APU rotor assembly balancing method and system

Abstract

This application discloses an APU rotor assembly balancing method and system, belonging to the field of aerospace technology. It includes: determining initial sample points based on initial imbalance test data; optimizing the initial sample points to determine test results under recommended assembly angles; if the requirements are not met, constructing expanded sample points by combining current and historical test data; estimating the probability density of the expanded sample points to obtain updated sample points with the highest probability density; and iteratively optimizing based on the updated sample points until the balancing requirements are met. This application, by constructing expanded sample points and using probability density estimation to screen optimal parameters, achieves dynamic and accurate correction of component imbalance, effectively overcoming the shortcomings of existing technologies that rely on predicted quantities or blind trial and error, and improving parameter estimation accuracy and convergence speed.

Description

Technical Field

[0001] This application relates to the field of aerospace technology, and in particular to a method and system for balancing an APU rotor assembly. Background Technology

[0002] The APU (Auxiliary Power Unit) is a crucial auxiliary power source on an aircraft, and the dynamic balance of its rotor assembly directly affects flight safety and equipment lifespan. An APU rotor assembly is typically assembled from multiple toothed parts that mesh and are axially tensioned. According to maintenance manuals, rotor balancing is only permitted to reduce residual imbalance by rearranging the assembly angles of the components within the rotor assembly; traditional balancing methods such as adjusting blade sequence, adding counterweights, or grinding parts are strictly prohibited.

[0003] In existing technologies, due to the large number of parts constituting the rotor (e.g., a certain type of APU rotor consists of 13 parts), the imbalance of each part is random, resulting in a complex distribution of the overall imbalance after assembly. Traditional balancing methods mainly rely on the operator's experience for repeated disassembly and reassembly and trial and error, which is highly unreliable, inefficient, and makes it difficult to effectively utilize test data. Although existing technologies attempt to introduce genetic algorithms, particle swarm optimization, or gradient descent for optimization, all have significant drawbacks: for example, genetic algorithms require pre-measuring a large number of individual component parameters, which is time-consuming and subject to equipment conversion errors; particle swarm optimization, based on differences, lacks convergence capability for errors; and gradient descent is greatly affected by the learning rate parameter, easily getting trapped in local optima or producing fluctuations.

[0004] In summary, existing technologies suffer from problems such as low efficiency in the balancing process, reliance on experience or the need to obtain precise individual parameters in advance, and difficulty in effectively utilizing limited test data during the iteration process to improve parameter estimation accuracy and achieve rapid convergence. Summary of the Invention

[0005] This application provides an APU rotor assembly balancing method and system to at least solve the problems of low efficiency in the balancing process, reliance on experience or the need to obtain accurate individual parameters in advance in the prior art, and difficulty in effectively utilizing limited test data during the iteration process to improve parameter estimation accuracy and achieve fast convergence.

[0006] In a first aspect, this application provides an APU rotor assembly balancing method, the method comprising: Based on the acquired initial unbalance test data, initial sample points are determined. The initial unbalance test data is used to represent the remaining unbalance vector of the APU rotor assembly on the support surface under different assembly angle combinations. The initial sample points are used to represent the parameter estimate of the unbalance of each component. Based on the optimization of the initial sample points, the test results under the recommended assembly angle are determined; When the test results do not meet the preset balance requirements, an amplified sample point is constructed based on the current test data and historical test data. The current test data is used to indicate the data obtained in the most recent assembly and balance test, and the historical test data is used to indicate the test dataset accumulated before the current test. The probability density of the amplified sample points is estimated to obtain updated sample points; Based on the optimization of the updated sample points, the test results under the latest recommended assembly angle are determined until the test results that meet the preset balance requirements are obtained. The updated sample points correspond to the points with the highest probability density of the amplified sample points.

[0007] The above technical solution constructs a parameter estimation model for the imbalance of each component and accumulates historical test data to build expanded sample points during the iteration process. It uses a probability density estimation algorithm to select the point with the highest probability density as the updated sample point, thereby achieving dynamic and accurate correction and optimization of the imbalance parameters of the components. This method effectively overcomes the shortcomings of existing technologies that rely on pre-measuring individual component parameters or blind trial and error. By fully exploring the value of previous test data, it suppresses the influence of measurement errors and significantly improves the accuracy and convergence speed of parameter estimation, thus achieving efficient and accurate balancing of the APU rotor assembly.

[0008] Secondly, this application provides an APU rotor assembly balancing system, the system comprising: The initial sample point determination module is used to determine the initial sample point based on the acquired initial unbalance test data. The initial unbalance test data is used to represent the remaining unbalance vector of the APU rotor assembly on the support surface under different assembly angle combinations. The initial sample point is used to represent the parameter estimate of the unbalance of each part. The optimization module is used to optimize based on the initial sample points and determine the test results under the recommended assembly angle; The amplified sample point construction module is used to construct amplified sample points based on the current test data and historical test data when the test results do not meet the preset balance requirements. The current test data is used to indicate the data obtained in the most recent assembly and balance test, and the historical test data is used to indicate the test dataset accumulated before the current test. The sample point update module is used to estimate the probability density of the amplified sample points, obtain updated sample points, and feed the updated sample points back to the optimization module for iterative optimization until a test result that meets the preset balance requirements is obtained.

[0009] Thirdly, this application provides an apparatus comprising one or more processors and one or more memories, wherein at least one piece of program code is stored in the one or more memories, the program code being loaded and executed by the one or more processors to implement the operations performed by the APU rotor assembly balancing method.

[0010] Fourthly, this application also provides a computer-readable storage medium storing at least one piece of program code, which is loaded and executed by a processor to implement the operations performed by the APU rotor assembly balancing method.

[0011] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described APU rotor assembly balancing methods. Attached Figure Description

[0012] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0013] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 1 ; Figure 2 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 2 ; Figure 3 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 3 ; Figure 4 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 4 ; Figure 5 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 5 ; Figure 6 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 6 ; Figure 7A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 7 ; Figure 8 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 8 ; Figure 9 A flowchart illustrating an APU rotor assembly balancing method provided in this application embodiment. Figure 9 ; Figure 10 A schematic diagram of the structure of an APU rotor assembly balancing system provided in this application embodiment; Figure 11 This is a schematic diagram of the structure of a device provided in an embodiment of this application; Figure 12 This application provides a schematic diagram illustrating the generation of estimated values ​​from sample values ​​in an embodiment of the present application. Figure 13 A schematic diagram of the overall calculation process for an APU rotor assembly balancing method provided in this application embodiment; Figure 14 A flowchart of a particle swarm optimization algorithm is provided for an embodiment of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0016] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0017] The APU (Auxiliary Power Unit) is a crucial auxiliary power source on an aircraft, and the dynamic balance of its rotor assembly directly affects flight safety and equipment lifespan. An APU rotor assembly is typically assembled from multiple toothed parts that mesh and are axially tensioned. According to maintenance manuals, rotor balancing is only permitted to reduce residual imbalance by rearranging the assembly angles of the components within the rotor assembly; traditional balancing methods such as adjusting blade sequence, adding counterweights, or grinding parts are strictly prohibited.

