Online Performance Estimation and Health Assessment Method for Aero-engines Based on Kalman Filtering and Iterative Backpropagation

Abstract

This invention relates to the field of aero-engine health management technology, and discloses an online performance estimation and health assessment method for aero-engines based on Kalman filtering and iterative back-engineering. This method comprehensively combines physical consistency and data adaptability in performance estimation, enabling real-time correction of sensor data during flight, iterative back-engineering of key aerodynamic parameters, and calculation of the engine health index, thus achieving performance degradation detection and early warning. The method includes: acquiring multi-dimensional parameter signals of the aero-engine through airborne sensors; dynamically fusing and suppressing noise using an extended Kalman filter algorithm; establishing a one-dimensional performance calculation model and integrating compressor and turbine characteristic models to construct an aero-engine performance model; using an iterative back-engineering method based on isentropic relationships and introducing a neural network correction model; calculating the flow rate and efficiency degradation rate of major components to construct a health index; and triggering an edge alarm when the health index falls below a set threshold.

Description

Technical Field

[0001] This invention relates to fields such as aero-engine health management technology, and more specifically, to a method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation. Background Technology

[0002] As the core power unit of aircraft, the performance degradation, component deterioration, and measurement anomalies of aero-engines directly affect flight safety and economy. Currently, in-service performance monitoring of aero-engines mainly relies on ground-based analysis software to process flight-collected data offline, which suffers from problems such as response lag, low estimation accuracy, and poor model adaptability.

[0003] During flight, key parameters such as temperature, pressure, and flow rate collected by airborne sensors are generally affected by airflow disturbances, measurement noise, and drift errors, resulting in uncertainties in the input of the performance calculation model and causing the total temperature at the compressor outlet to fluctuate. T t3 and total pressure P t3 Key intermediate quantities are difficult to obtain accurately. In addition, traditional one-dimensional performance equation calculations are often based on the assumption of ideal isentropicity, which makes it difficult to reflect the actual variation of component characteristic curves, resulting in performance estimation errors.

[0004] In recent years, researchers have attempted to introduce neural networks and data-driven models into engine performance evaluation. However, pure data models lack physical constraints, have poor extrapolation capabilities, and are susceptible to interference from anomalous data. On the other hand, models based entirely on physical mechanisms are highly sensitive to input accuracy and are not suitable for airborne real-time computing environments. Summary of the Invention

[0005] The purpose of this invention is to overcome the reliance on ground post-processing in the existing technology for estimating the performance of aero-engines. T t3 and P t3 To address the issues of inability to directly measure performance degradation and delayed performance degradation detection, this paper provides an online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation. When performing performance estimation, this method can comprehensively combine physical consistency and data adaptability. It can correct sensor data in real time during flight, iteratively backpropagate key aerodynamic parameters, and calculate engine health index, thereby achieving performance degradation detection and early warning under edge computing conditions.

[0006] This invention is achieved through the following technical solution: an online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation, comprising the following steps: 1) Collect multi-parameter signals of the aero-engine through airborne sensors. These multi-parameter signals include temperature data, pressure data, flow rate data, and speed signals. 2) The extended Kalman filter algorithm is used to dynamically fuse and suppress noise in multi-parameter signals; 3) Establish a one-dimensional performance calculation model and integrate the compressor and turbine characteristic models to construct an aero-engine performance model; 4) Regarding the total temperature at the compressor outlet T t3 Total pressure at compressor outlet P t3 For problems that cannot be directly measured, an iterative back-calculation method based on isentropic relations is proposed, and a neural network correction model is introduced to achieve [the desired result]. T t3 , P t3 High robustness estimation; 5) Calculate the flow rate and efficiency degradation rate of the main components to construct a health index; when the health index is lower than the set threshold, trigger an edge alarm.

[0007] To further improve the online performance estimation and health assessment method for aero-engines based on Kalman filtering and iterative backpropagation described in this invention, the following settings are specifically adopted: the temperature data includes the compressor inlet total temperature. T t2 Total temperature at compressor outlet T t3 Turbine inlet total temperature T t41 Turbine outlet total temperature T t5 The pressure data includes the total compressor inlet pressure. P t2 Total pressure at compressor outlet P t3 The speed signal includes the low-pressure rotor speed. N 1. High-voltage rotor speed N 2; The flow data includes airflow at each stage and fuel flow. W f .

[0008] To further improve the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation described in this invention, the following configuration is specifically adopted: Step 2) includes the following steps: 2.1) To suppress the impact of measurement noise and sensor drift on performance calculation results, this invention uses the Extended Kalman Filter (EKF) algorithm to perform real-time preprocessing on the acquired multivariate parameter signals; 2.2) Dynamically updating the covariance matrix PThis enables smooth estimation of the measurement signal; wherein the measurement signal is selected, mapped or combined from the acquired multivariate parameter signals and used as the observation input to the observation equation of the Kalman filter process.

