Aero-engine generation design method and system based on fusion implicit empirical reasoning

Abstract

The invention relates to the technical field of aero-engine overall design, and discloses an aero-engine generation design method and system based on fusion implicit empirical reasoning, and the method comprises the steps: generating an initial thermal scheme population at a thermal parameter space sampling boundary, and carrying out the optimization through a multi-objective evolutionary algorithm, decision feature vectors including multi-dimensional physical and engineering attributes such as thermal load, cost and service life are constructed, and a related engineering knowledge base is retrieved in combination with an RAG technology, so that downstream structure design is not blind trial and error any more, but can actively adapt to upstream top layer decisions, and the engineering rationality of a design scheme is greatly improved; besides, standard actions of'positive and negative experience focusing / protocol + multi-target optimization 'are introduced into a performance layer and a structure layer, and invalid spaces are greatly eliminated in the initial stage of design, so that the problem of'dimensionality disaster' in a high-dimensional design space is solved, and the generation efficiency is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of overall design technology for aero-engines, and discloses a generative design method and system for aero-engines based on fusion implicit empirical reasoning. Background Technology

[0002] Existing aero-engine design technologies mainly rely on human expert experience or traditional numerical optimization algorithms. Although these technologies have been widely applied in engineering practice, they have revealed the following significant problems when facing the development requirements of high-performance and highly complex next-generation aero-engines:

[0003] Question 1: Traditional design methods struggle to address the optimization challenges in high-dimensional design spaces. The existing aero-engine design process is essentially a trial-and-error method relying on human expert experience. Designers determine the initial configuration (such as the number of stages and the flow channel form) based on experience, and then perform sequential iterations. However, the overall design of an aero-engine involves hundreds of coupled variables, such as thermodynamic parameters, flow channel geometry, and structural dimensions, forming an extremely high-dimensional non-convex design space. Relying on human experience can only explore a very small part of this space, and it is very easy to get trapped in local optima; while directly using traditional optimization algorithms (such as genetic algorithms) to search in such a high-dimensional space will encounter a severe "curse of dimensionality," with extremely slow or even non-convergent computational convergence.

[0004] Question 2: Generative AI lacks guarantees of physical consistency While generative AI (such as GANs and LLMs) performs exceptionally well in areas like image generation, its direct application to precision mechanical design suffers from a serious "illusion" problem. AI-generated solutions often violate fundamental physical laws (such as non-conservation of mass and excessive stress), rendering subsequent engineering simulations impossible and resulting in solutions with virtually no practical engineering value.

[0005] Question 3: Unidirectional information flow and intent distortion in hierarchical design In the traditional "overall → structural" design process, the information flow is unidirectional. The upstream overall design team only transmits parameters to the downstream team, while the implicit "design intent" (such as limiting the use of expensive materials to reduce costs) cannot be explicitly perceived by the downstream team. When the downstream structural team finds that the strength constraints cannot be met, due to the lack of an automated backtracking mechanism, they can only rely on manual offline coordination and rework, resulting in long design iteration cycles and low collaboration efficiency. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for generating and designing aero-engines based on implicit empirical reasoning. By introducing the standard action of "positive and negative empirical focusing / reduction + multi-objective optimization" at both the performance and structural levels, and by significantly eliminating invalid space in the early stage of design, the "curse of dimensionality" problem in high-dimensional design space is solved, and the generation efficiency is significantly improved.

[0007] To achieve the above-mentioned technical effects, the technical solution adopted by the present invention is as follows: A generative design method for aero-engines based on implicit empirical reasoning includes: S1. According to the requirements of the project task book for aero-engine design, extract the engine performance and structural index parameters, including thrust, fuel consumption rate, cost, lifespan, and size limitations. S2. Preprocess the thermodynamic parameter space in the historical database that satisfies the performance structure index parameters to obtain the thermodynamic parameter space sampling boundary based on weighted kernel density; the thermodynamic parameters include engine inlet flow rate, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency; S3. Using a preset sampling strategy, an initial thermodynamic scheme population of the aero-engine is generated within the spatial sampling boundary of the thermodynamic parameters. The thermodynamic cycle simulation program is called to calculate the performance index parameters. With the performance index parameters meeting the corresponding preset requirements as the optimization objective, a multi-objective evolutionary algorithm is used to optimize and obtain the Pareto front solution set of the thermodynamic scheme. S4. Select a set of initial thermodynamic schemes from the Pareto front solution set, and construct an engineering decision feature vector containing engine thermal load, cost sensitivity, and life requirements based on the requirements of the project task book and the thermodynamic parameters in the initial thermodynamic schemes. S5. Based on the engineering decision feature vector, retrieve the engineering knowledge base to generate a structural design domain specification containing logical basis. Based on the design domain specification and thermodynamic parameters, calculate the flow channel geometry of each component section, and use an optimization algorithm to find a structural parameter design scheme that satisfies multi-objective constraints. The multi-objective constraints include maximum diameter constraints and rotor tip velocity constraints. The design domain specification includes the range of compressor stage values, the range of turbine stage values, and the range of compressor hub ratio values.