[0018] In related technologies, due to the large number of parts constituting the rotor (e.g., a certain type of APU rotor consists of 13 parts), the imbalance of each part is random, resulting in a complex distribution of the overall imbalance after assembly. Traditional balancing methods mainly rely on the operator's experience for repeated disassembly and reassembly and trial and error, which is highly unreliable, inefficient, and makes it difficult to effectively utilize test data. Although existing technologies attempt to introduce genetic algorithms, particle swarm optimization, or gradient descent methods for optimization, all have significant drawbacks. A review of relevant literature and patent information reveals the following specific problems: In May 2022, Shenyang Aerospace University published a paper titled "Research on APU Rotor Balancing Optimization Scheme Based on Genetic Algorithm". This method determines the impact of the unbalance parameters of a single part on the component parameters by pre-testing the unbalance of a single part and converting the support dimensions between different balancing machines. This method requires pre-measuring most parts, which consumes some time. At the same time, changing the equipment during testing will cause a large error. This method does not have a specific solution.

[0019] In December 2023, China Southern Airlines Shenyang Maintenance Base published a paper titled "Application of Particle Swarm Optimization Algorithm in APU Rotor Balancing Process". This method determines the unbalance parameter of a part by rotating a single part and using the difference between two measurements. However, it lacks convergence effect on unbalance fluctuations caused by measurement errors and the part's own errors.

[0020] In April 2024, China Southern Airlines Co., Ltd. applied for an invention patent, application number 202410495665.8, entitled "A method, device, equipment and medium for balancing APU rotor assembly". This method estimates the parameters of each component by using the gradient descent method based on the difference between the previous test data and the current test data. However, the estimation of component parameters by this method is greatly affected by the learning rate parameter, which is prone to fluctuations and cannot make full use of past data to achieve effective convergence of errors.

[0021] In summary, existing technologies suffer from problems such as low efficiency in the balancing process, reliance on experience or the need to obtain precise individual parameters in advance, and difficulty in effectively utilizing limited test data during the iteration process to improve parameter estimation accuracy and achieve rapid convergence.

[0022] To address the aforementioned technical problems, this application provides an APU rotor assembly balancing method. This method determines initial sample points by establishing a mapping relationship model, constructs expanded sample points based on iteratively accumulated test data, and uses probability density estimation to select the point with the highest probability density as the updated sample point. This solves the technical problems of existing technologies that rely on pre-measuring individual component parameters, suffer from low efficiency due to blind trial and error, and find it difficult to effectively utilize previous test data to improve parameter estimation accuracy and convergence speed.

[0023] The technical solutions provided in the embodiments of this application are described below. (See also...) Figure 1 Taking the APU rotor assembly balancing device, such as an industrial computer or a dedicated balancing computing terminal processor, as an example, the method includes the following steps.

[0024] Step S101: Determine the initial sample points based on the obtained initial imbalance test data.

[0025] The initial unbalance test data represents the residual unbalance vector of the APU rotor assembly on the support surface under different assembly angle combinations, while the initial sample points represent the parameter estimates of the unbalance of each component. Specifically, the initial test data is first read. This test data is typically obtained by randomly selecting several different assembly angle combinations, performing trial assembly, and measuring using a dynamic balancing machine. The test data includes the residual unbalance vector of the APU rotor assembly on two support surfaces, such as the front and rear support surfaces, under each assembly angle combination, including its amplitude and phase. Subsequently, based on the actual assembly structure of the APU rotor assembly, the transmission coefficient of the unbalance of each component to the support surface is determined. According to the principle of vector superposition, combining different assembly angle combinations, transmission coefficients, and the overall residual unbalance vector of the assembly, a mapping relationship model is established, such as constructing a system of linear equations. Since the initial test data usually contains multiple sets of measurement results, the constructed system of linear equations is an overdetermined system. The processor uses the least squares method to solve this system of linear equations, calculating the initial estimate of the unbalance of each component, which is then used as the initial sample points. The initial sample points reflect the optimal estimation results of the imbalance of each part under the current limited amount of data.

[0026] Step S102: Based on the initial sample points, optimize and determine the test results under the recommended assembly angle.

[0027] Specifically, the processor inputs the initial sample points obtained in step S101 as known parameters into the optimization algorithm module. Using the assembly angles of each part as variables and minimizing the overall residual imbalance modulus of the assembly as the objective, an objective function is constructed. An optimization algorithm, such as a genetic algorithm, particle swarm optimization, or simulated annealing, is used to search the angle solution space to find the combination of assembly angles that minimizes the objective function value; this is then considered the recommended assembly angles. Subsequently, the operator assembles the APU rotor assembly according to these recommended assembly angles and performs physical testing on a balancing machine. The actual residual imbalance vector of the assembly in this state is measured and used as the test result.

[0028] Step S103: When the test results do not meet the preset balance requirements, construct amplified sample points based on the current test data and historical test data.

[0029] The current test data indicates the data obtained from the most recent assembly and balancing test, while the historical test data indicates the test dataset accumulated before the current test. Specifically, the processor compares the residual unbalance magnitude value in the test result obtained in step S102 with a preset balancing threshold. If the test result meets the requirements, the process ends; otherwise, a sample expansion mechanism is triggered. The assembly angle combination and the corresponding residual unbalance vector data obtained from the current test are added to the historical test dataset, thereby updating and expanding the test dataset. Based on the updated expanded test dataset, the vector superposition equation system is reconstructed and solved again using the least squares method to redetermine the parameter estimates of the unbalance of each part. Due to the increase in data volume, the solution process may generate a set of solutions that satisfy the constraints, or generate a sample set through multiple solutions, thereby constructing expanded sample points. Compared with the initial sample points, the expanded sample points utilize more measured information and theoretically have higher estimation accuracy.

[0030] Step S104: Estimate the probability density of the amplified sample points to obtain updated sample points.

