A data-driven method for calculating the service loads of aircraft engine turbine disks

Abstract

The present invention belongs to the field of aviation engine technology, and specifically relates to a data-driven aviation engine turbine disk service load calculation method, including collecting service history data of the engine to be evaluated; using the service history data to infer and correct the thermodynamic performance history data of the engine to be evaluated based on the engine thermal cycle principle; extracting the turbine disk working condition history data based on the service history data of the engine to be evaluated and the corrected thermodynamic performance history data of the engine to be evaluated; using a data-driven method to establish and verify a turbine disk transient load proxy model, using the turbine disk working condition history data as input, and calculating the turbine disk service load through the turbine disk transient load proxy model. Compared with traditional numerical simulation methods, the present invention improves the calculation efficiency while ensuring the calculation accuracy and retains the stress and temperature response characteristics under transient loads.

Description

Technical Field

[0001] The invention belongs to the technical field of aero-engines, and in particular relates to a data-driven method for calculating the service load of an aero-engine turbine disk. Background Art

[0002] The turbine disk of an aircraft engine is responsible for converting the internal energy of the combustion gas into mechanical energy and transferring it to the supercharging working components. The turbine disk structure fixes the turbine blades and spans the high-temperature flow channel and the second flow channel where the cooling gas is located. The load environment is harsh and the load types are complex. The durability of the turbine disk is an important constraint on the life of the entire machine. The industry generally defines all turbine disks as key parts for the life of the engine. Durability assessment of the turbine disk, especially damage analysis under service loads, requires the load history at key positions of the disk as input. The loads at key positions involve aspects such as temperature or stress.

[0003] The performance and service life of modern advanced aircraft engines are constantly improving, and the diversity of engine missions and the time-varying nature of flight conditions are becoming increasingly apparent. In this context, on the one hand, the traditional paradigm of analyzing and determining the life of the turbine disk loads based on a few design working conditions is difficult to reflect the actual use of the engine. On the other hand, the structural characteristics of the turbine disk result in obvious time-lag characteristics of the local loads under time-varying engine conditions. In the classical method, the loads at key positions of the turbine disk are generally obtained through numerical analysis methods with high computational requirements, which means unacceptable computational costs in practice under the conditions of massive engine service data and relatively large parameter dimensions.

[0004] Patent CN116384012A "A method for predicting the stress of turbine disks in aircraft engines" is based on the simulation of turbine disk loads under steady-state conditions, with reduced-dimensional engine overall parameters as input and turbine disk stress under steady-state conditions as output, to establish a random forest prediction model to predict the stress of turbine disks under other conditions. However, this invention does not consider the time domain response characteristics of turbine disk loads under transient flight spectra, and is not suitable for situations where the engine operating state changes frequently or drastically.

[0005] Considering the current technical background, it is necessary to establish a calculation method that can efficiently calculate the service load of turbine disks, which can be used to calculate the load of turbine disks under complex service flight spectra and provide effective input for disk durability evaluation. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a data-driven method for calculating the service load of an aircraft engine turbine disk. Through the data-driven method, a direct mapping of the temperature and mechanical boundary conditions of the turbine disk with the load response at the local position is established. Compared with the traditional numerical simulation method, the calculation efficiency is improved while ensuring the calculation accuracy and the stress and temperature response characteristics under transient loads are retained. The present invention is applied to calculate the load history of the key position of the turbine disk damage when the service history of the engine is known.

[0007] The present invention is implemented by providing a data-driven method for calculating the service load of an aircraft engine turbine disk, comprising the following steps:

[0008] Step A1: collecting service history data of the engine to be evaluated, including operating environment history data, power level history data and thermodynamic performance parameters of several sections of the engine during the service history;

[0009] Step A2: Based on the engine thermodynamic cycle principle, using the service history data in step A1, the thermodynamic performance history data of the engine to be evaluated is estimated and corrected;

[0010] Step A3: extracting turbine disk working condition history data according to the service history data of the engine to be evaluated obtained in step A1 and the corrected thermodynamic performance history data of the engine to be evaluated obtained in step A2;

[0011] Step A4: Use a data-driven approach to establish and verify a turbine disk transient load proxy model, use the turbine disk operating condition history data obtained in step A3 as input, and calculate the turbine disk service load through the turbine disk transient load proxy model.

[0012] Preferably, the turbine disk service load includes temperature and stress.

[0013] Further preferably, the cross section is the interface between the engine compressor, combustion chamber and turbine.