[0008] Furthermore, methods for obtaining the spatial sampling boundary of thermodynamic parameters based on weighted kernel density include: The cosine similarity between the current task requirement feature vector and the historical case feature vector is calculated as the confidence weight; the feature vector includes fuel consumption rate and thrust. Based on the confidence weights, kernel density estimation is performed on the key design variables to generate probability density functions; Based on the probability density function, the pre-set information interval is used as the sampling domain.

[0009] Furthermore, the method for retrieving the engineering knowledge base based on the engineering decision feature vector in step S5 and generating a structural design domain specification containing logical basis includes: Construct a query instruction containing the engineering decision feature vector, and use retrieval enhancement generation technology to retrieve matching engineering specifications and design principles from a vector database containing design experience and engineering knowledge; The large language model generates a chain of evidence in natural language form based on the retrieval results and outputs the range of values ​​for key structural parameters, including the range of values ​​for the number of compressor stages, the number of turbine stages, and the compressor hub ratio in the design domain specification.

[0010] Furthermore, in step S5, the flow channel geometry of each component cross-section is calculated, and an optimization algorithm is used to find a structural parameter design scheme that satisfies multi-objective constraints. This includes: Based on thermodynamic and aerodynamic design parameters, the diameter of the critical cross-section of the flow path is deterministically calculated using one-dimensional aerodynamic formulas. The critical cross-sections include the engine inlet cross-section, compressor outlet cross-section, combustion chamber outlet cross-section, and turbine outlet cross-section. Based on the determined key cross-section, the flow channel geometry is constructed, and the optimal structural parameters are sought with the goal of minimizing cost or weight, while satisfying the maximum diameter constraint and rotor tip velocity constraint.

[0011] Furthermore, if step S5 fails to generate a structural parameter design scheme that satisfies the multi-objective constraints, the sampling boundary of the thermodynamic parameter space is adjusted, and steps S3 to S5 are repeated until a structural parameter design scheme that satisfies the multi-objective constraints is obtained.

[0012] Furthermore, attribution analysis is performed to adjust the spatial sampling boundaries of the thermodynamic parameters, specifically including: Calculate the sensitivity gradient of the structural constraint index relative to the thermodynamic parameters, lock the parameter with the largest sensitivity gradient magnitude as the key thermodynamic parameter, and give the adjustment direction and magnitude of the key thermodynamic parameter to meet the structural constraints based on the sign and magnitude of the sensitivity gradient. The key thermodynamic parameters are jointly analyzed with the engineering decision feature vector, and the inference engine is used to determine whether there is a hard constraint conflict between the adjustment of key thermodynamic parameters and the engineering decision features. If there is no hard constraint conflict, the direction and magnitude of adjustment of the key thermodynamic parameters are given, the sampling boundary of the key thermodynamic parameters is corrected, and the optimization is re-executed to find a design scheme that meets the structural constraints. If there is a hard constraint conflict, that is, adjusting the key thermodynamic parameters will violate the cost or life constraints in the engineering decision characteristics, the size or weight requirements in the structural constraints are relaxed.