[0031] Specifically, considering the existence of measurement errors, component manufacturing errors, and calculation errors, the amplified sample points are not isolated points in the parameter space, but rather exhibit certain probability distribution characteristics. Probability density estimation methods (such as kernel density estimation (KDE) or Gaussian mixture model (GMM)) are used to statistically analyze the distribution of the amplified sample points in the parameter space, generating a probability density distribution function. By analyzing this probability density distribution function, the point with the highest probability density is found. This point with the highest probability density represents the most likely true value location of the imbalance of each component under all current accumulated data constraints. The parameter value corresponding to this point is extracted and used as the updated sample point. This step effectively eliminates outliers with large errors, improving the robustness of parameter estimation.

[0032] Step S105: Based on the updated sample points, optimize and determine the test results under the latest recommended assembly angle until the test results that meet the preset balance requirements are obtained.

[0033] Among them, the updated sample point corresponds to the point with the highest probability density of the amplified sample point.

[0034] Specifically, the processor substitutes the updated sample points, filtered by probability density estimation (i.e., the corrected part imbalance parameters), into the optimization algorithm to reconstruct the objective function and perform optimization calculations to obtain a new theoretically optimal assembly angle. Assembly and testing are performed according to this angle to obtain the latest test results. If the results meet the preset balance requirements, the current assembly scheme is output; otherwise, the iterative process from steps S103 to S105 is repeated. By continuously accumulating test data and correcting parameter estimates, the actual part imbalance is gradually approximated until the balance standard is finally reached.

[0035] This embodiment solves for the initial sample points by constructing a mapping relationship model, determines the recommended assembly angle using an optimization algorithm, and accumulates historical test data to construct expanded sample points during the iteration process. A probability density estimation algorithm is then used to select the points with the highest probability density as the updated sample points. This technique achieves dynamic and precise correction and optimization of the unbalance parameters of the parts. This method effectively overcomes the shortcomings of existing technologies that rely on pre-measuring individual component parameters or blind trial and error. By fully utilizing the value of previous test data, it suppresses the influence of measurement errors, solves the problem of difficult parameter estimation convergence, and improves the accuracy and convergence speed of parameter estimation, thereby achieving efficient and precise balancing of the APU rotor assembly.

[0036] It should be noted that the above steps S101-S105 are a simplified description of the embodiments provided in this application.

[0037] The following will provide a more detailed explanation of how the embodiments of this application determine the initial sample points based on the acquired initial imbalance test data, using some examples. (See also...) Figure 2 It includes the following steps: Step S201: Based on the obtained initial imbalance test data, establish a mapping relationship model.

[0038] The mapping model is used to represent the vector superposition relationship between the unbalance of each part and the overall remaining unbalance of the assembly; the initial unbalance test data is used to represent the vector of the remaining unbalance of the APU rotor assembly on the support surface under different assembly angle combinations.

[0039] Specifically, the initial test data is first read. This data is typically obtained by randomly selecting several different combinations of assembly angles, performing trial assembly, and measuring using a dynamic balancing machine. The test data includes the vector of the remaining unbalance of the rotor assembly on two support surfaces, such as the front and rear support surfaces, under each combination of assembly angles. Subsequently, based on the actual assembly structure of the APU rotor assembly, such as the axial mounting position of each component and the end tooth meshing angle relationship, the transmission coefficient of the unbalance of each component to the support surface is determined. According to the principle of vector superposition, the unbalance of each component, after being rotated by the assembly angle and mapped by the transmission coefficient, is vector-superimposed on the support surface to form the overall remaining unbalance of the assembly. Based on this, a set of mathematical equations characterizing the relationship between the unbalance of each component and the overall remaining unbalance of the assembly is established as a mapping relationship model.

[0040] Step S202: Determine the initial sample points based on the mapping relationship model.

[0041] Specifically, a system of linear equations is constructed based on a mapping relationship model. This system of linear equations indicates the vector superposition relationship between the unbalance of each component and the overall remaining unbalance of the assembly. In this system of equations, the unknowns are the unbalance vectors of each component, and the knowns are the overall remaining unbalance of the assembly and the coefficient matrix composed of assembly angles and transfer coefficients. Since multiple trial assembly tests are usually conducted in the initial stage, the number of equations in the constructed system of linear equations is often greater than the number of unknowns, forming an overdetermined system of equations. The least squares method is used to solve this overdetermined system of equations to calculate the estimated value of the unbalance of each component, which is then used as the initial sample points. These initial sample points reflect the optimal estimation results of the unbalance of each component under the current limited amount of data, providing a parameter basis for the subsequent optimization process.

[0042] This embodiment achieves the function of inverting the imbalance of each part from the overall imbalance of the component by establishing a mapping relationship model and using the least squares method to solve the initial sample points. It solves the technical problems of traditional methods that require pre-measuring individual parameters, which is time-consuming and labor-intensive and has conversion errors. It achieves the beneficial effects of simplifying the process, reducing measurement costs, and providing an accurate parameter basis for subsequent optimization.

[0043] In some embodiments, see Figure 3 The step of optimizing based on the initial sample points to determine the test results under the recommended assembly angle includes: Step S301: Based on the initial sample points, determine the recommended assembly angle using an optimization algorithm.

[0044] Specifically, the initial sample points determined in step S202 are used as known parameters input into the optimization algorithm. First, the vector parameters of the imbalance of each part are determined based on the initial sample points. Then, with the assembly angle of each part as the optimization variable and minimizing the magnitude of the overall remaining imbalance of the assembly as the objective, an objective function is constructed. Using an optimization algorithm, such as a genetic algorithm, particle swarm optimization, or simulated annealing, a global search is performed in the angle solution space to find a set of assembly angles that minimizes the objective function value, which is then used as the recommended assembly angle. This recommended assembly angle is theoretically the optimal solution that minimizes the remaining imbalance of the rotor assembly.

[0045] Step S302: Based on the recommended assembly angle, perform assembly and balancing tests on the rotor assembly to obtain the test results.

[0046] Specifically, based on the recommended assembly angle calculated in step S301, the operator assembles each part of the APU rotor assembly to ensure that the relative angular position of each part is consistent with the recommended assembly angle. After assembly, the rotor assembly is placed on a dynamic balancing machine for physical testing. The actual remaining unbalance vector of the assembly on the support surface in this state is measured as the test result. This test result is used to verify the actual balancing effect of the recommended assembly angle and serves as the basis for determining whether the preset balancing requirements are met or for proceeding to the next iteration.