[0014] Further preferably, in step A2, the method for calculating and correcting the thermodynamic performance history data of the engine to be evaluated is:

[0015] A21: Based on the general method for solving the thermal state of the engine, the engine to be evaluated without performance degradation is taken as the benchmark, the engine model is established according to the thermodynamic cycle principle, and the service history data of the engine to be evaluated without performance degradation is used as the input of the engine model to determine the engine thermodynamic performance history data of the engine to be evaluated without performance degradation under the service history;

[0016] A22: Correct the thermodynamic performance history data of the engine to be evaluated without performance degradation obtained in step A21 according to the thermodynamic performance parameters of several sections in the engine service history data. The correction method is specifically as follows:

[0017] The relationship between the unmeasurable and measurable thermodynamic performance parameters of the engine is expressed as a vector function :

[0018]

[0019] in: represents the unmeasurable engine thermodynamic performance parameter, Represents the measurable engine thermodynamic performance parameters, taking the thermodynamic performance parameters of the engine without performance degradation in the design state as the reference value, and Taking the first-order Taylor expansion, the deviation of the engine thermodynamic performance parameters is Deviations from unmeasurable and measurable thermodynamic property parameters The linear relationship between them is expressed as:

[0020] ;

[0021] ;

[0022] In the formula, HOT is a high-order term that can be ignored. It is defined as the influence coefficient matrix, where the subscript 0 represents the reference value of the thermodynamic performance parameter. The deviation of the engine thermodynamic performance parameter is calculated by the following formula: , according to the deviation Correct the thermodynamic performance history data of the engine to be evaluated:

[0023]

[0024] When the influence coefficient matrix When is a reversible matrix, for The inverse matrix of; When the number of unmeasurable and measurable thermodynamic performance parameters of the engine is unequal, Pick The pseudo-inverse matrix of .

[0025] Further preferably, in step 4, establishing the turbine disk transient load proxy model using a data-driven method comprises the following steps:

[0026] Step B1: Investigate the actual use of the engine to be evaluated and collect typical mission profiles representing the engine's operating state range within the engine's service mission range;

[0027] Step B2: Under typical tasks, based on the engine thermodynamic cycle principle and the grid-based numerical simulation method, calculate the transient load history data of the turbine disk test position;

[0028] Step B3: using the turbine disk working condition history data as input and the transient load history data of the test position as output to form a data set required for establishing a turbine disk transient load proxy model, and preprocessing the data set;

[0029] Step B4: Select a time series-oriented data-driven theoretical method to establish a turbine disk transient load proxy model, train the turbine disk transient load proxy model based on the preprocessed data set, and verify the model effect.

[0030] Further preferably, the typical tasks are obtained through actual usage statistics of external field engines.

[0031] Further preferably, in step B4, the data-driven theoretical method for time series includes system identification theory, sequence-oriented convolutional neural network and recurrent neural network.

[0032] Compared with the prior art, the advantages of the present invention are:

[0033] A data-driven method for calculating the service load of aircraft engine turbine disks is provided to obtain the load spectrum of turbine disks under actual service history in the field, which provides a basis for damage assessment and predictive maintenance of turbine disks based on usage. Specifically:

[0034] (1) Under massive flight conditions, the present invention greatly improves the efficiency of load calculation compared with the traditional grid-based numerical simulation method;

[0035] (2) The prediction effect of the present invention retains the time domain characteristics of the thermal response of the turbine disk and provides more accurate temperature and stress results under complex flight missions where the engine state changes dramatically. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is the thermal boundary condition of the turbine disk under typical missions;

[0037] Figure 2 It is the structure of the roulette transient load proxy model based on the LSTM network;

[0038] Figure 3 This is the comparison of the calculated results of turbine disk wheel center temperature;

[0039] Figure 4 This is a comparison of the calculated results of Mises stress at the turbine disk center. DETAILED DESCRIPTION

[0040] The present invention is now further described in conjunction with the accompanying drawings and specific implementation examples. It should be understood that the specific implementation examples described herein are only used to explain the present invention and are not used to limit the present invention.

[0041] The present invention provides a data-driven method for calculating the service load of an aircraft engine turbine disk, comprising the following steps:

[0042] Step A1: collecting the service history data of the engine to be evaluated recorded by the flight history recorder, including the operating environment history, power level history and thermodynamic performance parameters of several sections of the engine in the service history; the operating environment history involves the engine inlet temperature and inlet pressure directly recorded or obtained by converting the ground temperature, Mach number and altitude; the power level history includes the engine fuel flow, throttle lever angle and rotor speed, etc.; the thermodynamic performance parameter of the engine section is the turbine outlet temperature history data;

[0043] In this embodiment, the engine service history data includes the engine inlet temperature, inlet pressure, turbine outlet temperature history data, and engine rotor speed;

[0044] Step A2: Using the engine service history data collected in step A1, according to the engine thermodynamic cycle principle, analyze the thermodynamic performance history data of the engine to be evaluated without performance degradation, in this embodiment, including the thermodynamic performance history data of the key cross section of the main flow channel and the cooling gas path. Further, according to the engine flow channel temperature in the thermodynamic performance parameters of the engine flow channel cross section, the remaining flow channel thermodynamic parameters are corrected to obtain the thermodynamic performance history data of several cross sections of the main flow channel of the evaluated engine that actually considers the performance difference of a single machine. Among them, several cross sections refer to the interfaces between engine compressors, combustion chambers, turbines and other components.

