Aero-engine main fuel regulating system state estimation and health management method and system

Abstract

The invention discloses a state estimation and health management method and system for a main fuel regulating system of an aero-engine. The method comprises the following steps: constructing a state space model of a main fuel regulating system; estimating the fuel flow and the engine speed based on the state space model so as to obtain optimal estimation values of the fuel flow and the engine speed; optimizing a system parameter matrix of a state space model according to the optimal estimation value; the system parameter matrix comprises a state transition matrix and an input control matrix; constructing an innovation sequence based on the optimal estimation value and the sensor observation value, and calculating a residual error between the optimal estimation value and the sensor observation value; and performing fault diagnosis on the state of the main fuel regulating system and performing trend analysis on the health state according to the innovation sequence and the residual error. According to the method, the state estimation precision of the full service cycle of the engine is guaranteed, and the fault diagnosis reliability is greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of aero-engine technology, and in particular relates to a method and system for state estimation and health management of the main fuel regulation system of an aero-engine. Background Technology

[0002] Aero-engines are the core power units of aircraft, and their operational reliability directly determines flight safety and operational economy. The main fuel regulation system is the core execution unit of the full authority digital electronic control system (FADEC) of an aero-engine. It is responsible for precisely adjusting the fuel flow supplied to the engine combustion chamber based on flight conditions, environmental conditions, and control commands, thereby controlling the engine's core operating parameters such as speed and thrust in a closed loop. The stability and control accuracy of its operating state directly affect the engine's power performance, fuel efficiency, and operational safety. Therefore, accurate and real-time state estimation and full-cycle health management of the main fuel regulation system are core technologies for ensuring the safe, reliable, and efficient operation of aero-engines.

[0003] Currently, the main fuel regulation system of aero engines is based on a mechanical-hydraulic architecture, equipped with electronic sensors for monitoring operating status such as speed, fuel flow, and pressure. Its health management mostly adopts the traditional mode of sensor threshold alarm and regular offline maintenance. This model has been applied in engineering for a long time and has basic operational stability. However, as the requirements for high reliability, long service life and low maintenance costs of aero engines continue to increase, its inherent limitations are becoming increasingly prominent: First, aero engines operate under complex conditions of high temperature, high pressure, strong vibration and strong electromagnetic interference. The sampling signals of speed and fuel flow sensors are easily affected by multi-source noise, resulting in data delay, fluctuation and even distortion, which cannot provide a stable and accurate data source for status judgment. Second, over-limit alarms based on fixed thresholds are passive and reactive monitoring methods. They can only identify serious faults that have already occurred. They cannot effectively identify and provide early warnings for potential faults such as early minor wear and gradual performance degradation of core components such as metering valves, slide valves and actuators, which can easily lead to unplanned downtime risks. Third, the offline maintenance mode based on fixed cycles cannot match the actual performance degradation state of a single engine. It is easy to cause increased maintenance costs due to over-maintenance or increased failure risks due to delayed maintenance, which is difficult to adapt to the development needs of modern aero engine condition-based maintenance.

[0004] To address the shortcomings of traditional monitoring schemes, state estimation algorithms have been gradually introduced into the field of aero-engine fuel system health management. Among them, the Kalman filter algorithm has been widely used due to its ability to estimate the optimal state of linear systems. For example, Chinese invention patent CN115903738B discloses a diagnostic method and device for the main fuel control system of an aero-engine based on AMESim. This method establishes a Simulink controller model by collecting aero-engine operating data, obtains controller command signals, and then establishes a main fuel control system model and a Kalman filter model based on the AMESim platform. The same command signal is injected into both models, and the residual is calculated by comparing the fault-free characteristic data output by the Kalman filter model with the actual operating data. Finally, fault diagnosis is achieved by comparing the residual with a threshold. This scheme improves the accuracy of main fuel system fault diagnosis to a certain extent and provides a feasible technical path for fuel system health management. However, it still has unavoidable technical defects in engineering applications, airborne real-time adaptability, and long-term operational reliability.

[0005] The aforementioned existing technologies employ a single Kalman filter model with fixed parameters, lacking an adaptive parameter correction mechanism. This makes them unsuitable for adapting to the performance degradation characteristics of systems during long-term service, resulting in insufficient estimation accuracy and reliability over extended periods. In this approach, core parameters such as the state transition matrix and control matrix of the Kalman filter model are fixed values, determined based on initial simulation or calibration data. They cannot be dynamically updated as components wear and gradually degrade in performance during system operation. Furthermore, during long-term service, the mechanical and hydraulic characteristics of the core components of the main fuel regulation system undergo irreversible gradual changes. The fixed-parameter model will gradually mismatch with the dynamic characteristics of the actual system, leading to a continuous increase in state estimation errors, and even misjudgments or omissions, thus failing to guarantee the reliability of health management throughout the entire service life. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for state estimation and health management of the main fuel regulation system of an aero-engine, which partially solves or alleviates the above-mentioned deficiencies in the prior art and can iterate the system parameter matrix in the main fuel system model of the engine to ensure the robustness of the model.

[0007] To solve the aforementioned technical problems, the present invention specifically adopts the following technical solution: A first aspect of the present invention is to provide a method for state estimation and health management of an aircraft engine main fuel regulation system, comprising: Based on aero-engine test data, a state-space model of the main fuel regulation system is constructed. Based on the state-space model, fuel flow rate and engine speed are estimated to obtain the optimal estimated values ​​of fuel flow rate and engine speed. The system parameter matrix of the state-space model is optimized based on the optimal estimate; the system parameter matrix includes the state transition matrix and the input control matrix. Based on the optimal estimate and sensor observations, an innovation sequence is constructed, and the residual between the optimal estimate and sensor observations is calculated. Fault diagnosis of the main fuel conditioning system status is performed based on the aforementioned information sequence and residuals; The changes in the statistical system parameter matrix, the parameter estimation covariance matrix formed during the optimization of the system parameter matrix, and the deviation between the optimal estimated value and the design standard value of the main fuel regulation system are used to conduct trend analysis on the health status of the main fuel regulation system.

[0008] Furthermore, the state-space model includes: Equations of state , Observation equations ; Among them, X k Let Q be the state vector at time k, which includes the fuel flow rate Q. e_k and engine speed N e_k ;Y k The observation vector at time k is the fuel flow rate Q sampled by the sensor. s_k and engine speed N s_k U k-1 The control input at time k-1 is the piston linear displacement L; W k For process noise; V k denoted as observation noise; A is the state transition matrix; B is the input control matrix; and H is the observation matrix.