[0047] This embodiment constructs an objective function based on parameter estimation and performs optimization calculations. Combined with physical testing and verification techniques, it achieves the theoretical optimal solution and experimental feedback for the assembly angle. This solves the problem of low efficiency caused by relying on manual experience and blind trial and error in traditional methods. It achieves the beneficial effects of improving the first-time assembly success rate, reducing the number of disassembly and assembly operations, and lowering process costs.

[0048] In some embodiments, see Figure 4 When the test results do not meet the preset balance requirements, an amplified sample point is constructed based on the current test data and historical test data, including: Step S401: When the test results do not meet the preset balance requirements, add the current test data to the historical test data to obtain the amplified test dataset.

[0049] Specifically, the residual unbalance magnitude value in the test result obtained in step S102 is compared with a preset balance threshold, for example, the preset balance threshold is the maximum allowable unbalance specified in the maintenance manual. If the test result is greater than the preset balance threshold, it is determined that the preset balance requirement is not met, and the sample amplification mechanism is triggered. The assembly angle combination and the corresponding residual unbalance vector data (including amplitude and phase) obtained from the current test are appended and stored in the historical test dataset. This historical test dataset contains the initial test data and the data from each iteration test. The accumulation of data increases the sample size for subsequent parameter estimation, providing a data foundation for improving estimation accuracy.

[0050] Step S402: Based on the amplified test dataset, redetermine the parameter estimates of the self-imbalance of each part to obtain the amplified parameter estimate sample set, which is used as the amplified sample points.

[0051] Specifically, based on the updated and expanded test dataset, a new system of vector superposition equations is constructed. Due to the dataset expansion, the number of equations in the system increases further, and the degree of overdeterminacy is enhanced. The updated system of equations is solved using the least squares method to recalculate the parameter estimates of the imbalance of each component. To obtain a sample set for probabilistic statistics, resampling techniques, such as the Bootstrap method, can be used to sample and solve the expanded dataset multiple times, or to solve for the solution set under different constraints, thereby obtaining a set of parameter estimates that constitute the expanded parameter estimate sample set, i.e., the expanded sample points. The expanded sample points utilize more measured data information and are theoretically closer to the true values ​​than the initial sample points, and the form of the sample set can reflect the distribution characteristics of the parameters.

[0052] This embodiment achieves effective utilization of test data and dynamic updating of parameter estimation samples by accumulating historical test data and resolving the problem of inaccurate estimation due to error interference from a single or small amount of test data. It achieves the beneficial effects of improving parameter estimation accuracy, suppressing the influence of measurement error, and providing data support for subsequent probability density estimation.

[0053] In some embodiments, see Figure 5 The probability density of the amplified sample points is estimated to obtain updated sample points, including: Step S501: Estimate the probability density of the amplified sample points and determine the probability density distribution function.

[0054] Specifically, considering the existence of measurement errors, component manufacturing errors, and calculation errors, the amplified sample points are not isolated points in the parameter space, but rather exhibit certain probability distribution characteristics. Probability density estimation methods, such as kernel density estimation (KDE) or Gaussian mixture model (GMM), are used to statistically analyze the distribution of the amplified sample points in the parameter space. By smoothing the sample point set or fitting a model, a probability density distribution function that describes the distribution law of the unbalance parameters of each component is generated. This probability density distribution function intuitively reflects the likelihood of different parameter values ​​under the constraints of the current accumulated data.

[0055] Step S502: Based on the probability density distribution function, find the point of maximum probability density.

[0056] Specifically, an extremum analysis is performed on the generated probability density distribution function to search for the point with the maximum probability density value within the parameter space. According to statistical principles, the point with the maximum probability density represents the most likely true value location of the imbalance of each part under all current cumulative data constraints, i.e., the parameter combination with the highest probability of occurrence.

[0057] Step S503: Based on the point with the maximum probability density, obtain the updated sample points.

[0058] Specifically, the parameter coordinates corresponding to the point of maximum probability density are extracted, i.e., the amplitude and phase of the imbalance of each part, and the corresponding parameter coordinates are used as update sample points. Compared with directly taking the average value or the value of a single solution, the update sample points extracted based on the maximum probability density can effectively suppress the interference of random errors and eliminate outliers with large deviations, thereby achieving the optimal estimation of the imbalance parameters of the parts.

[0059] This embodiment achieves probabilistic statistical screening and optimal estimation of the unbalance parameters of parts by estimating the probability density of the amplified sample points and extracting the maximum points as updated sample points. It solves the technical problems of sample dispersion and difficulty in determining the true value due to measurement and calculation errors, and achieves the beneficial effects of effectively filtering noise interference, eliminating outliers, and significantly improving the robustness and accuracy of parameter estimation.

[0060] In some embodiments, see Figure 6 Based on the updated sample points, the test results under the latest recommended assembly angle are determined until test results that meet the preset balance requirements are obtained, including: Step S601: Optimize based on updated sample points to determine the target assembly angle.

[0061] The target assembly angle represents the assembly angle corresponding to the minimum theoretical residual imbalance. Specifically, the updated sample points, filtered by probability density estimation (i.e., the corrected imbalance parameters of each part), are substituted into the optimization algorithm. Using the assembly angle of each part as the optimization variable and minimizing the modulus of the overall residual imbalance of the assembly as the objective, the objective function is reconstructed. The optimization algorithm performs a global search in the angle solution space to calculate the set of assembly angle combinations that minimizes the objective function value, which is taken as the target assembly angle. This target assembly angle represents the theoretically optimal solution that minimizes the residual imbalance of the rotor assembly under the current corrected parameter model.

[0062] Step S602: Determine the latest recommended assembly angle based on the target assembly angle.

[0063] Specifically, the processor directly sets the target assembly angle determined in step S601 as the latest recommended assembly angle. This angle will serve as a guiding parameter for the next round of physical assembly, used to adjust the relative meshing positions of each part on the rotor assembly.

[0064] Step S603: Based on the latest recommended assembly angle, obtain the test results under the latest recommended assembly angle.

[0065] Specifically, the operator assembles and adjusts the APU rotor assembly according to the latest recommended assembly angle, and performs physical testing on a dynamic balancing machine to measure the actual remaining imbalance vector of the assembly in this state, which is taken as the test result under the latest recommended assembly angle. This test result is compared with the preset balance requirements. If the requirements are met, the current assembly scheme is output, and the process ends; if not, the process returns to step S103 and subsequent steps, continuously iterating and optimizing to gradually approach the actual imbalance of the parts until the final balance standard is achieved.

