Method for predicting reliability of reaction wheel in degradation process

Abstract

The invention provides a reaction wheel reliability prediction method in a degradation process, and the method comprises the steps: collecting the historical degradation data of the bearing temperature of a satellite reaction wheel, and carrying out the preprocessing of the data; establishing a reaction wheel temperature performance degradation model by using a Wiener process and performing parameter estimation; establishing a reaction wheel reliability model and calculating the reliability of the reaction wheel; a Kalman filtering algorithm and the reaction wheel temperature performance degradation model are combined to establish a prediction model of the reaction wheel; and predicting the reliability of the reaction wheel by using the prediction model. According to the invention, a more accurate and reliable fault prediction result can be obtained, so that the purpose of fault prediction of the satellite attitude control system is achieved, and improvement of reliability and performance of the satellite attitude control system is promoted; according to the method, the random item is added in the Kalman filtering algorithm, the advantages of each model and algorithm are fully played, the accuracy and stability of overall prediction are improved, and the problem that the real-time performance is poor in the method for predicting the reliability by using the random degradation process is solved.

Description

Technical Field

[0001] The invention relates to the technical field of spacecraft fault prediction, and in particular to a method for predicting the reliability of a reaction wheel in a degradation process. Background Art

[0002] With the rapid development of aerospace, attitude control systems play a vital role in satellite systems, and reaction wheels are indispensable key actuators. As a key actuator of the attitude control system, the reaction wheel undertakes the important tasks of moment of inertia adjustment and angular momentum control. However, because the attitude control system needs to work for a long time, it has become one of the systems in the satellite that is prone to failure, and the degradation failure of the reaction wheel is also an important part of it.

[0003] The existing technology mainly predicts the reliability of the product by measuring the degradation data of a certain performance parameter. For example, by establishing a temperature degradation model of the momentum wheel bearing, using the historical temperature data of the same type of momentum wheel bearing, the parameters in the degradation model are estimated, and the sample path of the bearing temperature random process is simulated according to the modeling. Finally, the reliability of the momentum wheel is predicted. The results show that this method works well in the reliability assessment of the momentum wheel. However, this method relies on historical telemetry data and has poor real-time performance. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for predicting the reliability of a reaction wheel under a degradation process. The present invention can provide more accurate and reliable fault prediction results, thereby achieving the purpose of fault prediction of a satellite attitude control system and promoting the reliability and performance improvement of the satellite attitude control system.

[0005] The technical solution of the present invention is: a method for predicting the reliability of a reaction wheel under a degradation process, comprising the following steps:

[0006] S1), collecting the historical degradation data of the satellite reaction wheel bearing temperature and preprocessing the data;

[0007] S2), using the Wiener process to establish a reaction wheel temperature performance degradation model and perform parameter estimation;

[0008] S3), establishing a reaction wheel reliability model and calculating the reaction wheel reliability;

[0009] S4), establishing a prediction model of the reaction wheel by combining the Kalman filter algorithm and the reaction wheel temperature performance degradation model;

[0010] S5), using the prediction model to predict the reliability of the reaction wheel.

[0011] Preferably, in step S2), the degradation trend model of the reaction wheel bearing temperature is modeled using a Wiener process with drift to obtain a reaction wheel temperature performance degradation model:

[0012] Y(t)=a0+a1t+σ B B(t); (1)

[0013] Where Y(t) represents the temperature degradation of the reaction wheel at time t; a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, σ B is the diffusion coefficient, and B(t) is the standard Wiener process.

[0014] Preferably, in step S2), under different working conditions, the bearing temperature telemetry drift parameter a1 and the diffusion coefficient σ in the reaction wheel temperature performance degradation model are B The bearing temperature telemetry data under working load is combined with the maximum likelihood estimation method to estimate the unknown parameters a1 and σ in the degradation model. B Make an estimate.

[0015] Preferably, in step S3), when the degradation trend Y(t) of the reaction wheel bearing temperature reaches a certain level, namely, the failure threshold Xm, the reaction wheel fails. Therefore, the probability density function f(t|a0,a1,Xm,Y(t)) of the reaction wheel is obtained as follows:

[0016]

[0017] Where a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, σ B is the diffusion coefficient.