[0009] Furthermore, the steps to obtain the optimal estimate include: Using the optimal estimates of fuel flow rate and speed at time k-1 and the piston line displacement, the fuel flow rate and speed at time k are predicted to obtain the prior estimates of fuel flow rate and speed at time k. Based on the error covariance matrix of the process at time k-1 and the error covariance matrix of the system's optimal estimate, the error covariance matrix of the system's prior estimate at time k is predicted. The confidence level of the observed values ​​is calculated based on the error covariance matrix of the prior estimates of the system at time k, and the Kalman gain is obtained. The optimal estimates of fuel flow rate and speed at time k are calculated using Kalman gain, prior estimates of fuel flow rate and speed at time k, and observed values ​​of fuel flow rate and speed at time k; and the error covariance matrix of the optimal estimates at time k is calculated.

[0010] Furthermore, using the formula:

[0011] Calculate the prior estimates of fuel flow rate and engine speed at time k; where, Let k be the prior estimate. Let A be the optimal estimate at time k-1, B be the state transition matrix, and U be the input control matrix. k The control input of the main fuel conditioning system at time k is the piston linear displacement L. Using the formula:

[0012] Calculate the error covariance matrix of the prior estimate at time k; where, Let k be the error covariance matrix of the prior estimate. Let A be the error covariance matrix of the optimal estimate at time k-1. T S is the transpose of the state transition matrix, and S is the error covariance matrix at time k-1. Using the formula:

[0013] Calculate the Kalman gain; where K k Let H be the Kalman gain at time k, and H be the observation matrix. T R is the transpose of the observation matrix, and R is the error covariance matrix of the observations; Using the formula:

[0014] Calculate the optimal estimates of oil flow rate and rotational speed at time k; where, Y represents the optimal estimate of oil flow rate and rotational speed at time k. k Let be the observation vector at time k; Using the formula:

[0015] Calculate the error covariance matrix of the optimal estimate at time k; where, Let I be the error covariance matrix of the optimal estimate at time k, and let I be the identity matrix.

[0016] Furthermore, the steps for optimizing the system parameter matrix include: Construct a regression vector. The regression vector at time k includes the optimal estimates of fuel flow rate and engine speed at time k-1 and the piston line displacement. The system parameter matrix updated at time k-1 and the regression vector at time k are used to calculate the predicted values ​​of fuel flow rate and speed at time k, thus obtaining the predicted values ​​of fuel flow rate and speed at time k. Calculate the residuals between the predicted and observed values ​​of fuel flow rate and engine speed at time k; The covariance matrix is ​​estimated using the updated parameters at time k-1, the regression vector at time k, and the forgetting factor to calculate the gain at time k. The updated system parameter matrix at time k is calculated using the system parameter matrix at time k-1, the gain at time k, and the residuals between the predicted and observed values ​​of fuel flow and speed. The parameter estimation covariance matrix updated at time k is calculated using the forgetting factor, the parameter estimation covariance matrix updated at time k-1, the gain at time k, and the regression vector.

[0017] Furthermore, using the formula:

[0018] Calculate the predicted values ​​of fuel flow rate and engine speed at time k; where, Let K be the predicted values ​​of fuel flow rate and engine speed at time k. Let k be the transpose of the regression vector at time k. The system parameter with index i is updated at time k-1; Using the formula: ,

[0019] Calculate the gain at time k; where G i (k) represents the gain at time k with index i. The covariance matrix of the parameter estimate with index i updated at time k-1. Let k be the regression vector. Forgetting factor; Using the formula:

[0020] Calculate the system parameter matrix updated at time k; where, Let i be the system parameter matrix updated at time k. Let i be the residual between the predicted value and the observed value at time k with index i. Using the formula:

[0021] Calculate the parameter estimation covariance matrix updated at time k; where, The parameter estimate covariance matrix with index i is updated at time k.

[0022] Furthermore, the steps for fault diagnosis of the main fuel conditioning system state based on the information sequence and residuals include: If both the fuel flow information sequence and the engine speed information sequence are abnormal, and their correlation is higher than the correlation threshold, it is determined to be a mechanical fault. If only one item in the fuel flow information sequence and the speed information sequence is abnormal, and the residual corresponding to the abnormal item exceeds the residual threshold, the corresponding sensor is judged to be faulty; in the case of sensor failure, the optimal estimate is used as the sensor feedback.

[0023] Furthermore, if the mean and / or variance of the innovation sequence exceeds the corresponding threshold, it is determined to be abnormal.

[0024] Furthermore, the steps for trend analysis of the health status of the main fuel conditioning system include: When the change in the state transition matrix exceeds the state change threshold, the system dynamic behavior is determined to be abnormal; when the change in the input control matrix exceeds the control change threshold, the system control characteristics are determined to be abnormal. When the trace value of the parameter estimation covariance matrix shows a continuous increasing trend, it is determined that the model uncertainty is increasing and the system has the risk of performance degradation. When the deviation continues to increase beyond the preset trend threshold, or when the optimal estimated value fluctuates abnormally, the system is determined to have entered the performance degradation stage.

[0025] The present invention also provides a state estimation and health management system for an aero-engine main fuel conditioning system, comprising: The model building module is used to construct a state-space model of the main fuel regulation system based on aero-engine test data; The optimal estimate prediction module is used to estimate the fuel flow rate and engine speed based on the state space model, so as to obtain the optimal estimate of the fuel flow rate and engine speed. The system parameter matrix optimization module is used to optimize the system parameter matrix of the state-space model based on the optimal estimated value; the system parameter matrix includes a state transition matrix and an input control matrix. The calculation module is used to construct an innovation sequence based on the optimal estimate and the sensor observation, and to calculate the residual between the optimal estimate and the sensor observation; The fault diagnosis module is used to diagnose faults in the status of the main fuel conditioning system based on the information sequence and residuals. The trend analysis module is used to statistically analyze the changes in the system parameter matrix, the parameter estimation covariance matrix formed during the optimization process of the system parameter matrix, and the deviation between the optimal estimated value and the design standard value of the main fuel regulation system, so as to perform trend analysis on the health status of the main fuel regulation system.

[0026] Beneficial effects: In terms of model construction, existing technologies generally use indirect signals such as throttle lever commands and altitude Mach numbers as model inputs, which are easily affected by changes in flight conditions and system characteristics, resulting in inherent deviations between the input benchmark and the actual executed actions. This invention is the first to use the linear displacement of the servo piston connecting rod measured by an LVDT sensor as the sole control input. Relying on the rigid linear physical correlation between the linear displacement of the main fuel regulation system, the metering valve opening, and the fuel flow, the input signal is a direct physical quantity of the system's executed actions. Under the complex operating conditions of high temperature, high pressure, and strong vibration of aero engines, the measurement stability and anti-interference capability are far superior to traditional indirect command inputs, ensuring the benchmark reliability of state estimation and model prediction from the source.