[0066] This embodiment achieves closed-loop iterative optimization of the balancing process by using the technique of optimization calculation based on updated sample points and guiding physical testing. It solves the problem that the optimization cannot meet the target in one step due to inaccurate initial parameter estimation, and achieves the beneficial effect of gradually approaching the true optimal solution and ensuring that the final assembly result meets the balancing quality requirements.

[0067] In some embodiments, see Figure 7 Based on the obtained initial imbalance test data, a mapping relationship model is established, including: Step S701: Based on the acquired initial imbalance test data, extract different assembly angle combinations and the corresponding overall remaining imbalance vector of the components.

[0068] Specifically, the test data file stored in the initial stage is read, and each set of test records is parsed out. Each set of records contains two key pieces of information: first, the combination of assembly angles for each part, i.e., the relative meshing positions of each part in the circumferential direction; and second, the vector of the remaining unbalance of the entire assembly on the two support surfaces, measured by the dynamic balancing machine under that angle combination. This information is organized into "input-output" data pairs as the basic data for modeling.

[0069] Step S702: Determine the transmission coefficient of the unbalance of each part according to the assembly structure of the APU rotor assembly.

[0070] Specifically, based on the design drawings or 3D model of the APU rotor assembly, geometric parameters such as the axial mounting position, mass distribution, and span of the front and rear support surfaces of each component are obtained. Based on rotor dynamics principles or finite element analysis, the vibration response of each component's unit unbalance to the front and rear support surfaces is calculated; this is known as the transmission coefficient or influence coefficient. This transmission coefficient reflects the ratio and phase shift relationship between the component's unbalance and the overall unbalance of the assembly.

[0071] Step S703: Based on the principle of vector superposition, and combining different assembly angle combinations, transmission coefficients, and the overall remaining unbalance vector of the component, establish a mapping relationship model.

[0072] Specifically, a mapping model is constructed based on the principle of vector superposition in mechanics. For each assembly test, the processor treats the unbalance of each part as a vector with direction and magnitude. First, the actual direction of the unbalance vector in space is determined based on the assembly angle of the part. Second, the vector is corrected using a transfer coefficient to simulate the magnitude change and phase shift of the unbalance when it is transmitted to the support surface. Finally, the corrected vectors of all parts are synthesized. A logical relationship is established: the synthesized total vector should be consistent with the overall remaining unbalance vector of the assembly obtained from the test. This logical relationship, which links the parameters of each part with the overall test results of the assembly, constitutes the mapping model, providing a mathematical basis for the subsequent reverse calculation of the unbalance of each part.

[0073] This embodiment uses the technical means of extracting test data, calculating transfer coefficients and establishing a mapping relationship model by combining the principle of vector superposition to realize the mathematical description of the complex assembly balance relationship of APU rotor assembly. It solves the problem of fuzzy and difficult-to-quantify relationship between part parameters and overall assembly performance, and achieves the beneficial effect of providing a mathematical model basis for subsequent use of the least squares method to solve part parameters.

[0074] In some embodiments, see Figure 8 The determination of initial sample points based on the mapping relationship model includes: Step S801: Construct a system of linear equations based on the mapping relationship model.

[0075] The linear equations are used to indicate the vector superposition relationship between the unbalance of each component and the overall remaining unbalance of the assembly; specifically, the established vector superposition relationship is transformed into a specific algebraic form. For each set of assembly data obtained in the initial testing phase, a linear equation describing the relationship between the vector of unbalance of each component and the vector of the overall remaining unbalance of the assembly is generated based on the assembly angle and transfer coefficient at that time. Since multiple trial assembly tests with different angle combinations are usually conducted in the initial phase, the equations corresponding to all test data are combined to construct a linear equation system containing multiple equations. In this equation system, the known quantity is the overall remaining unbalance of the assembly measured in each test, and the unknown quantity is the original unbalance of each component.

[0076] Step S802: Solve the linear equation system using the least squares method to determine the estimated value of the unbalance of each part.

[0077] Specifically, because test data inevitably contains measurement errors and random interference, the constructed linear equation system is usually an overdetermined system, meaning the number of equations exceeds the number of unknown parameters, and there is no exact unique solution. The processor uses the least squares method to numerically solve this system of equations, searching for an optimal solution that minimizes the sum of squared residuals. This solution process calculates the vector values ​​of the imbalance of each component, including the magnitude and phase angle, which is the best estimate of the true value of the component under the current data conditions.

[0078] Step S803: Determine the initial sample points based on the estimated value of the imbalance of each part.

[0079] Specifically, the calculated estimates of the imbalance of each component are integrated to form a dataset containing the imbalance parameters of all components. This dataset serves as the initial sample points, representing the optimal estimation result of the imbalance state of each component under the constraints of initial limited test data, providing the necessary parameter input for subsequent assembly angle optimization calculations.

[0080] This embodiment achieves the function of inverting the unbalance of each part from the overall unbalance test data of the assembly by constructing a system of linear equations and using the least squares method to solve the part parameters. It solves the problem that the existing technology relies on pre-measuring the parameters of individual parts, which leads to cumbersome processes and large errors. It achieves the beneficial effect of effectively obtaining the initial estimate of the unbalance of the parts without disassembling and measuring individual parts.

[0081] In some embodiments, see Figure 9 Based on the initial sample points, an optimization algorithm is used to determine the recommended assembly angle, including: Step S901: Based on the initial sample points, determine the vector parameters of the unbalance of each part.

[0082] Specifically, estimates of the imbalance of each component are extracted from the initial sample points. These estimates include vector parameters in two dimensions: amplitude and phase. In subsequent optimization processes, these vector parameters are fixed as known constants, representing the imbalance characteristics of each component at the current estimation level.

[0083] Step S902: Construct the objective function based on the vector parameters.

[0084] The objective function characterizes the mapping relationship between the magnitude of the overall residual unbalance of the component and the assembly angles. Specifically, the objective function is constructed based on the principle of vector superposition. Using the assembly angles of each part as optimization variables, and combining determined vector parameters, the vector sum of the unbalance of each part after angle rotation and transmission coefficient mapping is calculated to obtain the theoretical value of the overall residual unbalance of the component. The objective function is then constructed, with its function value being the magnitude of the overall residual unbalance of the component. This objective function characterizes the mapping relationship between the combination of assembly angles and the overall vibration level of the component; the optimization objective is to find the assembly angle combination that minimizes this magnitude.

[0085] Step S903: Based on the objective function, determine the recommended assembly angle using an optimization algorithm.

