A method and apparatus for correcting APU characteristics

Abstract

This application provides an APU characteristic correction method and apparatus, belonging to the field of aerospace electromechanical systems. The method includes: establishing an APU system simulation model; determining multiple correction parameters and correction targets; calculating the calculated value corresponding to the correction target based on the APU system simulation model and the correction parameters; constructing a correction error function based on the calculated value of the correction target and experimental values ​​of the correction target obtained through experiments; determining the updated components in the APU, and determining an optimization strategy based on the identity of the updated components; executing a particle swarm optimization algorithm based on the optimization strategy to obtain the optimal correction parameters corresponding to the correction error function; and updating the APU system simulation model based on the optimal correction parameters to obtain a corrected APU characteristic model. The APU characteristic correction method and apparatus provided in this application can improve the accuracy of model correction results after updating local components of the APU.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of aviation electromechanical technology, and in particular to an APU characteristic correction method and device. BACKGROUND

[0002] An auxiliary power unit (APU) is a small gas turbine engine on an aircraft, and its design and verification highly depend on simulation technology. At present, the simulation modeling of the APU is usually based on physical mechanisms or test data, but in the modeling process, the flow loss, transition process and the like need to be simplified, resulting in a deviation between the model and the real system. In addition, component manufacturing tolerances, performance degradation in long-term use, and local component replacement and the like will further deviate the simulation model from the actual situation.

[0003] At present, for the deviation between the APU model and the test data, a parameter correction method based on test data is usually used for calibration. However, when the APU has a local component update, there is a problem of more historical data and less new data. The original system accumulates a plurality of test data, while the test data of the new system is limited. If only a small amount of new data is used to correct all parameters, it is easy to lead to insufficient correction; if new and old data are mixed, the change in characteristics of the new component may not be accurately reflected. In addition, the existing correction method mainly focuses on fitting the current data, and lacks active optimization of the overall performance of the system, especially the stability margin. The APU needs to ensure that the compressor does not enter a stall or surge state in actual operation, and therefore, how to further improve the system stability margin on the basis of the corrected model is also a problem to be solved at present. SUMMARY

[0004] Therefore, the present application provides an APU characteristic correction method and device to solve the problems of low correction accuracy and poor efficiency caused by insufficient test data of the new system after the local component of the APU is updated.

[0005] Specifically, the present application is realized by the following technical solutions:

[0006] The first aspect of the present application provides an APU characteristic correction method, which comprises:

[0007] establishing an APU system simulation model comprising a compressor, a combustor and a power turbine;

[0008] determining a plurality of correction parameters and a correction target;

[0009] calculating a calculation value corresponding to the correction target based on the APU system simulation model and the correction parameters;

[0010] construct a correction error function based on the calculated value of the correction target and a test value of the correction target obtained through a test;

[0011] determine an updated component in the APU, determine an optimization strategy based on the identity of the updated component; the optimization strategy comprises optimizing a first correction parameter subset based on a first test data set before component update and optimizing a second correction parameter subset based on a second test data set after component update;

[0012] based on the optimization strategy, execute a particle swarm algorithm to obtain optimal correction parameters corresponding to the correction error function;

[0013] update the APU system simulation model based on the optimal correction parameters to obtain a corrected APU characteristic model.

[0014] The second aspect of the application provides an APU characteristic correction device, the device comprises an establishment module, a determination module, a calculation module, a construction module, a processing module and an update module;

[0015] The establishment module is used to establish an APU system simulation model comprising a compressor, a combustor and a power turbine;

[0016] The determination module is used to determine a plurality of correction parameters and a correction target;

[0017] The calculation module is used to calculate a calculated value corresponding to the correction target based on the APU system simulation model and the correction parameters;

[0018] The construction module is used to construct a correction error function based on the calculated value of the correction target and a test value of the correction target obtained through a test;

[0019] The determination module is also used to determine an updated component in the APU, and determine an optimization strategy based on the identity of the updated component; the optimization strategy comprises optimizing a first correction parameter subset based on a first test data set before component update and optimizing a second correction parameter subset based on a second test data set after component update;

[0020] The processing module is used to execute a particle swarm algorithm based on the optimization strategy to obtain optimal correction parameters corresponding to the correction error function;

[0021] The update module is used to update the APU system simulation model based on the optimal correction parameters to obtain a corrected APU characteristic model.

[0022] The APU characteristic correction method and device provided in the application can realize accurate and efficient correction of the system simulation model after the local component of the APU is updated by intelligently selecting the optimization strategy matched with the updated component and respectively optimizing different subsets of correction parameters based on different test data sets. The method can overcome the problem of reduced correction accuracy of the model caused by insufficient new data after the component is updated, fully utilize the breadth of historical data and the pertinence of new data, and improve the consistency of the model and the real system. Specifically, the optimization strategy is determined based on the identity of the updated component, the new data is used to correct the replaced component, the historical data is used to calibrate the non-replaced component, the parameter correction is more physically reasonable and engineering targeted, and the false correction of the characteristics of the non-updated component is avoided. By respectively using the test data sets before and after the component is updated to optimize the corresponding subsets of correction parameters, the problem of more historical data and less new data is effectively solved, the statistical stability of a large amount of historical data is utilized, the accuracy of a small amount of new data is fully utilized, and the optimal model fitting is realized under the condition of unbalanced data. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The flowchart of the APU characteristic correction method provided in the application is shown in Embodiment 1.

[0024] Figure 2 The structural schematic diagram of the APU characteristic correction device provided in the application is shown in Embodiment 2. DETAILED DESCRIPTION

[0025] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The following description is with reference to the drawings, in which like numerals represent like elements, unless otherwise specified. The embodiments described in the following exemplary embodiments are not representative of all embodiments consistent with the application.

