Satellite attitude control system event-triggered stf fault detection method

Abstract

The application discloses a satellite attitude control system event-triggered STF fault detection method, and belongs to the technical field of detection, which comprises the following steps: establishing a fault model of a satellite attitude control system; introducing an event-triggered mechanism; adopting an STF method to design an event-triggered fault detection filter; obtaining an estimation error dynamic equation and a residual signal through linearization processing of the fault detection filter, realizing complete decoupling of the residual signal and the event-triggered error; converting a design problem of a residual generator into solving a gain matrix of the fault detection filter; calculating a residual evaluation function and comparing the residual evaluation function with a threshold value to judge whether a fault occurs or not. The application utilizes the STF method to design the fault detection filter, can enable the system to keep the tracking ability to a sudden state when reaching a stable state, adopts a brand-new event-triggered framework, realizes complete decoupling of the residual signal and the event-triggered error, and effectively improves the resource utilization rate under the premise of guaranteeing the system performance.

Description

Technical Field

[0001] This invention belongs to the field of detection technology, specifically relating to a method for detecting event-triggered STF faults in a satellite attitude control system. Background Technology

[0002] With the further development and utilization of space resources, satellite systems adapted for long-term, large-scale space operations have experienced rapid development in both military and civilian fields. Due to the complex structure of satellites, numerous moving parts, harsh operating environments, and the requirement for long-term on-orbit operation, various malfunctions are inevitable. As one of the most critical subsystems of a satellite system, the health of the satellite attitude control system is a prerequisite for the successful completion of space missions, and sensor malfunctions have the most significant impact on this system. Therefore, timely and accurate detection of sensor malfunctions in the system is of great practical importance. Analytical model-based fault detection methods are the earliest developed and most widely studied approach. Depending on the residual generation method, they can be divided into observer-based methods, equivalent space methods, and parameter estimation methods. Among them, observer-based fault detection methods establish state-space equations and measurement equations describing the system, construct filters / state observers to estimate or reconstruct the system's measurable variables, and simultaneously subtract the estimated output from the actual system output to obtain the output estimation error, further constructing residuals. Observer-based fault detection methods are simple to implement, mainly involving residual generation and evaluation, and have been widely applied.

[0003] In 2016, E. Farjah et al.'s paper, "Extended Kalman filter based method for inter-turn fault detection of the switched reluctance motors," published in the IET Electric Power Applications journal, proposed a fault detection method based on an extended Kalman filter for inter-turn short-circuit faults in switched reluctance motor windings. However, the extended Kalman filter method has poor robustness to model uncertainties and loses its ability to track abrupt changes when the system reaches a steady state.

[0004] Furthermore, due to the complexity of the system and the limitations of physical equipment, it is essential to design an effective transmission method to improve resource utilization. Traditional time-triggered mechanisms inevitably lead to wasted network resources due to frequent data transmission, while event-triggered mechanisms are a non-uniformly triggered method that operates on demand. Data transmission is determined by an event generator, and events only occur when predefined trigger conditions are met.

[0005] To save communication resources and reduce unnecessary data transmission, event-triggered mechanisms have attracted increasing attention in the research of fault detection in complex systems. In 2018, the International Journal of Adaptive Control and Signal Processing published a paper by A. Golabi et al., "Event-triggered fault detection for discrete-time LPV systems with application to a laboratory tank system," which investigated a novel event-triggered fault detection method for discrete dynamic systems with linear variable parameter models. However, because the event-triggered mechanism is an "on-demand" non-uniform periodic triggering method, the loss of system information and the introduction of time-varying modes with non-uniform sampling periods mean that the residual signal generated by the aforementioned event-triggered system is affected not only by interference and faults but also by event-triggered errors. For the fault detection problem of nonlinear systems, existing event-triggered fault detection methods focus on reducing the impact of event-triggered errors on the residual signal, but have not achieved complete decoupling between the two. Summary of the Invention

[0006] To address the aforementioned technical problems in existing technologies, this invention proposes a fault detection method for event-triggered STF (Strong Tracking Filter) in satellite attitude control systems. The method is rationally designed, overcomes the shortcomings of existing technologies, and has good performance.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for detecting event-triggered STF faults in a satellite attitude control system includes the following steps:

[0009] Step 1: Establish a fault model for the satellite attitude control system;

[0010] Step 2: Introduce an event triggering mechanism;

[0011] Step 3: Design an event-triggered fault detection filter using the STF method;

[0012] Step 4: By linearizing the fault detection filter, the dynamic equation of the estimation error and the residual signal are obtained, thus achieving complete decoupling between the residual signal and the event triggering error;

[0013] Step 5: Transform the design problem of the residual generator into solving the gain matrix of the fault detection filter;

[0014] Step 6: Calculate the residual evaluation function and compare it with the threshold to determine whether a fault has occurred.

[0015] Preferably, in step 1, the specific content is as follows:

[0016] The dynamic equations of the satellite attitude control system:

[0017]

[0018] Among them, I i (i = x, y, z) represents the satellite's moments of inertia on its three principal axes; ω i (i = x, y, z) represents the projection of the attitude velocity vector onto the three principal axes of inertia; T i (i = x, y, z) represents the control torque components along the principal axis of inertia; This represents the time derivative of the moment of inertia. Represents the angular acceleration components;

[0019] The transformation relationship between the satellite's body coordinate system and its orbital coordinate system is defined as the satellite's attitude. When the attitude angle is small, the dynamic equations are rewritten as follows:

[0020]

[0021] in, θ and ψ are the roll angle, pitch angle, and yaw angle, respectively; ω0 is the constant orbital speed.

[0022] Define state vector and control input vector u(t)=[T x (t)T y (t)T z (t)] T Without considering faults, the dynamics of the satellite attitude control system are described by the following nonlinear model:

[0023]

[0024] Where y(t) is the measurement output vector, d(t) is the process noise, and v(t) is the measurement noise.

[0025] q(x(t)) = [q1 q2 q3 q4 q5 q6] T

[0026] q1=ω0ψ(t)+ω x (t),

[0027] q2=ω0+ω y (t)

[0028]

[0029] q4=(I y -I z )ω y (t)ω z (t) / I x

[0030] q5=(I z -I x )ω x (t)ω z (t) / I y ,

[0031] q6=(I x -I y )ω y (t)ω x (t) / I z

[0032]

[0033] Without loss of generality, assuming the sensor fault is an additive signal, the satellite attitude control system with additive faults can be further modeled as follows:

[0034]

[0035] Among them, D f Given a matrix of appropriate dimension;

[0036] Let the sampling period be T. s The following nonlinear discretized fault model for the satellite attitude control system can be obtained:

[0037]

[0038] Where Φ(x(k))=x(k)+T s q(x(k)) is the nonlinear function of the system; B u B d C,D v All are known matrices; d(k)∈R n v(k)∈R m All are Gaussian white noise with the following distribution characteristics; R n Represents an n-dimensional Euclidean space; R m Describes m-dimensional Euclidean space;

[0039] E{d(i)d T (j)}=σ ij R

[0040] E{v(i)v T (j)}=σij Q

[0041] E{d(i)v T (j)}=0

[0042] Where R > 0, Q > 0 are weighting matrices, and when i = j, σ ij =1, otherwise σ ij =0;

[0043] Assume that the initial states x(k0) and d(k), v(k) are independent and have the following statistical properties:

[0044]

[0045] Where E{x(k0)} is the mathematical expectation of x(k0), and P(k0|k0) is the covariance matrix of the initial state.

[0046] Preferably, in step 2, an event-triggered mechanism is used to check whether the current measurement output meets the following event conditions:

[0047] ξ(k)=e y T (k)Ωe y (k)-δy T (k)Ωy(k)≥0,k≥k i

[0048] Among them, e y (k)=y(k)-y(k i ) represents the event triggering error; y(k) i (k) represents the latest event trigger time. i The transmission value, Ω∈R q×q This is a weighted matrix; δ > 0 represents the event trigger threshold.