[0086] Specifically, a pre-defined optimization algorithm is invoked to solve the objective function. The algorithm iteratively searches within the solution space, encompassing all possible combinations of assembly angles, continuously trying different angle combinations and calculating the corresponding objective function values. Through iterative convergence, it ultimately finds the set of assembly angle combinations that theoretically minimizes the overall residual imbalance modulus of the component. This set is output as the recommended assembly angles to guide subsequent physical assembly operations.

[0087] This embodiment achieves automated and accurate recommendation of assembly angles by constructing an objective function based on parameter estimates and solving it using an optimization algorithm. This solves the problem of low efficiency caused by blind trial and error due to reliance on human experience in traditional methods, and achieves the beneficial effects of significantly improving the efficiency of the balancing process, reducing the number of disassembly and assembly operations, and reducing reliance on operator experience.

[0088] In another embodiment, the overall calculation process is illustrated with an example. This embodiment provides a specific APU rotor assembly balancing method, using a certain type of APU rotor assembly as an example. (See also...) Figure 12 and Figure 13 The rotor assembly consists of 13 parts. In this example, we will only use the adjustment of three of these parts as an example to illustrate the process. The remaining parts are regarded as a whole, so a total of four parts need to be balanced.

[0089] The imbalance quantity A is a vector parameter consisting of its magnitude and angle, and therefore can be expressed as a complex polynomial or a complex exponential form. The two can be equivalently transformed using Euler's formula. That is: , in, , in, For amplitude, For angle, Here, 'a' represents the real part of the complex number A, which is the component of the unbalanced quantity corresponding to the real number axis in the complex representation; 'b' represents the imaginary part of the complex number A, which is the component of the unbalanced quantity corresponding to the imaginary number axis in the complex representation; and 'e' represents the base of the natural logarithm (Euler's number, approximately equal to 2.71828), a constant used in Euler's formula for the exponential form of complex numbers, used to convert the trigonometric form (magnitude and angle) of complex numbers into exponential form.

[0090] The specific steps of this method are as follows: Step (1) Mark the initial position on each set of key teeth, combine the rotor parts with any initial position, and record the key tooth rotation matrix [N] at this time. 11 N 12 N 13 N 14 and balancing machine test results A T1 , B T1 ; Step (2) Randomly reassemble the parts and record the key tooth rotation matrix [N] at this time. 21 N 22 N 23 N 24 and balancing machine test results A T2 , B T2 ; Step (3) Randomly reassemble the parts again and record the key tooth rotation matrix [N] at this time. 31 N 32 N 33 N 34 and balancing machine test results A T3 , B T3 ; Step (4) Randomly rearrange the parts again and record the key tooth rotation matrix [N] at this time. 41 N 42N 43 N 44 and balancing machine test results A T4 , B T4 ; Steps (1) to (4) are all for obtaining initial unbalance test data, recording the key tooth rotation matrix under different assembly angle combinations and the remaining unbalance vector of the entire component on the support surface; Step (5) Initial Sample Point Solution: Based on the obtained initial unbalance test data, establish a mapping relationship model and construct a system of linear equations, that is, construct the system of equations using the above four sets of test data. Based on the least squares method, determine the estimated value of the unbalance of each part, thereby determining the initial sample points. Specifically, solve for the unbalance parameters of the four parts. The solution process can be transformed from the complex exponential expression to the complex polynomial form to simplify the calculation. Thus, a set of part unbalance parameter sample points are obtained and recorded in the coordinate system;

[0091] Step (6) Initial Optimization: Based on the initial sample points, the recommended assembly angle is determined using an optimization algorithm. Specifically, based on the parameter sample values ​​obtained in the previous step, the particle swarm optimization (PSO) algorithm is used to perform optimization, resulting in a set of recommended key tooth rotation number matrices [N]. 51 N 52 N 53 N 54 ]; Step (7) Testing and Verification: Based on the recommended assembly angle, perform assembly and balancing tests on the rotor assembly using a balancing machine to obtain the test results. A T5 , B T5 If the result is satisfactory, the preset balance requirement is met, and the test ends; if it is unsatisfactory, the preset balance requirement is not met, and the next step is performed. Step (8) Sample amplification: Based on the current test data and historical test data, the current test data and the most recent key tooth rotation number matrix [N] are used to amplify the sample data. 51 N 52 N 53 N 54 Add the data to the historical test data, and obtain the amplified test dataset from the aforementioned four sets of key tooth rotation number matrices. In the aforementioned four sets of key tooth rotation number matrices [N... 11 N 12 N 13 N 14 ]、[N 21 N 22 N 23 N24 ]、[N 31 N 32 N 33 N 34 ]、[N 41 N 42 N 43 N 44 In the matrix, take any three sets of key tooth rotation numbers and the matrix [N] of the most recent key tooth rotation. 51 N 52 N 53 N 54 Combining these elements to form a four-dimensional matrix, there are a total of [number] combination methods. The combination of these combinations yields four sets of samples: Based on the amplified test dataset, the parameter estimates of the self-imbalance of each part are re-determined to obtain the amplified parameter estimate sample set, which serves as the amplified sample points.

[0092]

[0093]

[0094] , Step (9) Amplify sample records: Record the four sets of sample points obtained in the coordinate system ( Figure 12 As shown by the solid black dots in the middle), at this time the coordinate system contains a set of sample points obtained in step (5). and the four sets of sample points obtained in step (8) , , , There are a total of five sample points; Step (10) Probability density estimation update: The probability density of the amplified sample points is estimated to obtain updated sample points. Specifically, the probability density distribution function is determined using the two-dimensional kernel density estimation method based on the obtained five sets of sample points. Based on the probability density distribution function, the point with the maximum probability density, i.e., the maximum probability density point, is determined as the estimated value of the unbalance parameter of the part. ( Figure 12 (as shown by the hollow point); based on the point with the maximum probability density, the updated sample point is obtained; Step (11) Iterative optimization: Based on the updated sample points, the latest recommended assembly angle is determined. Specifically, based on the estimated value in step (10), the particle swarm optimization (PSO) algorithm is used to determine the target assembly angle. Based on the target assembly angle, the latest recommended assembly angle is determined, thus obtaining a new set of key tooth rotation number matrices [N]. 61 N 62 N 63 N 64 ]; Step (12) Iterative testing: Based on the latest recommended assembly angle, obtain the test results under the latest recommended assembly angle. Specifically, run the balancing machine to obtain the test results. A T6 , B T6 If the result is satisfactory, the test ends; if it is unsatisfactory, refer to step (8) and randomly select three of the aforementioned five sets of key tooth rotation matrices, and then combine them with [N]. 61 N 62 N 63 N 64 The components are combined to form a four-dimensional matrix, and the total number of combination schemes is [number missing]. The group continuously expands the sample using this method, estimates the probability density of the expanded sample points to obtain updated sample points, and uses an optimization algorithm based on the updated sample points to obtain a new key tooth rotation matrix. Specifically, a two-dimensional kernel density estimation method is used to update the estimated values ​​of the part parameters, and a particle swarm optimization algorithm is used based on these estimated values ​​to obtain a new key tooth rotation matrix, i.e., the latest recommended assembly angle, until the test results meet the preset balance requirements.