[0018] Preferably, in step S3), the reliability model of the reaction wheel at time t is:

[0019]

[0020] Where P(X(t)) is the probability distribution of the temperature degradation amount X(t) at time t, and f(t|Xm) is the probability density function of the temperature degradation amount of the reaction wheel when the degradation amount threshold is Xm.

[0021] Preferably, in step S4), a prediction model of the reaction wheel is established by combining the Kalman filter algorithm and the reaction wheel performance degradation model, which specifically includes the following steps:

[0022] S41), establish the state of the estimated parameters and the observed variables and the discrete time linear state description equation:

[0023] u(k)=A(k)U(k)+E(k);

[0024] Z(k)=H(k)U(k)+W(k);

[0025] Where U(k) is the state of the estimated parameters at time k, A(k) is the state transfer matrix, E(k) and W(k) are zero mean and positive definite covariance matrices, E(k) and W(k) obey the normal distribution of N(0,q) and N(0,r) respectively, and E(k) and W(k) are independent of each other; Z(k) is the observation vector at time k, which is the mean of the degradation amount of the reaction wheel within Δt time, and H(k) is the measurement matrix at time k;

[0026] S42), establish the prediction equation and covariance of state estimation:

[0027] U(k+1|k)=A(k)U(k|k)+δ(k)+E(k);

[0028] G(k+1|k)=A(k)G(k|k)+E(k);

[0029] δ(k)=a1+σ B (B(t+1)-B(t));

[0030] In the formula, U(k+1|k) is the state estimation value, which represents the prior estimation of the degradation amount at time k+1; G(k|k) and G(k+1|k) are the covariance matrix at time k and the prior estimation of the covariance matrix at time k+1 predicted based on the state estimation at time k and the state transfer matrix; δ(k) is a random term;

[0031] S42), combine the residual sequence and Kalman gain to establish the state update equation and error covariance update equation of the Kalman filter algorithm:

[0032] U(k+1|k+1)=U(k+1|k)+K(k+1)γ(k+1)

[0033] =U(k+1|k)+K(k+1)[Z(k+1)-H(k+1)U(k+1|k)];

[0034] G(k+1|k+1)=[1-K(k+1)H(k+1)]G(k+1|k);

[0035] Where U(k+1|k+1) is the a posteriori estimate of the state estimate at time k+1; G(k+1|k+1) is the a posteriori estimate of the covariance matrix at time k+1; K(k+1) is the Kalman gain at time k+1;

[0036] γ(k+1) is the residual sequence at time k+1 and k; Z(k+1) is the observation value at time k+1; H(k+1) is the measurement matrix at time k+1;

[0037] S43), after the degradation state U(k+1|k) is estimated a priori at time k+1, the reliability at time k+1 can be predicted;

[0038]

[0039] Where a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, σ B is the diffusion coefficient; Xm is the failure threshold; U(k+1|k) is the prior estimate of the state, which represents the prior estimate of the temperature degradation in this Kalman filter.

[0040] Preferably, in step S5), the reliability of the reaction wheel is predicted using a prediction model, which specifically includes the following steps:

[0041] S51), initialization parameters, including failure threshold Xm, initial value a0 of reaction wheel bearing temperature telemetry, drift parameter a1 of reaction wheel bearing temperature telemetry value, diffusion coefficient σ B , the initial state values ​​U(0|0) and G(0|0) are the initial covariance values;

[0042] S52), using the state prediction equation U(k+1|k)=A(k)U(k|k)+δ(k)+E(k) to predict the mean value at the next moment, and obtain the predicted value of the optimal estimate at moment k for moment k+1;

[0043] S53) Utilization Calculate the probability density function at this time, and then use The calculation expression of reliability is used to calculate the predictable reliability P (X(t)<Xm) at the k+1th moment;

[0044] S54), collecting n samples within the k+1th Δt moment, and calculating the moment estimate as the observation value Z(k+1) at the k+1th moment;

[0045] S55), prediction covariance G(k+1|k)=A(k)G(k|k)+E(k);

[0046] S56), update the gain K(k+1), the optimal state estimate U(k+1|k+1) and the covariance G(k+1|k+1);

[0047] S56), k=k+1, return to step S52).