[0013] To achieve the above-mentioned technical effects, the present invention also provides an aero-engine generation design system based on fused implicit empirical reasoning, comprising: The mission analysis module is used to extract engine performance and structural index parameters according to the requirements of the project mission statement for aero-engine design. These performance and structural index parameters include thrust, fuel consumption rate, cost, lifespan, and size limitations. The sampling boundary analysis module is used to preprocess the thermodynamic parameter space that satisfies the performance structure index parameters in the historical database to obtain the sampling boundary of the thermodynamic parameter space based on weighted kernel density; the thermodynamic parameters include engine inlet flow rate, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency. The first optimization analysis module is used to generate an initial thermodynamic scheme population of the aero-engine within the spatial sampling boundary of thermodynamic parameters using a preset sampling strategy, call the thermodynamic cycle simulation program to calculate performance index parameters and structural index parameters, take the performance index parameters meeting the corresponding preset requirements as the optimization objective, and use a multi-objective evolutionary algorithm to optimize and obtain the Pareto front solution set of the thermodynamic scheme. The feature vector construction module is used to select a set of initial thermodynamic schemes from the Pareto front solution set, and construct an engineering decision feature vector containing engine thermal load, cost sensitivity, and life requirements based on the requirements of the project task book and the thermodynamic parameters in the initial thermodynamic schemes. The second optimization analysis module is used to retrieve the engineering knowledge base based on the engineering decision feature vector, generate a structural design domain specification containing logical basis, and calculate the flow channel geometry of each component section according to the design domain specification and thermodynamic parameters. The optimization algorithm is used to find a structural parameter design scheme that meets multiple objective constraints. The multiple objective constraints include maximum diameter constraints and rotor blade tip velocity constraints. The design domain specification includes the range of compressor stage values, the range of turbine stage values, and the range of compressor hub ratio values.

[0014] Furthermore, it also includes a parameter adjustment module, which is used to adjust the thermal parameter space sampling boundary when a structural parameter design scheme that satisfies the multi-objective constraints cannot be generated, until a structural parameter design scheme that satisfies the multi-objective constraints is obtained.

[0015] Furthermore, the parameter adjustment module includes: The sensitivity gradient analysis unit is used to calculate the sensitivity gradient of the structural constraint index relative to the thermodynamic parameters, lock the parameter with the largest sensitivity gradient magnitude as the key thermodynamic parameter, and give the adjustment direction and magnitude of the key thermodynamic parameter to meet the structural constraints based on the sign and magnitude of the sensitivity gradient. The constraint conflict analysis unit is used to jointly analyze the key thermodynamic parameters and the engineering decision feature vector, and use the inference engine to determine whether there is a hard constraint conflict between the adjustment of key thermodynamic parameters and the engineering decision features. The adjustment unit is used to provide the direction and magnitude of adjustment for key thermodynamic parameters when there are no hard constraint conflicts, correct the sampling boundaries of key thermodynamic parameters, and re-execute the optimization to find a design scheme that meets the structural constraints. When there are hard constraint conflicts, that is, when adjusting key thermodynamic parameters will violate the cost or life constraints in the engineering decision characteristics, the size or weight requirements in the structural constraints are relaxed.

[0016] Compared with the prior art, the beneficial effects of this invention are: This invention constructs a decision feature vector that includes multi-dimensional physical and engineering attributes such as heat load, cost, and lifespan, and combines it with RAG technology to retrieve relevant standards. This enables downstream structural design to move beyond blind trial and error and proactively adapt to upstream top-level decisions (such as automatically matching inexpensive materials with low safety factors for low-cost target machines), greatly improving the engineering rationality of the design scheme.

[0017] This invention introduces the standard action of "positive and negative empirical focusing / reduction + multi-objective optimization" in both the performance and structural layers, and solves the "curse of dimensionality" problem in high-dimensional design space by significantly eliminating invalid space in the early stage of design, thereby significantly improving generation efficiency. Attached Figure Description

[0018] Figure 1 This is a flowchart of the aero-engine generation design method based on fused implicit empirical reasoning in Example 1 or 2; Figure 2 This is a block diagram of the aero-engine generation design system based on fused implicit empirical reasoning in Example 1; The module includes: 1. Task analysis module; 2. Sampling boundary analysis module; 3. First optimization analysis module; 4. Feature vector construction module; 5. Second optimization analysis module; 6. Parameter adjustment module; 601. Sensitivity gradient analysis unit; 602. Constraint conflict analysis unit; 603. Adjustment unit. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0020] Example 1 See Figure 1 and Figure 2 A generative design method for aero-engines based on implicit empirical reasoning includes: S1. According to the requirements of the project task book for aero-engine design, extract the engine performance and structural index parameters, including thrust, fuel consumption rate, cost, lifespan, and size limitations. S2. Preprocess the thermodynamic parameter space in the historical database that satisfies the performance structure index parameters to obtain the thermodynamic parameter space sampling boundary based on weighted kernel density; the thermodynamic parameters include engine inlet flow rate, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency; S3. Using a preset sampling strategy, an initial thermodynamic scheme population of the aero-engine is generated within the spatial sampling boundary of the thermodynamic parameters. The thermodynamic cycle simulation program is called to calculate the performance index parameters. With the performance index parameters meeting the corresponding preset requirements as the optimization objective, a multi-objective evolutionary algorithm is used to optimize and obtain the Pareto front solution set of the thermodynamic scheme. S4. Select a set of initial thermodynamic schemes from the Pareto front solution set, and construct an engineering decision feature vector containing engine thermal load, cost sensitivity, and life requirements based on the requirements of the project task book and the thermodynamic parameters in the initial thermodynamic schemes. S5. Based on the engineering decision feature vector, retrieve the engineering knowledge base to generate a structural design domain specification containing logical basis. Based on the design domain specification and thermodynamic parameters, calculate the flow channel geometry of each component section, and use an optimization algorithm to find a structural parameter design scheme that satisfies multi-objective constraints. The multi-objective constraints include maximum diameter constraints and rotor tip velocity constraints. The design domain specification includes the range of compressor stage values, the range of turbine stage values, and the range of compressor hub ratio values.

