Method and system for optimizing all working conditions of axial flow compressor

Abstract

The invention belongs to the technical field of pneumatic shape data mining and design optimization, and relates to an axial flow compressor full-working-condition optimization method and system. According to the method, the design variables having significant influence on the performance indexes and the influence directions and the influence degrees of the significant variables on the performance indexes are analyzed through the constructed mapping relation between the design variables and the performance indexes, so that the optimization problem has interpretability, and meanwhile, by analyzing the relation of the change trends among the multiple performance indexes, the optimization efficiency is improved. A performance index having a conflict relationship with a main optimization target is used as a constraint condition of the main optimization target, and a multi-working-condition multi-target optimization problem is converted into a form of one optimization target and a plurality of optimization constraints, so that the optimization problem is simplified; the restrictive relation between the stall margin under the partial rotating speed and the peak efficiency under the designed rotating speed in traditional gas compressor optimization is effectively solved, and comprehensive improvement of the performance of the gas compressor is achieved.

Description

Technical Field

[0001] This invention belongs to the field of aerodynamic shape data mining and design optimization technology, and relates to a method and system for optimizing axial flow compressors under all operating conditions. Background Technology

[0002] Modern multistage axial compressors require higher stability over a wide speed range and higher efficiency at the design speed in their full-range performance profile. Therefore, compressor optimization often needs to be performed on the full-range performance curves, requiring consideration of the stall margin range formed by multiple near-stall points at different speeds, as well as the peak efficiency across all speeds. This constitutes a complex multi-condition, multi-objective problem. However, current publicly available research on compressor optimization often fails to simultaneously address the different requirements for stall points, peak efficiency points, and clogging points at different speeds. Furthermore, since stall margin at certain speeds and peak efficiency at the design speed are often negatively correlated, overcoming this constraint to achieve a comprehensive improvement in compressor performance remains a challenging problem for designers.

[0003] Inspired by the human ability to extract useful information from experience when solving new problems, the technology of "data mining" has been introduced into the optimization design of compressors. On the one hand, data mining can make full use of the large amount of data generated during the optimization design process, explain the correlation between target performance and design variables, and summarize the common characteristics of the optimization solution set. On the other hand, data mining technology can be used to extract design criteria in the design space or Pareto front to reduce the high dependence of the compressor design process on the designer's experience and reduce the uncertainty of the design work. However, applying data mining technology to compressor optimization design still faces the following challenges: (1) How to fully apply data mining technology before the start of optimization work, simplify the complex compressor optimization problem into a single-objective multi-constraint optimization problem by exploring the design space and analyzing the correlation between performance indicators, and formulate an effective optimization strategy; (2) After applying data mining technology to the optimization problem, how to handle the relationship between the optimization objective and the optimization constraints, while avoiding the non-convergence problem of near stall point calculation that often exists in the compressor optimization process. Summary of the Invention

[0004] The purpose of this invention is to solve the problem that the existing compressor optimization process has complex multi-objective and multi-condition optimization problems, which leads to a mutual constraint between the stall margin at some speeds and the peak efficiency at the design speed, affecting the optimization efficiency and overall performance of the compressor. The invention provides a method and system for optimizing axial flow compressors under all operating conditions.

[0005] To achieve the above objectives, the present invention employs the following technical solution: A method for optimizing an axial compressor under all operating conditions includes the following steps: Select the design variables that affect the compressor performance and the corresponding performance indicators; Construct a mapping relationship between design variables and performance indicators, calculate the feature importance of each design variable based on the mapping relationship, identify design variables that have a significant impact on performance indicators based on feature importance, use the identified design variables as significant variables, and obtain the direction and degree of influence of significant variables on each performance indicator based on significant variables. Based on the performance indicators corresponding to the design variables, the relationship between the changing trends of multiple performance indicators is analyzed, the results of the performance indicator conflict relationship analysis are obtained, the main optimization objective is selected, and based on the performance indicator conflict relationship analysis results, the performance indicators that conflict with the main optimization objective are used as the constraints of the main optimization objective. Based on the main optimization objective and the constraints, the optimization problem of the axial compressor is constructed, the optimization problem is solved, and the optimized design variables are obtained. The optimization results were obtained based on the optimized design variables and the influence direction and degree of significant variables on each performance index.

[0006] A further improvement of the present invention is that: The selection of design variables affecting compressor performance and the corresponding performance indicators includes: A CFD performance evaluation model was constructed, and key operating point calculations were performed on the selected design variables based on the CFD performance evaluation model. Based on the calculation results, the stall margin at different speeds and the peak efficiency at the design speed of the axial compressor were selected as performance indicators.