[0026] The terms used in the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms "a", "an" and "the" used in the present application are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein means and includes any or all possible combinations of one or more associated listed items.

[0027] It should be understood that, although the terms first, second, third, etc. can be employed in this application to describe various information, the information is not to be limited to these terms. These terms are only used to distinguish one particular information from another. For example, a first information can also be termed a second information, and, similarly, a second information can also be termed a first information, without departing from the scope of the present application. Depending on the context, the word "if' as used herein can be interpreted to mean "when" or "in response to determining".

[0028] The following specific embodiments are given to introduce the technical solutions of the present application in detail.

[0029] Embodiment one

[0030] Figure 1 The flow chart of the APU characteristic correction method provided by the present application is shown in the following. Figure 1 The method provided by the present application can include:

[0031] S101, an APU system simulation model containing a compressor, a combustor and a power turbine is established.

[0032] It should be noted that the APU (auxiliary power unit) is essentially a small gas turbine engine, and its core aerodynamic and thermodynamic process is mainly completed by three components in series: compressor (air pressure), combustor (fuel addition and combustion), and power turbine (expansion to do work and output energy). These three components jointly determine the key performance of the APU, such as power, efficiency, flow rate, etc. Therefore, to accurately simulate the system behavior of the APU, accurate physical modeling of these three core components must be performed. Each of the three core components is abstracted as a non-linear mathematical model. The input of the non-linear mathematical model is the inlet boundary condition and control parameter (such as rotational speed, fuel quantity) of the corresponding component, and the output is the outlet state. The model is used to represent the characteristic relationship of the corresponding component, which describes how the key parameters such as efficiency, flow rate, pressure ratio / expansion ratio of the component change under a specific input. These characteristic relationships are usually fitted from component design data or bench test data.

[0033] Specifically, the APU system simulation model containing the compressor, the combustor and the power turbine is established, including:

[0034] (1) a compressor model is established, which represents the characteristic relationship between the compressor outlet flow rate, outlet temperature, outlet pressure and consumed power and the shaft rotational speed, inlet pressure, inlet temperature and pressure ratio.

[0035] It should be noted that the compressor model is a non-linear mathematical model representing the characteristic map of the compressor component. This model takes the shaft rotational speed , inlet pressure , inlet temperature and pressure ratio As input, through the internal encapsulated characteristic relationship (such as the efficiency line, flow line fitted based on experimental data), the corresponding outlet flow is output , outlet temperature , outlet pressure and consumed power . Its mathematical form can be abstracted as a function:

[0036] .

[0037] (2) A combustion chamber model is established, which represents the characteristic relationship between the outlet flow, outlet temperature and outlet pressure of the combustion chamber and the inlet pressure, inlet temperature, inlet flow and fuel quantity.

[0038] It should be noted that the combustion chamber model is used to describe the energy conversion and flow process of fuel and air mixed combustion, which takes the inlet pressure , inlet temperature , inlet flow and fuel quantity as input, and calculates the outlet gas flow , outlet temperature and outlet pressure based on characteristic parameters such as combustion efficiency and total pressure recovery coefficient. Its mathematical form is:

[0039] .

[0040] (3) A power turbine model is established, which represents the characteristic relationship between the outlet flow, outlet temperature, outlet pressure and output power of the power turbine and the shaft speed, inlet pressure, inlet temperature and expansion ratio.

[0041] It should be noted that the power turbine model represents the characteristics of the turbine converting high-temperature gas energy into mechanical power, and the input is the shaft speed , inlet pressure , inlet temperature and expansion ratio , and the output is the outlet flow , outlet temperature , outlet pressure and output power . Its functional relationship is:

[0042] .

[0043] (4) Based on the compressor model, the combustion chamber model and the power turbine model, and according to the flow conservation, pressure conservation and energy conservation relationship, the APU system simulation model is constructed.

[0044] It should be noted that the three independent component models are integrated into a complete APU system simulation model, which must meet the physical consistency of the component connection, and the APU system simulation model is constructed and solved based on the flow conservation, pressure conservation and energy conservation relationship. Among them, the flow conservation refers to the continuity of the working fluid flowing through each component under steady state; the pressure conservation and energy conservation refer to the natural connection of the pressure at the component interface, and the system shaft power balance, that is, the output power of the power turbine is equal to the sum of the power consumption of the compressor, the net output power (such as power generation) of the system and the mechanical loss.

[0045] Specifically, based on the above determined component model, and introducing the following system balance conditions based on the law of conservation of physics, the APU system simulation model is jointly constituted:

[0046] Flow conservation: ;

[0047] Pressure conservation: ;

[0048] Energy conservation: ;

[0049] Wherein, represents the nozzle expansion coefficient fitting formula, represents the system power generation.

[0050] In the simulation calculation, the environmental conditions and the running setting are given, and the model is solved by simultaneously solving the component characteristic equation (f1, f 2， f3) and the above system balance conditions, and the unique steady state working point of the whole system is determined by iteration.

[0051] S102, determine a plurality of correction parameters and correction targets.

[0052] It should be noted that there is a deviation between the theoretical characteristics of the APU system simulation model and the real characteristics of the APU actual components due to manufacturing, wear and update. It is relatively difficult to directly correct the complex component characteristic map, therefore, a group of correction parameters are introduced as the calibration coefficients of the core component characteristics (such as flow, pressure ratio and efficiency), and these coefficients are adjusted to make the model output approximate to the real system. At the same time, in order to evaluate the correction effect, a group of correction targets which can be directly and accurately measured by test are selected, and the error between the simulation value and the test value is calculated to drive the correction process.

