Aero-engine assembly parameter optimization method, equipment, medium and product
By evaluating the importance of aero-engine assembly parameters using a random forest regression model and conducting feature-guided search, the problem of assembly parameter optimization under high-dimensional and small-sample conditions is solved, achieving efficient and accurate assembly parameter optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-06
- Publication Date
- 2026-04-14
AI Technical Summary
Existing technologies in the aero-engine assembly process suffer from high computational complexity in high-dimensional parameter spaces, insufficient modeling accuracy under small sample conditions, and low search efficiency, making it impossible to effectively optimize assembly parameters.
A random forest regression model is used to evaluate the importance of assembly parameters, and feature-guided search is performed. By combining discrete grids and evaluation functions, efficient assembly parameter optimization is achieved.
Under the condition of small sample and high dimension, it improves modeling accuracy and search efficiency, ensures the acquisition of the global optimal solution, and reduces computational complexity.
Smart Images
Figure CN121859285A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aero-engine assembly parameter technology, and more specifically, to a method, equipment, medium, and product for optimizing aero-engine assembly parameters. Background Technology
[0002] In the manufacturing process of aero-engines, the setting of process parameters or assembly parameters (hereinafter collectively referred to as "assembly parameters") directly determines the final performance of the product. Currently, the methods for determining the optimal parameters mainly rely on trial and error based on engineers' experience or numerical optimization algorithms. The trial and error method based on experience is inefficient and the results are unreliable; while traditional optimization algorithms (such as genetic algorithms and particle swarm optimization algorithms) are prone to getting trapped in local optima when solving such inverse problems, and it is difficult to guarantee that the solution satisfies all performance constraints.
[0003] In the prior art, there exists a method for inversely solving assembly parameters based on global search (or exhaustive search) (which can be considered the closest existing technology). This method establishes a forward model between assembly parameters and performance indicators, and performs a systematic traversal in the discretized parameter space to find all parameter combinations that meet the requirements, thereby ensuring the global optimality of the solution.
[0004] However, this existing technology has the following inherent limitations, which prevent it from being applied to the complex industrial scenarios of aero-engine assembly. These limitations are as follows:
[0005] Existing technologies typically involve a small number of process parameters (e.g., 3-5), with the number of parameter combinations within an acceptable range, classifying it as a low-dimensional computable problem. However, in the assembly process of aero-engines, the number of key parameters often exceeds 10, even reaching more than 20, with the number of parameter combinations increasing exponentially. Therefore, directly applying a global search is computationally infeasible.
[0006] Existing technologies typically employ regression models based on physical mechanisms (such as multinomial response surfaces). As the dimensionality of parameters increases, the amount of experimental data required for such models increases dramatically, making them unsuitable for the current situation of small sample data resulting from the small-batch, high-cost production of high-end equipment.
[0007] Existing global search methods lack effective acceleration mechanisms and generate a large number of invalid calculations that clearly do not meet the constraints, even in medium-dimensional spaces. The search efficiency is extremely low and they are not practical for engineering applications. Summary of the Invention
[0008] The purpose of this invention is to provide a method, device, medium, and product for optimizing aero-engine assembly parameters, which solves the problems of high computational complexity, insufficient modeling accuracy under small sample conditions, and low search efficiency of existing global search technologies when dealing with high-dimensional parameter spaces.
[0009] The above-mentioned technical objective of the present invention is achieved through the following technical solution:
[0010] In a first aspect, the present invention provides a method for optimizing assembly parameters of an aero-engine, the method comprising:
[0011] Use a pre-trained random forest regression model to obtain the importance evaluation vector for each assembly parameter;
[0012] The assembly parameters of the aero-engine are sorted according to the importance assessment vector to obtain the key parameter index set;
[0013] The key parameter index set is discretized to obtain a discrete mesh; wherein the discrete mesh describes the combination result of the assembly parameters.
[0014] The random forest regression model is called to calculate the predicted performance of each assembly parameter combination obtained by traversing the discrete grid, and the predicted performance of each assembly parameter combination is judged according to the preset constraints on the performance of the assembly parameter combination, so as to obtain the discrete grid that satisfies the constraints.
[0015] A feasible solution set is obtained by searching the discrete grid that satisfies the constraints.
[0016] Each assembly parameter combination in the feasible solution set is input into a preset evaluation function to obtain the evaluation value of each assembly parameter combination. The assembly parameter combination corresponding to the minimum evaluation value is selected as the assembly parameter optimization result.