[0048] The beneficial effects of the present invention are:

[0049] 1. The present invention can provide more accurate and reliable fault prediction results, thereby achieving the purpose of fault prediction of satellite attitude control system and promoting the reliability and performance improvement of satellite attitude control system;

[0050] 2. The present invention uses the Wiener process to establish a reaction wheel temperature performance degradation model, and then uses maximum likelihood estimation to estimate unknown parameters in the established model;

[0051] 3. The present invention adds a random term δ(k) containing a diffusion coefficient and a drift coefficient to the prediction equation of state estimation in the Kalman filter algorithm, giving full play to the advantages of each model and algorithm, improving the accuracy and stability of the overall prediction, and improving the problem of poor real-time performance in the reliability prediction method using random degradation processes. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the process of the present invention;

[0053] Figure 2 It is a schematic diagram of the process of predicting the reliability of the reaction wheel of the present invention. DETAILED DESCRIPTION

[0054] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings:

[0055] like Figure 1 As shown, this embodiment provides a method for predicting the reliability of a reaction wheel in a degradation process, comprising the following steps:

[0056] S1), collecting historical degradation data of the reaction wheel bearing temperature of multiple different models of satellites; and preprocessing the historical degradation data; in this embodiment, considering that the temperature degradation of the reaction wheel is a slow process, the preprocessing includes taking the average of the degradation amount of a reaction wheel for which historical data is collected within one day, simplifying the degradation amount data within this day into single data, and performing parameter estimation after simplifying these data.

[0057] S2) Use the Wiener process to establish a reaction wheel temperature performance degradation model and perform parameter estimation; the details are as follows:

[0058] This embodiment uses the Wiener process with drift to model the degradation trend of the reaction wheel bearing temperature, and obtains the reaction wheel temperature performance degradation model:

[0059] Y(t)=a0+a1t+σ B B(t); (1)

[0060] Where Y(t) represents the temperature degradation of the reaction wheel at time t; a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, σ B is the diffusion coefficient, and B(t) is the standard Wiener process.

[0061] In addition, in this embodiment, under different working conditions, the bearing temperature telemetry drift parameter a1 and the diffusion coefficient σ in the reaction wheel temperature performance degradation model are B The bearing temperature telemetry data under working load is combined with the maximum likelihood estimation method to estimate the unknown parameters a1 and σ in the degradation model. B The estimation process includes the following steps:

[0062] S21), collect n+1 bearing temperature data (t0, y0) in sequence at a certain time interval,

[0063] (t1,y1), ..., (t n ,y n ), where t i represents the sampling time, y i Indicates the corresponding time t i The sampled bearing temperature value, 0≤i≤n;

[0064] S22) Calculate the bearing temperature difference Δy according to the following formula i ,Right now:

[0065] Δy i =a1Δt i +σ B ΔB(t i ),i=1,2,3,...,n;(2)

[0066] Where a1 is the bearing temperature remote sensing drift parameter; σ B is the diffusion coefficient; Δt i =t i -t i-1 is the time difference; Δy i =y i -y i-1 is the bearing temperature difference; ΔB(t i )=B(t i )-B(t i-1 ) is the standard Wiener process difference, which conforms to the normal distribution;

[0067] S23), from the properties of the standard Wiener process, we know that ΔB(t i ):N(0,Δt i ),therefore,

[0068] And according to the property of the Wiener process with stable independent increments, the joint probability density function of the temperature increment is obtained as follows:

[0069] f(Δy1,Δy2,Δy3,...Δy n )=f(Δy1)f(Δy2)f(Δy3)...f(Δy n ); (3)

[0070] The resulting likelihood function is:

[0071] L(a1,σ B )=f(Δy1)f(Δy2)f(Δy3)...f(Δy n ); (4)

[0072] Among them, the bearing temperature difference Δy i The probability density function of is expressed as:

[0073]

[0074] S24), the likelihood function L(a1,σ B ) and f(Δy i ) Take the logarithm and get:

[0075] lnL(a1,σ B )=lnf(Δy1)f(Δy2)f(Δy3)...f(Δy n )=lnf(Δy1)+lnf(Δy2)+....+lnf(Δy n ); (6)

[0077]

[0078] S25) and lnf(Δy i ) into lnL(a1,σ B )get:

[0079]

[0080] S26), for a1 and σ in equation (8) B Taking partial derivatives we get:

[0081]

[0082] S27), let (9) and (10) equal to 0, so that the parameter estimates are:

[0083]

[0084] In the formula, is the estimated value of the bearing temperature telemetry drift parameter a1; is the diffusion coefficient σ B The estimated value of .

[0085] S3) Establish a reaction wheel reliability model and calculate the reaction wheel reliability; the details are as follows:

[0086] In this embodiment, when the degradation trend Y(t) of the reaction wheel bearing temperature reaches a certain level, namely, the failure threshold Xm, the reaction wheel fails. Therefore, the probability density function f(t|a0,a1,Xm,Y(t)) of the reaction wheel is obtained as follows:

[0087]

[0088] Where a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, σ B is the diffusion coefficient;

[0089] Therefore, the reliability model of the reaction wheel at time t is:

[0090]

[0091] Where P(X(t)) is the probability distribution of the temperature degradation amount X(t) at time t, and f(t|Xm) is the probability density function of the temperature degradation amount of the reaction wheel when the degradation amount threshold is Xm.

[0092] S4), combining the Kalman filter algorithm and the reaction wheel temperature performance degradation model to establish a reaction wheel prediction model; specifically comprising the following steps:

[0093] S41), establish the state of the estimated parameters and the observed variables and the discrete time linear state description equation:

[0094] U(k)=A(k)U(k)+E(k);

[0095] Z(k)=H(k)U(k)+W(k);

[0096] Where U(k) is the state of the estimated parameters at time k, A(k) is the state transfer matrix, E(k) and W(k) are zero-mean and positive-definite covariance matrices, E(k) and W(k) obey the normal distribution of N(0, q) and N(0, s), respectively, that is, the Gaussian distribution observation noise and measurement noise of the normally distributed errors Q(k) and S(k), respectively; E(k) and W(k) are independent of each other; Z(k) is the observation vector at time k, which is the mean of the degradation amount of the reaction wheel within Δt time, and H(k) is the measurement matrix at time k;

[0097] S42), establish the prediction equation and covariance of state estimation:

[0098] U(k+1|k)=A(k)U(k|k)+δ(k)+E(k);

[0099] G(k+1|k)=A(k)G(k|k)+E(k);

[0100] δ(k)=a1+σ B (B(t+1)-B(t));

[0101] In the formula, U(k+1|k) is the state estimation value, which represents the prior estimation of the degradation amount at time k+1; G(k|k) and G(k+1|k) are the covariance matrix at time k and the prior estimation of the covariance matrix at time k+1 predicted based on the state estimation at time k and the state transfer matrix; δ(k) is a random term;

[0102] S42), combine the residual sequence and Kalman gain to establish the state update equation and error covariance update equation of the Kalman filter algorithm:

[0103] U(k+1|k+1)=U(k+1|k)+K(k+1)γ(k+1)

[0104] =U(k+1|k)+K(k+1)[Z(k+1)-H(k+1)U(k+1|k)];

[0105] G(k+1|k+1)=[1-K(k+1)H(k+1)]G(k+1|k);

[0106] Where U(k+1|k+1) is the a posteriori estimate of the state estimate at time k+1; G(k+1|k+1) is the a posteriori estimate of the covariance matrix at time k+1;

[0107] K(k+1) is the Kalman gain at time k+1; γ(k+1) is the residual sequence between time k+1 and time k; Z(k+1) is the observation value at time k+1; H(k+1) is the measurement matrix at time k+1;