[0009] To further improve the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative back-engineering described in this invention, the following configuration is specifically adopted: In step 2.1), during real-time preprocessing, the Kalman filtering process includes the construction of state equations and observation equations: x k = A x k-1 + B u k + w k ; z k = H x k + v k ; in, x Let A be the engine state vector, containing temperature, pressure, and flow characteristics; B be the state transition matrix; and u be the control input matrix. w For process noise; z is the initial measurement vector of the sensor; H is the observation matrix; v For measuring noise; x k This represents the engine state vector at the current moment; x k-1 This is the engine state vector from the previous moment; u k This is the control input vector at the current moment; w k This represents the process noise at the current moment. v k The measurement noise at the current moment; z k This represents the initial measurement vector of the sensor at the current moment; In step 2.2), the smoothing estimation is achieved using the following formula: ;in, k Represents the current moment; k -1 represents the previous moment; P k Let be the state estimation error covariance matrix, and take the expected value of the product of the state estimation error vector and its transpose; E For expectations; T This is the transpose symbol.

[0010] To further improve the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative back-calculation described in this invention, the following setup is specifically adopted: Step 3) specifically involves: constructing a one-dimensional performance calculation model of the aero-engine based on the filtered multivariate parameter data of the aero-engine, and introducing compressor and turbine characteristic models into the one-dimensional performance calculation model. Parameter constraints and closed-loop solutions are applied to the one-dimensional performance calculation model. By coupling the compressor and turbine characteristic models with the one-dimensional performance calculation model, a unified description between system-level conservation relationships and actual component performance characteristics is achieved, thereby constructing an aero-engine performance model. The compressor and turbine characteristic models are used to characterize the actual working characteristics of the compressor and turbine under different operating conditions, including the flow characteristic models and efficiency characteristic models of the compressor and turbine. Both the flow characteristic models and efficiency characteristic models are represented by characteristic curves or mathematical functions and are embedded as constraints into the one-dimensional performance calculation model for solution.

[0011] To further improve the online performance estimation and health assessment method for aero-engines based on Kalman filtering and iterative backpropagation as described in this invention, the following configuration is specifically adopted: The one-dimensional performance calculation model is established based on the mass conservation equation, energy conservation equation, momentum conservation equation, and ideal gas law, and is used to describe the thermodynamic and aerodynamic relationships between the main components of the aero-engine in the axial airflow direction; The one-dimensional performance calculation model divides the aero-engine into the inlet section, compressor section, combustion chamber section, turbine section, and exhaust section, and forms a system-level calculation equation set describing the overall performance of the aero-engine by establishing the functional relationship between the inlet and outlet parameters of each adjacent component; When calculating compressor performance in the aforementioned aero-engine performance model, the compressor efficiency is determined by the compressor efficiency characteristic model, and the compressor efficiency is calculated using the following formula: ; in, or c The subscript C refers to the compressor efficiency. T t2 This refers to the compressor inlet total temperature in the temperature data. T t3 This refers to the compressor outlet total temperature in the temperature data. π c For the compressor pressure ratio, and , p t2 This refers to the total compressor inlet pressure in the pressure data. p t3 This refers to the total compressor outlet pressure in the pressure data. c The specific heat ratio of air; The aero-engine performance model uses the following formula to calculate the compressor flow rate: ; in, P s3 This refers to the static pressure at the compressor outlet. S R is the compressor outlet cross-sectional area, and R is the air gas constant. W c This refers to the compressor flow rate; The aero-engine performance model uses the following formula to calculate turbine efficiency: ; in, For turbine efficiency, the subscript T refers to the turbine. T t41 This refers to the turbine inlet total temperature in the temperature data. T t5 This refers to the total turbine outlet temperature in the temperature data. π T The turbine pressure ratio, and , p t5 This refers to the total turbine outlet pressure in the pressure data. c g Specific heat ratio of fuel gas; The aero-engine performance model uses the following formula to calculate turbine flow rate: ; in, W T For turbine flow, W c This refers to the compressor flow rate.

[0012] This aero-engine performance model can adaptively adjust the calculation coefficients under changing flight conditions, enabling consistent performance calculations of the aero-engine under different thrust commands, altitudes, and temperature conditions.