[0017] In one implementation, the training process of the random forest regression model includes:
[0018] Create the first vector of aero-engine assembly parameters;
[0019] Create a second vector of aero-engine assembly parameter performance;
[0020] Construct a small sample dataset based on the first and second vectors;
[0021] The random forest regression algorithm was used to train a small sample dataset to obtain a trained random forest regression model.
[0022] In one implementation, the assembly parameters of the aero-engine are sorted according to the importance evaluation vector to obtain a key parameter index set, including:
[0023] The assembly parameters of the aero-engine are sorted in descending order using the importance assessment vector to obtain the descending order ranking result;
[0024] The top k assembly parameters in descending order are selected to form a key parameter index set; where k is a positive integer.
[0025] In one implementation, before discretizing the key parameter index set to obtain a discrete grid, the method further includes: performing dimensionality reduction processing on the key parameter index set.
[0026] In one implementation scheme, the predicted performance of each assembly parameter combination is determined according to the preset constraints on the performance of the assembly parameter combination, and a discrete mesh that satisfies the constraints is obtained. This includes: if the predicted performance of the current assembly parameter combination is not within the constraints, then jumping to the next assembly parameter combination for judgment, until all assembly parameter combinations have completed the performance judgment, and a discrete mesh that satisfies the constraints is obtained.
[0027] In one implementation, a search is performed on the discrete grid that satisfies the constraints to obtain a feasible solution set, including:
[0028] The first discrete step size is preset, and a search is performed on the discrete grid that meets the constraints to obtain the search solution set;
[0029] A second step is preset to perform a second search on the search solution set to obtain a feasible solution set; wherein, the second step is smaller than the first step.
[0030] In one implementation, the expression for the evaluation function is: ;in, The weight of the j-th assembly parameter, Let j be the predicted value of the j-th performance index. Let j be the target value of the j-th performance metric. Let j be the allowable deviation for the j-th performance index. Let $x$ be the total number of performance metrics, and $Cost(x)$ be a penalty function related to assembly efficiency.
[0031] A second aspect of the present invention provides an electronic device, including a memory and a processor;
[0032] A memory for storing computer programs, the computer programs including program instructions;
[0033] A processor is configured to execute the program instructions to cause the electronic device to perform the steps of an aircraft engine assembly parameter optimization method as provided in the first aspect of the present invention.
[0034] A third aspect of the present invention provides a computer program product comprising program instructions that, when executed by an electronic device, cause the electronic device to perform the steps of an aero-engine assembly parameter optimization method as provided in the first aspect of the present invention.
[0035] A fourth aspect of the present invention provides a computer-readable storage medium comprising a computer program that, when executed by one or more processors, implements a method for optimizing assembly parameters of an aero-engine as provided in the first aspect of the present invention.
[0036] Compared with the prior art, the present invention has the following beneficial effects:
[0037] In the technical solution provided by this invention, firstly, a random forest regression model is used to evaluate the importance of assembly parameters, thereby enabling the establishment of a high-precision and robust positive mapping model under small sample conditions, providing key feature importance information. Secondly, a data-driven dimensionality reduction method based on the intrinsic feature importance of the random forest regression model is used, employing feature importance as a quantification criterion to accurately identify and focus on key dimensions, realizing the transformation from global traversal to feature-oriented search. Then, the predictive function of the random forest regression model is deeply embedded into the search loop, achieving real-time, proactive pruning, improving search efficiency and reducing computational complexity. Finally, an evaluation function is used to comprehensively evaluate the feasible solution set after the search, obtaining the globally optimal solution as the optimization result, and predicting the performance of the optimized assembly parameter combination. Therefore, this invention solves the problems of high computational complexity, insufficient modeling accuracy under small sample conditions, and low search efficiency of existing global search techniques when dealing with high-dimensional parameter spaces. Attached Figure Description
[0038] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and form part of this application, do not constitute a limitation thereof. In the drawings:
[0039] Figure 1 This is a flowchart illustrating a method for optimizing assembly parameters of an aero-engine, as provided in an embodiment of the present invention. Detailed Implementation
[0040] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.
[0041] It should be noted that the terms "comprising" or "may include" used in the various embodiments of this application indicate the presence of the claimed function, operation, or element, and do not limit the addition of one or more functions, operations, or elements. Furthermore, as used in the various embodiments of this application, the terms "comprising," "having," and their cognates are intended only to indicate a specific feature, number, step, operation, element, component, or combination of the foregoing, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing, or adding one or more combinations of the foregoing.