[0095] In another embodiment, the specific implementation of two-dimensional Gaussian kernel density estimation is described in detail. This embodiment provides a detailed explanation of the two-dimensional kernel density estimation method in step (10) above. Two-dimensional Gaussian kernel density estimation is a nonparametric statistical method used to estimate an unknown probability density function (PDF) from two-dimensional sample data. Each unbalanced quantity sample includes two parameters: amplitude and angle, and its distribution law is unknown. The core idea of ​​kernel density estimation is to apply a Gaussian kernel function of appropriate bandwidth to each sample point and then superimpose them to obtain the density distribution over the entire plane.

[0096] Taking the imbalance on surface A of part number 1 as an example, and based on the five samples mentioned above, the process is explained as follows: (1) Input sample: Input amplified sample points Its amplitude A mi With angle φ iAs two-dimensional sample data.

[0097] (2) Generate the estimated density function, i.e. determine the probability density distribution function: , In the formula: n is the number of samples, which is 5 in this example, h Am For amplitude bandwidth, The angular bandwidth is used to adjust the smoothness of the estimated density function. Its selection can be based on the Silverman rule of thumb: , in These are the variances of the amplitude and angle samples, respectively.

[0098] (3) Parameter update: Based on the probability density distribution function, determine the point of maximum probability density, obtain the updated sample points based on the point of maximum probability density, and solve the problem. The maximum value of the independent variable corresponding to this maximum value is the maximum estimated probability point. This serves as the optimal estimate of the unbalance parameter for the part.

[0099] In another embodiment, the specific implementation of the Particle Swarm Optimization (PSO) algorithm is explained in detail; see [link to documentation]. Figure 14 This embodiment provides a detailed explanation of the particle swarm optimization algorithm used in the step of determining the recommended assembly angle or the latest recommended assembly angle based on the initial sample points or updated sample points using an optimization algorithm: (1) Randomly initialize the particle swarm: Based on the known number of parameter components k, establish a set of position variables of each component as a particle in the particle swarm to be solved, combine multiple particles to construct a particle swarm, and randomly initialize each variable within the range of values.

[0100] (2) Calculating Fitness: The fitness function is the standard for evaluating the quality of each particle in the particle swarm. It corresponds to constructing an objective function based on vector parameters. The objective function is used to characterize the mapping relationship between the magnitude of the overall residual imbalance of the component and the assembly angle. In this example, the fitness is the magnitude of the residual imbalance. Since there are two measurement surfaces and the optimization objectives are coupled, the sum of the magnitudes of the two vectors is taken as the fitness function. At the same time, the two fitness values ​​are constrained according to the required limits.

[0101] (3) Update swarm velocity and position: Based on the best historical position P of the particles obtained by the particle swarm during the optimization process. best The historical best position G of all particles in the particle swarm best And set the weight coefficients c1, c2, c3, α The particle swarm is updated with random numbers r1 and r2.

[0102] (4) Termination Condition Determination: The termination condition of the particle swarm optimization algorithm is based on the set number of iterations. When the set number of iterations is reached, the program automatically terminates. At this time, G... best The solution obtained by the algorithm determines the recommended assembly angle or the target assembly angle / latest recommended assembly angle based on the solution of the objective function. The rotor assembly is then recombined according to the angle obtained.

[0103] Figure 10 This is a schematic diagram of the structure of an APU rotor assembly balancing system provided in an embodiment of this application. See also... Figure 10 The APU rotor assembly balancing system 10 includes: an initial sample point determination module 1001, an optimization module 1002, an amplified sample point construction module 1003, and a sample point update module 1004.

[0104] The initial sample point determination module 1001 is used to determine the initial sample points based on the acquired initial unbalance test data. The initial unbalance test data is used to represent the remaining unbalance vector on the support surface of the APU rotor assembly under different assembly angle combinations. The initial sample points are used to represent the parameter estimates of the unbalance of each part.

[0105] Specifically, the module reads the vector data of the overall remaining imbalance of the component measured under different assembly angle combinations in the initial stage, determines the transmission coefficient based on the rotor structure, establishes a mapping relationship model between the imbalance of each part and the overall imbalance of the component, such as a system of linear equations, and obtains the initial estimated value of the imbalance of each part, i.e. the initial sample point, by solving the least squares method.

[0106] The optimization module 1002 is used to optimize based on the initial sample points and determine the test results under the recommended assembly angle. Specifically, the module uses the initial sample points as input parameters, constructs an objective function with the goal of minimizing the overall residual unbalance modulus of the component, uses optimization algorithms such as genetic algorithm and particle swarm optimization algorithm to search in the solution space, outputs the recommended assembly angle for operators to assemble and test, and receives the feedback test results.

[0107] The expanded sample point construction module 1003 is used to construct expanded sample points based on current test data and historical test data when the test results do not meet the preset balancing requirements. The current test data indicates the data obtained in the most recent assembly and balancing test, and the historical test data indicates the test dataset accumulated before the current test. Specifically, when the module determines that the test results do not meet the requirements, it merges the current test data into the historical dataset, reconstructs and solves the equation system using the expanded dataset, and generates a parameter estimation sample set containing more observation information, which serves as the expanded sample points.

[0108] The sample point update module 1004 is used to estimate the probability density of the amplified sample points, obtain updated sample points, and feed the updated sample points back to the optimization module 1002 for iterative optimization until a test result that meets the preset balance requirements is obtained.

[0109] The system provided in this embodiment achieves a complete closed loop from data modeling, parameter estimation, optimization recommendation to error convergence through the collaborative work of various modules. It solves the problem of slow convergence and low efficiency caused by the inability of existing technologies to effectively utilize test data from previous tests, and significantly improves the automation and accuracy of the APU rotor assembly balancing process.

[0110] It should be noted that the APU rotor assembly balancing system provided in the above embodiments is only illustrated by the division of the functional modules described above during the balancing process. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the APU rotor assembly balancing system and the APU rotor assembly balancing method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0111] This application embodiment also provides a device 11, Figure 11 This is a schematic diagram of the structure of a device provided in an embodiment of this application.

