Aircraft propeller design method and system based on sensitivity analysis and intelligent multi-objective collaborative optimization

Abstract

This invention proposes a design method and system for aircraft propellers based on sensitivity analysis and intelligent multi-objective collaborative optimization. The method includes: sampling propeller blade sample points under different conditions to obtain design parameters; performing hydrodynamic and aeroacoustic numerical simulations to obtain aerodynamic lift and aerodynamic noise values; calculating the sensitivity of each design parameter to the target value using range analysis to identify a subset of design variables strongly correlated with aerodynamic lift and aerodynamic noise; sampling using the optimal Latin hypercube method, generating sample geometric models through automatic mesh deformation and simulating them to obtain sample data; constructing a Kriging approximation model for aerodynamic lift and aerodynamic noise based on the sample data; using a multi-island genetic algorithm to perform single-objective optimization on the approximation model to obtain the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise; and constructing a two-layer optimization framework based on the maximum and minimum values ​​to obtain a collaborative optimization solution set. This invention reduces computational costs by leveraging sensitivity analysis for dimensionality reduction and combining it with surrogate model technology.

Description

Technical Field

[0001] This invention relates to the field of aircraft propeller design and optimization technology, and in particular to an aircraft propeller design method and system based on sensitivity analysis and intelligent multi-objective collaborative optimization. Background Technology

[0002] As a key propulsion component of aircraft, the propeller's aerodynamic and acoustic performance directly determines the aircraft's propulsion efficiency, range, and noise characteristics. With the rapid development of electric vertical takeoff and landing (EVTOL) aircraft and the low-altitude economy, the composition of aircraft noise sources has changed significantly. After traditional engine noise has been greatly reduced or eliminated, the aerodynamic noise generated by propeller rotation has become the dominant factor affecting cabin passenger comfort, aircraft stealth characteristics, and community noise compliance.

[0003] There is an inherent coupling and conflict between the aerodynamic and acoustic performance of propellers. Increasing the blade angle or rotational speed can effectively improve lift and thrust, but this often exacerbates tip vortices, uneven blade loads, and flow separation, leading to strong broadband vortex noise and discrete frequency noise. Therefore, effectively suppressing aerodynamic noise while maintaining sufficient aerodynamic lift has become a core challenge in modern aircraft propeller design.

[0004] Traditional propeller design heavily relies on designers' experience and repeated "design-manufacturing-testing" cycles. This approach is costly, time-consuming, and struggles to systematically approximate the global optimum at the Pareto front of performance. While high-fidelity numerical simulation techniques based on computational fluid dynamics and computational aeroacoustics provide effective tools for accurately evaluating propeller performance, directly embedding them into the optimization loop still faces bottlenecks such as excessive computational cost and low optimization efficiency. Furthermore, propeller design involves numerous geometric parameters, which exhibit significantly different sensitivities to different performance objectives. Traditional optimization methods or multi-objective algorithms typically treat all design variables equally, failing to effectively differentiate the strength of the correlation between variables and objectives. This leads to increased dimensionality of the optimization problem, low search efficiency, and a tendency to get trapped in local optima. Summary of the Invention

[0005] This invention aims to at least solve the technical problems of excessively high computational cost and low optimization efficiency in the prior art, and innovatively proposes an aircraft propeller design method and system based on sensitivity analysis and intelligent multi-objective collaborative optimization.

[0006] To achieve the above-mentioned objectives of this invention, this invention provides a method for designing aircraft propellers based on sensitivity analysis and intelligent multi-objective collaborative optimization, the method comprising: S1. Sample points on the propeller blades under different conditions to obtain design parameters; S2. Based on the design parameters, perform numerical simulations of fluid dynamics and aeroacoustics to obtain the aerodynamic lift and aerodynamic noise values ​​at each sample point. S3. Based on the aerodynamic lift value and aerodynamic noise value, use the range analysis method to calculate the sensitivity of each design parameter to the target aerodynamic lift value and the target aerodynamic noise value, and based on the sensitivity, separate the design variable subset that is strongly correlated with aerodynamic lift and the design variable subset that is strongly correlated with aerodynamic noise. S4. Based on the subset of design variables strongly correlated with aerodynamic lift and the subset of design variables strongly correlated with aerodynamic noise, sample data is obtained by sampling using the optimal Latin hypercube method, generating sample geometric models through the automatic mesh deformation method, and performing numerical simulation. S5. Based on the sample data, construct Kriging approximate models of aerodynamic lift and aerodynamic noise with respect to their respective design variables to obtain the aerodynamic lift Kriging approximate model and the aerodynamic noise Kriging approximate model. S6. Based on the aforementioned Kriging approximation model for aerodynamic lift and Kriging approximation model for aerodynamic noise, a multi-island genetic algorithm is used for single-objective optimization to obtain the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise. S7. Based on the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise, construct a system-level and discipline-level two-layer optimization framework to obtain a collaborative optimization solution set of aerodynamic lift and aerodynamic noise.