[0027] Existing technologies, such as CN115903738B, which construct idealized models based on the AMESim simulation platform, have inherent deviations that cannot be eliminated from the actual engine's machining and assembly errors, actual component characteristics, and dynamic changes in operating conditions. Moreover, they can only be used for offline ground simulations. This invention constructs a discrete state-space model based on ground test data of aero-engines under all operating conditions. This model perfectly matches the actual dynamic characteristics of a single engine. At the same time, the model has a discrete architecture, and the sampling period and iteration logic perfectly match the millisecond-level operation requirements of the airborne control system. This solves the problem of mismatch between existing simulation models and real systems, and the inability to operate online on airborne systems.

[0028] In terms of algorithm construction, existing technologies employ a single Kalman filter scheme with fixed parameters. The state transition matrix and input control matrix are calibrated only once using initial data, which cannot adapt to the characteristics of component wear and gradual performance degradation during long-term engine service. The mismatch between the model and the real system continues to increase, and the state estimation accuracy continues to decline over service time. This invention uses the optimal estimate value output by the Kalman filter as the identification benchmark and periodically updates the core system matrix online using an RLS algorithm with a forgetting factor, allowing the model to continuously fit the time-varying characteristics of the system. Bench tests have verified that the state estimation accuracy of fuel flow and speed remains stable throughout the entire service life of the engine, completely solving the industry problem of long-term accuracy failure of fixed models.

[0029] Existing technologies are open-loop designs, with no linkage between model parameters and state estimation optimization. In this invention, the high-precision optimal estimate of Kalman filtering provides reliable identification samples that eliminate noise interference for the RLS algorithm. The high-precision model optimized by RLS further improves the state estimation accuracy of Kalman filtering. The two form a mutually reinforcing positive adaptive closed loop, which, unlike the open-loop design of existing technologies with fixed models, achieves accuracy assurance throughout the entire lifecycle.

[0030] In terms of fault diagnosis, existing technologies can only determine whether a system has a fault through a single residual threshold, and cannot distinguish whether the fault originates from sensor failure or mechanical component damage, thus failing to provide an effective basis for maintenance decisions. This invention constructs a three-layer progressive diagnostic logic, which includes initial judgment of full-dimensional anomalies in the information sequence, differentiation of fault types through dual-channel correlation analysis, and accurate verification through residual thresholds. It can accurately distinguish three major categories of faults: mechanical component faults, fuel flow sensor faults, and engine speed sensor faults, directly locating the faulty link, allowing maintenance personnel to carry out targeted repairs.

[0031] In addition, existing technologies can only achieve offline post-fault diagnosis in ground laboratories, without the ability to manage the health status throughout the entire cycle. The real-time diagnosis layer of this invention achieves millisecond-level fault identification and fault-tolerant processing, and the trend analysis layer quantifies the degree of system performance degradation through changes in model parameters, trace values ​​of the RLS covariance matrix, and the deviation trend between estimated and design values, thereby achieving early warning of potential faults. This upgrades the traditional periodic post-fault maintenance of aero engines to predictive condition-based maintenance, significantly reducing maintenance costs and avoiding over-maintenance and maintenance delays. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0033] Figure 1 This is a flowchart of the present invention; Figure 2 This is a flowchart illustrating the updating of the state transition matrix and the input control matrix in this invention; Figure 3 This is a system structure diagram of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0035] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0036] In this document, the terms "upper," "lower," "inner," "outer," "front," "rear," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the present invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0037] In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0038] In this document, "and / or" includes any and all combinations of one or more of the listed related items.

[0039] In this article, "multiple" means two or more, that is, it includes two, three, four, five, etc.

[0040] Example 1: like Figure 1 As shown, this embodiment provides a method for state estimation and health management of the main fuel regulation system of an aircraft engine, specifically including the following steps: S1 constructs a state-space model of the main fuel regulation system based on aero-engine test data.

[0041] In the main fuel regulation system of an aero-engine, the differential pressure regulator maintains a constant fuel pressure difference across the fuel metering valve. According to the fluid orifice outflow formula, the fuel flow rate is strictly linearly related only to the flow opening of the metering valve. Furthermore, the opening of the metering valve and the linear displacement of the follower piston are rigidly linked through a mechanical linkage structure, and are also linearly related. This forms the core physical link of the model of this invention: the linear displacement of the follower piston, the opening of the metering valve, the fuel flow rate, and the engine speed. This linear mapping relationship provides a stable and reliable physical basis for the construction of the state-space model and is also the core basis for this invention's use of piston linear displacement as the core input.

[0042] Meanwhile, the model satisfies the basic assumptions of the Kalman filter algorithm: both the system process noise and the sensor observation noise follow a zero-mean Gaussian white noise distribution, the system is a linear discrete dynamic system, and it conforms to the discrete sampling operation characteristics of the airborne control system of aero-engines.

[0043] The difference between the target speed set by the throttle lever and the current speed fed back by the centrifugal flyweight is amplified by a hydraulic amplifier and then drives the follower piston. The follower piston, through a rotating shaft and connecting rod, drives the fuel metering valve to change its opening degree, thereby regulating the fuel supply. The differential pressure regulator in the system is used to maintain a basically constant oil pressure difference before and after the metering valve.

[0044] When the target fuel flow rate increases, the pressure in the upper chamber of the follower piston increases accordingly, causing the follower piston to move downwards and drive the rotating shaft to control the metering valve to increase its opening, thus increasing the fuel flow rate. Conversely, when the target fuel flow rate decreases, the above process proceeds in reverse.

[0045] The relationship between the fuel flow rate supplied to the engine through the metering valve and the opening area of ​​the metering valve is as follows:

[0046] In the formula, Q is the fuel flow rate; The flow coefficient for measuring valves; This refers to the opening area of ​​the metering valve. The density of fuel; To measure the oil pressure difference before and after the metering valve, the differential pressure controller keeps the oil pressure difference almost constant, so the fuel flow rate is only related to the opening of the metering valve.

[0047] The opening degree of the fuel metering valve is linearly related to the displacement of the follower piston rod. Therefore, the relationship between fuel flow rate and follower piston rod displacement can be substituted into the above formula:

[0048] in For compensation coefficient, This refers to the linear displacement of the piston connecting rod. Since fuel flow is directly related to engine speed, information about fuel flow and engine speed can be obtained through linear displacement.