[0021] In this embodiment, by preprocessing the thermodynamic parameter space that meets the performance and structural index parameters of aero-engines in the historical database, a sampling boundary of the thermodynamic parameter space based on weighted kernel density is obtained. Then, an initial thermodynamic scheme population is generated on the sampling boundary of the thermodynamic parameter space, and optimization is performed through a multi-objective evolutionary algorithm to construct a decision feature vector containing multi-dimensional physical and engineering attributes such as heat load, cost, and lifespan. Combined with RAG technology to retrieve relevant standards, the downstream structural design is no longer a blind trial and error, but can actively adapt to the top-level decisions of the upstream (such as automatically matching inexpensive materials and low safety factors for low-cost target drones), which greatly improves the engineering rationality of the design scheme. In addition, by introducing the standard action of "positive and negative experience focusing / reduction + multi-objective optimization" at both the performance and structural layers, the invalid space is greatly eliminated in the early stage of design, solving the "curse of dimensionality" problem in high-dimensional design space and significantly improving the generation efficiency.

[0022] Based on the same inventive concept, this embodiment also provides an aero-engine generation design system based on fused implicit empirical reasoning, including: Task analysis module 1 is used to extract engine performance and structural index parameters according to the requirements of the project task book for aero-engine design. The performance and structural index parameters include thrust, fuel consumption rate, cost, lifespan, and size limitations. The sampling boundary analysis module 2 is used to preprocess the thermodynamic parameter space that satisfies the performance structure index parameters in the historical database to obtain the sampling boundary of the thermodynamic parameter space based on the weighted kernel density; the thermodynamic parameters include engine inlet flow rate, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency. The first optimization analysis module 3 is used to generate an initial thermodynamic scheme population of the aero-engine within the spatial sampling boundary of the thermodynamic parameters using a preset sampling strategy, call the thermodynamic cycle simulation program to calculate the performance index parameters, take the performance index parameters meeting the corresponding preset requirements as the optimization objective, and use a multi-objective evolutionary algorithm to optimize and obtain the Pareto front solution set of the thermodynamic scheme. Feature vector construction module 4 is used to select a set of initial thermodynamic schemes from the Pareto front solution set, and construct an engineering decision feature vector containing engine thermal load, cost sensitivity, and life requirements based on the requirements of the project task book and the thermodynamic parameters in the initial thermodynamic schemes. The second optimization analysis module 5 is used to retrieve the engineering knowledge base based on the engineering decision feature vector, generate a structural design domain specification containing logical basis, and calculate the flow channel geometry of each component section according to the design domain specification and thermodynamic parameters, and use optimization algorithms to find a structural parameter design scheme that satisfies multi-objective constraints; the multi-objective constraints include maximum diameter constraints and rotor blade tip velocity constraints; the design domain specification includes the range of compressor stage values, the range of turbine stage values, and the range of compressor hub ratio values.

[0023] The aero-engine generation design system in this embodiment also includes a parameter adjustment module 6, which is used to adjust the thermal parameter space sampling boundary when a structural parameter design scheme that meets the multi-objective constraints cannot be generated, until a structural parameter design scheme that meets the multi-objective constraints is obtained.