[0007] The design variables include three-dimensional modeling parameters and blade suction surface control points that affect compressor performance.

[0008] The process involves constructing a mapping relationship between design variables and performance indicators, calculating the feature importance of each design variable based on this mapping relationship, identifying design variables that have a significant impact on performance indicators based on feature importance, designating these identified design variables as significant variables, and obtaining the direction and degree of influence of these significant variables on each performance indicator based on these significant variables. The SHAP method is introduced to calculate the average of the absolute values ​​of the SHAP values ​​of each design variable, and the average of the absolute values ​​of the SHAP values ​​is used as a quantitative indicator of feature importance. A ring heatmap is constructed based on the quantitative index of feature importance. The contribution of design variables to performance indicators is identified based on the ring heatmap. Significant variables are selected based on the contribution identification results. Based on significant variables, an improved SHAP cell diagram is generated, and the direction and degree of influence of significant variables on each performance index are obtained from the improved SHAP cell diagram.

[0009] The calculation of the average of the absolute values ​​of the SHAP values ​​of each design variable includes:

[0010]

[0011] In the formula, i This is the sample count value. j For the count values ​​of the feature variables, N The total number of samples, M It is the total number of characteristic variables; S It is a set {1,2..j-1,j+1,..M} a subset of, | S |Yes S The total number of elements in; Indicates when only The features in the model are the predicted values. Indicates only S The feature in the model is the predicted value, and the difference between the two is the first value. j The feature variables in the subset S The marginal contribution below.

[0012] The optimization problem of constructing the axial flow compressor includes: A self-organizing map method is introduced to analyze multiple performance indicators and generate SOM maps; Based on the generated SOM (System Oscillator) plot, determine whether the changing trends of the various performance indicators are consistent or conflicting. Determine the main optimization objective, and use other performance indicators that conflict with the main optimization objective as constraints on the main optimization objective to construct a single-objective combined with multiple constraints optimization problem.

[0013] When solving the optimization problem, a penalty mechanism is constructed, including: If the performance index of the generated candidate solution exceeds the constraint range, a penalty factor is added to the optimization objective value; If the performance index calculation process of the candidate solution fails to meet the preset convergence criterion, a penalty factor is added to the optimization target value.

[0014] A full-condition optimization system for an axial compressor includes: The module for obtaining design variables and performance indicators is used to select the design variables that affect the compressor performance and the corresponding performance indicators. The module for analyzing the relationship between design variables and performance indicators is used to construct the mapping relationship between design variables and performance indicators. Based on the mapping relationship, the feature importance of each design variable is calculated. Based on the feature importance, design variables that have a significant impact on performance indicators are identified. The identified design variables are used as significant variables. Based on the significant variables, the direction and degree of influence of the significant variables on each performance indicator are obtained. The design variable optimization module is used to analyze the relationship between the changing trends of multiple performance indicators based on the performance indicators corresponding to the design variables, obtain the performance indicator conflict relationship analysis results, select the main optimization objective, and, based on the performance indicator conflict relationship analysis results, take the performance indicators that conflict with the main optimization objective as the constraints of the main optimization objective, construct the optimization problem of the axial compressor based on the main optimization objective and the constraints, solve the optimization problem, and obtain the optimized design variables. The optimization result acquisition module is used to obtain optimization results based on the direction and degree of influence of the optimized design variables and significant variables on each performance index.

[0015] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of any of the methods described above.

[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the methods described herein.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention discloses a full-condition optimization method for axial compressors. First, data mining analysis is performed on the data. By analyzing the mapping relationship between design variables and performance indicators, the design variables that significantly affect the performance indicators and the direction and degree of influence of these significant variables on each performance indicator are identified, making the optimization problem interpretable. Simultaneously, by analyzing the changing trends among multiple performance indicators, performance indicators that conflict with the main optimization objective are used as constraints on the main optimization objective. This transforms the multi-condition, multi-objective optimization problem into a single optimization objective plus several optimization constraints, simplifying the optimization problem. Combining the optimized design variables obtained on this basis with the results of the previous data mining analysis, full-condition optimization of the axial compressor is achieved. This effectively resolves the constraint relationship between stall margin at certain speeds and peak efficiency at the design speed in traditional compressor optimization, achieving a comprehensive improvement in compressor performance.

