An aircraft fault identification method based on the integral of normalized fault degree density

Abstract

A method for aircraft fault identification based on the integral of normalized fault degree density. The purpose of the present invention is to solve the problems of backward judgment strategies and slow fault identification speed in existing aircraft fault identification technologies. By researching an online fault identification method with advanced, fast, and accurate judgment strategies, the speed and accuracy of fault identification technology are improved. The present invention designs an exponential function that maps from [0, +inf) to [0, 1], normalizes the estimated residuals and covariances corresponding to different fault modes, and eliminates the magnitude differences. A more sensitive fault degree index is obtained through integration, and at the same time, it can cope with measurement noise interference, improving the identification speed and accuracy.

Description

Technical Field

[0001] The present invention relates to a low-time-consuming and high-precision aircraft online fault identification method based on an exponential normalized fault degree density integral criterion, which can identify the fault location, fault mode and fault degree according to the current aircraft state. Background Art

[0002] The probability of failure of aircraft power system / actuator components increases dramatically under the influence of complex flight environments, and online fault identification is the basis for effective response. In view of the difficulties in accurately modeling aircraft systems, the lack of prior knowledge of fault information, and the limited quantity and quality of measurement data, the existing fault identification technology has backward judgment strategies, slow fault identification speed, and insufficient fault degree identification accuracy, and it is urgent to develop new fast and accurate identification methods. Summary of the invention

[0003] The purpose of the present invention is to solve the problems of backward judgment strategy and slow fault identification speed of existing aircraft fault identification technology, and to provide an aircraft fault identification method based on normalized fault degree density integral. By studying the advanced, fast and accurate online fault identification method with judgment strategy, the speed of fault identification technology and the identification accuracy are improved.

[0004] The objective of the present invention is achieved through the following technical solutions:

[0005] An aircraft fault identification method based on normalized fault degree density integral, the method steps are as follows:

[0006] Step 1: Analyze the aircraft status before and after the failure: x-axis position x, y-axis position y, z-axis position z, x-axis speed v x , y-axis speed v y , z-axis speed v z , x-axis attitude angular velocity ω x , y-axis attitude angular velocity ω y , z-axis attitude angular velocity ω z , the thrust T of the engine at the ith position i , the swing angle θ of the jth servo mechanism j According to the changes of

[0007] Step 2: Design multiple sets of extended Kalman filter (EKF) observers to observe the aircraft state x, y, z, v x , v y , v z ,ω x ,ω y ,ω z , Ti ,θ j Make estimates;

[0008] Step 3: Design a normalized exponential function Where c is the function variable, e is the e-exponential function, α and β are adjustable parameters of the exponential function, and normalized exponential functions of different shapes are designed by adjusting α and β to process the estimated state quantity with input [0, +inf) into the observed value of the fault degree with a value in the interval [0, 1];

[0009] Step 4: Comprehensively compare the observed values ​​ε of each fault i , σ i , calculate the fault severity index density f of each engine / servo mechanism fault severity i (t);

[0010] Step 5: Based on the idea of ​​error accumulation, the tracking degree index density f i (t) Integration Obtain the fault severity index F(t) and the prior index threshold F m The comparison results in fast and accurate recognition.

[0011] Furthermore, in step 1, the specific process of establishing the system state equation and the measurement equation is as follows:

[0012] Step 1: Set the system state vector x = [ω x ω y ω z ] T The output of the system

[0013] Step 1 and 2: Set the system output matrix C including the measurement error and design the system matrix A to satisfy: Where q represents the dynamic pressure, S m represents the frontal area, V represents the aircraft speed, Respectively represent the moment of inertia of the three axes of the aircraft, represents the x-axis torque coefficient, represents the y-axis torque coefficient, represents the z-axis torque coefficient, l represents the distance from the thrust vector to the vehicle body axis; the disturbance vector d(t) of the designed system satisfies: Among them, M BX ,M BY ,M BZ Represent the known disturbance torques of the three axes, r z Represents the cross-sectional radius of the aircraft, M KY ,M KZ denote the known control torques on the y-axis and z-axis, δ γ,δ ψ , Represents the roll angle, yaw angle and pitch angle, m R represents the disturbance torque coefficient, l R represents the distance from the disturbance vector to the vehicle body axis, J R represents the moment of inertia of the aircraft with disturbance, x R represents the distance from the disturbance vector to the y-axis of the aircraft, x T represents the distance from the thrust vector to the body axis of the aircraft; the control matrix B of the designed system is written as Where r represents the longitudinal radius of the aircraft, T 1 ,T 2 Represent the thrust of the two engines respectively; the input matrix u of the design system = [δ 1 δ 2 δ 3 δ 4 ], where δ 1 ,δ 2 ,δ 3 ,δ 4 Respectively represent the swing angles in four directions;