[0007] As an optional embodiment of the present invention, optionally, in step S7, a two-layer optimization framework at the system level and discipline level is constructed to obtain a synergistic optimization solution set for aerodynamic lift and aerodynamic noise, including: S701, the system-level optimizer takes a weighted normalized function that maximizes aerodynamic lift and minimizes aerodynamic noise as the optimization objective, sets the weight coefficients for aerodynamic lift and aerodynamic noise, and balances the importance of the two objectives by adjusting the weight coefficients; S702, The system-level optimizer uses the aerodynamic lift target value and aerodynamic noise target value under the current weight coefficients as consistency constraints for discipline-level optimization, and passes them to the aerodynamic lift discipline optimizer and the aerodynamic noise discipline optimizer respectively. S703, the aerodynamic lift optimization tool, is based on the Kriging approximation model of aerodynamic lift. Under the constraint of the aerodynamic lift target value transmitted at the system level, it optimizes a subset of design variables that are strongly related to aerodynamic lift and seeks to satisfy the optimal combination of design variables. S704, the aerodynamic noise optimization tool, is based on the Kriging approximation model of aerodynamic noise. Under the constraint of the aerodynamic noise target value transmitted at the system level, it optimizes a subset of design variables that are strongly correlated with aerodynamic noise to seek the optimal combination of design variables. S705, the aerodynamic lift optimizer and the aerodynamic noise optimizer feed back the design variable combinations and corresponding aerodynamic lift and aerodynamic noise values ​​obtained by their respective optimizations to the system-level optimizer. S706. The system-level optimizer determines whether the preset convergence condition is met based on the feedback result. If it is met, it outputs the co-optimization solution under the current weight coefficient. If it is not met, it adjusts the weight coefficient and returns to step S702 until it traverses the preset weight coefficient range and finally obtains the co-optimization solution set of aerodynamic lift and aerodynamic noise.

[0008] As an optional embodiment of the present invention, optionally, the expression of the system-level optimizer in step S701 is: in, This represents the system-level optimization objective function value. Indicates the weighting coefficient. This represents the system-level aerodynamic lift target value. Indicates aerodynamic lift. This indicates the maximum aerodynamic lift. This represents the minimum aerodynamic lift. This represents the optimal value for aerodynamic noise. This represents the target value for the system-level aerodynamic noise sound pressure level. This indicates the maximum aerodynamic noise level. This indicates the minimum aerodynamic noise level.

[0009] As an optional embodiment of the present invention, the consistency constraint in step S702 may be relaxed by introducing a relaxation factor to improve convergence.

[0010] As an optional embodiment of the present invention, the weighting coefficient may be... The value range is from 0.001 to 0.01.

[0011] As an optional embodiment of the present invention, the design parameters in step S1 may include at least four of the following: blade diameter, blade mounting angle, chord length distribution, relative airfoil thickness, airfoil camber, and leading edge radius.

[0012] As an optional embodiment of the present invention, in step S5, the prediction accuracy of the aerodynamic lift Kriging approximation model and the aerodynamic noise Kriging approximation model are evaluated by cross-validation. When the root mean square error of the corresponding Kriging approximation model is less than a preset threshold, the model is deemed to be qualified. If the error exceeds the threshold, sample data is added and the model is retrained until the preset accuracy is met.

[0013] As an optional embodiment of the present invention, the method may further include: S8. Based on the co-optimized solution set of aerodynamic lift and aerodynamic noise, perform CFD numerical simulation verification, compare the error between the optimization result and the predicted value of the approximate model. If the error is within the preset range, the optimized solution set is determined to be valid; if the error exceeds the preset range, return to step S4, adjust the sampling strategy or increase the number of samples and reconstruct the Kriging approximate model and optimize it until the verification is passed.

[0014] In another aspect, the present invention also provides an aircraft propeller design system based on sensitivity analysis and intelligent multi-objective collaborative optimization, the system comprising: processor; Memory used to store processor-executable instructions; The processor is configured to implement the aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization when executing the executable instructions.

[0015] The beneficial effects of this invention are as follows: This invention achieves dimensionality reduction of the problem by means of sensitivity analysis, and combines it with surrogate model technology. While ensuring optimization accuracy, the computational cost is reduced by one to two orders of magnitude compared with traditional methods.

[0016] The core optimization framework constructed in this invention has good universality and can be widely applied to multidisciplinary design optimization problems of various complex equipment. Furthermore, given the complex and tightly coupled relationship between variables and aerodynamic / acoustic targets in aircraft propeller design, a sensitivity-based subset partitioning mechanism for design variables is used for targeted processing. This allows the optimization process to focus more on key influencing factors, thereby achieving better specific optimization results while ensuring the method's versatility.

[0017] The entire process can achieve a high degree of automation and can be seamlessly integrated with commercial CFD software and the intelligent optimization toolkit Vaeopy. Automated operation scripts can be written in both pre-processing and post-processing to meet automation needs, making it easy for applications to adopt and effectively shortening the development cycle.

[0018] Practice has shown that the propeller optimized by the method of this invention can significantly improve lift under design conditions, while effectively reducing far-field aerodynamic noise, achieving a synergistic leap in performance.