[0042] It should be understood that terms such as "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.
[0043] Figure 1 This is a flowchart illustrating a method for optimizing assembly parameters of an aero-engine according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:
[0044] S101, call the pre-trained random forest regression model to obtain the importance evaluation vector for each assembly parameter.
[0045] Specifically, the assembly parameters of an aero-engine are the core technical indicators that ensure its aerodynamic performance, structural strength, and operational stability. They need to cover key modules such as rotor system, stator system, fluid circuit, sealing protection, and transmission control, including basic geometric and form and position tolerance parameters, clearance and compensation parameters, etc.
[0046] Basic geometric and form-position tolerance parameters are used as assembly datums to determine the alignment and fit accuracy of components. These generally include rotor alignment parameters, stator component form-position tolerances, and fit tolerances. Rotor alignment parameters include the coaxiality of each shaft section, rotor runout, and the coaxiality and parallelism of the bearing housing bores. Stator component form-position tolerances include the roundness of the casing, the flatness of the combustion chamber mounting datum surface, and the concentricity of the turbine guide vane ring. Fit tolerances include interference fit parameters (e.g., 0.01~0.03mm interference between the shaft and hub to ensure torque transmission) and clearance fit parameters (e.g., 0.02~0.08mm clearance between the bushing and shaft sleeve to allow for thermal expansion).
[0047] Clearance and compensation parameters, used to accommodate thermal expansion, avoid friction, and ensure aerodynamic efficiency, generally include clearance and thermal expansion compensation parameters. Clearance includes tip clearance (cold state: compressor 0.3~0.8mm, turbine 1.0~2.5mm, thermal expansion compensation must be considered); axial clearance (rotor-stator end face clearance 0.5~1.5mm, bearing axial clearance 0.01~0.05mm); and gear meshing clearance (0.1~0.3mm, including tooth flank clearance and tip clearance).
[0048] To address the challenges of small sample sizes, high dimensionality, and black-box processing, a random forest regression algorithm was creatively adopted as the core modeling tool, replacing existing regression models suitable for low-dimensional, well-defined scenarios. Random forest is an ensemble learning algorithm composed of multiple decision trees. Its core advantages are: 1) Strong resistance to overfitting, effectively improving the model's generalization ability on small sample data through bootstrap sampling and random feature selection; 2) Built-in feature importance evaluation mechanism, automatically quantifying the impact of each input parameter on output performance, providing a direct basis for subsequent intelligent dimensionality reduction; 3) Natural adaptability to nonlinear relationships and interaction effects in high-dimensional data, without requiring pre-assumptions about the specific functional form between assembly data and test performance, making it particularly suitable for black-box process modeling. Therefore, choosing random forest enables the establishment of a high-precision, robust forward mapping model under small sample conditions and provides crucial feature importance information for subsequent steps. The specific steps are as follows:
[0049] Define the input / output space: Define the first vector of aero-engine assembly parameters. Where m≥10; define the second vector of aero-engine assembly parameter performance. ,in, express, express.
[0050] Using small sample historical datasets Training a random forest regression model .
[0051] Cross-validation and the coefficient of determination (R²) are used to ensure model accuracy. After training, the intrinsic feature importance score vectors of the model are obtained simultaneously. This forms the core basis and data foundation for the intelligent dimensionality reduction achieved in this invention.
[0052] S102, Sort the assembly parameters of the aero-engine according to the importance assessment vector to obtain the key parameter index set.
[0053] Specifically, existing technologies rely on empirical screening or full parameter traversal. This invention proposes a data-driven dimensionality reduction method based on the importance of model-endogenous features. This mechanism uses feature importance as a quantification criterion to accurately identify and focus on key dimensions, realizing a shift from full-domain traversal to feature-oriented search. First, the assembly parameters of the aero-engine are sorted in descending order using importance evaluation vectors to obtain the descending order ranking result. Then, the top k assembly parameters in the descending order ranking result are selected to form a key parameter index set; where k is a positive integer.
[0054] Specifically, the first k parameters (k≤m, usually k≤5) are selected to form the key parameter index set. The key parameter index set is then dimensionality-reduced to obtain the dimensionality-reduced key parameter index set: .