[0018] Furthermore, in this invention, when solving the optimization problem, a penalty mechanism is constructed to address the problem of instability in the near-stall point calculation of the compressor. The convergence of key performance parameters during the calculation process is monitored, and by adding a penalty factor, the non-convergent solution is made to deviate from the optimal solution. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the optimization process based on data mining in an embodiment of the present invention.

[0021] Figure 2a This invention utilizes data mining methods to analyze ring heatmaps in an embodiment of the invention. Figure 2b This invention utilizes data mining methods to analyze SHAP cellar diagrams in an embodiment of the invention. Figure 3 This is a SOM diagram analyzed using data mining methods in an embodiment of the present invention.

[0022] Figure 4a This is a full spectrum performance curve of the reference design for an embodiment of the present invention; Figure 4b The full spectrum performance curve is optimized for an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0024] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0025] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0026] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper," "lower," "horizontal," or "inner" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of the invention is in use, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, terms such as "first" and "second" are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0027] Furthermore, the use of the term "horizontal" does not imply that the component must be absolutely horizontal, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0028] In the description of the embodiments of the present invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention according to the specific circumstances.

[0029] See Figure 1 This invention discloses a full-condition optimization method for axial compressors. This method can improve the stability and efficiency of compressors across different speed ranges and solves the negative correlation between stall margin at certain speeds and peak efficiency at the design speed in the prior art. The method includes the following steps: Step 1: Establish a full-spectrum CFD (Computational Fluid Dynamics) performance evaluation model.

[0030] Step 2: Establish the target task CFD performance evaluation model (i.e., use the reference design calculation results as the initial field and directly calculate the corresponding operating point of the new sample).

[0031] Step 3: Establish a design space, obtain several uniformly distributed compressor geometry samples in the design space, and perform CFD calculations for the target task to obtain performance indicators such as stall margin at different speeds and peak efficiency at the design speed.

[0032] Step 4: Apply the data mining method SHAP to decompose and quantify the significance and mechanism of each variable's impact on the performance index, making the research question interpretable and obtaining relevant knowledge.

[0033] First, the SHAP feature importance map needs to be modified to use a circular heatmap to more intuitively present the feature importance of each design variable under multiple objectives. The average absolute value of the SHAP for each variable is calculated as an indicator of feature importance. The circular heatmap represents different performance indicators of interest from the outside to the inside along the radial direction, while different sectors along the tangential direction represent different design variables. The average absolute value of the normalized SHAP is marked with color, and the contribution of a design variable to a performance indicator can be obtained based on the color distribution of any sector. By synthesizing the entire circular heatmap, the identification of significant variables under multi-objective, high-dimensional design variables can be completed.

[0034] Secondly, the SHAP honeycomb diagram was further modified to better adapt it to multi-objective problems. The original honeycomb diagram uses SHAP values ​​on the horizontal axis and lists significant variables for a specific performance indicator on the vertical axis. It can only be used to analyze the degree and direction of the influence of significant variables on performance indicators under a single objective. To extend the application of the honeycomb diagram to multi-objective problems, a significant design variable that we want to analyze further is selected from the SHAP eigenvalue plot. Its SHAP values ​​for multiple performance indicators are then calculated and normalized. The honeycomb diagram is then modified so that the horizontal axis uses SHAP values ​​and the vertical axis lists the multiple performance indicators of interest. This allows us to obtain the degree and direction of the influence of a significant variable on multiple objectives, thus completing the analysis of multi-objective problems.

[0035] Step 5: Apply the data mining method Self-Organizing Map (SOM) to analyze the consistency / conflict relationship of the improvement of multiple performance indicators of interest. After determining the optimization goal, the performance indicators that are consistent with the trend of the goal can be temporarily ignored, while the performance indicators that are inconsistent with the trend of the goal are used as optimization constraints. In this way, the multi-condition multi-objective optimization problem is transformed into the form of one optimization goal plus several optimization constraints for subsequent optimization work.

[0036] Step 6: Couple efficient global optimization algorithms, parametric modeling, and CFD flow field analysis modules to carry out optimization work. Given the expected range of variation of the optimization constraints, if the value exceeds this range, a penalty will be added to the optimization objective, causing the solution to deviate from the optimal solution. Considering that the compressor is prone to computational instability and convergence curve divergence near the stall point, the convergence of key performance indicators is monitored during the calculation process. If the calculation fails to meet the convergence criteria, a large penalty will be added to the objective, similarly causing the non-convergent solution to deviate from the optimal solution.