[0019] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of an aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization according to the present invention; Figure 2 This is a flowchart of an aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization according to the present invention; Figure 3 This is the L9(34) orthogonal experimental table of the present invention; Figure 4 This is a flowchart of the collaborative optimization strategy based on sensitivity analysis of the present invention; Figure 5 This is a flowchart of the single-objective optimization process based on the Kriging approximation model of this invention; Figure 6 This is a flowchart of the multi-island genetic algorithm of the present invention. Detailed Implementation

[0021] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0022] Example 1 like Figure 1-6 As shown, an aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization is presented, the method comprising: S1. Sample points on the propeller blades under different conditions to obtain design parameters; S2. Based on the design parameters, perform numerical simulations of fluid dynamics and aeroacoustics to obtain the aerodynamic lift and aerodynamic noise values ​​at each sample point. S3. Based on the aerodynamic lift value and aerodynamic noise value, use the range analysis method to calculate the sensitivity of each design parameter to the target aerodynamic lift value and the target aerodynamic noise value, and based on the sensitivity, separate the design variable subset that is strongly correlated with aerodynamic lift and the design variable subset that is strongly correlated with aerodynamic noise. S4. Based on the subset of design variables strongly correlated with aerodynamic lift and the subset of design variables strongly correlated with aerodynamic noise, sample data is obtained by sampling using the optimal Latin hypercube method, generating sample geometric models through the automatic mesh deformation method, and performing numerical simulation. S5. Based on the sample data, construct Kriging approximate models of aerodynamic lift and aerodynamic noise with respect to their respective design variables to obtain the aerodynamic lift Kriging approximate model and the aerodynamic noise Kriging approximate model. S6. Based on the aforementioned Kriging approximation model for aerodynamic lift and Kriging approximation model for aerodynamic noise, a multi-island genetic algorithm is used for single-objective optimization to obtain the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise. S7. Based on the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise, construct a system-level and discipline-level two-layer optimization framework to obtain a collaborative optimization solution set of aerodynamic lift and aerodynamic noise.

[0023] This invention proposes a design method for aircraft propeller systems based on intelligent multi-objective collaborative optimization. Its core is to integrate single-objective variables to carry out collaborative optimization, decomposing the complex design problem into three logically clear stages.

[0024] Variable selection and problem dimensionality reduction stage Leveraging the efficiency of orthogonal experimental design, this invention minimizes the number of numerical simulations and systematically determines the independent and interactive effects of multiple design variables on the two objectives of aerodynamic lift and aerodynamic noise. Through sensitivity negotiation analysis, key variables with greater weighting are identified, forming a negotiation set of the two variable parameters. This step is a crucial prerequisite for this invention, enabling subsequent optimization to focus on the core variables and significantly improving optimization efficiency.

[0025] Single-discipline performance boundary exploration stage For the selected subset of variables, high-precision Kriging surrogate models were constructed for aerodynamic lift and aerodynamic noise respectively. These models can approximate time-consuming CFD+CAA simulation results with extremely low computational cost, forming the foundation for rapid optimization. Based on this, a multi-island genetic algorithm with powerful global search capabilities was used to perform single-objective optimization within each target discipline, accurately depicting the performance boundaries (i.e., maximum aerodynamic lift and noise). This not only provides a benchmark for objective normalization for subsequent collaborative optimization but also supports the standard for design limits.

[0026] System-level collaborative optimization phase An improved collaborative optimization strategy framework is constructed. The core improvement of this framework lies in: At the system level, the weighted summation method is used to hierarchically transform multi-objective problems into single-objective problems, and the weighting factors can be flexibly adjusted according to design preferences and actual needs.

[0027] By introducing a relaxation factor ε, the equality consistency constraint (J=0) that is difficult to satisfy in the traditional CO method is relaxed into an inequality constraint (J≤ε) that is easier to converge, which significantly improves the robustness and convergence of the algorithm.

[0028] Subject-level optimizers run in parallel, with each optimizer processing only a subset of design variables strongly related to its own subject. Their optimization objective is to minimize the difference between the subject-specific design and the system-level instructions. Through iterative coordination between the system-level and subject-level optimizers, they eventually converge to a globally optimal solution that is consistent across subjects.