[0045] Step A2 specifically includes:

[0046] A21: Based on the general method for solving the thermal state of the engine, the engine to be evaluated without performance degradation is taken as the benchmark, the engine model is established according to the thermodynamic cycle principle, and the service history data of the engine to be evaluated without performance degradation is used as the input of the engine model to determine the engine thermodynamic performance history data of the engine to be evaluated without performance degradation under the service history;

[0047] A22: Correct the thermodynamic performance history data of the engine to be evaluated without performance degradation obtained in step A21 according to the thermodynamic performance parameters of several sections in the engine service history data. The correction method is specifically as follows:

[0048] The relationship between the unmeasurable and measurable thermodynamic performance parameters of the engine is expressed as a vector function :

[0049]

[0050] in: represents the unmeasurable engine thermodynamic performance parameter, Represents the measurable engine thermodynamic performance parameters, taking the thermodynamic performance parameters of the engine without performance degradation in the design state as the reference value, and Taking the first-order Taylor expansion, the deviation of the engine thermodynamic performance parameters is Deviations from unmeasurable thermodynamic property parameters and measurable thermodynamic property parameters The linear relationship between them is expressed as:

[0051] ;

[0052] ;

[0053] In the formula, HOT is a high-order term that can be ignored. It is defined as the influence coefficient matrix, where the subscript 0 represents the reference value of the thermodynamic performance parameter. The deviation of the engine thermodynamic performance parameter is calculated by the following formula: , according to the deviation Correct the thermodynamic performance history data of the engine to be evaluated:

[0054]

[0055] When the influence coefficient matrix When is a reversible matrix, for The inverse matrix of; When the number of unmeasurable and measurable thermodynamic performance parameters of the engine is unequal, Pick The pseudo-inverse matrix of .

[0056] Step A3: extracting turbine disk working condition history data from the service history data of the engine obtained in step A1 and the modified engine thermodynamic performance history data obtained in step A2, including turbine disk thermal load boundary condition history data and turbine disk speed. The turbine disk thermal load boundary condition history data includes the turbine disk inlet temperature and the cooling airflow inlet temperature. The turbine disk speed is equivalent to the speed of the rotor where the turbine disk is located.

[0057] Step A4: Based on the simulation of the transient temperature and stress field of the turbine disk under typical mission conditions, a data-driven method is used to establish a turbine disk transient load proxy model. The turbine disk service load is calculated using the turbine disk working condition history data obtained in step A3 using the turbine disk transient load proxy model.

[0058] In this step, the establishment of the load proxy model specifically includes the following steps, which are explained by taking the high-pressure turbine disk of a certain type of turbofan engine as an object:

[0059] Step B1: Investigate the actual use of the engine and collect typical mission profiles of the engine; the typical mission refers to representative engine service history data within the scope of the engine service mission, covering the range of working conditions during the engine service process;

[0060] Step B2: using the turbine disk working condition history data and the thermal boundary conditions as constraints, using a grid-based numerical simulation method to calculate the transient temperature field of the turbine disk under a typical task, and using the transient temperature field of the wheel disk and the turbine disk speed in the turbine disk working condition history data as boundary conditions, using a grid-based numerical simulation method to calculate the stress field of the turbine disk under a typical task; Figure 1 The data are the boundary condition history data of thermal load on the turbine disk of a certain type of engine under 9 typical flight missions.

[0061] Step B3: Taking the recording interval of the engine history recorder as the time step, extract the load history data at the test position from the calculation results of the transient temperature field and stress field of the turbine disk in step B2. The load history data includes temperature and stress state history. The stress state reflects the mechanical load intensity at the test position. According to different durability assessment requirements, the reference range includes but is not limited to the stress component in a certain direction, the stress calculated according to a certain strength theory, and the effective stress that comprehensively considers the regional stress distribution. In this example, the Mises equivalent stress at the wheel center is extracted with a time step of 1s to demonstrate the establishment of the load proxy model.

[0062] Step B4: Establishment and verification of turbine disk transient load proxy model. This example establishes a turbine disk transient load proxy model based on LSTM neural network. The neural network structure is as follows: Figure 2 . Taking the turbine disk working condition history data as input and the load history data at the test position as output, a data set required for establishing a turbine disk transient load proxy model is formed, and the data set is preprocessed as necessary. Preferably, the preprocessing includes data normalization and forward filling. Taking tasks 1#-8# as training data sets and task 9# as test data sets, adjust network parameters and train the neural network. In this example, the number of units in the input layer, LSTM layer, fully connected layer and output layer are [4, 16, 8, 2] respectively, and the activation function of the fully connected layer is the tanh function. The parameters of the neural network are trained using the Adam method. Figure 3-4The comparison between the turbine disk transient load proxy model prediction and the numerical simulation results on the wheel center temperature and Mises stress on the test set is given. The maximum temperature prediction error is 22°C and the maximum stress prediction error is 8MPa, which meets the load prediction error requirements in structural service damage analysis.