[0049] More specifically, the state-space model used to describe the main fuel regulation system in this embodiment includes: Equations of state , Observation equations ; Among them, X k Let Q be the state vector at time k, which includes the fuel flow rate Q. e_k and engine speed N e_k ;Y kThe observation vector at time k is the fuel flow rate Q sampled by the sensor. s_k and engine speed N s_k U k-1 The control input at time k-1 is the piston linear displacement L; W k For process noise; V k denoted as observation noise; A is the state transition matrix; B is the input control matrix; and H is the observation matrix.

[0050] State vector X k It is the estimation object of the model, which is a 2-dimensional column vector. , Q e_k Let N be the actual fuel flow rate at time k. e_k The actual engine speed at time k. These two parameters are the core controlled parameters of the main fuel regulation system, directly determining the engine's thrust, operating efficiency, and safety status, and are also the monitoring objects of the health management in this invention.

[0051] Control input vector U k 1 is the model's only external control input, which is the linear displacement L of the follower piston connecting rod measured by the LVDT displacement sensor at time k-1. Unlike existing technologies that use throttle lever commands and altitude Mach numbers as indirect inputs, linear displacement is a direct mechanical actuation quantity of the main fuel regulation system. Its anti-interference capability under high temperature, high pressure, and strong vibration conditions is far superior to traditional command signals and flow / speed sensors, ensuring the robustness of the model from the input source.

[0052] Observation vector Y k It is the measured feedback input of the model, which is a 2D column vector. , Q s_k Let N be the measured value of the fuel flow sensor at time k. s_k The measured value of the engine speed sensor at time k is the observable feedback of the system's true state and the core basis for subsequent Kalman filter correction of state estimation.

[0053] The noise term includes process noise W k With observation noise V k , where W is a 2D zero-mean Gaussian white noise vector. k The covariance matrix of a random disturbance characterizing the system itself, such as small fluctuations in oil pressure or changes in mechanical friction characteristics, is S; V k The covariance matrix of a sensor is R, which characterizes measurement errors such as electromagnetic interference and sampling drift.

[0054] The system parameter matrix includes the state transition matrix A, the input control matrix B, and the observation matrix C.

[0055] The state transition matrix A is a 2×2 square matrix that describes the dynamic inertial characteristics of the main fuel regulation system and characterizes the influence of the system state at time k-1 on the state at time k. It is the core mathematical representation of the inherent dynamic characteristics of the system.

[0056] The input control matrix B is a 2×1 dimensional column vector that describes the control effect of the linear displacement input on the system state. It represents the change in fuel flow and engine speed per unit length change in the linear displacement of the follower piston, and directly maps the linear relationship between linear displacement, flow, and speed.

[0057] The observation matrix C is a 2×2 dimensional identity matrix, representing a one-to-one mapping between the actual state of the system and the sensor observations.

[0058] S2 estimates the fuel flow rate and engine speed based on the state-space model, thereby obtaining the optimal estimates of the fuel flow rate and engine speed.

[0059] In this embodiment, the Kalman filter algorithm is used to estimate the optimal value.

[0060] The Kalman filter (KF) algorithm is an optimization algorithm for estimating the state of dynamic systems. Originally developed to solve spacecraft navigation problems, its advantage lies in efficiently handling the uncertainty propagation problem of linear Gaussian systems. The core idea of ​​the KF algorithm is to fuse the system's dynamic model and sensor measurement data, predict and correct based on the principle of minimizing the covariance of the estimation error, fully utilizing the statistical characteristics of process noise and measurement noise to obtain the optimal estimate of the state variable in the sense of minimum mean square error. Through recursion, the state estimate is updated with each sensor measurement, and the next state is predicted. Essentially, the KF algorithm weights and fuses the predicted and observed values. Combining this algorithm can yield a fuel flow rate that more closely reflects the actual system's regulation patterns. Its specific steps include a prediction phase and a correction phase.

[0061] The prediction phase does not rely on the current sensor measurements; instead, it uses the optimal estimate from the previous time step and the linear displacement input to make a priori prediction of the system state at the current moment through a state-space model, simultaneously predicting the error covariance of this prior estimate. This phase fully utilizes the physical dynamics of the main fuel regulation system, providing a baseline prediction value for subsequent corrections. The prediction phase includes: S21 uses the optimal estimates of fuel flow rate and speed at time k-1, as well as the piston line displacement, to predict the fuel flow rate and speed at time k, thus obtaining the prior estimates of fuel flow rate and speed at time k.

[0062] Specifically, this embodiment utilizes the following formula:

[0063] Calculate the prior estimates of fuel flow rate and engine speed at time k; where, Let k be the prior estimate. Let A be the optimal estimate at time k-1, B be the state transition matrix, and U be the input control matrix. k The control input of the main fuel conditioning system at time k is the piston linear displacement L.

[0064] S22 predicts the error covariance matrix of the prior estimate of the system at time k based on the error covariance matrix during the process at time k-1 and the error covariance matrix of the optimal estimate of the system.

[0065] Specifically, this embodiment utilizes the following formula:

[0066] Calculate the error covariance matrix of the prior estimate at time k; where, Let k be the error covariance matrix of the prior estimate. Let A be the error covariance matrix of the optimal estimate at time k-1. T Let S be the transpose of the state transition matrix, and let S be the error covariance matrix at time k-1.

[0067] The correction phase incorporates the current sensor measurements to correct the prior estimates obtained in the prediction phase, ultimately outputting the optimal estimate for the current moment. This phase uses Kalman gain to dynamically allocate weights, balancing the confidence levels of model predictions and sensor observations, ultimately yielding the optimal result in the sense of minimum mean square error. The correction phase includes: S23 calculates the confidence level of the observed values ​​based on the error covariance matrix of the prior estimates of the system at time k, and obtains the Kalman gain.

[0068] Specifically, this embodiment utilizes the following formula:

[0069] Calculate the Kalman gain; where K k Let H be the Kalman gain at time k, and H be the observation matrix. T R is the transpose of the observation matrix, and R is the error covariance matrix of the observations.

[0070] S24 uses the Kalman gain, prior estimates of fuel flow rate and speed at time k, and observed values ​​of fuel flow rate and speed at time k to calculate the optimal estimates of fuel flow rate and speed at time k; and calculates the error covariance matrix of the optimal estimates at time k.