[0055] This invention fulfills the prerequisites and key assumptions for applying global search to high-dimensional problems, reducing the computational complexity from... Down to This improves computational efficiency.
[0056] S103, Discretize the key parameter index set to obtain a discrete mesh; wherein, the discrete mesh describes the combination result of the assembly parameters.
[0057] Specifically, the key parameter subspace Discretization, constructing a discrete mesh .
[0058] S104, call the random forest regression model to calculate the predicted performance of each assembly parameter combination obtained by traversing the discrete grid, and judge the predicted performance of each assembly parameter combination according to the preset constraints on the performance of the assembly parameter combination to obtain the discrete grid that satisfies the constraints.
[0059] Existing technologies calculate all combinations first and then make a unified decision. In contrast, this invention employs a strategy of predicting, deciding, and pruning simultaneously, deeply embedding the predictive function of the random forest regression model into the search loop. This achieves real-time, proactive pruning, avoiding redundant calculations on invalid solutions (those that do not meet constraints) and skipping most useless search paths, thus significantly improving search efficiency. Simultaneously, it inherently guarantees that all solutions found ultimately satisfy performance constraints.
[0060] Define the constraint condition for the j-th performance as follows: .
[0061] Traversing the discrete grid At that time, for the current assembly parameter combination The predictive performance of the random forest regression model F is calculated. .
[0062] The determination process is as follows: If , making or If the current search path is not found, the search will immediately terminate and jump to the next assembly parameter combination until all assembly parameter combinations have completed performance evaluation and a discrete mesh that meets the constraints is obtained.
[0063] S105 searches the discrete grid that satisfies the constraints to obtain a feasible solution set.
[0064] Specifically, the process involves searching the discrete grid that satisfies the constraints to obtain a feasible solution set, including: performing a first search on the discrete grid that satisfies the constraints with a preset first step size to obtain a search solution set; and performing a second search on the search solution set with a preset second step size to obtain a feasible solution set; wherein the second step size is smaller than the first step size.
[0065] During the search, the present invention performs two searches, namely a first search and a second search, wherein the first search is a coarse search and the second search is a fine search.
[0066] The coarse search phase uses a relatively large discrete step size to quickly scan the global parameter space. Its main purpose is not to directly obtain a high-precision solution, but rather to quickly identify and collect feasible solutions that satisfy performance constraints, forming a preliminary feasible solution set (Scoarse). The distribution of this solution set in the parameter space essentially identifies a series of "regions" with satisfactory performance and potential for further optimization. The fine search phase does not start afresh across the entire space; instead, it uses the region covered or densely distributed by the Scoarse (i.e., the Rcoarse) as the initial subspace for the fine search. Therefore, the feasible solution set of the coarse search directly determines the starting range of the fine search.
[0067] The parameter combinations that yield better performance predictions in the feasible solution set allow for more precise search steps in practical implementation. Further priority is given to searching regions with high concentrations of optimal solutions. In the coarse search phase, the pruning mechanism quickly eliminates most parameter combinations that do not meet the constraints, ensuring that the generated coarse search feasible solution set is already a valid result after rigorous constraint filtering. When the fine search refines this process, since the search space has been significantly reduced and the starting point is excellent, constraint pruning can continue to work efficiently within a smaller scope.
[0068] First layer (coarse search): within a discrete grid that satisfies the constraints. Using larger engineering distances from walking length Perform a global scan to quickly locate potential areas with superior performance. This yields a coarse search feasible solution set. .
[0069] Second layer (refined search): with Construct a smaller subspace centered on the current location and use a finer step size. A second round of search was conducted to obtain a feasible solution set. .
[0070] The constraint pruning described above plays a role in both coarse and fine search, enabling the rapid elimination of a large number of invalid regions even in the coarse search stage, which greatly optimizes the overall efficiency of hierarchical search.
[0071] S106. Input each assembly parameter combination in the feasible solution set into the preset evaluation function to obtain the evaluation value of each assembly parameter combination, and select the assembly parameter combination corresponding to the minimum evaluation value as the assembly parameter optimization result.
[0072] Specifically, the expression for the evaluation function is: ;in, The weight of the j-th assembly parameter, Let j be the predicted value of the j-th performance index. Let j be the target value of the j-th performance metric. Let j be the allowable deviation for the j-th performance index. Let $x$ be the total number of performance metrics, and $Cost(x)$ be a penalty function related to assembly efficiency.