[0013] To further improve the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative back-calculation described in this invention, the following setup is specifically adopted: In step 4), the iterative back-calculation method based on isentropic relations is used, and a neural network correction model is introduced, including the following specific steps: 4.1) Given the total inlet temperature, total outlet temperature, and initial pressure ratio of the compressor. π c 0 According to the isentropic equation Preliminary calculation of the total temperature at the compressor outlet T t3 ; 4.2) Iterative solution: Apply the preliminary calculations... T t3 Substitute into the one-dimensional performance equation and recalculate. π c i+1 until ( T t3 , P t3 )convergence: | T t3 i+1 - T t3 i |< e 1, of which i The current step of the iteration. i +1 indicates the next iteration. e 1 represents the preset convergence threshold; | P t3 i+1 - P t3 i |< e 2, of which i The current step of the iteration. i +1 indicates the next iteration. e 2 represents the preset convergence threshold; 4.3) Neural Network Correction: To correct aerodynamic deviations and nonlinear errors, a BP neural network model is established: | T t3 ’ , P t3 ’ | = NN(| T t2 , P t2 , T t41 , N 1, N 2, W f |), where NN(||) is a neural network function, which can learn the error correction relationship of the BP neural network model through training; P t2 This is the total pressure at the compressor inlet. T t41 This refers to the turbine inlet total temperature in the temperature data. N 1 represents the low-pressure rotor speed.N 2 represents the high-voltage rotor speed. W f This refers to fuel flow rate.

[0014] Supervised training using multivariate parameter signals ensures that the estimated values ​​remain highly robust under different operating conditions.

[0015] To further improve the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation described in this invention, the following settings are specifically adopted: Step 5) includes the following specific steps: 5.1) Based on the intermediate parameters obtained through reverse engineering, calculate the performance degradation index of the main components: ; In this context, "real" refers to the actual measured value, and "ref" refers to the design value. Traffic decay rate, To account for the efficiency degradation rate, the main components include two types of components: compressor and turbine; 5.2) Construct an overall health index by combining the flow and efficiency degradation rates of multiple components. HI : ; in , These are weighting coefficients, determined based on the importance of the components; W For traffic, or For efficiency, The component's flow decay rate. This refers to the efficiency degradation rate of the component.

[0016] 5.3) When HI When the value falls below a set threshold, it is automatically identified as a potential performance anomaly, and a health alarm signal is triggered at the airborne edge computing node.

[0017] To further improve the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation described in this invention, the following settings are adopted: In step 5), when an edge alarm is triggered, alarm information including the abnormal component number, decay type, and estimated confidence level will be uploaded to the ground monitoring center via a data link.

[0018] Compared with the prior art, the present invention has the following advantages and beneficial effects: This invention eliminates measurement noise and drift through Kalman filtering, enabling high-reliability data input.

[0019] This invention employs an iterative back-calculation combined with a neural network correction model to significantly improve... T t3 , Pt3 Estimation accuracy.

[0020] This invention integrates characteristic curves and physical equations to achieve state adaptation of the performance model.

[0021] This invention quantifies component degradation trends and provides early warnings through health index calculation.

[0022] This invention features low computational complexity and is suitable for airborne real-time performance monitoring systems.

[0023] This invention employs a core component airflow performance evaluation technology based on multiple measurement points in aero-engines, enabling the calculation of compressor, turbine, and overall engine performance based on airborne measurement data. Compared to the previous method of evaluating the performance of service engines through periodic maintenance and ground testing, this method saves time and costs, achieves component performance quantification, and has promising application prospects. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to embodiments, but the implementation of the present invention is not limited thereto.

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0026] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] Example 1: This invention presents an online performance estimation and health assessment method for aero-engines based on Kalman filtering and iterative back-engineering, overcoming the limitations of existing technologies that rely on ground post-processing for aero-engine performance estimation. T t3 and Pt3 To address the issues of inability to directly measure performance degradation and delayed performance degradation detection, this system comprehensively combines physical consistency and data adaptability in performance estimation. It can correct sensor data in real time during flight, iteratively deduce key aerodynamic parameters, and calculate engine health indices, achieving performance degradation detection and early warning under edge computing conditions. This includes the following steps: 1) Collect multi-parameter signals of the aero-engine through airborne sensors. These multi-parameter signals include temperature data, pressure data, flow rate data, and speed signals. 2) The extended Kalman filter algorithm is used to dynamically fuse and suppress noise in multi-parameter signals; 3) Establish a one-dimensional performance calculation model and integrate the compressor and turbine characteristic models to construct an aero-engine performance model; 4) Regarding the total temperature at the compressor outlet T t3 Total pressure at compressor outlet P t3 For problems that cannot be directly measured, an iterative back-calculation method based on isentropic relations is proposed, and a neural network correction model is introduced to achieve [the desired result]. T t3 , P t3 High robustness estimation; 5) Calculate the flow rate and efficiency degradation rate of the main components to construct a health index; when the health index is lower than the set threshold, trigger an edge alarm.