[0053] Specifically, the embodiment focuses on correcting the characteristics of the compressor and the power turbine, without including the combustion chamber related parameters in the correction range, mainly because the main problem to be solved is the model deviation caused by component updates, which directly affects the system stable working boundary (such as surge margin). The compressor and the power turbine are strongly coupled through shaft work and flow, and their characteristic matching directly determines the system stable working line. In contrast, the efficiency and pressure loss characteristics of the combustion chamber usually change smoothly and stably within the working envelope, and have a relatively indirect impact on system stability. Therefore, concentrating the correction resources on the compressor and turbine components that have the most direct and greatest possible impact on system stability can effectively control the complexity of the optimization problem while ensuring that the core problem is solved.

[0054] Specifically, a plurality of correction parameters and correction targets are determined, including:

[0055] (1) Based on the deviated performance characteristics of the components in the APU system simulation model, correction parameters for representing the performance changes of the components are determined; the correction parameters include compressor flow, compressor pressure ratio, compressor efficiency, power turbine flow, power turbine expansion ratio, and power turbine efficiency.

[0056] It should be noted that the six parameters are selected as the correction parameters because they collectively constitute the most basic and core aerodynamic and thermodynamic performance characteristics of the compressor and the power turbine, the two core components. Correcting these parameters is to directly adjust the fundamental elements that determine the performance of the components at the model level. Among them, the flow and pressure ratio / expansion ratio determine the work capacity of the component, and the efficiency reflects the quality of energy conversion. The deviation of these parameters can be directly related to actual engineering problems such as machining errors and performance degradation, making the correction results have good interpretability.

[0057] (2) Based on the characteristic quantities in the APU system simulation model that can be obtained through experiments and can reflect the output performance of the system, the correction targets are determined; the correction targets include compressor outlet pressure, compressor outlet temperature, fuel flow, and power turbine outlet temperature.

[0058] It should be noted that the four parameters are selected as the correction targets based on their measurability, system representativeness, and sensitivity to changes in the correction parameters. The compressor outlet pressure and outlet temperature are direct manifestations of the compression performance of the compressor component; the power turbine outlet temperature comprehensively reflects the effects of the combustion and expansion processes; and the fuel flow is an accurate measure of the energy input of the system. These parameters are all routine high-precision measurement items in APU bench tests, and the data is reliable. In addition, they are sensitive to changes in the aforementioned six correction parameters, for example, a decrease in compressor efficiency will directly lead to an increase in its outlet temperature, thereby providing a clear optimization gradient for the subsequent optimization algorithm, ensuring the effectiveness and convergence of the correction process.

[0059] S103, calculating a calculated value corresponding to the correction target based on the APU system simulation model and the correction parameters.

[0060] Specifically, the correction parameters determined in step S102 are applied to the corresponding component characteristics in the APU system simulation model as adjustment factors. For example, the correction coefficient of the compressor efficiency is multiplied by the original efficiency characteristic value in the APU system simulation model, thereby obtaining the corrected component efficiency used in this calculation.

[0061] Subsequently, the APU system simulation model with the injected correction parameters is run under a set operating condition (such as a given ambient pressure, ambient temperature). The model will converge to a system steady-state operating point by iteratively solving the flow, pressure, and energy conservation equations based on the corrected component characteristics.

[0062] Finally, the values of the correction targets defined in step S102 are extracted from the converged solution. These values are the system outputs predicted by the model under the current set of specific correction parameters, i.e., the calculated values corresponding to the correction targets.

[0063] S104, constructing a correction error function based on the calculated values of the correction targets and the experimental values of the correction targets obtained through experiments.

[0064] Specifically, constructing the correction error function includes:

[0065] (1) calculating the deviation of each correction target based on the calculated value and the corresponding experimental value of the correction target.

[0066] It should be noted that the deviation here can be absolute error, relative error, etc. Specifically, it is used to quantify the difference between simulation and experiment for a single correction target.

[0067] In specific implementation, each set of experimental data is read in sequence, and for each correction target, the operation of deviation = calculated value - experimental value or deviation = (calculated value - experimental value) / experimental value is performed. This step is completed by looping through all targets and all data sets in the computer program.

[0068] (2) weighting the deviations of each correction target according to a preset weight to obtain the corresponding target error.

[0069] It should be noted that the introduction of the preset weight is to consider that different correction targets may have different importance in engineering. For example, if it is believed that ensuring the accuracy of the compressor outlet temperature is more important than the pressure, a higher weight can be assigned to it, thereby enhancing the flexibility and engineering applicability of the method.

[0070] Specifically, during program initialization, a weight vector is predefined or input by the user to correspond to the correction target. Then, during calculation, each deviation obtained in step (1) is multiplied by the corresponding weight coefficient to obtain the corresponding target error.

[0071] (3) Accumulate all target errors to construct the correction error function.

[0072] It should be noted that the summation (usually a summation) operation combines the weighted biases on all correction targets into a single scalar value. The smaller this value, the higher the overall fit between the current model and the experimental data under that set of correction parameters.

[0073] Specifically, after weighting all objectives for a single set of data, the errors of all objectives in that set of data are summed to obtain the overall error of that set of data. Then, the overall errors of all N sets of experimental data are summed (or averaged) again, and finally a scalar value is output as the value of the correction error function.

[0074] Preferably, the construction of the correction error function includes:

[0075] (1) For the i-th set of test data, calculate the relative error between the calculated and test values ​​of compressor outlet pressure, compressor outlet temperature, fuel flow rate and power turbine outlet temperature.