[0049] Once the event triggering condition shown in the above formula is met, the current measurement value will be transmitted to the fault detection module; otherwise, the data packet will be discarded. Therefore, when the event generator releases the current measurement output y(k)... i When ), the next triggering time is determined by the following formula:

[0050]

[0051] Among them, e y (k i )=y(k i +j)-y(k i );τ M ≥0 represents the maximum event trigger interval;

[0052] Therefore, the input data of the fault detection module Updated by the following formula:

[0053]

[0054] Preferably, in step 3, the fault detection filter is triggered as a residual generator by the following event:

[0055]

[0056] in, This is the state estimate. For a one-step prediction, r(k) i (k) represents the event trigger time. i The generated residual signal; K(k i ) represents the gain matrix of the fault detection filter to be designed.

[0057] Preferably, in step 4, the specific content is as follows:

[0058] In k = k i time:

[0059]

[0060] For k∈[k i ,k i+1 )time:

[0061]

[0062] definition For the state estimation error, we get:

[0063]

[0064] The nonlinear part of the system Φ(x(k) i ))exist Performing a Taylor expansion at this point and ignoring higher-order terms, we get:

[0065]

[0066] in,

[0067]

[0068] F(k i ) is the Jacobian matrix;

[0069] Therefore, we can conclude that:

[0070] e(k i +1|k i )=F(k i )e(k i |k i )+Bd d(k i )

[0071] Where, e(k) i +1|k i ) represents the one-step prediction error of the state;

[0072] Based on the above formula, the event trigger time k is obtained. i+1 One-step forecast error:

[0073]

[0074] in,

[0075] A(k i+1 |k i )=F(k i+1 -1)F(k i+1 -2)...F(k i ),F(k i ,k i ) = I

[0076] s (k i )=[s T (k i ) s T (k i +1) ... s T (k i+1 -1)] T

[0077]

[0078] Where s represents d and v;

[0079] From the definition of state estimation error, we can obtain:

[0080]

[0081] in,

[0082]

[0083] From the definition of residual signal:

[0084]

[0085] This achieves complete decoupling between residual signals and event triggering errors.

[0086] Preferably, in step 5, the specific content is as follows:

[0087] Without considering system failures:

[0088]

[0089] By defining covariance, we obtain the state one-step prediction error covariance matrix P(k) i+1 |k i ) and the estimated error covariance matrix P(k i+1 |k i+1 The following are the details:

[0090]

[0091] in, and These are the weighted matrices;

[0092] By analyzing P(k) i+1 |k i+1 Taking the partial derivatives, we obtain the gain matrix of the fault detection filter:

[0093]

[0094] Setting the right side of the above equation to zero, we get:

[0095]

[0096] And because

[0097]

[0098] Therefore, we can conclude that:

[0099] P(k i+1 |k i+1 )=(IK(k i )C)P(k i+1 |k i )

[0100] Further applying the design concept of strong tracking filters, a suboptimal fading factor λ(k) is introduced. i+1 Real-time adjustment of the state estimation error covariance matrix P(k) i+1 |k i+1 ) and its corresponding gain matrix K(k) i+1 Based on this, the event-triggered fault detection filter can not only minimize the state estimation error covariance matrix, but also maintain the ability to track abrupt changes when the system reaches steady state; therefore, the state estimation error covariance matrix P(k i+1 |k i The update of ) can be given by the following formula:

[0101]

[0102] Where, λ(k) i+1It can be approximated by the following formula:

[0103]

[0104] Where β≥1 is a selected weakening factor. The purpose of introducing this weakening factor is to make the state estimate smoother. The covariance matrix of the residuals is V0(k i+1 The actual value of ) is in λ(k) i+1 The forgetting factor ρ∈(0,1) is unknown in the iterative solution of ).

[0105] Preferably, in step 6, the specific details are as follows:

[0106] Considering the system operating normally without faults, under the assumption of Gaussian white noise, the generated residual signal r(k) i They are mutually orthogonal and follow the x-axis. 2 Distribution; the finite-time residual evaluation function is selected as:

[0107]

[0108] Where N represents the length of the moving time window;

[0109] For the satellite attitude control system model, it is easy to see that under fault-free conditions, the residual evaluation function J r (k i ) obeys χ 2 The distribution is such that, after a failure in a certain part of the system, the residual sequence will no longer be zero-mean Gaussian white noise, and the residual evaluation function J... r (k i It also no longer obeys χ. 2 distributed;

[0110] Therefore, by selecting an appropriate false alarm rate ε for fault detection, a corresponding fault alarm threshold is given:

[0111]

[0112] Then, the following decision logic is applied to detect the occurrence of the fault:

[0113]

[0114] Among them, J r (k i ) is k i The residual evaluation function corresponding to time step J th This is the fault alarm threshold.