[0028] Example 2: This embodiment is a further optimization based on the above embodiment. The parts identical to the aforementioned technical solutions will not be repeated here. Furthermore, to better realize the online performance estimation and health assessment method for aero-engines based on Kalman filtering and iterative backpropagation described in this invention, the following settings are specifically adopted: The temperature data includes the compressor inlet total temperature. T t2 Total temperature at compressor outlet T t3 Turbine inlet total temperature T t41 Turbine outlet total temperature T t5 The pressure data includes the total compressor inlet pressure. P t2 Total pressure at compressor outlet P t3 The speed signal includes the low-pressure rotor speed. N 1. High-voltage rotor speed N 2; The flow data includes airflow at each stage and fuel flow. W f .

[0029] Example 3: This embodiment is a further optimization based on any of the above embodiments. The parts identical to the aforementioned technical solutions will not be repeated here. Furthermore, to better realize the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation described in this invention, the following settings are specifically adopted: Step 2) includes the following steps: 2.1) To suppress the impact of measurement noise and sensor drift on performance calculation results, this invention uses the Extended Kalman Filter (EKF) algorithm to perform real-time preprocessing on the acquired multivariate parameter signals; 2.2) Dynamically updating the covariance matrix P This enables smooth estimation of the measurement signal; wherein the measurement signal is selected, mapped or combined from the acquired multivariate parameter signals and used as the observation input to the observation equation of the Kalman filter process.

[0030] Example 4: This embodiment is a further optimization based on any of the above embodiments. The parts identical to the aforementioned technical solutions will not be repeated here. Furthermore, to better realize the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative back-engineering described in this invention, the following settings are specifically adopted: In step 2.1), during real-time preprocessing, the Kalman filtering process includes the construction of state equations and observation equations: x k = A x k-1 + B u k + w k ; z k = H x k + v k ; in, x Let A be the engine state vector, containing temperature, pressure, and flow characteristics; B be the state transition matrix; and u be the control input matrix. w For process noise; z is the initial measurement vector of the sensor; H is the observation matrix; v For measuring noise; x k This represents the engine state vector at the current moment; x k-1 This is the engine state vector from the previous moment; u k This is the control input vector at the current moment; wk This represents the process noise at the current moment. v k The measurement noise at the current moment; z k This represents the initial measurement vector of the sensor at the current moment; In step 2.2), the smoothing estimation is achieved using the following formula: ;in, k Represents the current moment; k -1 represents the previous moment; P k Let be the state estimation error covariance matrix, and take the expected value of the product of the state estimation error vector and its transpose; E For expectations; T This is the transpose symbol.

[0031] Example 5: This embodiment is a further optimization based on any of the above embodiments. The parts that are the same as those in the aforementioned technical solutions will not be repeated here. In order to better realize the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative back-engineering described in this invention, the following settings are specifically adopted: Step 3) is specifically as follows: Based on the filtered multivariate parameter data of the aero-engine, a one-dimensional performance calculation model of the aero-engine is constructed, and the compressor and turbine characteristic models are introduced into the one-dimensional performance calculation model. The one-dimensional performance calculation model is subjected to parameter constraints and closed-loop solution. By coupling the compressor and turbine characteristic models with the one-dimensional performance calculation model, a unified description between the system-level conservation relationship and the actual performance characteristics of the components is realized, thereby constructing the aero-engine performance model; wherein, the compressor and turbine characteristic models are used to characterize the actual working characteristics of the compressor and turbine under different operating conditions, including the flow characteristic model and the efficiency characteristic model of the compressor and turbine. The flow characteristic model and the efficiency characteristic model are both represented in the form of characteristic curves or mathematical functions, and are embedded as constraints into the one-dimensional performance calculation model to participate in the solution.

[0032] Example 6: This embodiment is a further optimization based on any of the above embodiments. The parts identical to the aforementioned technical solutions will not be repeated here. Furthermore, to better realize the online performance estimation and health assessment method for aero-engines based on Kalman filtering and iterative backpropagation described in this invention, the following configuration is specifically adopted: The one-dimensional performance calculation model is established based on the mass conservation equation, energy conservation equation, momentum conservation equation, and ideal gas law, used to describe the thermodynamic and aerodynamic relationships between the main components of the aero-engine in the axial airflow direction; the one-dimensional performance calculation model divides the aero-engine into the inlet section, compressor section, combustion chamber section, turbine section, and exhaust section, and forms a system-level calculation equation set describing the overall performance of the aero-engine by establishing functional relationships between the inlet and outlet parameters of each adjacent component; When calculating compressor performance in the aforementioned aero-engine performance model, the compressor efficiency is determined by the compressor efficiency characteristic model, and the compressor efficiency is calculated using the following formula: ; in, or c The subscript C refers to the compressor efficiency. T t2 This refers to the compressor inlet total temperature in the temperature data. T t3 This refers to the compressor outlet total temperature in the temperature data. π c For the compressor pressure ratio, and , p t2 This refers to the total compressor inlet pressure in the pressure data. p t3 This refers to the total compressor outlet pressure in the pressure data. c The specific heat ratio of air; The aero-engine performance model uses the following formula to calculate the compressor flow rate: ; in, P s3 This refers to the static pressure at the compressor outlet. S R is the compressor outlet cross-sectional area, and R is the air gas constant. W c This refers to the compressor flow rate; The aero-engine performance model uses the following formula to calculate turbine efficiency: ; in, For turbine efficiency, the subscript T refers to the turbine. T t41 This refers to the turbine inlet total temperature in the temperature data.T t5 This refers to the total turbine outlet temperature in the temperature data. π T The turbine pressure ratio, and , p t5 This refers to the total turbine outlet pressure in the pressure data. c g Specific heat ratio of fuel gas; The aero-engine performance model uses the following formula to calculate turbine flow rate: ; in, W T For turbine flow, W c This refers to the compressor flow rate.