[0073] From the final feasible solution set In the process, select the assembly parameter combination with the optimal evaluation value: .
[0074] Output the global optimal solution and its predictive performance .
[0075] Based on the optimization method described above, and taking the small-batch assembly process of a high-pressure compressor rotor for a certain type of aero-engine as an example, this invention will be described in detail using data from the file "Assembly Data and Test Characteristics.xlsx". This implementation aims to highlight the unique advantages of this invention in solving high-dimensional, small-sample problems.
[0076] This implementation case focuses on the assembly process of a certain type of aero-engine. This scenario is characterized by high parameter dimensionality (m=20), small sample size (N=30), and many performance indicators (n=24).
[0077] Data preparation and preprocessing:
[0078] Data source: Data from the first 30 assembly and test runs in the file.
[0079] Inputs (assembly parameters): Inputs 1 to 20, a total of m=20. These are then standardized.
[0080] Outputs (performance metrics): Outputs 1 to 6 under states 1 to 4, totaling n=24.
[0081] Constructing a random forest regression model, under conditions of 20 parameters and only 25 training samples, reveals that existing regression models cannot establish reliable prediction models. This invention addresses this issue by training the random forest model using the first 25 data sets and validating it with the last 5 sets. The overall average coefficient of determination R0 of the model on the test set is... 2 =0.88, demonstrating its excellent performance in small-sample, high-dimensional scenarios. Furthermore, for the first time, a feature importance ranking method was obtained to guide dimensionality reduction.
[0082] Based on feature importance, the top k=4 key parameters are selected: input 4, input 7, input 12, and input 19. The search space is reduced from 20 dimensions to 4 dimensions. This step is not available in existing technologies and is essential in this scenario.
[0083] Discretize the four key parameters within their historical range (±10%) with a step size of 1%, generating a dataset containing 21... 4 =194481 discrete combinations. Set constraints for 24 performance indicators.
[0084] Coarse search: Traverses the discrete space and applies constraint pruning. Statistically, the actual number of evaluated combinations N is... actual With a pruning rate of 92.3%, the number of branches reached 15,000. This efficient mechanism is another key to solving this high-dimensional problem.
[0085] Fine search: In the optimal region, the step size is reduced to 0.2% for a fine search, and finally a feasible solution set is obtained. There are a total of 7 solutions that satisfy all constraints.
[0086] Efficiency Comparison: Total computation time is approximately 3 hours. Existing methods cannot achieve efficient search in this case due to the explosive computational load (more than 190,000 full calculations) and lack of pruning.
[0087] Construct an evaluation function to determine the global optimum from seven feasible solutions. .use Physical assembly and testing were conducted, and the first-pass yield rate reached 100%, with all performance indicators falling within the target range. This demonstrates the effectiveness and practicality of the invention in real-world industrial scenarios with high-dimensional, small-sample conditions.
[0088] This invention also provides an electronic device. The electronic device includes a processor, a memory, a communication interface, and at least one communication bus for connecting the processor, the memory, and the communication interface. The memory includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (PROM), or portable read-only memory (CD-ROM), and is used for related instructions and data.
[0089] The communication interface is used to receive and send data. The processor can be one or more CPUs; if the processor is a single CPU, it can be a single-core CPU or a multi-core CPU. The processor in the electronic device reads one or more programs stored in the memory and performs the following operations: It calls a pre-trained random forest regression model to obtain the importance evaluation vector for each assembly parameter; it sorts the assembly parameters of the aero-engine according to the importance evaluation vector to obtain a key parameter index set; it discretizes the key parameter index set to obtain a discrete grid; wherein the discrete grid describes the combination result of the assembly parameters; it calls the random forest regression model to calculate the predicted performance of each assembly parameter combination obtained by traversing the discrete grid, and judges the predicted performance of each assembly parameter combination according to preset constraints on the performance of the assembly parameter combinations to obtain a discrete grid that satisfies the constraints; it searches the discrete grid that satisfies the constraints to obtain a feasible solution set; it inputs each assembly parameter combination in the feasible solution set into a preset evaluation function to obtain the evaluation value of each assembly parameter combination, and selects the assembly parameter combination corresponding to the minimum evaluation value as the assembly parameter optimization result.
[0090] It should be noted that the specific implementation of each operation can be described above. Figure 1 The corresponding description of the method embodiments shown indicates that the electronic device can be used to execute an aero-engine assembly parameter optimization method according to the above method embodiments of this application, which will not be described in detail here.