[0076] It should be noted that the relative error used here, (calculated value - experimental value) / experimental value, can eliminate the influence of different physical quantities in terms of dimensions and orders of magnitude, making the error contributions of each correction target comparable. Specifically, the relative error is calculated one by one according to the formula Relative Error = (calculated value - experimental value) / experimental value, resulting in four relative error values.

[0077] (2) Squat each of the relative errors and sum them up, then divide by the number of correction targets to obtain the mean squared error.

[0078] It should be noted that squaring can convert all errors into positive numbers, preventing the cancellation of positive and negative values. It can also amplify the impact of larger errors, making the optimization process more sensitive to outliers and thus better reducing large deviations. Specifically, the four relative error values ​​are squared separately, then added together, and the sum is divided by 4 (the number of correction targets).

[0079] (3) Perform the square root operation on the mean square error to obtain the root mean square error of the i-th group of experimental data.

[0080] It should be noted that the square root operation restores the dimensions to a level close to the original error, and the resulting single-group root mean square error is an indicator of the overall error of the model under this specific experimental condition.

[0081] (4) Repeat the calculation of the root mean square error of each group for all N groups of experimental data to obtain N root mean square errors.

[0082] It should be noted that this is to comprehensively utilize the information from all available test conditions (a total of N groups) to ensure that the corrected model can perform well under various working conditions, rather than fitting only a single condition. Specifically, steps (1) to (3) are repeated for all N groups of test data to obtain N single-group root mean square errors.

[0083] (5) Sum the N single-group root mean square errors and divide by N to obtain the final value of the correction error function.

[0084] It should be noted that the root mean square error of the N operating conditions is finally averaged to obtain the global root mean square error, which is used as the output of the entire correction error function. This value comprehensively reflects the overall prediction accuracy of the model for all key output targets under all experimental operating conditions.

[0085] In summary, the final correction error function is:

[0086] .

[0087] S105. Determine the updated component in the APU, and determine an optimization strategy based on the identity of the updated component; the optimization strategy includes optimizing a first subset of correction parameters based on a first test dataset before the component update, and optimizing a second subset of correction parameters based on a second test dataset after the component update.

[0088] It should be noted that traditional correction methods, if using only a small amount of new data to correct all parameters, are prone to insufficient accuracy; if all new and old data are mixed, the inherent characteristic information of unupdated components contained in the old data will overwrite the learning of the characteristics of the new components. This step identifies the component update status and intelligently allocates the correspondence between data and parameters accordingly, so that new and old data can be reasonably distributed and corrected, thereby achieving optimal correction under the condition of data imbalance.

[0089] Specifically, the determination of the optimization strategy includes the following three parallel strategy schemes, one of which can be executed in practical applications and the optimal solution can be selected by comparison.

[0090] (1) Optimize all correction parameters using the first test dataset before component update.

[0091] It should be noted that this is based on the assumption of historical data consistency and is suitable for preliminary assessments or scenarios where the characteristics of the new component are not significantly different from those of the old component. It utilizes the statistical stability of a large amount of historical data to correct the entire model, but it may not accurately capture the unique characteristics of the new component.

[0092] (2) Optimize all correction parameters using the second test dataset after component update.

[0093] This is based on the assumption of the accuracy of new data and fully trusts the changes brought about by the new components. It directly uses the new system data for correction, ensuring the learning of the characteristics of the new components. However, since the amount of data is usually much smaller than the amount of historical data, correcting all parameters individually is prone to the risk of overfitting or insufficient correction due to insufficient data.

[0094] (3) Based on the identity of the updated component, the first test dataset and the second test dataset are respectively allocated to optimize the first modified parameter subset and the second modified parameter subset.

[0095] To overcome the problems of the two aforementioned approaches, a method is proposed that uses the most relevant and matching data source to correct the parameter set most likely to change.

[0096] Specifically, based on the identity of the updated component, the first test dataset and the second test dataset are respectively allocated to optimize the first and second subsets of correction parameters. This includes: if the updated component is a compressor, the second test dataset is allocated to optimize the first subset of correction parameters, including compressor flow rate, compressor pressure ratio, and compressor efficiency, because the second test dataset directly contains the real characteristic information of the new compressor. Simultaneously, the first test dataset is allocated to optimize the second subset of correction parameters, including power turbine flow rate, power turbine expansion ratio, and power turbine efficiency, because the power turbine has not been replaced, its characteristics should remain stable, and a large amount of historical data can be used for high-precision calibration of its model.

[0097] If the updated component is a power turbine, the second test dataset is allocated to optimize a second subset of correction parameters, including power turbine flow rate, power turbine expansion ratio, and power turbine efficiency, to learn new turbine characteristics, and the first test dataset is allocated to optimize a first subset of correction parameters, including compressor flow rate, compressor pressure ratio, and compressor efficiency, to maintain the accuracy of the compressor model.

[0098] It should be further clarified that the three strategies mentioned above are not simply selected in practice, but rather serve as three parallel candidate paths within a complete correction process. In actual operation, subsequent optimization algorithms can be executed based on each of the three strategies to obtain three different sets of optimal correction parameters. Then, by comparing the final values ​​of their corresponding correction error functions, the set of parameters with the smallest error is selected as the global optimal solution. This design ensures that the method can adaptively select the most effective correction strategy under specific data conditions.

[0099] S106. Based on the optimization strategy, execute the particle swarm optimization algorithm to obtain the optimal correction parameters corresponding to the correction error function.

[0100] It should be noted that the Particle Swarm Optimization (PSO) algorithm simulates the social behavior of bird flocks. Through the collaborative search and experience sharing of multiple particles (i.e., candidate solutions) in the population, it efficiently seeks optimization in a complex multi-parameter space and ultimately converges to the optimal solution to the problem. The purpose here is to transform the set of correction parameters found in the previous steps that minimizes the overall error between the simulation model output and the experimental data into an optimization problem that can be automatically solved by a computer.