[0024] The parameter adjustment module 6 includes: The sensitivity gradient analysis unit 601 is used to calculate the sensitivity gradient of the structural constraint index relative to the thermodynamic parameters, lock the parameter with the largest sensitivity gradient magnitude as the key thermodynamic parameter, and give the adjustment direction and magnitude of the key thermodynamic parameter to meet the structural constraints based on the sign and magnitude of the sensitivity gradient. The constraint conflict analysis unit 602 is used to jointly analyze the key thermodynamic parameters and the engineering decision feature vector, and use the inference engine to determine whether there is a hard constraint conflict between the adjustment of key thermodynamic parameters and the engineering decision features. The adjustment unit 603 is used to provide the adjustment direction and magnitude of key thermodynamic parameters when there is no hard constraint conflict, correct the sampling boundary of key thermodynamic parameters and re-execute the optimization to find a design scheme that meets the structural constraints; when there is a hard constraint conflict, that is, adjusting the key thermodynamic parameters will violate the cost or life constraints in the engineering decision characteristics, the size or weight requirements in the structural constraints are relaxed.

[0025] Example 2 See Figure 1 This embodiment takes the overall generative design of a high-bypass ratio engine for a single-aisle mainline passenger aircraft as an example to describe in detail the process of the aero-engine generative design method based on implicit empirical reasoning of the present invention. The specific process is as follows: S1. According to the requirements of the project task book for aero-engine design, extract the engine performance and structural index parameters, including thrust, fuel consumption rate, cost, lifespan, and size limitations. As required by the project task book in this embodiment: Design a high bypass ratio engine suitable for civil passenger aircraft.

[0026] Key performance indicators: engine power level 10-15 tons, extremely low fuel consumption rate (SFC), and lifespan > 20,000 cycles.

[0027] Constraints: For the wing-mounted layout, the maximum diameter and overall length of the engine nacelle are strictly limited to reduce cruise drag and meet ground clearance requirements.

[0028] S2. Preprocess the thermodynamic parameter space in the historical database that satisfies the performance structure index parameters to obtain the thermodynamic parameter space sampling boundary based on weighted kernel density; the thermodynamic parameters include engine inlet flow rate, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency; Analysis of historical databases of civil aviation engines (covering CFM56, LEAP, GTF, etc.) revealed that high bypass ratio (BPR>5) and high overall pressure ratio (OPR>30) are the mainstream trends for this engine's power stage. Using weighted kernel density estimation (W-KDE), the optimal parameter probability distribution domain for the SFC was constructed, automatically excluding aircraft-type parameter spaces with low bypass ratios. Details are as follows: Constructing the required feature vector of current high bypass ratio engine performance structural index parameters (Including thrust, fuel consumption rate requirements, etc.) and historical case feature vectors , define the first Confidence weights of historical cases For cosine similarity, ,in For the first Feature vectors of historical cases.

[0029] Key design variables in thermodynamic parameters (e.g., the pressure ratio of a certain compression component) ), fusion Data points from historical cases Generate a continuous probability density function Rather than a rigid range:

[0030] in For Gaussian kernel function, For the first Design variables from historical cases The value of , For the first The confidence weights of historical cases can be represented by cosine similarity. Therefore, the recommendation distribution of the design variables is no longer uniform, but follows a different pattern. A mixed distribution.

[0031] in accordance with Determine the confidence interval (e.g.) Confidence level Using this as the primary sampling domain ensures both dense searching of high-confidence regions and preserves the possibility of exploring low-probability regions (corresponding to possible "outlier" innovative solutions), thus completely eliminating the problem of empirical conflicts.

[0032] S3. Using a preset sampling strategy, an initial thermodynamic scheme population of the aero-engine is generated within the spatial sampling boundary of the thermodynamic parameters. The thermodynamic cycle simulation program is called to calculate the performance index parameters and structural index parameters. With the performance index parameters meeting the corresponding preset requirements as the optimization objective, a multi-objective evolutionary algorithm is used to optimize and obtain the Pareto front solution set of the thermodynamic scheme. In this embodiment, non-uniform Latin hypercube sampling (LHS) is performed to initialize the population within the sampling principal domain formed by the obtained spatial sampling boundary. The NSGA-III algorithm is then used for multi-objective optimization to generate a set of thermodynamic cycle schemes at the Pareto front. For example, the system outputs scheme A, which is characterized by a balanced bypass ratio and a moderate turbine inlet temperature, aiming to balance fuel consumption and maintenance costs.

[0033] It should be noted that this embodiment selects a group (rather than a single one, such as thrust-first, fuel consumption-first, or lifespan-first) of representative non-dominated solution schemes from the Pareto front and outputs them in parallel to the structural design layer. By preserving the diversity of thermodynamic parameters, the probability of the structural layer finding a feasible solution under physical constraints is significantly increased.