[0037] Step 7: Based on the above settings, optimize the target task, obtain a phased optimized solution, and then conduct a full-spectrum CFD performance evaluation.

[0038] Step 8: Repeat steps 6 and 7 until you obtain an individual that satisfies the multi-condition, multi-objective optimization task, then stop the optimization.

[0039] The present invention will now be described in further detail with reference to the accompanying drawings: Reference Figure 1 The present invention discloses a method for optimizing the stability and efficiency of an axial compressor under all operating conditions based on data mining, comprising the following steps: Step 1: Establishment of a full-condition CFD performance evaluation model.

[0040] Mesh generation and refinement of the near-wall region are performed to ensure mesh independence in the solution. A CFD solver and turbulence model suitable for turbomachinery flow are selected. From low to high speeds, the compressor's complete characteristic curves at different speeds are obtained by gradually increasing the outlet back pressure while maintaining given inlet total temperature and pressure, from the clogging point to near stall. This model needs to be validated against experimental data or high-precision simulation results to ensure its predictive accuracy.

[0041] Step 2: Determining the CFD computation strategy for the target task To improve optimization efficiency, a targeted CFD calculation strategy is established, focusing on the near-stall point and peak efficiency conditions that are of interest in the optimization process: for the calculation of the near-stall point at each speed, the flow field of the reference design at the corresponding speed is used as the initial condition to accelerate convergence and improve calculation stability; for the calculation of the peak efficiency point, the flow field of the high-efficiency condition can be used as the initial condition, which greatly reduces the number of operating points to be calculated and the calculation cost of the optimization sample.

[0042] Step 3: Design Space Construction and Sample Initialization Key geometric parameters affecting compressor performance are selected as design variables, and their reasonable physical variation range is determined. An experimental design method is used to generate N initial samples within the design space. Using the CFD model established in step 2, the key operating points of these samples are automatically calculated, and key performance indicators such as stall margin at different speeds and peak efficiency at the design speed are extracted to form an initial sample database.

[0043] Furthermore, key geometric parameters include three-dimensional shaping parameters such as blade bending, twisting, and sweeping, as well as control points on the blade suction surface.

[0044] Furthermore, the experimental design method employed the Latin hypercube sampling method.

[0045] Furthermore, the initial sample size N can be selected as 2 to 3 times the number of design variables.

[0046] Step 4: Optimization Problem Analysis and Restructuring Based on Data Mining Step 4.1: Identification of Significant Multi-Objective Variables and Analysis of Influence Mechanisms. A high-performance surrogate model is trained based on the initial sample database. Subsequently, the SHAP method is applied for interpretability analysis. The SHAP values ​​and their absolute averages are calculated as follows:

[0047] In the formula, i This is the sample count value. j For the count values ​​of the feature variables, N The total number of samples, M It is the total number of characteristic variables; S It is a set {1,2..j-1,j+1,..M} a subset of, | S |Yes S The total number of elements in; Indicates when only The features in the model are the predicted values. Indicates only S The feature in the model is the predicted value, and the difference between the two is the first value. j The feature variables in the subset S The marginal contribution below.

[0048] (1) Modify the circular heatmap to identify significant variables: Calculate the absolute average of the SHAP values ​​of each design variable for each performance index and normalize them as a quantitative indicator of feature importance. Construct a circular heatmap to intuitively present the global importance of each variable under multiple objectives through the shades of color.

[0049] (2) Modify the honeycomb diagram to analyze the influence mechanism: For the selected significant variables, calculate their SHAP values ​​for multiple performance indicators and normalize them. Construct the modified honeycomb diagram and use it to determine the direction of influence of a variable on different objectives.

[0050] Step 4.2: Performance Indicator Conflict Analysis Based on SOM. Self-Organizing Maps (SOM) are used to visualize and reduce the dimensionality of multiple performance indices of interest. Indicators that conflict with the main optimization objective (such as peak efficiency at design speed) (such as stall margin at certain speeds) are set as optimization constraints, and their allowable range of variation is specified. This simplifies the complex multi-objective optimization problem into a "single objective + constraint" form.

[0051] Step 5: Optimization process for integrating penalty mechanisms The optimization problem described above is searched by coupling efficient global optimization algorithms, parametric modeling, and CFD flow field analysis modules. Two penalty mechanisms are integrated into the algorithm's fitness function: performance constraint penalty and convergence penalty. The penalty factor is written as:

[0052] In the formula, P(x) To apply to the solution x Total penalty value; g j (x) For the first j A constraint function, when g j (x) A value greater than 0 indicates that the constraint has been violated; λ j For the first j The penalty coefficient for each constraint ( λ j >0), used to control the severity of different constraint penalties.