[0115] The beneficial technical effects of this invention are as follows:

[0116] This invention utilizes the STF method to design a fault detection filter, which, compared to the traditional extended Kalman filter method, enables the system to maintain its ability to track abrupt changes when it reaches a steady state.

[0117] This invention adopts a novel event-triggered framework, which completely decouples the residual signal from the event-triggered error, effectively improving resource utilization while ensuring system performance. Attached Figure Description

[0118] Figure 1 This is a flowchart of the satellite attitude control system event-triggered STF fault detection method of the present invention;

[0119] Figure 2 A graph showing the residual evaluation function, fault alarm threshold, and corresponding event trigger interval for a star sensor fault occurring on the X-axis of the satellite attitude control system. Figure 2 (a) in the figure shows the residual evaluation function and its fault alarm threshold when the star sensor fails on the X-axis. Figure 2 (b) in the diagram is the corresponding event trigger interval diagram;

[0120] Figure 3 A graph showing the residual evaluation function, fault alarm threshold, and corresponding event trigger interval for a gyroscope fault occurring on the X-axis of a satellite attitude control system. Figure 3 (a) in the figure represents the residual evaluation function and fault alarm threshold diagram when the system experiences a gyroscope failure on the X-axis. Figure 3 (b) in the diagram is the corresponding event trigger interval diagram; Detailed Implementation

[0121] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0122] like Figure 1 As shown, a method for detecting event-triggered STF faults in a satellite attitude control system includes the following steps:

[0123] Step 1: Establish a fault model for the satellite attitude control system; the specific content is as follows:

[0124] The dynamic equations of the satellite attitude control system:

[0125]

[0126] Among them, I i (i = x, y, z) represents the satellite's moments of inertia on its three principal axes; ω i (i = x, y, z) represents the projection of the attitude velocity vector onto the three principal axes of inertia; T i (i = x, y, z) represents the control torque components along the principal axis of inertia; This represents the time derivative of the moment of inertia. Represents the angular acceleration components;

[0127] The transformation relationship between the satellite's body coordinate system and its orbital coordinate system is defined as the satellite's attitude. When the attitude angle is small, the dynamic equations are rewritten as follows:

[0128]

[0129] in, θ and ψ are the roll angle, pitch angle, and yaw angle, respectively; ω0 is the constant orbital speed.

[0130] Define state vector and control input vector u(t)=[T x (t) T y (t) T z (t)] T Without considering faults, the dynamics of the satellite attitude control system are described by the following nonlinear model:

[0131]

[0132] Where y(t) is the measurement output vector, d(t) is the process noise, and v(t) is the measurement noise.

[0133] q(x(t)) = [q1 q2 q3 q4 q5 q6] T

[0134] q1=ω0ψ(t)+ω x (t),

[0135] q2=ω0+ω y (t)

[0136]

[0137] q4=(I y -I z )ω y (t)ω z (t) / I x

[0138] q5=(I z -I x )ω x (t)ω z (t) / I y ,

[0139] q6=(I x -I y )ω y (t)ωx (t) / I z

[0140]

[0141] Without loss of generality, assuming the sensor fault is an additive signal, the satellite attitude control system with additive faults can be further modeled as follows:

[0142]

[0143] Among them, D f Given a known matrix with appropriate dimensions.