[0112] Typically, device 11 includes one or more processors 1101 and one or more memories 1102.

[0113] Processor 1101 may include one or more processing cores, such as a quad-core processor, a hexa-core processor, etc. Processor 1101 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). Processor 1101 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1101 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, processor 1101 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0114] The memory 1102 may include one or more computer-readable storage media, which may be non-transitory. The memory 1102 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1102 are used to store at least one computer program, which is executed by the processor 1101 to implement the APU rotor assembly balancing method provided in the method embodiments of this application.

[0115] Those skilled in the art will understand that Figure 11 The structure shown does not constitute a limitation on device 11 and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0116] In addition, the device provided in the embodiments of this application may specifically be a chip, component or module. The chip may include a connected processor and a memory. The memory is used to store instructions. When the processor calls and executes the instructions, the chip can execute an APU rotor component balancing method provided in the above embodiments.

[0117] This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the above-described related method steps to implement the APU rotor assembly balancing method provided in the above embodiment.

[0118] This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to implement an APU rotor assembly balancing method provided in the above embodiment.

[0119] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0120] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0121] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0122] The above description is only a specific implementation of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application.

Claims

1. A method for balancing an APU rotor assembly, characterized in that, The method includes: Based on the acquired initial unbalance test data, initial sample points are determined. The initial unbalance test data is used to represent the remaining unbalance vector of the APU rotor assembly on the support surface under different assembly angle combinations. The initial sample points are used to represent the parameter estimate of the unbalance of each component. Based on the optimization of the initial sample points, the test results under the recommended assembly angle are determined; When the test results do not meet the preset balance requirements, an amplified sample point is constructed based on the current test data and historical test data. The current test data is used to indicate the data obtained in the most recent assembly and balance test, and the historical test data is used to indicate the test dataset accumulated before the current test. The probability density of the amplified sample points is estimated to obtain updated sample points; Based on the optimization of the updated sample points, the test results under the latest recommended assembly angle are determined until the test results that meet the preset balance requirements are obtained. The updated sample points correspond to the points with the highest probability density of the amplified sample points.

2. The APU rotor assembly balancing method according to claim 1, characterized in that, The determination of initial sample points based on the acquired initial imbalance test data includes: Based on the obtained initial imbalance test data, a mapping relationship model is established. The mapping relationship model is used to refer to the vector superposition relationship between the imbalance of each part and the remaining imbalance of the whole assembly. Based on the mapping relationship model, the initial sample points are determined.

3. The APU rotor assembly balancing method according to claim 1, characterized in that, The process of optimizing based on the initial sample points to determine the test results under the recommended assembly angle includes: Based on the initial sample points, an optimization algorithm is used to determine the recommended assembly angle; Based on the recommended assembly angle, the rotor assembly was assembled and balanced to obtain the test results.

4. The APU rotor assembly balancing method according to claim 1, characterized in that, When the test results do not meet the preset balance requirements, an amplified sample point is constructed based on the current test data and historical test data, including: When the test results do not meet the preset balance requirements, the current test data is added to the historical test data to obtain the expanded test dataset; Based on the amplified test dataset, the parameter estimates of the imbalance of each part are re-determined to obtain the amplified parameter estimate sample set, which is used as the amplified sample points.

5. The APU rotor assembly balancing method according to claim 1, characterized in that, The probability density of the amplified sample points is estimated to obtain updated sample points, including: The probability density of the amplified sample points is estimated to determine the probability density distribution function; Based on the probability density distribution function, the point of maximum probability density; The updated sample points are obtained based on the points with the highest probability density.

6. The APU rotor assembly balancing method according to any one of claims 1 to 5, characterized in that, The process of optimizing based on the updated sample points to determine the test results under the latest recommended assembly angle, until a test result that meets the preset balance requirements is obtained, includes: Based on the optimization of the updated sample points, the target assembly angle is determined. The target assembly angle is used to represent the assembly angle corresponding to the minimum theoretical remaining imbalance. Based on the target assembly angle, determine the latest recommended assembly angle; Based on the latest recommended assembly angle, the test results under the latest recommended assembly angle are obtained.

7. The APU rotor assembly balancing method according to claim 2, characterized in that, The process of establishing a mapping model based on the acquired initial imbalance test data includes: Based on the obtained initial imbalance test data, extract different assembly angle combinations and the corresponding overall remaining imbalance vector of the components; Based on the assembly structure of the APU rotor assembly, determine the transmission coefficient of the unbalance of each component; Based on the principle of vector superposition, a mapping relationship model is established by combining the different assembly angle combinations, the transmission coefficient, and the overall remaining unbalance vector of the component.

8. The APU rotor assembly balancing method according to claim 2, characterized in that, The determination of initial sample points based on the mapping relationship model includes: Based on the mapping relationship model, a linear equation system is constructed, which is used to indicate the vector superposition relationship between the unbalance of each part and the remaining unbalance of the whole assembly. The linear equations are solved using the least squares method to determine the estimated value of the unbalance of each part. Based on the estimated values ​​of the imbalance of each component, the initial sample points are determined.

9. The APU rotor assembly balancing method according to claim 3, characterized in that, The step of determining the recommended assembly angle based on the initial sample points using an optimization algorithm includes: Based on the initial sample points, determine the vector parameters of the imbalance of each part. Based on the vector parameters, an objective function is constructed, which is used to characterize the mapping relationship between the magnitude of the overall remaining unbalance of the component and the assembly angle. Based on the objective function, an optimization algorithm is used to determine the recommended assembly angle.

10. An APU rotor assembly balancing system, characterized in that, The system includes: The initial sample point determination module is used to determine the initial sample point based on the acquired initial unbalance test data. The initial unbalance test data is used to represent the remaining unbalance vector of the APU rotor assembly on the support surface under different assembly angle combinations. The initial sample point is used to represent the parameter estimate of the unbalance of each part. The optimization module is used to optimize based on the initial sample points and determine the test results under the recommended assembly angle; The amplified sample point construction module is used to construct amplified sample points based on the current test data and historical test data when the test results do not meet the preset balance requirements. The current test data is used to indicate the data obtained in the most recent assembly and balance test, and the historical test data is used to indicate the test dataset accumulated before the current test. The sample point update module is used to estimate the probability density of the amplified sample points, obtain updated sample points, and feed the updated sample points back to the optimization module for iterative optimization until a test result that meets the preset balance requirements is obtained.

Images