[0091] This invention also provides a computer-readable storage medium, which is a memory device in a computer device for storing programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of a terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the above-described method for optimizing aero-engine assembly parameters. Those skilled in the art should understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0092] This invention also provides a computer program product containing program instructions. The computer program product can be software or program products containing program instructions, capable of running on a computing device or stored on any usable medium. When the computer program product runs on at least one electronic device, it causes the at least one electronic device to perform an aero-engine assembly parameter optimization method.
[0093] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for optimizing assembly parameters of an aero-engine, characterized in that the method... include: Use a pre-trained random forest regression model to obtain the importance evaluation vector for each assembly parameter; The assembly parameters of the aero-engine are sorted according to the importance assessment vector to obtain the key parameter index set; The key parameter index set is discretized to obtain a discrete mesh; wherein the discrete mesh describes the combination result of the assembly parameters. The random forest regression model is called to calculate the predicted performance of each assembly parameter combination obtained by traversing the discrete grid, and the predicted performance of each assembly parameter combination is judged according to the preset constraints on the performance of the assembly parameter combination, so as to obtain the discrete grid that satisfies the constraints. A feasible solution set is obtained by searching the discrete grid that satisfies the constraints. Each assembly parameter combination in the feasible solution set is input into a preset evaluation function to obtain the evaluation value of each assembly parameter combination. The assembly parameter combination corresponding to the minimum evaluation value is selected as the assembly parameter optimization result.
2. The method for optimizing assembly parameters of an aero-engine according to claim 1, characterized in that, The training process of the random forest regression model includes: Create the first vector of aero-engine assembly parameters; Create a second vector of aero-engine assembly parameter performance; Construct a small sample dataset based on the first and second vectors; The random forest regression algorithm was used to train a small sample dataset to obtain a trained random forest regression model.
3. The method for optimizing assembly parameters of an aero-engine according to claim 1, characterized in that, The assembly parameters of the aero-engine are sorted according to the importance assessment vector to obtain a key parameter index set, including: The assembly parameters of the aero-engine are sorted in descending order using the importance assessment vector to obtain the descending order ranking result; The top k assembly parameters in descending order are selected to form a key parameter index set; where k is a positive integer.
4. The method for optimizing assembly parameters of an aero-engine according to claim 3, characterized in that, Before discretizing the key parameter index set to obtain a discrete grid, the method further includes: performing dimensionality reduction processing on the key parameter index set.
5. The method for optimizing assembly parameters of an aero-engine according to claim 1, characterized in that, The predicted performance of each assembly parameter combination is determined according to the preset constraints on the performance of the assembly parameter combination, and a discrete mesh that satisfies the constraints is obtained. This includes: if the predicted performance of the current assembly parameter combination is not within the constraints, then jumping to the next assembly parameter combination for judgment, until all assembly parameter combinations have completed the performance judgment and a discrete mesh that satisfies the constraints is obtained.
6. The method for optimizing assembly parameters of an aero-engine according to claim 1, characterized in that, A search is performed on the discrete grid that satisfies the constraints to obtain a feasible solution set, including: The first discrete step size is preset, and a search is performed on the discrete grid that meets the constraints to obtain the search solution set; A second step is preset to perform a second search on the search solution set to obtain a feasible solution set; wherein, the second step is smaller than the first step.
7. The method for optimizing assembly parameters of an aero-engine according to claim 1, characterized in that, The expression for the evaluation function is: ;in, The weight of the j-th assembly parameter, Let j be the predicted value of the j-th performance index. Let j be the target value of the j-th performance metric. Let j be the allowable deviation of the j-th performance index. Let $x$ be the total number of performance metrics, and $Cost(x)$ be a penalty function related to assembly efficiency.
8. An electronic device, characterized in that, Including memory and processor; A memory for storing computer programs, the computer programs including program instructions; A processor is configured to execute the program instructions to cause the electronic device to perform the steps of the aircraft engine assembly parameter optimization method as described in any one of claims 1 to 7.
9. A computer program product containing program instructions, characterized in that, When the program instructions are executed by the electronic device, the electronic device performs the steps of the method for optimizing aero-engine assembly parameters as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a computer program that, when executed by one or more processors, implements a method for optimizing aero-engine assembly parameters as described in any one of claims 1 to 7.