[0144] Let the sampling period be T. s The following nonlinear discretized fault model for the satellite attitude control system can be obtained:

[0145]

[0146] Where Φ(x(k))=x(k)+T s q(x(k)) is the nonlinear function of the system; B u B d C,D v All are known matrices; d(k)∈R n v(k)∈R m ( express All (in 12-dimensional Euclidean space) are Gaussian white noise with the following distribution characteristics:

[0147] E{d(i)d T (j)}=σ ij R

[0148] E{v(i)v T (j)}=σ ij Q

[0149] E{d(i)v T (j)}=0

[0150] Where R > 0, Q > 0 are weighting matrices, and when i = j, σ ij =1, otherwise σ ij =0;

[0151] Assume that the initial states x(k0) and d(k), v(k) are independent and have the following statistical properties:

[0152]

[0153] Step 2: Introduce an event triggering mechanism;

[0154] An event-triggered mechanism is used to verify whether the current measurement output meets the following event conditions:

[0155] ξ(k)=e y T (k)Ωe y (k)-δy T (k)Ωy(k)≥0,k≥k i

[0156] Among them, e y (k)=y(k)-y(k i ) represents the event triggering error; y(k) i (k) represents the latest event trigger time. i The transmission value, Ω∈R q×q This is a weighted matrix; δ > 0 represents the event trigger threshold.

[0157] Once the event triggering condition shown in the above formula is met, the current measurement value will be transmitted to the fault detection module; otherwise, the data packet will be discarded. Therefore, when the event generator releases the current measurement output y(k)... i When ), the next triggering time is determined by the following formula:

[0158]

[0159] Among them, e y (k i )=y(k i +j)-y(k i );τ M ≥0 represents the maximum event trigger interval;

[0160] Therefore, the input data of the fault detection module Updated by the following formula:

[0161]

[0162] Where E{x(k0)} is the mathematical expectation of x(k0), and P(k0|k0) is the covariance matrix of the initial state.

[0163] Step 3: Design an event-triggered fault detection filter using the STF method;

[0164] The following events trigger the fault detection filter as a residual generator:

[0165]

[0166] in, This is the state estimate. For a one-step prediction, K(k) i ) represents the gain matrix of the fault detection filter to be designed.

[0167] Step 4: By linearizing the fault detection filter, the dynamic equation of the estimation error and the residual signal are obtained, thus achieving complete decoupling between the residual signal and the event triggering error; the specific details are as follows:

[0168] In k = k i time:

[0169]

[0170] For k∈[k i ,k i+1 )time:

[0171]

[0172] definition For the state estimation error, we get:

[0173]

[0174] The nonlinear part of the system Φ(x(k) i ))exist Performing a Taylor expansion at this point and ignoring higher-order terms, we get:

[0175]

[0176] in,

[0177]

[0178] It is a Jacobian matrix.

[0179] Therefore, we can conclude that:

[0180] e(k i +1|k i )=F(k i )e(k i |k i )+B d d(k i )

[0181] Where, e(k) i +1|k i ) represents the one-step prediction error of the state.

[0182] Based on the above formula, the event trigger time k is obtained. i+1 One-step forecast error:

[0183]

[0184] in,

[0185] A(ki+1 |k i )=F(k i+1 -1)F(k i+1 -2)...F(k i ),F(k i ,k i ) = I

[0186] s (k i )=[s T (k i ) s T (k i +1) ... s T (k i+1 -1)] T

[0187]

[0188] Where s represents d and v;

[0189]

[0190] From the definition of state estimation error, we can obtain:

[0191]

[0192] in,

[0193]

[0194] From the definition of residual signal:

[0195]

[0196] This achieves complete decoupling between residual signals and event triggering errors.

[0197] Step 5: Transform the design problem of the residual generator into solving the gain matrix of the fault detection filter; the specific steps are as follows:

[0198] Without considering system failures:

[0199]

[0200] By defining covariance, we obtain the state one-step prediction error covariance matrix P(k) i+1 |k i ) and the estimated error covariance matrix P(k i+1 |k i+1 The following are the details:

[0201]

[0202] in, and These are weighted matrices, respectively.