[0101] Specifically, the step of executing the particle swarm optimization algorithm based on the optimization strategy to obtain the optimal correction parameters corresponding to the correction error function includes:

[0102] (1) Initialize the correction parameters and set the running parameters of the particle swarm algorithm.

[0103] It should be noted that the operating parameters include population size, learning factor, inertia weight range, particle position range, and particle velocity range. The population size determines the number of particles participating in the search simultaneously, affecting the breadth of the search and computational efficiency. The learning factor typically includes individual learning factor and social learning factor, which respectively adjust the degree to which particles learn from their own historical best experience and the group's historical best experience. The inertia weight controls the influence of the particle's current velocity on the velocity of the next generation, balancing global exploration and local development capabilities. The particle position range and velocity range are introduced physical constraints; the position range limits the feasible value range of each correction parameter. For example, based on physical meaning, the upper limit of the correction parameters for compressor efficiency and power turbine efficiency must be set to 1.0, because component efficiency cannot exceed 100%. The remaining parameters can be set with reasonable fluctuation ranges based on engineering experience (e.g., [0.8, 1.2]), while the velocity range limits the maximum magnitude of parameter adjustment in a single iteration, ensuring the stability of the search.

[0104] It should also be noted that the correction parameters are initialized to determine the dimension of the optimization problem, i.e., the number of parameters to be corrected (e.g., 6). Each particle represents a complete vector of candidate correction parameters. During algorithm initialization, a certain number of particles (determined by the population size) are randomly generated using a chaotic initialization method in the program, and each particle is assigned a random initial position (parameter value) and an initial velocity (direction of change and step size).

[0105] (2) Based on the experimental data usage method determined by the optimization strategy, when evaluating each particle in each iteration, the corresponding experimental dataset is called to calculate the value of the correction error function as the fitness of the particle.

[0106] It should be noted that when the algorithm needs to evaluate any particle in the population, the program will call the experimental data according to the optimization strategy determined in step S105. Specifically, the program parses the correction parameter vector represented by the particle and splits it into corresponding parameter subsets according to the strategy. For example, if the strategy is to optimize the compressor parameters using the second experimental dataset and optimize the turbine parameters using the first experimental dataset, the program will associate the three components of the particle's position vector corresponding to the compressor flow rate, pressure ratio, and efficiency with the second experimental dataset (new data) to calculate the model error under these parameters; at the same time, it will associate the three components of the corresponding turbine parameters with the first experimental dataset (old data).

[0107] Subsequently, following the correction error function constructed in step S104, the program runs the model using the associated experimental dataset and calculates the overall error value of the parameter set represented by the particle. This error value is defined as the fitness of the particle. The smaller the fitness value, the higher the fit between the model and the experimental data under this set of correction parameters, and the better the quality of the particle.

[0108] (3) Update the individual optimal position and the global optimal position of the population of the particle according to the fitness, and update the velocity and position of the particle.

[0109] It's important to note that the program compares each particle's current fitness with its best fitness found so far (recorded in its individual best-in-history position). If the current fitness is better (i.e., with smaller error), its individual best-in-history position is updated with the current position. Simultaneously, the program identifies the position with the best fitness among all particles' individual best-in-history positions and updates it as the global best-in-history position for the entire population. Subsequently, each particle, based on its current velocity, its own best-in-history position, the population's global best-in-history position, and preset learning factors and inertia weights, calculates its next-generation velocity and new position according to the standard velocity-position update formula of the particle swarm optimization algorithm. The new position represents an optimized new exploration point in the parameter space, potentially yielding smaller errors.

[0110] (4) Repeat the iterative process until the termination condition is reached, and use the final global optimal position as the optimal correction parameter.

[0111] It should be noted that the termination condition can be set to reach a preset maximum number of iterations, or the improvement in the global optimal fitness of the population over multiple consecutive iterations is less than a preset minimum threshold (indicating convergence). When the termination condition is met and the loop ends, the global historical optimal position vector currently recorded by the algorithm is the set of specific correction parameter values ​​that minimize the correction error function value after repeated iterative optimization. This vector is the optimal correction parameter output for this step.

[0112] S107. Update the APU system simulation model based on the optimal correction parameters to obtain the corrected APU characteristic model.

[0113] Specifically, in the calculation program, the optimal correction parameter vector output in step S106 is used as a set of determined coefficients, which are multiplied (or applied to) the corresponding component characteristics in the APU system simulation model by other preset mathematical relationships. After this coefficient replacement or calculation is completed, the updated model is obtained. The corrected APU characteristic model obtained in this way makes the results of subsequent simulation calculations closer to the real-world model version.

[0114] It should be noted that after obtaining a high-precision, corrected APU characteristic model, this method further provides a performance optimization stage. The goal of this stage is not to revise the model again to match the current situation, but to use this accurate model as a virtual experimental platform to explore directions for improving the characteristics of APU components, in order to solve the problem of how to design the next generation of components to achieve better overall system performance.

[0115] Specifically, after obtaining the modified APU characteristic model, the process includes:

[0116] (1) Based on the modified APU characteristic model, determine the component characteristic parameters to be optimized and the system-level optimization target.

[0117] It should be noted that the component characteristic parameters to be optimized are different from the correction parameters adjusted during the model correction stage. The object of optimization here is the design value of the component characteristic parameters themselves, such as the design flow rate and design pressure ratio of the compressor, or the design expansion ratio of the power turbine. Optimizing these parameters can redesign the characteristics of the component. This application provides three optimization parameter combination schemes, such as simultaneously optimizing the compressor flow rate and pressure ratio, or simultaneously optimizing the turbine flow rate and expansion ratio, or jointly optimizing all four parameters.