[0033] This aero-engine performance model can adaptively adjust the calculation coefficients under changing flight conditions, enabling consistent performance calculations of the aero-engine under different thrust commands, altitudes, and temperature conditions.

[0034] Example 7: This embodiment is a further optimization based on any of the above embodiments. The parts identical to the aforementioned technical solutions will not be repeated here. Furthermore, to better realize the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative back-calculation described in this invention, the following setting is specifically adopted: In step 4), the iterative back-calculation method based on isentropic relations, and the introduction of a neural network correction model, includes the following specific steps: 4.1) Based on the known total compressor inlet temperature T t2 Total temperature at compressor outlet T t3 and initial value of pressure ratio π c 0 According to the isentropic equation Preliminary calculation of the total temperature at the compressor outlet T t3 ; 4.2) Iterative solution: Apply the preliminary calculations... T t3 Substitute into the one-dimensional performance equation and recalculate. π c i+1 until ( T t3 , P t3 )convergence: | T t3 i+1- T t3 i |< e 1, of which i The current step of the iteration. i +1 indicates the next iteration. e 1 represents the preset convergence threshold; | P t3 i+1 - P t3 i |< e 2, of which i The current step of the iteration. i +1 indicates the next iteration. e 2 represents the preset convergence threshold; 4.3) Neural Network Correction: To correct aerodynamic deviations and nonlinear errors, a BP neural network model is established: | T t3 ’ , P t3 ’ | = NN(| T t2 , P t2 , T t41 , N 1, N 2, W f |), where NN(||) is a neural network function, which can learn the error correction relationship of the BP neural network model through training; P t2 This is the total pressure at the compressor inlet. T t41 This refers to the turbine inlet total temperature in the temperature data. N 1 represents the low-pressure rotor speed. N 2 represents the high-voltage rotor speed. W f This refers to fuel flow rate.

[0035] Supervised training using multivariate parameter signals ensures that the estimated values ​​remain highly robust under different operating conditions.

[0036] Example 8: This embodiment is a further optimization based on any of the above embodiments. The parts identical to the aforementioned technical solutions will not be repeated here. Furthermore, to better realize the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation described in this invention, the following settings are specifically adopted: Step 5) includes the following specific steps: 5.1) Based on the intermediate parameters obtained through reverse engineering, calculate the performance degradation index of the main components: ; In this context, "real" refers to the actual measured value, and "ref" refers to the design value. Traffic decay rate, To account for the efficiency degradation rate, the main components include two types of components: compressor and turbine; 5.2) By combining the flow and efficiency degradation rates of multiple components, an overall health index (HI) is constructed: ; in , These are weighting coefficients, determined based on the importance of the components; W For traffic, or For efficiency, The component's flow decay rate. This refers to the efficiency degradation rate of the component.

[0037] 5.3) When HI When the value falls below a set threshold, it is automatically identified as a potential performance anomaly, and a health alarm signal is triggered at the airborne edge computing node.

[0038] Example 9: This embodiment is a further optimization based on any of the above embodiments. The parts that are the same as those in the aforementioned technical solutions will not be repeated here. In order to better realize the online estimation and health assessment method for aero-engine performance based on Kalman filtering and iterative backpropagation described in this invention, the following setting method is adopted: In step 5), when the edge alarm is triggered, alarm information including the abnormal component number, decay type and estimated confidence level will also be uploaded to the ground monitoring center through the data link.

[0039] Example 10: A method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation includes the following steps: I. Data Acquisition and Characteristic Curve Fusion Modeling: S1. In this embodiment, the onboard sensors of the airborne health management system for aero-engines collect multi-source monitoring data (i.e., multi-parameter signals) in real time during flight. The multi-source monitoring data includes: covering the compressor inlet total temperature. T t2 Total temperature at compressor outlet T t3 Turbine inlet total temperature T t41 Turbine outlet total temperature T t5 Temperature data; covering the total pressure at the compressor inletP t2 Total pressure at compressor outlet P t3 Pressure data; covering low-pressure rotor speed N 1. High-voltage rotor speed N 2. Speed ​​signal; covering airflow and fuel flow at various levels. W f Traffic data.