[0071] The optimal estimate of the calculation system, in the context of the fuel system, means that the fuel flow rate is calculated from the displacement sampled by the LVDT displacement sensor, and the predicted fuel flow rate is corrected based on the calculated fuel flow rate. Kalman gain K k This can be understood as trust level, K k A smaller value indicates greater confidence in the predicted fuel flow rate, K. k A higher Kalman gain indicates greater confidence in the fuel flow rate observed by the sensor. As the sensor measurement error increases, the Kalman gain decreases, meaning the confidence in the observation decreases; conversely, as the sensor measurement error decreases, the Kalman gain increases, meaning the confidence in the observation increases. This can be adjusted according to the actual situation.

[0072] Specifically, this embodiment utilizes the following formula:

[0073] Calculate the optimal estimates of oil flow rate and rotational speed at time k; where, Y represents the optimal estimate of oil flow rate and rotational speed at time k. k Let be the observation vector at time k.

[0074] Using the formula:

[0075] Calculate the error covariance matrix of the optimal estimate at time k; where, Let I be the error covariance matrix of the optimal estimate at time k, and let I be the identity matrix.

[0076] S3 optimizes the system parameter matrix of the state-space model based on the optimal estimate; the system parameter matrix includes the state transition matrix and the input control matrix.

[0077] Existing technologies, including the Chinese patent application CN115903738B mentioned in the background section, all employ fixed-parameter state-space models. The state transition matrix and input control matrix of these models are calibrated only once using initial simulation or ground test data and remain unchanged throughout the engine's entire service life. In contrast, this invention innovatively uses the optimal estimates of fuel flow and engine speed output by the Kalman filter as identification benchmarks. Through recursive least squares (RLS) with a forgetting factor, it periodically performs online iterative optimization on the core parameter matrices, state transition matrix A and input control matrix B, which determine the dynamic characteristics of the system. This allows the state-space model to continuously adapt to the changes in the dynamic characteristics of the main fuel regulation system caused by component wear and gradual performance degradation. This fundamentally solves the industry problem of long-term mismatch between fixed models and real systems, while also providing quantifiable core characteristic indicators for early warning of system performance degradation.

[0078] This invention does not perform parameter matrix updates for every sampling period. Instead, it initiates an update process after a preset sampling period. The trigger period can be flexibly set according to the engine's operating conditions. For example, during ground test bench testing, it can be set to update once every 1000 sampling periods, while during airborne flight testing, it can be set to update once after each flight hour / flight. This design avoids the airborne computing power consumption caused by high-frequency iterations and can promptly track the gradual performance degradation of the system, making it perfectly suited for airborne engineering scenarios.

[0079] like Figure 2 As shown, the steps for optimizing the system parameter matrix in this embodiment include: S31 parameter initialization specifically includes: The parameters to be estimated are initialized by expanding the state transition matrix A and input control matrix B, which are identified based on the full-condition test data of the new machine, into a vector of parameters to be estimated by rows. .

[0080] Covariance matrix initialization: The parameter estimation covariance matrix P(0) is initialized as a large-value unit diagonal matrix to ensure high confidence in the new running data during the initial iteration of the algorithm, and to quickly converge to the true parameters of the system.

[0081] Forgetting factor initialization: The forgetting factor λ is set, with a value range of 0 < λ ≤ 1. In the scenario of the main fuel system of aero-engine, 0.98 is preferred, which not only ensures the convergence stability of the algorithm, but also has sufficient time-varying tracking capability.

[0082] This invention employs a lightweight design that updates row by row, treating each row of the state transition matrix A and the input control matrix B as an independent set of parameters to be estimated and iterating accordingly. This significantly reduces the computational load and adapts to the computing power limitations of airborne microcontrollers. The complete iterative steps are as follows: S32 constructs a regression vector. The regression vector at time k includes the optimal estimates of fuel flow and speed at time k-1 and the piston line displacement.

[0083] For the current matrix row to be updated, a regression vector is constructed using the optimal estimates of fuel flow and engine speed output by the Kalman filter at time k-1, and the servo piston linear displacement measured by the LVDT sensor at time k-1. , The state-space model structure of this invention is matched to provide an input benchmark for parameter identification. Wherein, Let be the regression vector at time k. This is the optimal estimate of the engine speed at time k-1. This is the optimal estimate of the fuel flow rate at time k-1. The control input at time k-1 is the piston linear displacement.

[0084] In this embodiment, the parameters in matrices A and B are updated row by row.

[0085]

[0086]

[0087] , The parameter estimation vector at the k-th sampling time corresponds to the first and second rows of parameters in the state-space model, respectively. This represents the element in the i-th row and j-th column of the state transition matrix A(k) estimated at the k-th sampling time. This represents the element in the i-th row of the state transition matrix B(k) estimated at the k-th sampling time. make . , This represents the identification output at the k-th sampling time.

[0088] S33 uses the system parameter matrix updated at time k-1 and the regression vector at time k to calculate the prediction of fuel flow and speed at time k, and obtains the predicted values ​​of fuel flow and speed at time k.

[0089] Specifically, this embodiment utilizes the following formula:

[0090] Calculate the predicted values ​​of fuel flow rate and engine speed at time k; where, Let K be the predicted values ​​of fuel flow rate and engine speed at time k. Let k be the transpose of the regression vector at time k. The system parameters with index i updated at time k-1 include fuel flow and engine speed.

[0091] S34 calculates the residuals between the predicted and observed values ​​of fuel flow rate and engine speed at time k.

[0092] In this embodiment, the formula is used:

[0093] Calculate the residuals between the predicted and observed values. Among them, The residuals between the predicted and observed values ​​of fuel flow and engine speed at time k are called identification residuals. The residuals directly reflect the degree of matching between the current parameter matrix and the actual characteristics of the system. The larger the residuals, the more severe the model mismatch, and the greater the magnitude of parameter correction.

[0094] S35 uses the updated parameter estimate covariance matrix at time k-1, the regression vector at time k, and the forgetting factor to calculate the gain at time k.

[0095] Specifically, this embodiment utilizes the following formula: ,

[0096] Calculate the gain at time k; where G i (k) represents the gain at time k with index i. The covariance matrix of the parameter estimate with index i updated at time k-1. Let k be the regression vector. It is a forgetting factor.

[0097] The RLS gain determines the correction magnitude of the parameter to be estimated; the forgetting factor weakens the weight of historical data through exponential decay, allowing the algorithm to prioritize the recent operating characteristics of the engine.

[0098] S36 uses the system parameter matrix at time k-1, the gain at time k, and the residuals between the predicted and observed values ​​of fuel flow and speed to calculate the updated system parameter matrix at time k.