[0118] System-level optimization objectives are one or more higher-level system performance metrics. For example, a system-level optimization objective could be to minimize the total number of times a system's surge margin is consistently maintained above a safe threshold (e.g., 5%) under P different operating conditions covering the entire typical operating range of the APU, while simultaneously minimizing the total number of times key parameters such as exhaust temperature and compressor outlet pressure exceed specified limits. This system-level optimization objective directly relates to the APU's safety, reliability, and operating envelope.

[0119] (2) The particle swarm optimization algorithm is used, with the system-level optimization objective as the fitness function, to iteratively optimize the characteristic parameters of the component to be optimized.

[0120] The particle swarm optimization algorithm is used again, but this time the optimization object is changed to the component characteristic design parameters determined in step (1), and the optimization target is changed to the system-level performance index calculated based on the modified model.

[0121] Specifically, the algorithm execution flow is similar to S106, but when evaluating a particle (i.e., a set of candidate component characteristic design parameters), the program substitutes these candidate design parameters into the modified APU characteristic model, and then runs the model under all P preset operating conditions. Based on the model output, it calculates the value of the system-level optimization objective function (e.g., the sum of the number of margin deficiencies) and uses it as the fitness of the particle. The algorithm iterates to find the component characteristic parameter combination that makes the fitness function optimal (e.g., the minimum number of total violations).

[0122] (3) When the iterative optimization reaches the termination condition, output a set of optimal component characteristic parameters as the actual optimized characteristics.

[0123] When the particle swarm optimization algorithm meets the termination condition (such as reaching the maximum number of iterations or convergence), its output global optimal position vector is a set of component characteristic target values ​​that enable the system to achieve optimal comprehensive performance under preset multi-conditions. These are the optimized actual characteristics, which represent the ideal component characteristic targets for future performance improvement that can be derived based on the current actual system model.

[0124] The method provided in this embodiment establishes an APU system simulation model including a compressor, combustion chamber, and power turbine. After determining the correction parameters and objectives, it constructs a correction error function and formulates differentiated optimization strategies based on the identity of the updated components. This enables intelligent allocation and targeted optimization of test data before and after component updates. When a local component of the APU is updated, the system can correct the parameters of the unreplaced components based on the first test dataset before the update, and simultaneously use the second test dataset after the update to accurately correct the characteristics of the new components. This effectively solves the model bias problem caused by insufficient test data for the new system. Furthermore, this method uses a particle swarm optimization algorithm to efficiently optimize the correction error function, obtain the optimal correction parameters, and update the model. Ultimately, this improves the prediction accuracy and overall system matching degree of the corrected APU characteristic model under various operating conditions, providing a reliable foundation for subsequent performance optimization and reliable operation of the APU.

[0125] For better illustration, this application also provides a specific embodiment.

[0126] This embodiment updates the compressor in the APU, including 100 sets of APU1 system test data before component replacement (first test data set), and 15 sets of APU2 test data after component replacement (second test data set).

[0127] Step one: First, mathematical modeling is performed based on the characteristics of the compressor, combustion chamber, and power turbine components used. Then, based on the various equilibrium relationships, a simulation model of the APU system is established.

[0128] Flow conservation: ;

[0129] Pressure conservation: ;

[0130] Energy conservation: ;

[0131] Step two involves determining the APU's correction parameters and objectives. The correction parameters include six parameters: compressor flow rate, compressor pressure ratio, compressor efficiency, power turbine flow rate, power turbine expansion ratio, and power turbine efficiency. The model correction objective includes the compressor outlet pressure. Compressor outlet temperature Fuel flow and the outlet temperature of the power turbine The corrected objective error function is:

[0132] , Where N represents the total number of correction iterations.

[0133] Step 3: Select test data. Choose three options from the test data: 1) Select 100 sets of APU1 test data to correct the 6 parameters in Step 2; 2) Select 15 sets of APU2 test data to correct the above 6 parameters; 3) Select 100 sets of APU1 test data to correct the three parameters of power turbine flow rate, power turbine expansion ratio, and power turbine efficiency, and select 15 sets of APU2 test data to correct the three parameters of compressor flow rate, compressor pressure ratio, and compressor efficiency.

[0134] Step 3: Model refinement based on particle swarm optimization (PSO) parameters. The six refinement parameters are initialized, given a population size of 100 particles, two learning factors of 1.3, maximum and minimum inertia weights of 0.8 and 0.4 respectively, and maximum and minimum refinement parameter velocities of 0.01 and -0.01 respectively. Additionally, the maximum refinement parameter for compressor efficiency and turbine efficiency is 1, and the minimum refinement parameter is 0.8; for the remaining refinement parameters, the maximum refinement parameter is 1.2, and the minimum refinement parameter is 0.8. A chaotic initialization method is used to randomly generate an initial population of the same dimension based on the refinement parameter dimensions. After initialization, iterative optimization begins. For each particle i in the nth iteration, the fitness is first evaluated, and the individual's historical best position is updated. The global best position is updated by comparing the local optimum and the global historical optimum. Then, the particle velocity and position are updated, and finally, the current global historical best position and its corresponding fitness value are output.

[0135] Step four: Obtain the corrected model based on the corrected results. According to the different methods of selecting experimental data, three sets of optimal solutions for the corrected parameters are obtained. Comparing the correction error functions corresponding to these three sets of corrected parameters, it is found that the correction error function corresponding to the third set of experimental data is the lowest. Therefore, the corrected parameters obtained based on this experimental data are the corrected results of this embodiment. The error between the original simulation model and the experimental data is 9.8%, and the error between the corrected model and the experimental data is reduced to 4.3%.