[0034] S4. Select a set of initial thermodynamic schemes from the Pareto front solution set, and construct an engineering decision feature vector containing engine thermal load, cost sensitivity, and life requirements based on the requirements of the project task book and the thermodynamic parameters in the initial thermodynamic schemes. In this embodiment, by comprehensively utilizing the project task requirements of S1 and the thermal parameters in the initial thermal scheme, a system including heat load is constructed. Cost sensitivity Lifespan requirements Engineering decision feature vector .

[0035] S5. Based on the engineering decision feature vector, retrieve the engineering knowledge base to generate a structural design domain specification containing logical basis. Then, based on the design domain specification and thermodynamic parameters, calculate the flow channel geometry of each component cross-section and use an optimization algorithm to find a structural parameter design scheme that satisfies multi-objective constraints. The multi-objective constraints include maximum diameter constraints and rotor tip velocity constraints. The design domain specification includes the range of compressor stage values, turbine stage values, and compressor hub ratio values. Specifically, it includes: 5.1 Based on the constructed engineering decision feature vector And it integrates the structural context (Contextstruct) to form a query command. (If the compressor pressure ratio is known, look up the commonly used / recommended number of compressor stages; if the turbine expansion ratio is known, look up the commonly used / recommended number of turbine stages).

[0036] 5.2 Utilize retrieval enhancement generation techniques to retrieve relevant sets of engineering specifications from vector databases containing design experience and knowledge (such as the empirical relationship between compressor stages and compressor pressure ratio, and the empirical relationship between turbine stages and turbine expansion ratio). And lock the design constraints.

[0037] In this embodiment, it is required that when the large model outputs the design domain reduction range, it must be accompanied by clear logical basis, such as: Standard basis: Referencing specific industry standards (e.g., "Based on Article 5.3 of xxxx").

[0038] Qualitative physical derivation: Qualitative judgment based on physical principles. For example: "Because..." The heat load is extremely high. To ensure the lifespan of the blades, space must be reserved for air cooling channels, thus limiting the lower limit of the blade thickness.

[0039] Formal output:

[0040] in For the structural design domain after specification, This is a thought process described in natural language. This step does not involve specific numerical iterations; its aim is to leverage AI's reasoning capabilities to quickly eliminate invalid spaces that clearly violate common sense and norms, providing a high-quality initial domain for subsequent refined calculations, such as the number of compressor stages and turbine stages.

[0041] 5.3 Based on thermodynamic parameters and aerodynamic design parameters, the diameter of the critical section of the flow path is deterministically calculated using one-dimensional aerodynamic formulas; 5.4 Based on the determined key cross-section, construct the flow channel geometry and, under the premise of satisfying the speed margin and surge margin constraints, find the optimal structural parameters with the goal of minimizing the axial length or the weight.

[0042] S6. When step S5 fails to generate a structural parameter design scheme that satisfies multi-objective constraints, adjust the thermal parameter space sampling boundary and repeat steps S3 to S5 until a structural parameter design scheme that satisfies multi-objective constraints is obtained; specifically including: 6.1 Calculate the sensitivity gradient of the structural constraint index relative to the thermodynamic parameters, and identify the parameter with the largest sensitivity gradient magnitude as the key thermodynamic parameter. Based on the sign and magnitude of the sensitivity gradient (positive indicates a positive correlation between the structural constraint index and the thermodynamic parameter, and negative indicates a negative correlation; based on the desired direction / magnitude of improvement in the structural constraint index, combined with the sign / magnitude of the sensitivity gradient, the adjustment direction / magnitude of the key thermodynamic parameter is given), the adjustment direction and magnitude of the key thermodynamic parameter to meet the structural constraints are given. For example, in this embodiment, structural constraint indices are calculated near the infeasible solution of the current Pareto front. (e.g., diameter) Thermodynamic parameters relative to the upstream performance design layer (such as engine speed) N Flow factor Sensitivity gradient The parameter with the largest gradient magnitude is selected as the key thermodynamic parameter. For example, it was found that... The maximum value indicates that the excessively low rotation speed is the root cause of the diameter exceeding the limit.

[0043] 6.2 The key thermodynamic parameters are jointly analyzed with the engineering decision feature vector, and the inference engine is used to determine whether there is a hard constraint conflict between the adjustment of key thermodynamic parameters and the engineering decision features; In this embodiment, the increased rotational speed was detected. N It can solve the problem of diameter exceeding the limit, but at the same time it will significantly increase the centrifugal stress of rotating parts.

[0044] examine (Cost sensitivity) and (Lifespan requirement), if Extremely high stress levels (limiting long lifespan) impose a strict upper limit on stress levels, leading to a reduction in rotational speed. Unable to improve.