[0099] Specifically, this embodiment utilizes the following formula:

[0100] Calculate the system parameter matrix updated at time k; where, Let i be the system parameter matrix updated at time k. Let i be the residual between the predicted value and the observed value at time k with index i. S37 uses the forgetting factor, the parameter estimation covariance matrix updated at time k-1, the gain at time k, and the regression vector to calculate the parameter estimation covariance matrix updated at time k.

[0101] Specifically, this embodiment utilizes the following formula:

[0102] Calculate the parameter estimation covariance matrix updated at time k; where, The parameter estimate covariance matrix with index i is updated at time k.

[0103] The updated state transition matrix A and input control matrix B are fed back to the Kalman filter module to replace the original fixed parameter matrix for optimal system state estimation in the next cycle, thus completing the complete adaptive closed loop of state estimation, model correction, and accuracy improvement.

[0104] S4 constructs an innovation sequence based on the optimal estimate and the sensor observation, and calculates the residual between the optimal estimate and the sensor observation.

[0105] In this invention, innovation refers to the difference between the sensor's observation at the current moment and the prior estimate obtained during the Kalman filter prediction stage; arranging the innovations at each discrete sampling moment in chronological order constitutes an innovation sequence.

[0106] Within each airborne sampling cycle, after the Kalman filter completes the prediction phase and obtains the prior estimate, it immediately collects the sensor observations at the current moment and calculates the current information. The information at all moments is stored in chronological order, thus forming two independent information time series: fuel flow and engine speed.

[0107] The innovation values ​​and statistical characteristics directly reflect the system's operating state. When the main fuel regulation system is fault-free and the sensors are functioning normally, the innovation consists only of inherent system process noise, such as minute fluctuations in fuel line pressure and sensor measurement noise. Furthermore, the Kalman filter algorithm assumes that these noises are all zero-mean Gaussian white noise. Therefore, the innovation sequence of a fault-free system strictly follows a zero-mean, Gaussian distribution, with values ​​fluctuating randomly around 0 without a fixed offset, and the fluctuation amplitude conforming to Gaussian probability laws.

[0108] When a system malfunctions, it disrupts the match between model predictions and actual conditions, leading to significant changes in the statistical characteristics of the innovation sequence. If the malfunction causes systematic biases such as sensor zero-point drift or wear of the metering valve causing continuous flow deviation, the mean of the innovation sequence will deviate significantly from 0. If the malfunction causes intermittent or sudden disturbances such as sensor signal jumps or piston jamming, the variance of the innovation sequence will fluctuate drastically. If the malfunction originates from mechanical components such as the metering valve or differential pressure regulator failure, it will simultaneously affect the actual state of fuel flow and engine speed, causing the innovation sequences of both fuel flow and engine speed channels to synchronously become abnormal and exhibit a high degree of linear correlation.

[0109] In this invention, residual refers to the difference between the sensor's current observation value and the optimal estimate obtained during the Kalman filter correction stage.

[0110] Within each airborne sampling period, after the Kalman filter completes the correction phase and obtains the optimal estimate, it immediately subtracts the optimal estimate from the current sensor observation to obtain the residual at the current time. Unlike the innovation, the residual does not need to construct a long sequence; it mainly focuses on the numerical changes at the current time and within a short time window.

[0111] The residual is the deviation between the sensor observation and the actual state of the system after removing model errors and process noise. It is a specific indicator that reflects the performance of the sensor itself.

[0112] When the sensor is functioning normally, the optimal estimate from the Kalman filter has already undergone prediction and correction processes, eliminating the influence of inherent model errors and system process noise, and best reflects the actual operating state of the system. At this point, the residual consists only of the sensor's minute measurement noise, which is small and stable within the preset normal threshold range.

[0113] When a sensor malfunctions, such as zero-point drift, decreased sensitivity, signal distortion, or complete failure, its observed values ​​will deviate significantly from the true system state. However, the optimal estimate, relying on linear displacement input and an adaptively updated state-space model, remains unaffected by the faulty sensor and accurately reflects the true system state. In this case, the single-channel residual corresponding to the faulty sensor will continuously and significantly exceed the normal threshold, while the residual of the other channel remains normal. This is a key characteristic for subsequent steps to distinguish between sensor faults and mechanical component faults.

[0114] S5 performs fault diagnosis on the status of the main fuel conditioning system based on the information sequence and residual.

[0115] Specifically, if both the fuel flow information sequence and the engine speed information sequence are abnormal, and their correlation is higher than the correlation threshold, it is determined to be a mechanical fault; more specifically, if the mean and / or variance of a certain information sequence (fuel flow or engine speed) exceeds the corresponding threshold, it is determined to be abnormal.

[0116] The mean index is the average value of the news sequence within a certain time window. If it exceeds the threshold, it is judged as a mean anomaly, reflecting a systematic bias in the system. The variance index is the variance of the news sequence within the same time window. If it exceeds the threshold or the variance fluctuates violently and irregularly, it is judged as a variance anomaly, reflecting sudden and intermittent disturbances in the system. The dual-channel correlation coefficient is the Pearson correlation coefficient between the fuel flow news sequence and the engine speed news sequence. If it is higher than the threshold, it is judged as a correlation anomaly, reflecting a common fault source in both channels.

[0117] If only one item in the fuel flow information sequence and the speed information sequence is abnormal, and the residual corresponding to the abnormal item exceeds the residual threshold, the corresponding sensor is judged to be faulty; in the case of sensor failure, the optimal estimate is used as the sensor feedback.

[0118] The specific judgment process for this step is as follows: The first step is to determine all anomalies in the new information sequence to identify potential system malfunctions.

[0119] For the information sequences of fuel flow and engine speed in both channels, the mean and variance are detected simultaneously. If the mean or variance of either channel exceeds the preset threshold, the information sequence is determined to be abnormal, indicating that there is a potential fault or a sudden change in the operating condition of the system, and the next step is to preliminarily distinguish the fault type. If both indicators are normal, the system is determined to be normal, and real-time monitoring continues without further steps.

[0120] The second step is dual-channel correlation detection to initially distinguish the fault types.

[0121] In cases of abnormal information sequences, the correlation coefficient between the fuel flow and engine speed information sequences is calculated. Based on whether the correlation coefficient exceeds the threshold, mechanical component failure and sensor failure are preliminarily distinguished.