[0040] S2. To suppress the impact of measurement noise and sensor drift on performance calculation results, this invention employs the Extended Kalman Filter (EKF) algorithm to perform real-time preprocessing on the acquired multivariate parameter signals. During real-time preprocessing, the Kalman filtering process includes the construction of state equations and observation equations: x k = A x k-1 + B u k + w k (1); z k = H x k + v k (2); in, x Let A be the engine state vector, containing temperature, pressure, and flow characteristics; B be the state transition matrix; and u be the control input matrix. w For process noise; z is the initial measurement vector of the sensor; H is the observation matrix; v For measuring noise; x k This represents the engine state vector at the current moment; x k-1 This is the engine state vector from the previous moment; u k This is the control input vector at the current moment; w k This represents the process noise at the current moment. v k The measurement noise at the current moment; z k This is the initial measurement vector of the sensor at the current moment.

[0041] S3. Dynamically update the covariance matrix using the following formula. P This enables smooth estimation of the measured signal. (3); in,k Represents the current moment; k -1 represents the previous moment; P k Let be the state estimation error covariance matrix, and take the expected value of the product of the state estimation error vector and its transpose; E For expectations; T This is the transpose symbol.

[0042] The measurement signal is obtained by selecting, mapping, or combining the acquired multivariate parameter signals and is used as the observation input to the observation equation of the Kalman filtering process.

[0043] S4. Based on the filtered multivariate parameter data of the aero-engine, a one-dimensional performance calculation model of the aero-engine is constructed. Compressor and turbine characteristic models are introduced into the one-dimensional performance calculation model. Parameter constraints and closed-loop solutions are applied to the one-dimensional performance calculation model. By coupling the compressor and turbine characteristic models with the one-dimensional performance calculation model, a unified description between system-level conservation relationships and the actual performance characteristics of components is achieved, thereby constructing the aero-engine performance model. The compressor and turbine characteristic models characterize the actual working characteristics of the compressor and turbine under different operating conditions, including the flow characteristic models and efficiency characteristic models of the compressor and turbine. Both the flow characteristic models and efficiency characteristic models are represented in the form of characteristic curves or mathematical functions and are embedded as constraints into the one-dimensional performance calculation model for solution.

[0044] The one-dimensional performance calculation model is based on the mass conservation equation, energy conservation equation, momentum conservation equation, and ideal gas law, and is used to describe the thermodynamic and aerodynamic relationships between the main components of the aero-engine in the axial airflow direction. The one-dimensional performance calculation model divides the aero-engine into the inlet section, compressor section, combustion chamber section, turbine section, and exhaust section, and forms a system-level calculation equation set describing the overall performance of the aero-engine by establishing the functional relationship between the inlet and outlet parameters of each adjacent component. When calculating compressor performance in the aforementioned aero-engine performance model, the compressor efficiency is determined by the compressor efficiency characteristic model, and the compressor efficiency is calculated using the following formula: (4); in, or c The subscript C refers to the compressor efficiency. T t2 This refers to the compressor inlet total temperature in the temperature data. T t3 This refers to the compressor outlet total temperature in the temperature data. π c For the compressor pressure ratio, and , p t2 This refers to the total compressor inlet pressure in the pressure data. p t3 This refers to the total compressor outlet pressure in the pressure data. c The specific heat ratio of air; The aero-engine performance model uses the following formula to calculate the compressor flow rate: (6); in, P s3 This refers to the static pressure at the compressor outlet. S R is the compressor outlet cross-sectional area, and R is the air gas constant. W c This refers to the compressor flow rate; The aero-engine performance model uses the following formula to calculate turbine efficiency: (7); in, For turbine efficiency, the subscript T refers to the turbine. T t41 This refers to the turbine inlet total temperature in the temperature data. T t5 This refers to the total turbine outlet temperature in the temperature data. π T The turbine pressure ratio, and (8), p t5 This refers to the total turbine outlet pressure in the pressure data. c g Specific heat ratio of fuel gas; The aero-engine performance model uses the following formula to calculate turbine flow rate: (9); in, W T For turbine flow, W c This refers to the compressor flow rate.

[0045] This aero-engine performance model can adaptively adjust the calculation coefficients to achieve consistent performance calculations of aero-engines under different thrust commands, altitudes, and temperature conditions.

[0046] two, T t3 and P t3 Iterative back-calculation and neural network correction: S5, for compressor outlet total temperature Tt3 Total pressure at compressor outlet P t3 To address the issue of inability to directly measure, this embodiment proposes an iterative back-calculation method based on isentropic relationships and introduces a neural network correction model to achieve [the desired measurement]. T t3 , P t3 High robustness estimation: S5.1, Given the compressor inlet total temperature, compressor outlet total temperature, and initial pressure ratio... π c 0 According to the isentropic equation (10) Preliminary calculation of the total temperature at the compressor outlet T t3 .