[0045] Conclusion: The main contradiction at this point is not simply the selection of parameters, but the conflict between the "long lifespan requirement" and structural constraints.

[0046] 6.3 If there is no hard constraint conflict, the direction and magnitude of adjustment of the key thermodynamic parameters are given. After the sampling boundary of the key thermodynamic parameters is corrected by step S2, the optimization is re-executed to find a design scheme that meets the structural constraints. If there is a hard constraint conflict, that is, adjusting the key thermodynamic parameters will violate the cost or life constraints in the engineering decision characteristics, the size or weight requirements in the structural constraints are relaxed. Based on the above analysis, a composite feedback message is generated that includes the adjustment direction of key thermodynamic parameters and design domain relaxation suggestions.

[0047] in, This indicates an increase in the key thermodynamic parameter, rotational speed. , Indicates "or", This indicates a relaxation of the maximum diameter constraint. In the next iteration, either the rotational speed must be significantly increased (if material stress allows), or the diameter limit must be relaxed (if lifespan is limited) to avoid ineffective infinite loop iterations. When the spatial sampling boundary of the thermodynamic parameters is generated, the receiver Correct the parameter boundaries in S2, and re-execute S3-S5 to start a new round of iterations until convergence.

[0048] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for aero-engine generative design based on fusion of implicit empirical reasoning, characterized in that, The method comprises the following steps: S1, extracting engine performance structure index parameters according to the requirements of the project task book of the aero-engine design, wherein the performance structure index parameters include thrust, specific fuel consumption, cost, life, and size limit; S2, preprocessing a thermodynamic parameter space in a historical database that meets the performance structure index parameters to obtain a thermodynamic parameter space sampling boundary based on a weighted kernel density; the thermodynamic parameters include engine inlet flow, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency; S3, generating an initial thermodynamic scheme population of the aero-engine in the thermodynamic parameter space sampling boundary by using a preset sampling strategy, calling a thermodynamic cycle simulation program to calculate performance index parameters, taking that the performance index parameters meet the corresponding preset requirements as an optimization target, and using a multi-objective evolutionary algorithm to obtain a thermodynamic scheme Pareto front solution set; S4, selecting a group of initial thermodynamic schemes from the Pareto front solution set, and constructing an engineering decision feature vector including engine thermal load, cost sensitivity, and life requirement according to the project task book requirements and the thermodynamic parameters in the initial thermodynamic schemes; S5, retrieving an engineering knowledge base based on the engineering decision feature vector, generating a structure design domain rule including a logical basis, and calculating the flow passage geometry of each component section according to the design domain rule and the thermodynamic parameters, and using an optimization algorithm to find a structure parameter design scheme that meets the multi-objective constraints; the multi-objective constraints include maximum diameter constraint and rotor tip speed constraint; the design domain rule includes compressor stage number value range, turbine stage number value range, and compressor hub ratio value range.

2. The aeroengine generative design method of claim 1, wherein, In step S2, the method for obtaining the thermodynamic parameter space sampling boundary based on the weighted kernel density comprises: calculating the cosine similarity between the current task demand feature vector and the historical case feature vector as a confidence weight; the feature vector includes specific fuel consumption and thrust; based on the confidence weight, kernel density estimation is performed on the key design variables to generate a probability density function; according to the probability density function, the preset confidence interval is determined as the sampling main domain.

3. The aeroengine generative design method of claim 1, wherein, In step S5, the method for retrieving the engineering knowledge base based on the engineering decision feature vector to generate the structure design domain rule including the logical basis comprises: constructing a query instruction containing the engineering decision feature vector, and using retrieval enhancement generation technology to retrieve matching engineering specifications and design criteria from a vector database containing design experience and engineering knowledge; a large language model generates an evidence chain in natural language form based on the retrieval result, and outputs the value range limit of the structure key parameters, including the compressor stage number value range, the turbine stage number value range, and the compressor hub ratio value range in the design domain rule.

4. The aeroengine generative design method of claim 1, wherein, In step S5, the method for calculating the flow passage geometry of each component section and using an optimization algorithm to find a structure parameter design scheme that meets the multi-objective constraints comprises: according to the thermodynamic parameters and aerodynamic design parameters, the key section diameters are determined by using one-dimensional aerodynamic formula, and the key sections include the engine inlet section, the compressor outlet section, the combustion chamber outlet section, and the turbine outlet section; Based on the determined key cross section, the flow channel geometry is constructed, and under the premise of meeting the maximum diameter constraint and rotor tip speed constraint, the optimal structural parameter is searched with the lowest cost or lightest weight as the target.