[0014] Step 13: According to the dynamic equation of the rocket around the center of mass and the kinematic equation, the following state equation and measurement equation are established:

[0015]

[0016] Furthermore, in step 2, the design steps of the multiple groups of extended Kalman filter observers are:

[0017] Step 21: Design the extended Kalman filter observer, the state variable time update equation is: and

[0018]

[0019] Where: is the estimated value of the state at step k, is the state equation of the kth step, T = t k+1 -t k is the filtering step size, t k ,t k+1 They represent the time corresponding to the kth step and the time corresponding to the k+1th step, respectively. k+1 is the k-th step state equation Estimated value of the state at step k The Jacobian matrix of ;

[0020] Step 22: The time update equation of the state error covariance matrix is:

[0021]

[0022] Where: Q k+1 is the noise variance matrix of the k+1th step system process, P k is the actual state error covariance matrix of the kth step, P k+1|k is the state error covariance matrix solved in the kth step, Φ k+1 is the state transfer matrix of the kth step, expressed as:

[0023]

[0024] Where: I is the unit matrix, T is the step size, F k+1 Represents the Jacobian matrix of the k+1th step state equation for state variables;

[0025] Step 2 and 3: The Kalman filter gain update equation is:

[0026]

[0027] Where: K k+1 is the gain matrix of the k+1th step, R k+1 is the measurement noise matrix of the k+1th step, H k+1 The measurement equation for the k+1th step is For state variables The Jacobian matrix of .

[0028] Step 24: The measurement of state variables is updated as follows:

[0029]

[0030] Where: Estimate the state variables for the k+1th step, Solve the state variables for the kth step, is the measurement value of the k+1th step;

[0031] Step 25: The measurement update of the state error covariance matrix is:

[0032] P k+1 =(IK k+1 H k+1 ) k+1|k

[0033] Where: P k+1 Estimated state error covariance matrix for the k+1th step;

[0034] The discretized thrust expanded state equation and servo mechanism expanded state equation are substituted into the extended Kalman filter. At each moment, the system state is updated and the measurement update is corrected through the information input by the navigation system. The observer finally realizes thrust identification and servo mechanism swing angle identification.

[0035] Furthermore, in step 4, the fault degree index density calculation step is:

[0036] Step 41: Obtain the state residual e through each filter observer i With covariance S i ;

[0037] Step 42: The state residual e i With covariance S i As the input of the normalized exponential function S(c), the output value is the residual fault degree observation value ε in the interval [0, 1] i and covariance of observed fault severity σ i ;

[0038] Step 43: Use the fault observation values ​​given by multiple fault observation models to calculate the fault degree index density f that can directly reflect the fault degree of each engine / servo mechanism / actuator i (t), the subscript i indicates different fault locations (i and j are only for the convenience of summation and have no difference in meaning):

[0039]

[0040] The beneficial effects of the present invention relative to the prior art are:

[0041] (1) An exponential function is designed to map from [0, +inf) to [0, 1] to normalize the estimated residuals and covariances corresponding to different fault modes and eliminate the magnitude differences.

[0042] (2) A more sensitive fault severity index is obtained through integration, which can also cope with measurement noise interference and improve identification speed and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a normalized exponential function curve graph of the present invention;

[0044] Figure 2 It is a schematic diagram of error accumulation determination of the present invention. DETAILED DESCRIPTION

[0045] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments, but is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention should be included in the protection scope of the present invention.

[0046] Embodiment 1:

[0047] A method for aircraft fault identification based on normalized fault degree density integral is designed with the following specific steps:

[0048] Step 1: Analyze the aircraft state (position, velocity, attitude angular velocity, thrust and swing angle) before and after the failure x, y, z, v x , v y , v z ,ω x ,ω y ,ω z , T i ,θ j According to the changes of

[0049] Step 2: Design multiple sets of extended Kalman filter (EKF) observers to observe the aircraft engine thrust T i 、Servo mechanism swing angle θ j Estimation of state quantities;

[0050] Step 3: Design a normalized exponential function Among them, α and β are adjustable parameters of the exponential function. By adjusting α and β, normalized exponential functions of different shapes are designed to process the estimated state quantity with input [0, +inf) into the observed value of the fault degree with a value in the interval [0, 1].