[0053] (1) Performance constraint penalty: If the performance index of the optimization solution violates the allowable constraint range set in step 4b (such as a decrease in stall margin), then a penalty factor positively correlated with the degree of violation is multiplied on its objective function value.

[0054] (2) Calculate convergence penalty: During the CFD calculation, monitor the residuals of key parameters such as flow rate and efficiency in real time. Set convergence criteria. If the calculation result does not converge, apply a very large penalty factor to the solution to eliminate it in the evolution. The convergence criterion disclosed in this embodiment is that the fluctuation is less than 1% in the last 200 iterations.

[0055] Furthermore, the optimization algorithm is the GSDE-surrogate model-assisted differential evolution algorithm; the parametric modeling method is the geometric parameter method; and the CFD flow field calculation and analysis is performed using the commercial software NUMECA.

[0056] Step 6: Iterative Optimization and Final Verification Run the optimization algorithm to obtain interim optimized solutions. For these optimized solutions, re-execute the "full-condition CFD performance evaluation" described in step 1 to further evaluate their performance under the full spectrum. If the full spectrum performance fails to meet the stopping condition, continue iterative optimization until a final design scheme that meets all operating condition requirements is obtained.

[0057] This invention discloses a specific embodiment: Step 1: Establishment of a full-spectrum CFD performance evaluation model This embodiment selects a 1.5-stage axial compressor as the research object and optimizes its three-row blade design, including one row of guide vanes, one row of moving vanes, and one row of stationary vanes. Four operating conditions—0.5 times the design speed (0.5 N), 0.75 times the design speed (0.75 N), 0.85 times the design speed (0.85 N), and the design speed (1.0 N)—are selected to form the full characteristic curve. A structured mesh is generated using NUMECA AotuGrid5, with finer meshing in the near-wall region to ensure mesh independence. The Navier-Stokes (RANS) solver NUMECA FINE, based on Reynolds averages, is selected, and the SA model is preferred as the turbulence model. Total temperature and total pressure are given at the inlet; the operating point is adjusted by giving the average static pressure at the outlet. The outlet back pressure is gradually increased at a certain speed, and the calculation results at the low back pressure operating point are used as the initial field at the high pressure operating point. The calculation proceeds from the clogging point to the near-stall point, thus obtaining the complete compressor characteristic curve.

[0058] Step 2: Establishing the CFD performance evaluation model for the target task The full-spectrum CFD performance evaluation model accurately obtains the compressor reference design's calculation results at all operating points. Subsequently, five operating points—1.0N peak efficiency, 1.0N near-stall, 0.85N near-stall, 0.75N near-stall, and 0.5N near-stall—are selected as optimization focus points to establish the target task CFD performance evaluation model. This means that each subsequent new sample only needs to be calculated under these five operating points. The calculation employs an "initial field inheritance" method. For the calculations at the four near-stall points, the flow field at the near-stall point of the reference design at the corresponding speed is used as the initial field to accelerate convergence and improve computational stability. For the calculation of the peak efficiency point at the design speed, the flow field at the high-efficiency point at the design speed is used as the initial field.

[0059] Step 3: Design Spatial Sampling and CFD Calculation Key geometric parameters affecting compressor performance were identified as design variables: For each blade row, three cross-sections were selected: the root section, the middle section, and the tip section. Five control points were set on the suction side of each cross-section for profile fine-tuning. Specifically, the control points included two fixed control points at the beginning and end, and three adjustable control points in the middle. In addition, rotation angle parameters were added to each cross-section to control blade twisting. Two circumferential translation parameters were set at the middle and tip of each of the three blade rows to control blade bending. One axial translation parameter at the tip controlled blade sweeping. A total of 45 design variables were defined for the three blade rows, each with a reasonable range of physical variation. Within the design space, 90 uniformly distributed geometric sample points were generated using the Latin Hypercube Sampling (LHS) method.

[0060] A Python script was written to drive parametric modeling and CFD solver, automatically performing "target task CFD calculations" for each sample. Key performance indicators were extracted from the calculation results, mainly including peak isentropic efficiency at the design speed, stall margin at various speeds, pressure ratio, and flow rate under various operating conditions. This resulted in a database containing geometric variables and their corresponding performance indicators, which will be used for subsequent data mining.