[0047] S5.2 Iterative Solution: The preliminary calculations obtained... T t3 Substitute into the one-dimensional performance equation and recalculate. π c i+1 until ( T t3 , P t3 )convergence: | T t3 i+1 - T t3 i |< e 1 (11); | P t3 i+1 - P t3 i |< e 2 (12); in, i The current step of the iteration. i +1 indicates the next iteration. e 1 and e 2 are both preset convergence thresholds.

[0048] S5.3 Neural Network Correction: To correct aerodynamic deviations and nonlinear errors, a BP neural network model is established: | T t3 ’ , P t3 ’ | = NN(| T t2 , Pt2 , T t41 , N 1, N 2, W f |) (13); where NN(||) is a neural network function, which can learn the error correction relationship of the BP neural network model through training; P t2 This is the total pressure at the compressor inlet. T t41 This refers to the turbine inlet total temperature in the temperature data. N 1 represents the low-pressure rotor speed. N 2 represents the high-voltage rotor speed. W f This refers to fuel flow rate.

[0049] Supervised training using multivariate parameter signals ensures that the estimated values ​​remain highly robust under different operating conditions.

[0050] III. Health Index and Abnormality Detection: S6. Based on the intermediate parameters obtained through reverse calculation, calculate the performance degradation index of the main components: (14); In this context, "real" refers to the actual measured value, and "ref" refers to the design value. Traffic decay rate, The main components, including the compressor and turbine, are responsible for the efficiency degradation rate.

[0051] S7. By combining the flow and efficiency degradation rates of multiple components, an overall health index (HI) is constructed: (15); in , These are weighting coefficients, determined based on the importance of the components; W For traffic, or For efficiency, The component's flow decay rate. This refers to the efficiency degradation rate of the component.

[0052] S8, when HI When the value falls below a set threshold, it is automatically identified as a potential performance anomaly and a health alarm signal is triggered at the airborne edge computing node. When the health alarm signal is triggered, alarm information, including the abnormal component number, degradation type, and estimated confidence level, is also uploaded to the ground monitoring center via the data link.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative back-engineering, characterized by: Includes the following steps: 1) Collect multi-parameter signals of the aero-engine through airborne sensors. These multi-parameter signals include temperature data, pressure data, flow rate data, and speed signals. 2) The extended Kalman filter algorithm is used to dynamically fuse and suppress noise in multi-parameter signals; 3) Establish a one-dimensional performance calculation model and integrate the compressor and turbine characteristic models to construct an aero-engine performance model; 4) An iterative back-calculation method based on isentropic relations, incorporating a neural network correction model, to achieve [the desired result]. T t3 , P t3 High robustness estimation; 5) Calculate the flow rate and efficiency degradation rate of the main components to construct a health index; when the health index is lower than the set threshold, trigger an edge alarm.

2. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to claim 1, characterized in that: The temperature data includes the compressor inlet total temperature. T t2 Total temperature at compressor outlet T t3 Turbine inlet total temperature T t41 Turbine outlet total temperature T t5 The pressure data includes the total compressor inlet pressure. P t2 Total pressure at compressor outlet P t3 The speed signal includes the low-pressure rotor speed. N 1. High-voltage rotor speed N 2; The flow data includes airflow at each stage and fuel flow. W f .

3. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to claim 1, characterized in that: Step 2) includes the following steps: 2.1) The acquired multivariate parameter signals are preprocessed in real time using the extended Kalman filter algorithm; 2.2) Dynamically updating the covariance matrix P This enables smooth estimation of the measurement signal; wherein the measurement signal is selected, mapped or combined from the acquired multivariate parameter signals and used as the observation input to the observation equation of the Kalman filter process.

4. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to claim 3, characterized in that: In step 2.1), during real-time preprocessing, the Kalman filtering process includes the construction of the state equation and the observation equation: x k = A x k-1 + B u k + w k ; z k = H x k + v k ； in, x Let A be the engine state vector, containing temperature, pressure, and flow characteristics; B be the state transition matrix; and u be the control input matrix. w For process noise; z is the initial measurement vector of the sensor; H is the observation matrix; v For measuring noise; x k This represents the engine state vector at the current moment; x k-1 This is the engine state vector from the previous moment; u k This is the control input vector at the current moment; w k This represents the process noise at the current moment. v k The measurement noise at the current moment; z k This represents the initial measurement vector of the sensor at the current moment; In step 2.2), the smoothing estimation is achieved using the following formula: ;in, k Represents the current moment; k -1 represents the previous moment; P k Let be the state estimation error covariance matrix, and take the expected value of the product of the state estimation error vector and its transpose; E For expectations; T This is the transpose symbol.

5. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to any one of claims 1 to 4, characterized in that: Step 3) specifically involves: based on the filtered multivariate parameter data of the aero-engine, constructing a one-dimensional performance calculation model of the aero-engine, and introducing compressor and turbine characteristic models into the one-dimensional performance calculation model. Parameter constraints and closed-loop solutions are then applied to the one-dimensional performance calculation model to obtain the aero-engine performance model. The compressor and turbine characteristic models include flow characteristic models and efficiency characteristic models of the compressor and turbine. Both the flow characteristic model and the efficiency characteristic model are represented by characteristic curves or mathematical functions and are embedded as constraints into the one-dimensional performance calculation model for solution.

6. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to claim 5, characterized in that: The one-dimensional performance calculation model is based on the mass conservation equation, energy conservation equation, momentum conservation equation and ideal gas law, and is used to describe the thermodynamic and aerodynamic relationships between the main components of the aero-engine in the axial airflow direction. When calculating compressor performance in the aforementioned aero-engine performance model, the compressor efficiency is determined by the compressor efficiency characteristic model, and the compressor efficiency is calculated using the following formula: ； in, η c The subscript C refers to the compressor efficiency. T t2 This refers to the compressor inlet total temperature in the temperature data. T t3 This refers to the compressor outlet total temperature in the temperature data. π c The compressor boost ratio, and , p t2 This refers to the total compressor inlet pressure in the pressure data. p t3 This refers to the total compressor outlet pressure in the pressure data. γ The specific heat ratio of air; The aero-engine performance model uses the following formula to calculate the compressor flow rate: ； in, P s3 This refers to the static pressure at the compressor outlet. S R is the compressor outlet cross-sectional area, and R is the air gas constant. W c This refers to the compressor flow rate; The aero-engine performance model uses the following formula to calculate turbine efficiency: ； in, For turbine efficiency, the subscript T refers to the turbine. T t41 This refers to the turbine inlet total temperature in the temperature data. T t5 This refers to the total turbine outlet temperature in the temperature data. π T The turbine pressure ratio, and , p t5 This refers to the total turbine outlet pressure in the pressure data. γ g Specific heat ratio of fuel gas; The aero-engine performance model uses the following formula to calculate turbine flow rate: ； in, W T For turbine flow, W c This refers to the compressor flow rate.

7. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative back-engineering according to any one of claims 1 to 4 and 6, characterized in that: In step 4), the iterative back-calculation method based on the isentropic relationship, and the introduction of a neural network correction model, includes the following specific steps: 4.1) Given the total inlet temperature, total outlet temperature, and initial pressure ratio of the compressor. π c 0 According to the isentropic equation Preliminary calculation of the total temperature at the compressor outlet T t3 ; 4.2) Iterative solution: The preliminary calculations obtained... T t3 Substitute into the one-dimensional performance equation and recalculate. π c i+1 until ( T t3 , P t3 )convergence: | T t3 i+1 - T t3 i | < ε 1, of which i The current step of the iteration. i +1 indicates the next iteration. ε 1 represents the preset convergence threshold; | P t3 i+1 - P t3 i | < ε 2, of which i The current step of the iteration. i +1 indicates the next iteration. ε 2 represents the preset convergence threshold; 4.3) Neural Network Correction: To correct aerodynamic deviations and nonlinear errors, a BP neural network model is established: | T t3 ’ , P t3 ’ | = NN(| T t2 , P t2 , T t41 , N 1, N 2, W f |), where NN(||) is a neural network function, which learns the error correction relationship of the BP neural network model through training; P t2 This refers to the total pressure at the compressor inlet. T t41 This refers to the turbine inlet total temperature in the temperature data. N 1 represents the low-pressure rotor speed. N 2 represents the high-voltage rotor speed. W f This refers to fuel flow rate.

8. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to any one of claims 1 to 4 and 6, characterized in that: Step 5) includes the following specific steps: 5.1) Based on the intermediate parameters obtained through reverse engineering, calculate the performance degradation index of the main components: ； In this context, "real" refers to the actual measured value, and "ref" refers to the design value. Traffic decay rate, To account for the efficiency degradation rate, the main components include two types of components: compressor and turbine; 5.2) Construct an overall health index by combining the flow and efficiency degradation rates of multiple components. HI : ； in , These are weighting coefficients, determined based on the importance of the components; W For traffic, η For efficiency, The component's flow decay rate. The efficiency degradation rate of the component; 3) When HI When the value falls below a set threshold, it is automatically identified as a potential performance anomaly, and a health alarm signal is triggered at the airborne edge computing node.

9. The method for online estimation and health assessment of aero-engine performance based on Kalman filtering and iterative backpropagation according to any one of claims 1 to 4 and 6, characterized in that: In step 5), when the edge alarm is triggered, alarm information, including the abnormal component number, decay type, and estimated confidence level, will be uploaded to the ground monitoring center via the data link.

Images