[0029] As an optional embodiment of the present invention, optionally, in step S7, a two-layer optimization framework at the system level and discipline level is constructed to obtain a synergistic optimization solution set for aerodynamic lift and aerodynamic noise, including: The S701 system-level optimizer uses a weighted normalized function that maximizes aerodynamic lift and minimizes aerodynamic noise as its optimization objective. It sets weighting coefficients for aerodynamic lift and aerodynamic noise, and adjusts these coefficients accordingly. To balance the importance of the two objectives; In step S701, it is necessary to explain in detail that the core objective of the system-level optimizer is to find the optimal balance between two conflicting performance indicators: aerodynamic lift and aerodynamic noise. To this end, a weighted normalization method is used to integrate these two objectives into a comprehensive system-level optimization objective function. Specifically, the original values ​​of aerodynamic lift and aerodynamic noise are first normalized to eliminate the influence of dimensional differences. For aerodynamic lift, the normalized value is obtained by subtracting the minimum aerodynamic lift value from the current aerodynamic lift value, and then dividing by the difference between the maximum and minimum aerodynamic lift values. This ensures that the normalized value of aerodynamic lift is within the range of [0,1], with a larger value indicating better lift performance. For aerodynamic noise, considering that its objective is minimization, its normalized value is obtained by subtracting the current aerodynamic noise value from the maximum aerodynamic noise value, and then dividing by the difference between the maximum and minimum aerodynamic noise values. Similarly, this ensures that the normalized value of aerodynamic noise is within the range of [0,1], with a larger value indicating better noise control. Then, the system-level optimization objective function is achieved by multiplying the normalized value of the aerodynamic lift by a weighting coefficient. In addition, the normalized value of the aerodynamic noise is multiplied by (1- The weighting coefficients are obtained from this. The value range is from 0.001 to 0.01. This range was determined based on extensive preliminary simulation experiments and engineering experience, aiming to ensure that adjusting the weights has a significant and controllable impact on the importance of the two objectives, avoiding overweighting or underweighting one objective to the point of losing practical significance. The weighting coefficients are adjusted continuously or discretely. It can simulate different preferences for lift and noise under different design scenarios, for example, when When the value is large, system-level optimization focuses more on improving aerodynamic lift performance; when... When the value is smaller, the focus is more on reducing the level of aerodynamic noise.

[0030] S702, the system-level optimizer will use the current weight coefficients The target values ​​for aerodynamic lift and aerodynamic noise are used as consistency constraints for discipline-level optimization and are passed to the aerodynamic lift discipline optimizer and the aerodynamic noise discipline optimizer, respectively. In step S702, it is necessary to explain in detail that the system-level optimizer determines the current weight coefficients. After setting the target values ​​for aerodynamic lift and aerodynamic noise, these are then used as hard indicators and assigned to each subject-level optimizer. This "consistency constraint" means that when performing their own optimization, each subject-level optimizer must use the target values ​​passed down from the system level as important boundary conditions or optimization directions. Specifically, for the aerodynamic lift optimizer, the target value for aerodynamic lift passed down from the system level is the benchmark it needs to strive to reach or even exceed during its optimization process; while for the aerodynamic noise optimizer, the target value for aerodynamic noise passed down from the system level is the upper limit it must control below. This constraint is not simply a matter of providing numerical values, but rather it is embedded into the objective function or constraint conditions of the subject-level optimizer, ensuring that the optimization direction of the subject level is consistent with the overall goal of the system level, and avoiding a disconnect between the subject-level optimization results and the system-level expectations. In this way, the top-level design intent of the system level can be effectively implemented in the specific optimization process of each sub-discipline.

[0031] S703, the aerodynamic lift optimization tool, is based on the Kriging approximation model of aerodynamic lift. Under the constraint of the aerodynamic lift target value transmitted at the system level, it optimizes a subset of design variables that are strongly related to aerodynamic lift and seeks to satisfy the optimal combination of design variables. In step S703, it is necessary to explain in detail that after receiving the aerodynamic lift target value from the system-level optimizer, the aerodynamic lift optimization optimizer uses this target value as the core constraint and conducts targeted optimization based on the constructed high-precision aerodynamic lift Kriging approximation model. First, the decision variables for optimization are defined as a subset of design variables strongly correlated with aerodynamic lift, identified during the variable selection and problem dimensionality reduction stages, such as key parameters like blade installation angle, chord distribution, and airfoil camber. These variables are considered to play a dominant role in aerodynamic lift performance, and optimizing them can more directly and efficiently affect lift output. During the optimization process, the primary task of the aerodynamic lift optimization optimizer is to explore the optimal combination of these strongly correlated design variables while satisfying the system-level aerodynamic lift target value. Specifically, the optimizer searches within the feasible range of variables, uses the Kriging model to quickly predict the aerodynamic lift value under different variable combinations, and determines whether it reaches or exceeds the system-level target value. If the predicted lift value corresponding to the current combination of variables fails to meet the target, the optimization algorithm (such as a multi-island genetic algorithm) will adjust the values ​​of each design variable based on the information from the model feedback. For example, it might increase the blade installation angle to obtain higher lift, or adjust the chord distribution to optimize the aerodynamic characteristics of the airfoil. Simultaneously, the optimization process does not only pursue maximizing lift, but also considers other potential, unmentioned, but important factors that affect the overall propeller performance while satisfying the lift target. In other words, it seeks a solution with better overall performance within the discipline. In this way, the aerodynamic lift optimizer can perform detailed optimization of core variables within its specialized field, ultimately outputting one or more optimal combinations of design variables that meet the system-level lift requirements.