[0061] Step 4: Data Mining and Knowledge Discovery Based on the Modified SHAP Method Step 4.1: Using data from the database, train a high-performance machine learning model, XGBoost, as a surrogate model to map the relationship between geometric variables and performance metrics.

[0062] Step 4.2: Calculate the SHAP value and its absolute average for each design variable for the five performance indices: peak efficiency at 1.0N (1.0PE_EF), stall margin at 1.0N near-stall (1.0NS_SM), stall margin at 0.85N near-stall (0.85NS_SM), stall margin at 0.75N near-stall (0.75NS_SM), and stall margin at 0.5N near-stall (0.5NS_SM).

[0063] Step 4.3: Construct a ring-shaped heatmap. The radial coordinates represent different performance indicators, from the outside in: 1.0PE_EF, 1.0NS_SM, 0.85NS_SM, 0.75NS_SM, and 0.5NS_SM. The tangential coordinates represent different design variables: x1~x6 are bending parameters, x7~x15 are torsional parameters, x16~x18 are sweep parameters, and x19~x45 are profile fine-tuning parameters. The color intensity of each sector (corresponding to one variable and one indicator) indicates the importance value of that design variable to the normalized SHAP of that performance indicator. For example, as shown in Figure 2(a), the variables that significantly affect different performance indicators are mainly concentrated in the torsional and sweep subspaces.

[0064] Step 4.4: Select a significant variable of interest from the annular heatmap. In this embodiment, the guide vane sweep parameter x16 is selected as the significant variable. The horizontal axis of the modified honeycomb diagram represents the normalized SHAP value, with the center at 0. The right side represents the positive impact (indicating improved performance), and the left side represents the negative impact (indicating reduced performance). The vertical axis lists the five performance indicators of interest from top to bottom. Each performance indicator corresponds to a horizontal band, and the points in the band represent the distribution of the SHAP value of that variable in different samples. Through this figure, the direction (including positive and negative impacts) and degree of influence of a significant variable on each performance indicator can be clearly determined. See Figure 2(b). It can be found that increasing the guide vane sweep parameter improves the peak efficiency at the design speed and the stall margin at each speed.

[0065] Step 5: Refactoring the optimization problem based on self-organizing map (SOM) Step 5.1: Analyze the aforementioned five performance indicators using a Self-Organizing Map (SOM). Observe the color distribution of different performance indicators on the SOM map. If the color distribution of different performance indicators at the same position on the SOM map is roughly the same, it indicates that their trends are consistent; if the color distribution is opposite, it indicates that there is a trade-off between them. See [link to relevant documentation] Figure 3 The stall margin changes in different speeds are generally consistent, while the stall margin and peak efficiency at all speeds are mutually restrictive at certain points.

[0066] Step 5.2: Based on the above analysis, the optimization problem is transformed, and the main optimization objective is determined to be maximizing the 1.0N peak efficiency. The 1.0N stall margin, 0.85N stall margin, 0.75N stall margin, and 0.5N stall margin are set as optimization constraints, requiring their changes to be no less than 0. In addition, to prevent an overall shift in the full spectrum characteristic curve, an additional constraint is added, requiring the flow rate change rate at the 1.0N peak efficiency operating point to be less than ±2%.

[0067] Step 6: Optimize processes and penalty mechanisms Step 6.1: Couple efficient global optimization algorithms, parametric modeling, CFD flow field analysis and other modules to search for the simplified optimization problem described above.

[0068] Step 6.2: Introduce a penalty term into the fitness function of the optimization algorithm: If the stall margin change is less than 0 at a certain speed and the flow rate change rate at the 1.0N peak efficiency operating point is greater than ±2%, then add a penalty factor related to the degree of violation to the optimization target value according to the degree of deviation.

[0069] Step 6.3: During the CFD calculation of each sample, monitor the residual curves of the outlet flow rate and efficiency in real time. Set the convergence criterion as follows: within the last 200 iterations, the fluctuation of both flow rate and efficiency is less than 1%. If the calculation result does not meet this convergence criterion, the calculation at that point is considered a failure, and a very large penalty factor is applied to its optimization objective value to ensure that the "non-convergent solution" is eliminated during the evolution process.