[0203] By analyzing P(k) i+1 |k i+1 Taking the partial derivatives, we obtain the gain matrix of the fault detection filter:

[0204]

[0205] Setting the right side of the above equation to zero, we get:

[0206]

[0207] And because

[0208]

[0209] Therefore, we can conclude that:

[0210] P(k i+1 |k i+1 )=(IK(k i )C)P(k i+1 |k i )

[0211] Further applying the design concept of strong tracking filters, a suboptimal fading factor λ(k) is introduced. i+1 Real-time adjustment of the state estimation error covariance matrix P(k) i+1 |k i+1 ) and its corresponding gain matrix K(k) i+1 Based on this, the event-triggered fault detection filter can not only minimize the state estimation error covariance matrix, but also maintain the ability to track abrupt changes when the system reaches steady state; therefore, the state estimation error covariance matrix P(k i+1 |k i The update of ) can be given by the following formula:

[0212]

[0213] Where, λ(k) i+1 It can be approximated by the following formula:

[0214]

[0215] Where β≥1 is a selected weakening factor. The purpose of introducing this weakening factor is to make the state estimate smoother. The covariance matrix of the residuals is V0(k i+1 The actual value of ) is in λ(k) i+1 The forgetting factor ρ∈(0,1) is unknown in the iterative solution of ).

[0216] Step 6: Calculate the residual evaluation function and compare it with the threshold to determine whether a fault has occurred;

[0217] The details are as follows:

[0218] Considering the system operating normally without faults, under the assumption of Gaussian white noise, the generated residual signal r(k) i They are mutually orthogonal and follow the x-axis. 2 Distribution; the finite-time residual evaluation function is selected as:

[0219]

[0220] Where N represents the length of the moving time window;

[0221] For the satellite attitude control system model, it is easy to see that under fault-free conditions, the residual evaluation function J r (k i ) obeys χ 2 The distribution is such that, after a failure in a certain part of the system, the residual sequence will no longer be zero-mean Gaussian white noise, and the residual evaluation function J... r (k i It also no longer obeys χ. 2 distributed;

[0222] Therefore, by selecting an appropriate false alarm rate ε for fault detection, a corresponding fault alarm threshold is given:

[0223]

[0224] Then, the following decision logic is applied to detect the occurrence of the fault:

[0225]

[0226] Among them, J r (k i ) is k i The residual evaluation function corresponding to time step J th This is the fault alarm threshold.

[0227] Depend on Figure 2 As can be seen from (a) and 2(b), the present invention can detect the star sensor fault that occurs on the X-axis of the system at k=42 steps. As can be seen from the event trigger interval diagram, compared with the traditional time triggering mechanism, the present invention reduces the amount of data transmission by 20%, and effectively improves the resource utilization rate while ensuring system performance.

[0228] Depend on Figure 3 As can be seen from (a) and 3(b), the present invention can also detect gyroscope malfunctions that occur in the system in a timely manner and save 16% of communication resources, effectively improving resource utilization.

[0229] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for detecting event-triggered STF faults in a satellite attitude control system, characterized in that: Includes the following steps: Step 1: Establish a fault model for the satellite attitude control system; Step 2: Introduce an event triggering mechanism; Step 3: Design an event-triggered fault detection filter using the STF method; Step 4: By linearizing the fault detection filter, the dynamic equation of the estimation error and the residual signal are obtained, thus achieving complete decoupling between the residual signal and the event triggering error; Step 5: Transform the design problem of the residual generator into solving the gain matrix of the fault detection filter; Step 6: Calculate the residual evaluation function and compare it with the threshold to determine whether a fault has occurred; In step 1, the specific content is as follows: The dynamic equations of the satellite attitude control system: ； in, Let be the moment of inertia of the satellite on its three principal axes; This is the projection of the attitude rate vector onto the three principal axes of inertia; This represents the control torque component along the principal axis of inertia. Represents the angular acceleration components; The transformation relationship between the satellite's body coordinate system and its orbital coordinate system is defined as the satellite's attitude. When the attitude angle is small, the dynamic equations are rewritten as follows: ； in, These are the roll angle, pitch angle, and yaw angle, respectively. The orbital speed is constant. Define state vector and control input vector Without considering faults, the dynamics of the satellite attitude control system are described by the following nonlinear model: ； in, To measure the output vector, For process noise, To measure noise, ； Without loss of generality, assuming the sensor fault is an additive signal, the satellite attitude control system with additive faults can be further modeled as follows: ； in, Given a matrix of appropriate dimension; Let the sampling period be The following nonlinear discretized fault model for the satellite attitude control system can be obtained: ； in, The system is a nonlinear function; All are known matrices; , All are Gaussian white noise with the following distribution characteristics; Represents an n-dimensional Euclidean space; Describes m-dimensional Euclidean space; ； in, For a weighted matrix, when hour, ,otherwise ; Assuming the initial state and They are independent of each other and have the following statistical properties: ； in, for The mathematical expectation, Let be the covariance matrix of the initial state.