[0032] S704, the aerodynamic noise optimization tool, is based on the Kriging approximation model of aerodynamic noise. Under the constraint of the aerodynamic noise target value transmitted at the system level, it optimizes a subset of design variables that are strongly correlated with aerodynamic noise to seek the optimal combination of design variables. In step S704, it is necessary to explain in detail that after receiving the aerodynamic noise target value from the system-level optimizer, the aerodynamic noise optimization tool uses this target value as the core constraint and conducts targeted optimization based on the constructed high-precision aerodynamic noise Kriging approximation model. The decision variables for optimization are a subset of design variables strongly correlated with aerodynamic noise, determined during the variable selection and problem dimensionality reduction stages. These include parameters that significantly affect noise generation, such as blade tip shape, thickness distribution, and trailing edge sweep angle. These variables directly relate to the magnitude of eddies, pressure pulsations, and aerodynamic noise generated by the propeller during rotation. During the optimization process, the core task of the aerodynamic noise optimization tool is to adjust and optimize these strongly correlated design variables while meeting the system-level aerodynamic noise target value (i.e., the noise does not exceed the target value). Specifically, the optimizer explores within the allowable range of each design variable, uses the Kriging model to quickly predict the aerodynamic noise value under different variable combinations, and determines whether it is controlled below the system-level target value. If the predicted noise value corresponding to the current combination of variables exceeds the target, the optimization algorithm (such as a multi-island genetic algorithm) will adjust the relevant design variables based on the sensitivity information provided by the model. For example, it might modify the blade tip to a swept shape to suppress tip vortex noise, or optimize the thickness distribution to reduce aerodynamic load fluctuations, thereby achieving the goal of reducing noise. This optimization process focuses on finding the combination of design variables that minimizes noise while meeting noise constraints. Furthermore, the adjustment of these variables must consider their feasibility in manufacturing and structural strength, rather than simply pursuing the theoretical lowest noise level. Through such meticulous optimization, the aerodynamic noise optimizer can efficiently optimize core noise variables within its professional scope, ultimately outputting one or more optimal combinations of design variables that meet system-level noise requirements.

[0033] S705, the aerodynamic lift optimizer and the aerodynamic noise optimizer feed back the design variable combinations and corresponding aerodynamic lift and aerodynamic noise values ​​obtained by their respective optimizations to the system-level optimizer. In step S705, it is necessary to explain in detail that after completing their respective optimization iterations, the aerodynamic lift optimizer and the aerodynamic noise optimizer will package and feed back the optimal combination of design variables determined during the optimization process, as well as the aerodynamic lift and aerodynamic noise values ​​predicted by their respective Kriging approximation models based on these variable combinations, to the system-level optimizer. Here, "combination of design variables" refers to the specific parameter values ​​determined after optimization for the subset of strongly correlated variables in each discipline. For example, the aerodynamic lift optimizer might feed back specific values ​​such as the blade installation angle and the chord length of a specific section, while the aerodynamic noise optimizer might feed back specific values ​​such as the blade tip sweep angle and the location of maximum thickness. The aerodynamic lift and aerodynamic noise values ​​fed back simultaneously are the direct performance reflections under these variable combinations and are used by the system-level optimizer to evaluate the current weight coefficients. This feedback is a key basis for evaluating the optimization effect. It is not a one-way data transmission, but rather constitutes an important link in the closed-loop iteration between the system-level and subject-level systems, enabling the system-level optimizer to promptly grasp the optimization progress and results of each subject.

[0034] S706. The system-level optimizer determines whether the preset convergence conditions are met based on the feedback results. If they are met, it outputs the current weight coefficients. The collaborative optimization solution is obtained; if it is not satisfied, the weight coefficients are adjusted. Return to step S702 until the preset weight coefficients are traversed. The range is then determined, ultimately yielding a co-optimized solution set for aerodynamic lift and aerodynamic noise.

[0035] In step S706, it is necessary to explain in detail that after receiving the design variable combinations and corresponding performance values ​​from the aerodynamic lift and aerodynamic noise disciplines, the primary task of the system-level optimizer is to determine whether the current optimization result meets the preset convergence conditions. These convergence conditions typically include several aspects: First, the change in the system-level objective function value. If, in several consecutive iterations, the fluctuation range of the objective function value is less than a very small threshold (e.g., 1e-6), then the objective function is considered to have stabilized. Second, the differences between the design variable combinations fed back from the disciplines. If there are overlapping variables (i.e., variables that affect both disciplines) among the design variables optimized from the aerodynamic lift and aerodynamic noise disciplines, then the differences in the values ​​of these overlapping variables need to be controlled within an acceptable engineering range, for example, their relative error should not exceed 5%. Furthermore, whether the performance values ​​fed back from each discipline meet the initial target requirements set by the system-level optimizer is also one of the judgment criteria, i.e., whether the aerodynamic lift value reaches or approaches the target value, and whether the aerodynamic noise value is controlled below the target value. If all the above convergence conditions are met, the system-level optimizer considers the current optimization result to be within the preset convergence range. An acceptable co-optimization solution has been found, and this solution (including the specific combination of design variables, the corresponding aerodynamic lift values, and aerodynamic noise values) has been recorded. If the convergence condition is not met, the system-level optimizer will adjust the weight coefficients according to the preset step size or strategy. The value can be set, for example, starting from 0.001 and gradually increasing to 0.01 in steps of 0.001, or a more intelligent adaptive adjustment strategy can be adopted to determine the weight coefficients based on the changing trend of the current objective function. The direction and magnitude of the adjustment. After the adjustment is complete, the system-level optimizer will assign the new weight coefficients. The target values ​​for aerodynamic lift and aerodynamic noise are then passed back to the respective subject-level optimizers, repeating steps S702 to S705. This process is iterated until all preset weight coefficients have been traversed. All values ​​within the range (or reaching the maximum number of iterations). By using different weighting coefficients... Through optimization, the system-level optimizer ultimately obtains a series of co-optimized solutions corresponding to different lift-noise tradeoffs. These solutions collectively constitute a co-optimized solution set for aerodynamic lift and aerodynamic noise. This solution set provides designers with a rich selection of options, allowing them to choose the most suitable propeller design scheme based on actual engineering needs and design preferences.