[0063] Through this example, the present invention provides a process for calculating the transient load of a turbine disk based on the actual flight history of the engine, and provides a method for establishing a data-driven proxy model of the transient load of the disk. The present invention realizes the efficient calculation of the service load of the disk based on the engine flight data record, and fully considers the time domain characteristics of the disk load when calculating the load, which is of great significance to the study of engine load and durability.

Claims

1. A data-driven method for calculating the service load of an aircraft engine turbine disk, characterized in that: The steps include: Step A1: collecting service history data of the engine to be evaluated, including operating environment history data, power level history data and thermodynamic performance parameters of several sections of the engine during the service history; Step A2: Based on the engine thermodynamic cycle principle, using the service history data in step A1, the thermodynamic performance history data of the engine to be evaluated is estimated and corrected; Step A3: extracting turbine disk working condition history data according to the service history data of the engine to be evaluated obtained in step A1 and the corrected thermodynamic performance history data of the engine to be evaluated obtained in step A2; Step A4: using a data-driven method to establish and verify a turbine disk transient load proxy model, using the turbine disk working condition history data obtained in step A3 as input, and calculating the turbine disk service load through the turbine disk transient load proxy model; In step A4, establishing a turbine disk transient load proxy model using a data-driven method includes the following steps: Step B1: Investigate the actual use of the engine to be evaluated and collect typical mission profiles representing the engine's operating state range within the engine's service mission range; Step B2: Under typical tasks, based on the engine thermodynamic cycle principle and the grid-based numerical simulation method, calculate the transient load history data of the turbine disk test position; Step B3: using the turbine disk working condition history data as input and the transient load history data of the test position as output to form a data set required for establishing a turbine disk transient load proxy model, and preprocessing the data set; Step B4: Select a time series-oriented data-driven theoretical method to establish a turbine disk transient load proxy model, train the turbine disk transient load proxy model based on the preprocessed data set, and verify the model effect.

2. The data-driven aircraft engine turbine disk service load calculation method according to claim 1, characterized in that: The turbine disk service loads include temperature and stress.

3. The data-driven aircraft engine turbine disk service load calculation method according to claim 1, characterized in that: The cross section is the interface between the engine compressor, combustion chamber and turbine.

4. The data-driven aircraft engine turbine disk service load calculation method according to claim 1, characterized in that: In step A2, the method for calculating and correcting the thermodynamic performance history data of the engine to be evaluated is: A21: Based on the general method for solving the thermal state of the engine, the engine to be evaluated without performance degradation is taken as the benchmark, the engine model is established according to the thermodynamic cycle principle, and the service history data of the engine to be evaluated without performance degradation is used as the input of the engine model to determine the engine thermodynamic performance history data of the engine to be evaluated without performance degradation under the service history; A22: Correct the thermodynamic performance history data of the engine to be evaluated without performance degradation obtained in step A21 according to the thermodynamic performance parameters of several sections in the engine service history data. The correction method is specifically as follows: The relationship between the unmeasurable and measurable thermodynamic performance parameters of the engine is expressed as a vector function : ； in: represents the unmeasurable engine thermodynamic performance parameter, Represents the measurable engine thermodynamic performance parameters, taking the thermodynamic performance parameters of the engine without performance degradation in the design state as the reference value, and Taking the first-order Taylor expansion, the deviation of the engine thermodynamic performance parameters is Deviations from unmeasurable and measurable thermodynamic property parameters The linear relationship between them is expressed as: ； ； In the formula, HOT is a high-order term that can be ignored. It is defined as the influence coefficient matrix, where the subscript 0 represents the reference value of the thermodynamic performance parameter. The deviation of the engine thermodynamic performance parameter is calculated by the following formula: , according to the deviation Correct the thermodynamic performance history data of the engine to be evaluated: ； When the influence coefficient matrix When is a reversible matrix, for The inverse matrix of; When the number of unmeasurable and measurable thermodynamic performance parameters of the engine is unequal, Pick The pseudo-inverse matrix of .

5. The data-driven aircraft engine turbine disk service load calculation method according to claim 1, characterized in that: Typical tasks are obtained through actual usage statistics of external field engines.

6. The data-driven aircraft engine turbine disk service load calculation method according to claim 1, characterized in that: In step B4, the data-driven theoretical methods for time series include system identification theory, sequence-oriented convolutional neural networks, and recurrent neural networks.

Images