[0051] Step 4: Comprehensively compare the observed values ​​ε of each fault i , σ i , calculate the fault severity index density f of each engine / servo mechanism fault severity i (t);

[0052] Step 5: Based on the idea of ​​error accumulation, the tracking degree index density f i (t) Integration Obtain the fault severity index F(t) and the prior index threshold F m The comparison results in fast and accurate recognition.

[0053] Furthermore, in step 1, the system state equation and measurement equation are established as follows:

[0054] Step 1.1: Set the system state vector x = [ω x ω y ω z ] T The output of the system

[0055] Step 1.2: Set the system output matrix C including the measurement error and design the system matrix A to satisfy: The disturbance vector d(t) of the designed system satisfies: The control matrix B of the designed system is written as The input matrix of the design system u=[δ 1 δ 2 δ 3 δ 4 ];

[0056] Step 1.3: According to the dynamic equation and kinematic equation of the rocket around the center of mass, the following state equation and measurement equation are established:

[0057]

[0058] Furthermore, the steps of designing multiple groups of extended Kalman filter observers in step 2 are as follows:

[0059] Step 2.1: Design the extended Kalman filter observer, and the state variable time update equation is:

[0060]

[0061] Where: T = t k+1 -t k is the filtering step size, F k+1 is the state equation Estimated value of the state The Jacobian matrix of ;

[0062] Step 2.2: The time update equation of the state error covariance matrix is:

[0063]

[0064] Where: Q k+1 is the system process noise variance matrix, Φ k+1 is the state transfer matrix, expressed as:

[0065]

[0066] Where: I is the unit matrix.

[0067] Step 2.3: The Kalman filter gain update equation is:

[0068]

[0069] Where: R k+1 is the measurement noise matrix, H k+1 The measurement equation For state variables The Jacobian matrix of .

[0070] Step 2.4: The measurement of state variables is updated as:

[0071]

[0072] Where: is the measured value.

[0073] Step 2.5: The measurement update of the state error covariance matrix is:

[0074] P k+1 =(IK k+1 H k+1 ) k+1|k

[0075] Substitute the discretized thrust expansion state equation and servo mechanism expansion state equation into the extended Kalman filter, x(k+1)=Φ(k)x(k)+w(k) as the state equation of the system, and y(k)=x(k)+v(k) as the measurement equation of the system. At each moment, the system is updated and the measurement update is corrected through the information input by the navigation system, and the observer finally realizes the thrust identification and servo mechanism swing angle identification.

[0076] Furthermore, the steps for calculating the fault degree index density in step 4 are as follows:

[0077] Step 4.1: Obtain the state residual e through each filter observer i With covariance S i ;

[0078] Step 4.2: The state residual e i With covariance S i As the input of the normalized exponential function S(c), the output value of the fault degree observation ε in the interval [0,1] is i , σ i ;

[0079] Step 4.3: Use the fault observation values ​​given by multiple fault observation models to calculate the fault degree index density f that can directly reflect the fault degree of each engine / servo mechanism / actuator i (t), the subscript i indicates different fault locations:

[0080]

Claims

1. A method for aircraft fault identification based on normalized fault degree density integration, Characterized in that: The steps of the method are as follows: Step 1: Analyze the changes in the state variables of the aircraft before and after the failure, including the x-axis position x, y-axis position y, z-axis position z, x-axis velocity v x , y-axis velocity v y , z-axis velocity v z , x-axis attitude angular velocity ω x , y-axis attitude angular velocity ω y , z-axis attitude angular velocity ω z , the thrust T of the i-th position engine i , the swing angle θ of the j-th servo mechanism j . For various fault models, establish multiple unbiased dynamic models with faults, and establish the system state equation and measurement equation Step 2: Design multiple sets of Extended Kalman Filter (EKF) observers to estimate the state variables x, y, z, v x , v y , v z , ω x , ω y , ω z , T i , θ j for estimation; Step 3: Design a normalized exponential function where c is the independent variable of the function, and e -βc is the exponential function, and α and β are adjustable parameters of the exponential function. By adjusting α and β, normalized exponential functions with different shapes are designed to process the estimated state quantity with an input of [0, +inf) into the fault degree observation value with a numerical value in the interval [0, 1]; Step 4: Comprehensively compare each fault observation value ε i , σ i , and calculate the fault degree index density f i (t) of the fault degree of each engine / servo mechanism; the calculation steps of the fault degree index density are as follows: Step 4-1: Obtain the state residual e through each filtering observer i and the covariance S i ; Step Four Two: Take the state residual e i and the covariance S i as the inputs of the normalized exponential function S(c), and output the residual fault degree observation value ε i and the covariance fault degree observation value σ i ; Step 43: Calculate the failure degree index density f that can directly reflect the failure degree of each engine / servo mechanism / actuator by using the failure observation values given by multiple failure observation models i (t), where the subscript i represents different failure positions: Step 5: Based on the idea of error accumulation, integrate the failure degree exponential density f i (t) to obtain the failure degree index F(t), and compare it with the prior index threshold F m to obtain a fast and accurate identification result.