[0070] Step 7: Iterative Optimization and Final Verification Run the optimization algorithm to obtain a preliminary set of optimized solutions. Repeat the "full-spectrum CFD performance evaluation" described in step 1 on the interim optimized solutions to comprehensively and accurately evaluate their performance across the entire working range and verify whether they truly meet the requirements of multi-condition optimization. If the full-spectrum performance fails to meet the stopping condition, iterative optimization continues until a final design scheme that meets all operating condition requirements is obtained.

[0071] See Figure 4a and Figure 4b The performance comparison of the optimized design and the reference design in the embodiment shows that the stall margin of the compressor along the entire stall line has been improved, and the peak efficiency at all speeds has also been significantly improved. Table 1 shows the specific improvement, which verifies the correctness and effectiveness of the present invention.

[0072] Table 1 Performance Comparison between Optimized Design and Reference Design

[0073] This embodiment also discloses a full-condition optimization system for axial compressors, including: The module for obtaining design variables and performance indicators is used to select the design variables that affect the compressor performance and the corresponding performance indicators. The module for analyzing the relationship between design variables and performance indicators is used to construct the mapping relationship between design variables and performance indicators. Based on the mapping relationship, the feature importance of each design variable is calculated. Based on the feature importance, design variables that have a significant impact on performance indicators are identified. The identified design variables are used as significant variables. Based on the significant variables, the direction and degree of influence of the significant variables on each performance indicator are obtained. The design variable optimization module is used to analyze the relationship between the changing trends of multiple performance indicators based on the performance indicators corresponding to the design variables, obtain the performance indicator conflict relationship analysis results, select the main optimization objective, and, based on the performance indicator conflict relationship analysis results, take the performance indicators that conflict with the main optimization objective as the constraints of the main optimization objective, construct the optimization problem of the axial compressor based on the main optimization objective and the constraints, solve the optimization problem, and obtain the optimized design variables. The optimization result acquisition module is used to obtain optimization results based on the direction and degree of influence of the optimized design variables and significant variables on each performance index.

[0074] The method disclosed in the embodiments of the present invention has the following advantages: (1) This method applies data mining technology to the optimization design problem of multi-stage axial compressors, especially SHAP and self-organizing map SOM. On the one hand, data mining technology is used to analyze and obtain knowledge such as variables that make significant contributions to the objective and their influence, making the optimization problem interpretable. On the other hand, the consistency / conflict of the improvement of multiple performance indicators is visualized. Consistent indicators are omitted and conflicting indicators are used as constraints. Thus, the multi-condition multi-objective optimization problem is transformed into the form of one optimization objective plus several optimization constraints, which simplifies the optimization problem.

[0075] (2) In addition to the application of data mining methods, this method will add a penalty to the optimization objective for solutions that violate the constraints; at the same time, in order to solve the problem of unstable calculation of compressor near stall point, the convergence of key performance during the calculation process is monitored. If a large convergence fluctuation occurs, a large penalty will also be added to the objective, which will also cause the non-convergent solution to deviate from the optimization solution.

[0076] (3) Based on the above technological innovations, the data mining-based compressor stability and efficiency optimization method established in this invention can effectively solve the constraint relationship between stall margin at some speeds and peak efficiency at the design speed in traditional compressor optimization, and realize the comprehensive improvement of compressor performance. Therefore, it has important engineering significance and broad application prospects.

[0077] A schematic diagram of a terminal device according to an embodiment of the present invention. The terminal device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the various method embodiments described above. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the various device embodiments described above.

[0078] The computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention.

[0079] The terminal device can be a desktop computer, laptop computer, cloud server, or other device with strong computing power. The terminal device may include, but is not limited to, a processor and memory.

[0080] The optimal choice for the processor is a multi-core high-speed central processing unit (CPU).

[0081] The memory can be used to store the computer program and / or module. The processor implements various functions of the terminal device by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory.

[0082] If the modules / units integrated into the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0083] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. 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 the operation of an axial compressor under all operating conditions, characterized in that, Includes the following steps: Select the design variables that affect the compressor performance and the corresponding performance indicators; Construct a mapping relationship between design variables and performance indicators, calculate the feature importance of each design variable based on the mapping relationship, identify design variables that have a significant impact on performance indicators based on feature importance, use the identified design variables as significant variables, and obtain the direction and degree of influence of significant variables on each performance indicator based on significant variables. Based on the performance indicators corresponding to the design variables, the relationship between the changing trends of multiple performance indicators is analyzed, the results of the performance indicator conflict relationship analysis are obtained, the main optimization objective is selected, and based on the performance indicator conflict relationship analysis results, the performance indicators that conflict with the main optimization objective are used as the constraints of the main optimization objective. Based on the main optimization objective and the constraints, the optimization problem of the axial compressor is constructed, the optimization problem is solved, and the optimized design variables are obtained. The optimization results were obtained based on the optimized design variables and the influence direction and degree of significant variables on each performance index.