2. The satellite attitude control system event-triggered STF fault detection method according to claim 1, characterized in that: In step 2, an event-triggered mechanism is used to verify whether the current measurement output meets the following event conditions: , ； in, This is for event triggering error; The latest event trigger time The transmitted value, It is a weighted matrix; This is the event trigger threshold; Once the event triggering conditions shown in the above formula are met, the current measurement value will be transmitted to the fault detection module; otherwise, the data packet will be discarded. Therefore, when the event generator releases the current measurement output... When the next triggering time is determined by the following formula: ； in, ; This represents the maximum event trigger interval. Therefore, the input data of the fault detection module Updated by the following formula: , 。 3. The satellite attitude control system event-triggered STF fault detection method according to claim 2, characterized in that: In step 3, the fault detection filter is triggered as a residual generator by the following event: ； in, This is the state estimate. This is a one-step prediction value. The moment the event is triggered The generated residual signal; This is the gain matrix of the fault detection filter to be designed.

4. The satellite attitude control system event-triggered STF fault detection method according to claim 3, characterized in that: In step 4, the specific details are as follows: exist time: ； for time: ； definition For the state estimation error, we get: ； The nonlinear part of the system exist Performing a Taylor expansion at this point and ignoring higher-order terms, we get: ； in, ； ； It is a Jacobian matrix; Therefore, we can conclude that: ； in, This represents the one-step prediction error for the state; Based on the above formula, the event trigger time can be obtained. One-step forecast error: ； in, ； in, represent ; From the definition of state estimation error, we can obtain: ； in, ； From the definition of residual signal: ； This achieves complete decoupling between residual signals and event triggering errors.

5. The satellite attitude control system event-triggered STF fault detection method according to claim 4, characterized in that: In step 5, the specific details are as follows: Without considering system failures: ； By defining covariance, we obtain the state one-step prediction error covariance matrix. and the estimated error covariance matrix They are as follows: ； in, These are the weighted matrices; Through the Taking the partial derivatives, we obtain the gain matrix of the fault detection filter: ； Setting the right side of the above equation to zero, we get: ； And because ； Therefore, we can conclude that: ； Further applying the design concept of strong tracking filters, a suboptimal fading factor is introduced. Real-time adjustment of state estimation error covariance matrix and its corresponding gain matrix Based on this, the event-triggered fault detection filter can not only minimize the state estimation error covariance matrix, but also maintain the ability to track abrupt changes when the system reaches steady state; therefore, the state estimation error covariance matrix The update can be given by the following formula: ； in, It can be approximated by the following formula: ； ； ； ； in, A selected weakening factor is introduced to make the state estimates smoother, and the covariance matrix of the residuals is... The actual value is The forgetting factor is unknown in the iterative solution and is estimated from the above formula; .

6. The satellite attitude control system event-triggered STF fault detection method according to claim 5, characterized in that: In step 6, the specific details are as follows: Considering the system operating normally without faults, under the assumption of Gaussian white noise, the generated residual signal... Mutually orthogonal and obedient Distribution; the finite-time residual evaluation function is selected as: ； in, Indicates the length of the moving time window; For the satellite attitude control system model, it is easy to see that, under fault-free conditions, the residual evaluation function... obey The residual sequence will no longer be zero-mean Gaussian white noise after a failure in a certain part of the system, and the residual evaluation function will change. They will no longer obey. distributed; Therefore, by selecting an appropriate fault detection false alarm rate Provide the corresponding fault alarm thresholds: ； Then, the following decision logic is applied to detect the occurrence of the fault: ； in, for The residual evaluation function corresponding to time step 1. This is the fault alarm threshold.

Images