[0036] As an optional embodiment of the present invention, optionally, the expression of the system-level optimizer in step S701 is: in, This represents the system-level optimization objective function value. Indicates the weighting coefficient. This represents the system-level aerodynamic lift target value. Indicates aerodynamic lift. This indicates the maximum aerodynamic lift. This represents the minimum aerodynamic lift. This represents the optimal value for aerodynamic noise. This represents the target value for the system-level aerodynamic noise sound pressure level. This indicates the maximum aerodynamic noise level. This indicates the minimum aerodynamic noise level.

[0037] As an optional embodiment of the present invention, the consistency constraint in step S702 may be relaxed by introducing a relaxation factor to improve convergence.

[0038] As an optional embodiment of the present invention, the weighting coefficient may optionally range from 0.001 to 0.01.

[0039] As an optional embodiment of the present invention, the design parameters in step S1 may include at least four of the following: blade diameter, blade mounting angle, chord length distribution, relative airfoil thickness, airfoil camber, and leading edge radius.

[0040] As an optional embodiment of the present invention, in step S5, the prediction accuracy of the aerodynamic lift Kriging approximation model and the aerodynamic noise Kriging approximation model are evaluated by cross-validation. When the root mean square error of the corresponding Kriging approximation model is less than a preset threshold, the model is deemed to be qualified. If the error exceeds the threshold, sample data is added and the model is retrained until the preset accuracy is met.

[0041] As an optional embodiment of the present invention, the method may further include: S8. Based on the co-optimized solution set of aerodynamic lift and aerodynamic noise, perform CFD numerical simulation verification, compare the error between the optimization result and the predicted value of the approximate model. If the error is within the preset range, the optimized solution set is determined to be valid; if the error exceeds the preset range, return to step S4, adjust the sampling strategy or increase the number of samples and reconstruct the Kriging approximate model and optimize it until the verification is passed.

[0042] Parametric modeling and variable sensitivity analysis; Geometric Model and Mesh: Taking the two-bladed propeller of a certain type of electric vertical takeoff and landing (EVTOL) aircraft as the optimization object, a parametric model was established using 3D CAD software. The computational domain was cylindrical, with a diameter of 5 times the propeller diameter and a length of 10 times the propeller diameter. The propeller surface was meshed using triangular facets, with local refinement at the leading edge, trailing edge, and tip regions. The first layer of mesh thickness was 0.01 mm to ensure y+≈1. The computational domain volume mesh used a polyhedral mesh, with a mesh size of 0.02 m in the rotating domain and 0.05 m in the stationary domain. The final total number of meshes was approximately 8.5 million, and mesh independence was verified (when the number of meshes increased from 6.5 million to 8.5 million, the changes in predicted lift and noise were both less than 1%).

[0043] Design variables were determined as follows: Four key geometric parameters that significantly affect propeller performance were selected as initial design variables: blade diameter (D), blade installation angle (β), chord length (c), and maximum relative thickness of the airfoil (t / c). Their baseline values ​​and optimization ranges are shown in Table 1.

[0044] Table 1 Design Variables and Initial Range Orthogonal Experiments and Simulations: Nine sets of experiments were arranged using an L9(34) orthogonal array. ANSA software was used for mesh deformation operations. Based on the orthogonal array parameters, the basic mesh was automatically deformed using scripts to generate nine sample models. In STAR-CCM+, the incoming flow velocity was set to 65 m / s (corresponding to cruise state), and the rotation speed was set to 2200 rpm. First, steady-state calculations (SSTk-ω turbulence model) were performed to obtain the aerodynamic lift force F. Then, transient calculations (DES model, with a time step corresponding to a 1° rotation) were performed. Using the FW-H acoustic analogy equation, the A-weighted total sound pressure level at the monitoring point 1.5 m away from the rotating plane was calculated as the noise target SPL.

[0045] Sensitivity analysis: Range analysis was performed on the results of the nine experimental groups. The average values ​​of aerodynamic lift and noise at different levels were calculated, and the range R was calculated. For ease of comparison, the range was normalized to obtain the sensitivity contribution S. The analysis results show that: The contribution of blade diameter (D) to aerodynamic lift F is 58.3%, chord length (c) is 28.7%, installation angle (β) is 11.2%, and relative thickness (t / c) has a small impact (1.8%).

[0046] For aerodynamic noise N, all four variables have a significant impact, with the following contributions: chord length (c) 38.5%, relative thickness (t / c) 29.2%, blade diameter (D) 19.8%, and installation angle (β) 12.5%.

[0047] Variable subset partitioning: Based on this, the variable subset XF={D,β,c} for aerodynamic lift is defined, and the variable subset XSPL={β,c,t / c} for noise is defined.