5. The aeroengine generative design method of claim 1, wherein, When step S5 fails to generate a structural parameter design scheme meeting the multi-objective constraint, the thermal parameter space sampling boundary is adjusted, and steps S3-S5 are repeated until a structural parameter design scheme meeting the multi-objective constraint is obtained.

6. The aeroengine generative design method of claim 5, wherein, The attribution analysis is performed to adjust the thermal parameter space sampling boundary, specifically including: The sensitivity gradient of the structural constraint index with respect to the thermal parameter is calculated, the parameter with the largest sensitivity gradient modulus is locked as the key thermal parameter, and the adjustment direction and amplitude of the key thermal parameter for meeting the structural constraint are given according to the positive and negative and size of the sensitivity gradient; The key thermal parameter and the engineering decision feature vector are jointly analyzed, and it is judged by the reasoning engine whether there is a hard constraint conflict between the key thermal parameter adjustment and the engineering decision feature; If there is no hard constraint conflict, the adjustment direction and amplitude of the key thermal parameter are given, the sampling boundary of the key thermal parameter is corrected, and the optimization is performed again to find a design scheme meeting the structural constraint; if there is a hard constraint conflict, i.e., adjusting the key thermal parameter will violate the cost or life constraint in the engineering decision feature, the size or weight requirement in the structural constraint is relaxed.

7. A gas turbine engine generative design system based on fusion of implicit empirical reasoning, characterized by, It includes: A task analysis module is configured to extract engine performance and structure index parameters according to the project task book requirements of the aero-engine design, wherein the performance and structure index parameters include thrust, specific fuel consumption, cost, life, and size limit. A sampling boundary analysis module is configured to preprocess a thermal parameter space meeting the performance and structure index parameters in a historical database to obtain a thermal parameter space sampling boundary based on a weighted kernel density; the thermal parameters include engine inlet flow, bypass ratio, compression component pressure ratio, efficiency, combustion chamber outlet temperature, and turbine efficiency. A first optimization analysis module is configured to generate an initial thermal scheme population of the aero-engine within the thermal parameter space sampling boundary by using a preset sampling strategy, to calculate performance index parameters by calling a thermal cycle simulation program, to take performance index parameters meeting corresponding preset requirements as optimization targets, and to obtain a thermal scheme Pareto frontier solution set by using a multi-objective evolutionary algorithm. A feature vector construction module is configured to select a group of initial thermal schemes from the Pareto frontier solution set, to construct an engineering decision feature vector including engine thermal load, cost sensitivity, and life requirement according to the project task book requirements and the thermal parameters in the initial thermal schemes. A second optimization analysis module is configured to retrieve an engineering knowledge base based on the engineering decision feature vector, to generate a structural design domain specification including logical basis, and to calculate flow channel geometries of component cross sections according to the design domain specification and the thermal parameters, and to find a structural parameter design scheme meeting multi-objective constraints by using an optimization algorithm; the multi-objective constraints include maximum diameter constraint and rotor tip speed constraint; the design domain specification includes compressor stage number value range, turbine stage number value range, and compressor hub ratio value range.

8. The aeroengine generative design system of claim 7, wherein, The parameter adjustment module is further configured to adjust the sampling boundary of the thermal parameter space until a structural parameter design scheme satisfying the multi-objective constraints is obtained when the structural parameter design scheme satisfying the multi-objective constraints cannot be generated.

9. The aeroengine generative design system of claim 8, wherein, The parameter adjustment module comprises: a sensitivity gradient analysis unit configured to calculate a sensitivity gradient of a structural constraint index with respect to the thermal parameters, lock a parameter with a maximum sensitivity gradient as a key thermal parameter, and give an adjustment direction and amplitude of the key thermal parameter for satisfying the structural constraint according to a sign and size of the sensitivity gradient; a constraint conflict analysis unit configured to jointly analyze the key thermal parameter and an engineering decision feature vector, and determine whether there is a hard constraint conflict between the adjustment of the key thermal parameter and the engineering decision feature by using an inference engine; an adjustment unit configured to give an adjustment direction and amplitude of the key thermal parameter when there is no hard constraint conflict, re-execute optimization after correcting the sampling boundary of the key thermal parameter, and find a design scheme satisfying the structural constraint; and relax a size or weight requirement in the structural constraint when there is a hard constraint conflict, i.e., the adjustment of the key thermal parameter will violate a cost or life constraint in the engineering decision feature.

Images