[0122] If the correlation coefficient is greater than or equal to the preset threshold, it is judged as a suspected mechanical component failure. The reason is that the dual-channel information synchronization is abnormal and highly correlated, indicating that there is a common source of failure. In the main fuel regulation system, the only common source of failure in both channels is the core mechanical components such as the metering valve, follow-up piston, and differential pressure regulator. Such failures will simultaneously affect the actual state of fuel flow and engine speed, directly triggering a mechanical failure warning and entering the performance degradation quantification stage of subsequent trend analysis. If the correlation coefficient is less than the preset threshold, the sensor is suspected of being faulty. This is because only one channel has abnormal information while the other channel is normal. The two channels do not have a common source of failure, which is consistent with the characteristic of the sensor working independently. At this time, the sensor cannot be directly determined to be faulty, and the process proceeds to the third step of residual accuracy verification.

[0123] The third step is residual threshold verification, which ultimately determines whether the problem is a sensor malfunction.

[0124] In cases where sensor malfunction is suspected, the residuals of the corresponding channels with abnormal information are calculated and threshold detection is performed. The other normal channel does not need to be verified. The final judgment is made based on whether the residuals are abnormal, which effectively distinguishes between real sensor malfunctions and sudden changes in operating conditions / model not yet adapted.

[0125] If the residual value of the corresponding channel continuously exceeds the preset threshold, it is ultimately determined that the sensor in that channel is faulty. At this time, the system automatically triggers a sensor fault alarm and immediately starts a fault-tolerant control strategy: reducing the confidence level of the faulty sensor and replacing the measured value of the faulty sensor with the optimal estimate output by the Kalman filter as the control feedback signal of the main fuel regulation system to ensure continuous and stable operation of the system.

[0126] If the corresponding channel residual value is within the preset threshold range: sensor faults are ruled out, and the system is determined to have experienced a sudden change in operating conditions or the model is not yet adapted. In this case, no fault alarm is triggered, but the operating condition data is only included in the parameter update sample of the RLS algorithm. The model parameters are then optimized and adapted to the current operating conditions in the next iteration to avoid misjudgment of faults due to changes in operating conditions.

[0127] S6 statistically analyzes the changes in the system parameter matrix, the parameter estimation covariance matrix formed during the optimization of the system parameter matrix, and the deviation between the optimal estimated value and the design standard value of the main fuel regulation system to conduct trend analysis on the health status of the main fuel regulation system.

[0128] Trend analysis aims to predict the future health status, performance degradation, and potential failures of a system through long-term data analysis, providing a quantitative understanding of system performance, supporting long-term predictive maintenance, and optimizing maintenance plans. Its core functions are as follows: Long-term trend recording and analysis of model parameters.

[0129] The optimal values ​​estimated by the Kalman oscillator are periodically input into the Recursive Least Squares (RLS) model to update and adapt the Kalman model parameters. In RLS, the state transition matrix A and the input control matrix B are updated, thereby improving the model's adaptability and accuracy. The state transition matrix A describes the system's state changes over time steps. It reflects how the system transitions from the current state to the next. In health management, significant changes in the state transition matrix may indicate changes in the system's dynamic behavior, suggesting potential problems. The input control matrix B describes the impact of external control inputs, such as throttle and fuel flow regulation, on the system state. Abnormal changes in the input control matrix may indicate problems with the control system or changes in the influence of external inputs on the system.

[0130] Monitoring the covariance matrix of the RLS model.

[0131] The Relative Strengths (RSS) model allows the Kalman model to adapt to system changes in real time by continuously updating its parameters. In this process, the parameter estimation covariance matrix is ​​crucial in the RLS model because it reflects the model's confidence in the parameters, as well as its stability and accuracy. Changes in the trace value of the covariance matrix reveal the model's stability; a continuously increasing trace value indicates increased uncertainty, potentially suggesting system performance degradation and requiring timely maintenance.

[0132] Monitoring the changing trends of system status.

[0133] Monitor the deviation trend between the optimal estimated value and the design standard value of the Kalman filter output. When the deviation between the estimated value and the design value continues to increase, it indicates that the system may be entering a degradation phase or that there are unstable factors. At the same time, by comparing historical data and current estimated values, it is possible to analyze whether the system has experienced performance degradation or abnormal fluctuations.

[0134] Example 2: like Figure 3As shown, this embodiment also provides a state estimation and health management system for an aero-engine main fuel regulation system, including: The model building module is used to construct a state-space model of the main fuel regulation system based on aero-engine test data; The optimal estimate prediction module is used to estimate the fuel flow rate and engine speed based on the state space model, so as to obtain the optimal estimate of the fuel flow rate and engine speed. The system parameter matrix optimization module is used to optimize the system parameter matrix of the state-space model based on the optimal estimated value; the system parameter matrix includes a state transition matrix and an input control matrix. The calculation module is used to construct an innovation sequence based on the optimal estimate and the sensor observation, and to calculate the residual between the optimal estimate and the sensor observation; The fault diagnosis module is used to diagnose faults in the status of the main fuel conditioning system based on the information sequence and residuals. The trend analysis module is used to statistically analyze the changes in the system parameter matrix, the parameter estimation covariance matrix formed during the optimization process of the system parameter matrix, and the deviation between the optimal estimated value and the design standard value of the main fuel regulation system, so as to perform trend analysis on the health status of the main fuel regulation system.

[0135] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0136] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0137] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for state estimation and health management of the main fuel regulation system of an aero-engine, characterized in that, include: Based on aero-engine test data, a state-space model of the main fuel regulation system is constructed. Based on the state-space model, fuel flow rate and engine speed are estimated to obtain the optimal estimated values ​​of fuel flow rate and engine speed. The system parameter matrix of the state-space model is optimized based on the optimal estimate; the system parameter matrix includes the state transition matrix and the input control matrix. Based on the optimal estimate and sensor observations, an innovation sequence is constructed, and the residual between the optimal estimate and sensor observations is calculated. Fault diagnosis of the main fuel conditioning system status is performed based on the aforementioned information sequence and residuals; The changes in the statistical system parameter matrix, the parameter estimation covariance matrix formed during the optimization of the system parameter matrix, and the deviation between the optimal estimated value and the design standard value of the main fuel regulation system are used to conduct trend analysis on the health status of the main fuel regulation system.

2. The method for state estimation and health management of the main fuel regulation system of an aero-engine according to claim 1, characterized in that, The state-space model includes: Equations of state , Observation equations ; Among them, X k Let Q be the state vector at time k, which includes the fuel flow rate Q. e_k and engine speed N e_k ;Y k The observation vector at time k is the fuel flow rate Q sampled by the sensor. s_k and engine speed N s_k U k-1 The control input at time k-1 is the piston linear displacement L; W k For process noise; V k denoted as observation noise; A is the state transition matrix; B is the input control matrix; and H is the observation matrix.