[0048] Single-discipline agent model construction and performance boundary exploration: DOE Sampling and Computation: Aerodynamic lift discipline: Within the 3D design space of XF, 40 sample points are generated using optimal Latin hypercube sampling.

[0049] Noise Science: Within the 3D design space of XSPL, 40 sample points are generated using optimal Latin hypercube sampling.

[0050] Similarly, ANSA was used to perform mesh deformation and STAR-CCM+ simulations to obtain the aerodynamic lift and aerodynamic noise values ​​for all samples.

[0051] Kriging Model Construction and Validation: Using the vaeopy intelligent optimization toolbox, a thrust model (ModelF) and a noise model (ModelSPL) were constructed based on 40 sets of data. Cross-validation showed that ModelF had a coefficient of determination (R²) of 0.932, and ModelSPL had an R² of 0.915, meeting engineering requirements.

[0052] Single-objective optimization: The surrogate model was optimized using the multi-island genetic algorithm component called by the vaeopy intelligent optimization toolbox. Algorithm parameters were set as follows: 10 islands, 20 individuals per island, 100 generations of genetic inheritance, crossover rate of 0.9, and mutation rate of 0.01. The optimization results determined the performance boundary. Lift range: Fmin=1250N, Fmax=1850N.

[0053] Sound pressure level range: SPLmin=88.3dB, SPLmax=96.7dB.

[0054] Solving for collaborative optimization strategies; Framework Setup: A collaborative optimization framework was built using the vaeopy intelligent optimization toolbox. Multi-island genetic algorithms were used as optimizers at both the system-level and the two subject-level levels.

[0055] Parameter settings: Weighting factor: To seek a balanced design, let w1=w2=0.5.

[0056] Relaxation factor: ε = 0.005.

[0057] System-level objective function: Optimization Execution and Results: After 32 iterations at the system and subject levels, the optimization process was executed, and the algorithm converged. The optimal solution at the system level was obtained, with the following specific parameter values ​​mapped back: blade diameter 1.25m, installation angle 23.8°, chord length 0.089m, and relative thickness 0.083. At this point, the surrogate model predicted a thrust of 1685N and a sound pressure level of 89.7dB.

[0058] Final design verification; Substituting the optimized parameters, the original mesh is automatically deformed using ANSA to generate the final design model. Transient CFD+CAA simulation verification is then performed in STAR-CCM+ with the same settings as in S1.

[0059] Results comparison (optimized design vs. original design): Aerodynamic lift: 1715N vs 1520N, an increase of 12.8%.

[0060] Weighted sound pressure level at far-field monitoring point A: 90.2dB vs 93.5dB, a decrease of 3.3dB.

[0061] Flow field analysis: The results show that the tip vortex intensity of the optimized model is significantly reduced, the pressure distribution on the propeller disk surface is more uniform, and flow separation is effectively suppressed. This is the main reason for the increase in thrust and the reduction in noise.

[0062] Sound field analysis: The sound pressure level spectrum shows that the optimized design has a significant reduction in sound pressure level in the low-to-mid frequency range of 200-800Hz and the high-frequency range of 2000-5000Hz, which is consistent with the analysis results of reduced vortex intensity and smoother pressure pulsation in the flow field.

[0063] The above embodiments fully demonstrate the effectiveness, advancement, and engineering practical value of the method of the present invention. The present invention is not only applicable to the optimization design of aircraft propellers, but its core ideas can also be extended to the optimization design of other fluid machinery and heat exchange equipment with multidisciplinary, multi-objective, and variable coupling characteristics.

[0064] Example 2 An aircraft propeller design system based on sensitivity analysis and intelligent multi-objective collaborative optimization includes: processor; Memory used to store processor-executable instructions; The processor is configured to implement an aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization when executing executable instructions.

[0065] It should be noted that the computer device includes a processor, a memory, and may also include one or more of a multimedia component, an input / output (I / O) interface, and a communication component.

[0066] The processor controls the overall operation of the computer device to complete all or part of the steps in the above-mentioned aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization.

[0067] Memory is used to store various types of data to support the operation of the computer device. This data may include, for example, instructions for any application or method used to operate on the computer device, as well as application-related data. Memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0068] The multimedia component may include a screen and an audio component, wherein the screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals; for example, the audio component may include a microphone for receiving external audio signals, the received audio signals may be further stored in memory or transmitted via a communication component; the audio component may also include at least one speaker for outputting audio signals.

[0069] I / O interfaces provide interfaces between the processor and other interface modules, such as keyboards, mice, buttons, etc.; these buttons can be virtual buttons or physical buttons.

[0070] The communication component is used for wired or wireless communication between the computer device and other devices; wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G or 5G, or one or more combinations thereof, and the corresponding communication component may include: Wi-Fi module, Bluetooth module, NFC module, mobile communication module.

[0071] As a preferred embodiment, the computer device may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the aforementioned aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization.

