoot; the safety constraint indicators include speed, temperature, surge margin, fuel flow rate, actuator amplitude, or actuator change rate.

9. The method for determining an aero-engine control scheme based on an intelligent agent according to claim 8, characterized in that, When the control scheme fails the verification, an evaluation report with a valid execution receipt is sent to the design agent. The design agent adjusts the control parameters, parameter search range, optimizes target settings, simulation condition configuration or control method selection based on the evaluation report, and generates the next round of basic tool call request.

10. A system for determining control schemes for aero-engines based on intelligent agents, characterized in that, include: The model library construction module is used to build a basic tool model library for determining the control scheme; The basic tool model library includes several basic tools; The agent invocation request generation module is used to provide the design agent with task information for control design. The task information includes control design objectives, a list of callable tools, constraints, and existing iteration records. The list of callable tools includes some basic tools in the basic tool model library. The module also obtains basic tool invocation requests generated by the design agent based on the task information. The local tool execution module is used to parse and verify the basic tool call request, call the basic tool based on the verified basic tool call request, and combine other basic tools in the basic tool model library to generate control schemes and control scheme execution receipts. The result verification module is used to verify the control scheme and execution receipts. The control scheme that passes the verification is taken as the final control scheme. The execution receipts and verification results of the control schemes that fail the verification are provided to the design agent as existing iteration records.

Images