2. The method for aircraft fault identification based on normalized fault degree density integration according to claim 1, Characterized in that: In step one, the specific establishment process of the system state equation and the measurement equation is as follows: Step 1: Set the state vector x of the system as x = [ω x ω y ω z T and the output quantity of the system ​ Step 1 and Step 2: Set the system output matrix C containing measurement errors, and design the system matrix A to satisfy: where q represents dynamic pressure, S m represents the frontal area, V represents the vehicle speed, J X1 , J Y1 , J Z1 represent the moments of inertia of the vehicle about its three axes respectively, represents the moment coefficient of the x-axis rotation, represents the moment coefficient of the y-axis rotation, represents the moment coefficient of the z-axis rotation, l represents the distance from the thrust vector to the vehicle body axis; design the disturbance vector d(t) of the system to satisfy: where M BX , M BY , M BZ represent the known disturbing torques about the three axes respectively, r z represents the cross-sectional radius of the vehicle, M KY , M KZ represent the known control torques about the y-axis and z-axis respectively, represent the roll angle, yaw angle and pitch angle respectively, m R represents the disturbance moment coefficient, l R represents the distance from the disturbance vector to the vehicle body axis, J R represents the moment of inertia of the vehicle with disturbance, x R represents the distance from the disturbance vector to the y-axis of the vehicle, x T represents the distance from the thrust vector to the vehicle body axis; design the control matrix B of the system and write it as where r represents the longitudinal radius of the vehicle, T 1 , T 2 represent the thrusts of two engines respectively; design the input matrix u = [δ 1 δ 2 δ 3 δ 4 , where δ 1 , δ 2 , δ 3 , δ 4 represent the swing angles in 4 directions respectively; Step 1-3: Establish the following state equation and measurement equation according to the rocket's dynamics equation and kinematics equation about the centroid:

3. The method for aircraft fault identification based on normalized fault degree density integration according to claim 1, Characterized in that: In step two, the design steps of the multiple extended Kalman filter observers are as follows: Step 2-1: Design an extended Kalman filter observer, and the time update equation of the state variable is: Wherein: is the state estimation value at the k-th step, is the state equation at the k-th step, T = t k+1 - t k is the filtering step size, t k , t k+1 respectively represent the time corresponding to the k-th step and the time corresponding to the (k + 1)-th step, F k+1 is the state equation at the k-th step for the state estimation value at the k-th step Jacobian matrix; Step 2-2: The time update equation of the state error covariance matrix is: where: Q k+1 is the system process noise variance matrix at the (k + 1)-th step, P k is the actual state error covariance matrix at the k-th step, P k+1|k is the calculated state error covariance matrix at the k-th step, Φ k+1 is the state transition matrix at the k-th step, expressed as: where: I is the identity matrix, T is the step size, and F k+1 represents the Jacobian matrix of the state equation at the (k + 1)-th step with respect to the state variables; Step 2-3: The Kalman filter gain update equation is: Where: K k+1 is the gain matrix at the (k + 1)-th step, R k+1 is the measurement noise matrix at the (k + 1)-th step, H k+1 is the measurement equation at the (k + 1)-th step for the state variable Jacobian matrix; Step 2-4: The measurement update of the state variable is: Wherein: is the estimated state variable at the (k + 1)-th step, is the calculated state variable at the k-th step, is the measurement value at the (k + 1)-th step; Step 2-5: The measurement update of the state error covariance matrix is: P k+1 = (I - K k+1 H k+1 )P k+1|k Where: P k+1 is the estimated state error covariance matrix at the (k + 1)-th step; Substitute the discretized thrust augmented state equation and the servo mechanism augmented state equation into the extended Kalman filter. At each moment, the system is updated in state and corrected in measurement update through the information input by the navigation system, and the observer finally realizes thrust identification and servo mechanism swing angle identification.

Images