[0136] Step 5: Determine the optimization parameters and optimization objectives. There are three ways to select the optimization parameters: 1) Select two parameters: compressor flow rate and compressor pressure ratio; 2) Select two parameters: power turbine flow rate and power turbine expansion; 3) Select four parameters: compressor flow rate, compressor pressure ratio, power turbine flow rate, and power turbine expansion. Given 42 operating conditions, the simulation model's calculation results for these 42 operating conditions before optimization show that the margin is negative in 9 conditions, and the margin is between 0% and 5% in 5 conditions. Based on the optimization parameters, simulation calculations are performed on these 42 operating conditions. The model optimization objective is to minimize the number of times the exhaust temperature, compressor outlet pressure, and compressor outlet temperature fail to meet system requirements, while ensuring a minimum margin of 5% in all operating conditions.

[0137] Step Six: Performance Optimization Based on Particle Swarm Optimization (PSO) Parameters. Initialize the two or four optimization parameters mentioned above. Given a population size (N=50), both learning factors are 1.3, the maximum and minimum inertia weights are 0.8 and 0.4 respectively, the maximum and minimum correction parameters are 1.2 and 0.9 respectively, and the maximum and minimum correction parameters are 0.01 and -0.01 respectively. A chaotic initialization method is used to randomly generate an initial population of the same dimension based on the optimization parameter dimension. After initialization, iterative optimization begins. For each particle i in the nth iteration, first evaluate its fitness, update its historical best position, update the global best position by comparing the local optimum and the global historical optimum, then update the particle velocity and position, and finally output the current global historical best position and its corresponding fitness value.

[0138] Step 7: Obtain the optimization model based on the optimization results. Three optimal solutions are obtained according to different selection methods for optimization parameters. Comparing the optimization error functions corresponding to these three sets of optimization parameters, it is found that the third optimization parameter, which simultaneously optimizes the compressor performance and the power turbine performance, has the lowest optimization error function. Therefore, the third optimization parameter represents the optimized actual characteristics in this embodiment.

[0139] Step 8: Obtain optimized design characteristics based on optimized actual characteristics. Referring to the relationship between the original characteristics and the original actual characteristics, the optimized design characteristics can be obtained based on the optimized actual characteristics: the corresponding press flow rate needs to be increased by 15.1%, the pressure ratio needs to be reduced by 12.9%, the power turbine flow rate should be increased by 12.97%, and the expansion ratio should be reduced by 6.15%. After optimization, the margin for all operating conditions is above 5%.

[0140] Example 2

[0141] Corresponding to the aforementioned embodiment of an APU characteristic correction method, this application also provides an embodiment of an APU characteristic correction device.

[0142] Figure 2 This is a schematic diagram of the structure of Embodiment 2 of the APU characteristic correction device provided in this application. Please refer to... Figure 2 The apparatus provided in this embodiment includes an establishment module 210, a determination module 220, a calculation module 230, a construction module 240, a processing module 250, and an update module 260.

[0143] The establishment module 210 is used to establish a simulation model of an APU system including a compressor, a combustion chamber, and a power turbine;

[0144] The determining module 220 is used to determine multiple correction parameters and correction targets;

[0145] The calculation module 230 is used to calculate the calculated value corresponding to the correction target based on the APU system simulation model and the correction parameters;

[0146] The construction module 240 is used to construct a correction error function based on the calculated value of the correction target and the experimental value of the correction target obtained through experiments;

[0147] The determining module 220 is further configured to determine the updated component in the APU, and determine an optimization strategy based on the identity of the updated component; the optimization strategy includes optimizing a first subset of correction parameters based on a first test dataset before the component update, and optimizing a second subset of correction parameters based on a second test dataset after the component update;

[0148] The processing module 250 is used to execute the particle swarm optimization algorithm based on the optimization strategy to obtain the optimal correction parameters corresponding to the correction error function.

[0149] The update module 260 is used to update the APU system simulation model based on the optimal correction parameters to obtain the corrected APU characteristic model.

[0150] The apparatus of this embodiment can be used to perform... Figure 1The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0151] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0152] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0153] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An APU characteristic correction method characterized by, The method comprises: establishing an APU system simulation model comprising a compressor, a combustor and a power turbine; determining a plurality of correction parameters and correction targets; calculating a calculated value corresponding to the correction target based on the APU system simulation model and the correction parameters; constructing a correction error function based on the calculated value of the correction target and a test value of the correction target obtained through testing; determining an updated component in the APU, and determining an optimization strategy based on the identity of the updated component; the optimization strategy comprises optimizing a first correction parameter subset based on a first test data set before component updating, and optimizing a second correction parameter subset based on a second test data set after component updating; performing a particle swarm algorithm based on the optimization strategy to obtain optimal correction parameters corresponding to the correction error function; updating the APU system simulation model based on the optimal correction parameters to obtain a corrected APU characteristic model; the optimization strategy comprises: optimizing all correction parameters using a first test data set before component updating; optimizing all correction parameters using a second test data set after component updating; allocating the first test data set and the second test data set to optimize the first correction parameter subset and the second correction parameter subset, respectively, based on the identity of the updated component; the allocating the first test data set and the second test data set to optimize the first correction parameter subset and the second correction parameter subset, respectively, based on the identity of the updated component, comprises: if the updated component is a compressor, the second test data set is allocated to optimize a first correction parameter subset comprising compressor flow rate, compressor pressure ratio and compressor efficiency, and the first test data set is allocated to optimize a second correction parameter subset comprising power turbine flow rate, power turbine expansion ratio and power turbine efficiency; if the updated component is a power turbine, the second test data set is allocated to optimize a second correction parameter subset comprising power turbine flow rate, power turbine expansion ratio and power turbine efficiency, and the first test data set is allocated to optimize a first correction parameter subset comprising compressor flow rate, compressor pressure ratio and compressor efficiency.