[0072] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization, characterized in that, The method includes: The design parameters were obtained by sampling various points on the propeller blades under different conditions. Based on the design parameters, computational fluid dynamics and computational aeroacoustics numerical simulations were performed to obtain the aerodynamic lift and aerodynamic noise values ​​at each sample point. Based on the aerodynamic lift and aerodynamic noise values, the range analysis method is used to calculate the sensitivity of each design parameter to the target values ​​of aerodynamic lift and aerodynamic noise, and based on the sensitivity, a subset of design variables strongly correlated with aerodynamic lift and a subset of design variables strongly correlated with aerodynamic noise are identified. Based on the subset of design variables strongly correlated with aerodynamic lift and the subset of design variables strongly correlated with aerodynamic noise, the optimal Latin hypercube method is used for sampling, and then the sample geometric model is generated by the automatic mesh deformation method and numerical simulation is performed to obtain sample data. Based on the sample data, Kriging approximate models of aerodynamic lift and aerodynamic noise with respect to their respective design variables are constructed to obtain the aerodynamic lift Kriging approximate model and the aerodynamic noise Kriging approximate model. Both the aerodynamic lift Kriging approximation model and the aerodynamic noise Kriging approximation model are optimized using a multi-island genetic algorithm to obtain the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise. Based on the maximum and minimum values ​​of aerodynamic lift and aerodynamic noise, a two-layer optimization framework at the system and discipline levels is constructed to obtain a collaborative optimization solution set for aerodynamic lift and aerodynamic noise.

2. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization of claim 1, wherein, A two-tiered optimization framework at both the system and discipline levels is constructed to obtain a synergistic optimization solution set for aerodynamic lift and aerodynamic noise, including: The system-level optimizer uses a weighted normalized function that maximizes aerodynamic lift and minimizes aerodynamic noise as its optimization objective. It sets weight coefficients for aerodynamic lift and aerodynamic noise and balances the importance of the two objectives by adjusting the weight coefficients. The system-level optimizer uses the aerodynamic lift target value and aerodynamic noise target value under the current weight coefficients as consistency constraints for discipline-level optimization, and passes them to the aerodynamic lift discipline optimizer and the aerodynamic noise discipline optimizer respectively. The aerodynamic lift optimization tool is based on the Kriging approximation model of aerodynamic lift. Under the constraint of the aerodynamic lift target value transmitted at the system level, it optimizes a subset of design variables that are strongly related to aerodynamic lift and seeks to satisfy the optimal combination of design variables. The aerodynamic noise optimization tool is based on the Kriging approximation model of aerodynamic noise. Under the constraint of the aerodynamic noise target value transmitted at the system level, it optimizes a subset of design variables that are strongly correlated with aerodynamic noise and seeks to satisfy the optimal combination of design variables. The aerodynamic lift optimizer and the aerodynamic noise optimizer feed back the design variable combinations and corresponding aerodynamic lift and aerodynamic noise values ​​obtained by their respective optimizations to the system-level optimizer. The system-level optimizer determines whether the preset convergence conditions are met based on the feedback results. If they are met, it outputs the co-optimization solution under the current weight coefficients. If they are not met, it adjusts the weight coefficients until it traverses the preset weight coefficient range, and finally obtains the co-optimization solution set of aerodynamic lift and aerodynamic noise.

3. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization of claim 2, wherein, The expression for the system-level optimizer is: in, This represents the value of the system-level optimization objective function. Indicates the weighting coefficient. This represents the system-level aerodynamic lift target value. Indicates aerodynamic lift. This indicates the maximum aerodynamic lift. This represents the minimum aerodynamic lift. This represents the optimal value for aerodynamic noise. This represents the target value for the system-level aerodynamic noise sound pressure level. This indicates the maximum aerodynamic noise level. This indicates the minimum aerodynamic noise level.

4. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization as described in claim 2, characterized in that, The consistency constraint is relaxed by introducing a relaxation factor to improve convergence.

5. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization as described in claim 2, characterized in that, The weighting coefficient ranges from 0.001 to 0.

01.

6. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization as described in claim 1, characterized in that, The design parameters include at least four of the following: blade diameter, blade installation angle, chord length distribution, relative airfoil thickness, airfoil camber, and leading edge radius.

7. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization as described in claim 1, characterized in that, The prediction accuracy of the aerodynamic lift Kriging approximation model and the aerodynamic noise Kriging approximation model was evaluated by cross-validation. When the root mean square error of the corresponding Kriging approximation model is less than a preset threshold, the model is deemed to be qualified. If the error exceeds the threshold, the model will be retrained with additional sample data until the preset accuracy is met.

8. The aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization as described in claim 1, characterized in that, The method further includes: CFD numerical simulations were performed to verify the optimized solution set of aerodynamic lift and aerodynamic noise. The error between the optimization results and the predicted values ​​of the approximate model was compared. If the error was within a preset range, the optimized solution set was determined to be valid. If the error exceeded the preset range, the sampling strategy was adjusted or the number of samples was increased before reconstructing the Kriging approximate model and optimizing it until the verification was successful.

9. An aircraft propeller design system based on sensitivity analysis and intelligent multi-objective collaborative optimization, characterized in that, The system includes: processor; Memory used to store processor-executable instructions; The processor is configured to implement the aircraft propeller design method based on sensitivity analysis and intelligent multi-objective collaborative optimization as described in any one of claims 1 to 8 when executing the executable instructions.

Images