2. The method for optimizing the operation of an axial compressor under all operating conditions according to claim 1, characterized in that, The selection of design variables affecting compressor performance and the corresponding performance indicators includes: A CFD performance evaluation model was constructed, and key operating point calculations were performed on the selected design variables based on the CFD performance evaluation model. Based on the calculation results, the stall margin at different speeds and the peak efficiency at the design speed of the axial compressor were selected as performance indicators.

3. The method for optimizing the operation of an axial compressor under all operating conditions according to claim 2, characterized in that, The design variables include three-dimensional modeling parameters and blade suction surface control points that affect compressor performance.

4. The method for optimizing the operation of an axial compressor under all operating conditions according to claim 1, characterized in that, The process involves constructing a mapping relationship between design variables and performance indicators, calculating the feature importance of each design variable based on this mapping relationship, identifying design variables that have a significant impact on performance indicators based on feature importance, designating these identified design variables as significant variables, and obtaining the direction and degree of influence of these significant variables on each performance indicator based on these significant variables. The SHAP method is introduced to calculate the average of the absolute values ​​of the SHAP values ​​of each design variable, and the average of the absolute values ​​of the SHAP values ​​is used as a quantitative indicator of feature importance. A ring heatmap is constructed based on the quantitative index of feature importance. The contribution of design variables to performance indicators is identified based on the ring heatmap. Significant variables are selected based on the contribution identification results. Based on significant variables, an improved SHAP cell diagram is generated, and the direction and degree of influence of significant variables on each performance index are obtained from the improved SHAP cell diagram.

5. The method for optimizing the operation of an axial compressor under all operating conditions according to claim 4, characterized in that, The calculation of the average of the absolute values ​​of the SHAP values ​​of each design variable includes: In the formula, i This is the sample count value. j For the count values ​​of the feature variables, N The total number of samples, M It is the total number of characteristic variables; S It is a set {1,2..j-1,j+1,..M} a subset of, | S |Yes S The total number of elements in; Indicates when only The features in the model are the predicted values. Indicates only S The feature in the model is the predicted value, and the difference between the two is the first value. j The feature variables in the subset S The marginal contribution below.

6. The method for optimizing the operation of an axial compressor under all operating conditions according to claim 1, characterized in that, The optimization problem of constructing the axial flow compressor includes: A self-organizing map method is introduced to analyze multiple performance indicators and generate SOM maps; Based on the generated SOM (System Oscillator) plot, determine whether the changing trends of the various performance indicators are consistent or conflicting. Determine the main optimization objective, and use other performance indicators that conflict with the main optimization objective as constraints on the main optimization objective to construct a single-objective combined with multiple constraints optimization problem.

7. The method for optimizing the operation of an axial compressor under all operating conditions according to claim 1, characterized in that, When solving the optimization problem, a penalty mechanism is constructed, including: If the performance index of the generated candidate solution exceeds the constraint range, a penalty factor is added to the optimization objective value; If the performance index calculation process of the candidate solution fails to meet the preset convergence criterion, a penalty factor is added to the optimization target value.

8. A full-condition optimization system for an axial compressor, characterized in that, include: The module for obtaining design variables and performance indicators is used to select the design variables that affect the compressor performance and the corresponding performance indicators. The module for analyzing the relationship between design variables and performance indicators is used to construct the mapping relationship between design variables and performance indicators. Based on the mapping relationship, the feature importance of each design variable is calculated. Based on the feature importance, design variables that have a significant impact on performance indicators are identified. The identified design variables are used as significant variables. Based on the significant variables, the direction and degree of influence of the significant variables on each performance indicator are obtained. The design variable optimization module is used to analyze the relationship between the changing trends of multiple performance indicators based on the performance indicators corresponding to the design variables, obtain the performance indicator conflict relationship analysis results, select the main optimization objective, and, based on the performance indicator conflict relationship analysis results, take the performance indicators that conflict with the main optimization objective as the constraints of the main optimization objective, construct the optimization problem of the axial compressor based on the main optimization objective and the constraints, solve the optimization problem, and obtain the optimized design variables. The optimization result acquisition module is used to obtain optimization results based on the direction and degree of influence of the optimized design variables and significant variables on each performance index.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

Images