2. The method of claim 1, wherein, the establishing an APU system simulation model comprising a compressor, a combustor and a power turbine, comprises: establishing a compressor model representing the characteristic relationship between compressor outlet flow rate, outlet temperature, outlet pressure and consumed power, shaft rotation speed, inlet pressure, inlet temperature and pressure ratio; establishing a combustor model representing the characteristic relationship between combustor outlet flow rate, outlet temperature and outlet pressure, inlet pressure, inlet temperature, inlet flow rate and fuel quantity; establishing a power turbine model representing the characteristic relationship between power turbine outlet flow rate, outlet temperature, outlet pressure and output power, shaft rotation speed, inlet pressure, inlet temperature and expansion ratio; constructing the APU system simulation model based on the compressor model, the combustor model and the power turbine model, and according to the flow conservation, pressure conservation and energy conservation relationship.

3. The method of claim 1, wherein, The determining multiple correction parameters and correction targets comprises: Based on the component performance characteristics existing deviation in the APU system simulation model, the correction parameters for characterizing the component performance variation are determined; the correction parameters include compressor flow, compressor pressure ratio, compressor efficiency, power turbine flow, power turbine expansion ratio and power turbine efficiency; Based on the characteristic quantities in the APU system simulation model which can be obtained through experiments and can reflect the system output performance, the correction targets are determined; the correction targets include compressor outlet pressure, compressor outlet temperature, fuel flow and power turbine outlet temperature.

4. The method of claim 1, wherein, The correction error function is constructed, comprising: Based on the calculated values and corresponding experimental values of the correction targets, the deviations of each correction target are calculated; The deviations of each correction target are weighted according to preset weights to obtain corresponding target errors; All target errors are accumulated to construct the correction error function.

5. The method of claim 1, wherein, The correction error function is constructed, comprising: For the i-th group of test data, the relative errors between the calculated values and experimental values of the compressor outlet pressure, the compressor outlet temperature, the fuel flow and the power turbine outlet temperature are calculated respectively; Each of the relative errors is squared and summed, and then divided by the number of correction targets to obtain the average squared error; The average squared error is square-rooted to obtain the single-group root mean square error of the i-th group of test data; The single-group root mean square errors of all N groups of test data are repeatedly calculated to obtain N single-group root mean square errors; The N single-group root mean square errors are summed and divided by N to obtain the final value of the correction error function.

6. The method of claim 1, wherein, Based on the optimization strategy, the particle swarm algorithm is executed to obtain the optimal correction parameters corresponding to the correction error function, comprising: The correction parameters are initialized, and the running parameters of the particle swarm algorithm are set; According to the test data usage mode determined by the optimization strategy, in each iteration, when each particle is evaluated, the value of the correction error function is calculated by calling the corresponding test data set, which is used as the fitness of the particle; The individual optimal position and the global optimal position of the population of the particle are updated according to the fitness, and the speed and position of the particle are also updated; The iteration process is repeated until the termination condition is reached, and the finally obtained global optimal position is taken as the optimal correction parameter.

7. The method of claim 1, wherein, After the corrected APU characteristic model is obtained, comprising: Based on the corrected APU characteristic model, the component characteristic parameters to be optimized and the system-level optimization target are determined; The particle swarm optimization algorithm is adopted, and the system-level optimization target is taken as the fitness function to iteratively optimize the component characteristic parameters to be optimized; When the iterative optimization reaches the termination condition, a set of optimal component characteristic parameters are output as the optimized actual characteristics.

8. An APU characteristic correction device characterized by comprising: The device comprises an establishment module, a determination module, a calculation module, a construction module, a processing module and an updating module; The establishment module is configured to establish an APU system simulation model comprising a compressor, a combustor and a power turbine; The determination module is configured to determine multiple correction parameters and correction targets; The computing module is configured to calculate a calculation value corresponding to the correction target based on the APU system simulation model and the correction parameters; The constructing module is configured to construct a correction error function based on the calculation value of the correction target and a test value of the correction target obtained through a test; The determining module is further configured to determine an updated component in the APU, and determine an optimization strategy based on an identity of the updated component; The optimization strategy comprises optimizing a first correction parameter subset based on a first test data set before component update, and optimizing a second correction parameter subset based on a second test data set after component update; The processing module is configured to execute a particle swarm algorithm based on the optimization strategy to obtain optimal correction parameters corresponding to the correction error function; The updating module is configured to update the APU system simulation model based on the optimal correction parameters to obtain a corrected APU characteristic model; The optimization strategy comprises: optimizing all correction parameters based on a first test data set before component update; optimizing all correction parameters based on a second test data set after component update; allocating the first test data set and the second test data set to optimize the first correction parameter subset and the second correction parameter subset, respectively, based on the identity of the updated component; The allocating the first test data set and the second test data set to optimize the first correction parameter subset and the second correction parameter subset, respectively, based on the identity of the updated component, comprises: if the updated component is a compressor, allocating the second test data set to optimize a first correction parameter subset comprising compressor flow, compressor pressure ratio and compressor efficiency, and allocating the first test data set to optimize a second correction parameter subset comprising power turbine flow, power turbine expansion ratio and power turbine efficiency; if the updated component is a power turbine, allocating the second test data set to optimize a second correction parameter subset comprising power turbine flow, power turbine expansion ratio and power turbine efficiency, and allocating the first test data set to optimize a first correction parameter subset comprising compressor flow, compressor pressure ratio and compressor efficiency.

Images