3. The method for state estimation and health management of the main fuel regulation system of an aero-engine according to claim 1, characterized in that, The steps to obtain the optimal estimate include: Using the optimal estimates of fuel flow rate and speed at time k-1 and the piston line displacement, the fuel flow rate and speed at time k are predicted to obtain the prior estimates of fuel flow rate and speed at time k. Based on the error covariance matrix of the process at time k-1 and the error covariance matrix of the system's optimal estimate, the error covariance matrix of the system's prior estimate at time k is predicted. The confidence level of the observed values ​​is calculated based on the error covariance matrix of the prior estimates of the system at time k, and the Kalman gain is obtained. The optimal estimates of fuel flow rate and speed at time k are calculated using Kalman gain, prior estimates of fuel flow rate and speed at time k, and observed values ​​of fuel flow rate and speed at time k; and the error covariance matrix of the optimal estimates at time k is calculated.

4. The method for state estimation and health management of the main fuel regulation system of an aero-engine according to claim 3, characterized in that, Using the formula: Calculate the prior estimates of fuel flow rate and engine speed at time k; where, Let k be the prior estimate. Let A be the optimal estimate at time k-1, B be the state transition matrix, and U be the input control matrix. k The control input of the main fuel conditioning system at time k is the piston linear displacement L. Using the formula: Calculate the error covariance matrix of the prior estimate at time k; where, Let k be the error covariance matrix of the prior estimate. Let A be the error covariance matrix of the optimal estimate at time k-1. T S is the transpose of the state transition matrix, and S is the error covariance matrix at time k-1. Using the formula: Calculate the Kalman gain; where K k Let H be the Kalman gain at time k, and H be the observation matrix. T R is the transpose of the observation matrix, and R is the error covariance matrix of the observations; Using the formula: Calculate the optimal estimates of oil flow rate and rotational speed at time k; where, Y represents the optimal estimate of oil flow rate and rotational speed at time k. k Let be the observation vector at time k; Using the formula: Calculate the error covariance matrix of the optimal estimate at time k; where, Let I be the error covariance matrix of the optimal estimate at time k, and let I be the identity matrix.

5. The method for state estimation and health management of the main fuel regulation system of an aero-engine according to claim 1, characterized in that, The steps for optimizing the system parameter matrix include: Construct a regression vector. The regression vector at time k includes the optimal estimates of fuel flow rate and engine speed at time k-1 and the piston line displacement. The system parameter matrix updated at time k-1 and the regression vector at time k are used to calculate the predicted values ​​of fuel flow rate and speed at time k, thus obtaining the predicted values ​​of fuel flow rate and speed at time k. Calculate the residuals between the predicted and observed values ​​of fuel flow rate and engine speed at time k; The covariance matrix is ​​estimated using the updated parameters at time k-1, the regression vector at time k, and the forgetting factor to calculate the gain at time k. The updated system parameter matrix at time k is calculated using the system parameter matrix at time k-1, the gain at time k, and the residuals between the predicted and observed values ​​of fuel flow and speed. The parameter estimation covariance matrix updated at time k is calculated using the forgetting factor, the parameter estimation covariance matrix updated at time k-1, the gain at time k, and the regression vector.

6. The method for state estimation and health management of the main fuel conditioning system of an aero-engine according to claim 5, characterized in that, Using the formula: Calculate the predicted values ​​of fuel flow rate and engine speed at time k; where, Let K be the predicted values ​​of fuel flow rate and engine speed at time k. Let k be the transpose of the regression vector at time k. The system parameter with index i is updated at time k-1; Using the formula: , Calculate the gain at time k; where G i (k) represents the gain at time k with index i. The covariance matrix of the parameter estimate with index i updated at time k-1. Let k be the regression vector. Forgetting factor; Using the formula: Calculate the system parameter matrix updated at time k; where, Let i be the system parameter matrix updated at time k. Let i be the residual between the predicted value and the observed value at time k with index i. Using the formula: Calculate the parameter estimation covariance matrix updated at time k; where, The parameter estimate covariance matrix with index i is updated at time k.

7. The method for state estimation and health management of the main fuel regulation system of an aero-engine according to claim 1, characterized in that, The steps for fault diagnosis of the main fuel conditioning system state based on the information sequence and residuals include: If both the fuel flow information sequence and the engine speed information sequence are abnormal, and their correlation is higher than the correlation threshold, it is determined to be a mechanical fault. If only one item in the fuel flow information sequence and the speed information sequence is abnormal, and the residual corresponding to the abnormal item exceeds the residual threshold, the corresponding sensor is judged to be faulty; in the case of sensor failure, the optimal estimate is used as the sensor feedback.

8. The method for state estimation and health management of the main fuel regulation system of an aero-engine according to claim 7, characterized in that, If the mean and / or variance of the new information sequence exceeds the corresponding threshold, it is determined to be abnormal.

9. The method for state estimation and health management of the main fuel conditioning system of an aero-engine according to claim 1, characterized in that, The steps for trend analysis of the health status of the main fuel conditioning system include: When the change in the state transition matrix exceeds the state change threshold, the system dynamic behavior is determined to be abnormal; when the change in the input control matrix exceeds the control change threshold, the system control characteristics are determined to be abnormal. When the trace value of the parameter estimation covariance matrix shows a continuous increasing trend, it is determined that the model uncertainty is increasing and the system has the risk of performance degradation. When the deviation continues to increase beyond the preset trend threshold, or when the optimal estimated value fluctuates abnormally, the system is determined to have entered the performance degradation stage.

10. A state estimation and health management system for an aircraft engine main fuel regulation system, characterized in that, include: The model building module is used to construct a state-space model of the main fuel regulation system based on aero-engine test data; The optimal estimate prediction module is used to estimate the fuel flow rate and engine speed based on the state space model, so as to obtain the optimal estimate of the fuel flow rate and engine speed. The system parameter matrix optimization module is used to optimize the system parameter matrix of the state-space model based on the optimal estimated value; the system parameter matrix includes a state transition matrix and an input control matrix. The calculation module is used to construct an innovation sequence based on the optimal estimate and the sensor observation, and to calculate the residual between the optimal estimate and the sensor observation; The fault diagnosis module is used to diagnose faults in the status of the main fuel conditioning system based on the information sequence and residuals. The trend analysis module is used to statistically analyze the changes in the system parameter matrix, the parameter estimation covariance matrix formed during the optimization process of the system parameter matrix, and the deviation between the optimal estimated value and the design standard value of the main fuel regulation system, so as to perform trend analysis on the health status of the main fuel regulation system.

Images