[0108] The expression of Kalman gain is:

[0109] K(k+1)=P(k+1|k)H T (k+1)+S -1 (k+1);

[0110] Among them, S(k+1) is the prediction error covariance;

[0111] S(k+1)=H(k+1)G(k+1|k)H T (k+1)+R(k+1);

[0112] The residual sequence is:

[0113] γ(k+1)=Z(k+1|k+1)-Z(k+1|k)=Z(k+1|k+1)-H(k+1|k)U(k+1|k);

[0114] Where Z(k+1|k+1) is the actual observation value at time k+1, Z(k+1|k) is the predicted observation value at k+1, H(k+1|k) is the measurement matrix, U(k+1|k) is the prior estimate of the state at time k+1; R(k+1) is the error at time k+1;

[0115] S43), after the degradation state U(k+1|k) is estimated a priori at time k+1, the reliability at time k+1 can be predicted;

[0116]

[0117] Where a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, σ B is the diffusion coefficient; xm is the failure threshold; U(k+1|k) is the prior estimate of the state, which represents the prior estimate of the temperature degradation in this Kalman filter.

[0118] S5), using the prediction model to predict the reliability of the reaction wheel, such as Figure 2 As shown, the specific steps include:

[0119] S51), initialization parameters, including failure threshold Xm, initial value a0 of reaction wheel bearing temperature telemetry, drift parameter a1 of reaction wheel bearing temperature telemetry value, diffusion coefficient σ B , the initial state values ​​U(0|0) and G(0|0) are the initial covariance values;

[0120] S52), using the state prediction equation U(k+1|k)=A(k)U(k|k)+δ(k)+E(k) to predict the mean value at the next moment, and obtain the predicted value of the optimal estimate at moment k for moment k+1;

[0121] S53) Utilization The probability density function of The calculation expression of reliability is used to calculate the predictable reliability P (X(t)<Xm) at the k+1th moment;

[0122] S54), collecting n samples within the k+1th Δt moment, and calculating the moment estimate as the observation value Z(k+1) at the k+1th moment;

[0123] S55), prediction covariance G(k+1|k)=A(k)G(k|k)+E(k);

[0124] S56), update the gain K(k+1), the optimal state estimate U(k+1|k+1) and the covariance G(k+1|k+1);

[0125] S56), k=k+1, return to step S52).

[0126] The above embodiments and descriptions are only for illustrating the principles and best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, all of which fall within the scope of the present invention to be protected.

Claims

1. A method for predicting the reliability of a reaction wheel under degradation process, characterized in that: The following steps are involved: S1), collecting satellite reaction wheel bearing temperature historical degradation data and preprocessing the data; S2), using the Wiener process to establish a reaction wheel temperature performance degradation model and perform parameter estimation; S3), establishing a reaction wheel reliability model and calculating the reaction wheel reliability; S4), establishing a prediction model for the reaction wheel by combining the Kalman filter algorithm and the reaction wheel temperature performance degradation model; S5) using the prediction model to predict the reliability of the reaction wheel.

2. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 1, characterized in that: In step S2), the degradation trend model of the reaction wheel bearing temperature is modeled using a Wiener process with drift to obtain a reaction wheel temperature performance degradation model: Y(t)=a0+a1t+σ B B(t); (1) Where Y(t) represents the temperature degradation of the reaction wheel at time t; a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry value, and σ B is the diffusion coefficient, and B(t) is the standard Wiener process.

3. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 2, characterized in that: Step S2) Under different working conditions, the bearing temperature telemetry drift parameter a1 and the diffusion coefficient σ in the reaction wheel temperature performance degradation model are compared. B They are all different. The bearing temperature telemetry data under working load is combined with the maximum likelihood estimation method to estimate the unknown parameters a1 and σ in the degradation model. B Make an estimate.

4. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 3, wherein: Step S2), unknown parameters a1 and σ B The estimated values ​​are: Where, is the estimated value of the bearing temperature telemetry drift parameter a1; is the diffusion coefficient σ B Estimated value of Δt i =t i -t i-1 is the time difference; Δy i =y i -y i-1 is the bearing temperature difference; n is the number of bearing temperature samples collected, t i represents the sampling time, y i Indicates the corresponding time t i The sampled bearing temperature value.

5. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 4, characterized in that: Step S2), the bearing temperature difference Δy i Expressed as: Δy i =a1Δt i +s B ΔB(t i ),i=1,2,3,...,n;(2) Where a1 is the bearing temperature remote sensing drift parameter; σ B is the diffusion coefficient; Δt i =t i -t i-1 is the time difference; ΔB(t i )=B(t i )-B(t i-1 ) is the standard Wiener process difference; According to the properties of the standard Wiener process, ΔB(t i ):N(0,Δt i ),therefore, And according to the property of the Wiener process with stable independent increments, the joint probability density function of the temperature increment is obtained as follows: f(Δy1,Δy2,Δy3,...Δy n )=f(Δy1)f(Δy2)f(Δy3)...f(Δy n ); (3) The resulting likelihood function is: L(a1,σ B )=f(Δy1)f(Δy2)f(Δy3)...f(Δy n ); (4) Among them, the bearing temperature difference Δy i The probability density function of is expressed as:

6. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 1, characterized in that: In step S3), when the degradation trend Y(t) of the reaction wheel bearing temperature reaches a certain level, namely, the failure threshold Xm, the reaction wheel fails. Therefore, the probability density function f(ta0, a1, Xm, Y(t)) of the reaction wheel is obtained as follows: Where a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry, σ B is the diffusion coefficient.

7. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 6, characterized in that: In step S3), the reliability model of the reaction wheel at time t is: Where P(X(t)) is the probability distribution of the temperature degradation X(t) at time t, and f(t|Xm) is the probability density function of the reaction wheel temperature degradation when the degradation threshold is Xm.

8. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 1, characterized in that: In step S4), a prediction model for the reaction wheel is established by combining the Kalman filter algorithm and the reaction wheel performance degradation model, which specifically includes the following steps: S41) Establish the state of the estimated parameters and the observed variables and the discrete time linear state description equation: U(k)=A(k)U(k)+E(k); Z(k)=H(k)U(k)+W(k); Where U(k) is the state of the estimated parameters at time k, A(k) is the state transfer matrix, E(k) and W(k) are zero-mean and positive-definite covariance matrices, E(k) and W(k) obey the normal distribution of N(0,q) and N(0,r), respectively, and E(k) and W(k) are independent of each other; Z(k) is the observation vector at time k, which is the mean of the reaction wheel degradation amount within time Δt, and H(k) is the measurement matrix at time k; S42) Establish the prediction equation and covariance of state estimation: U(k+1|k)=A(k)U(k|k)+δ(k)+E(k); G(k+1|k)=A(k)G(k|k)+E(k); δ(k)=a1+σ B (B(t+1)-B(t)); Where U(k+1|k) is the state estimate, which represents the prior estimate of the degradation amount at time k+1; G(k|k) and G(k+1k) are the covariance matrix at time k and the prior estimate of the covariance matrix at time k+1 predicted based on the state estimate and state transfer matrix at time k, respectively; δ(k) is a random term; S42) Combine the residual sequence and Kalman gain to establish the state update equation and error covariance update equation of the Kalman filter algorithm: U(k+1|k+1)=U(k+1|k)+K(k+1)γ(k+1) =U(k+1|k)+K(k+1)[Z(k+1)-H(k+1)U(k+1|k)]; G(k+1|k+1)=[1-K(k+1)H(k+1)]G(k+1|k); Where U(k+1|k+1) is the posterior estimate of the state estimate at time k+1; G(k+1|k+1) is the posterior estimate of the covariance matrix at time k+1; K(k+1) is the Kalman gain at time k+1; γ(k+1) is the residual sequence between k+1 and time k; Z(k+1) is the observation value at time k+1; H(k+1) is the measurement matrix at time k+1.

9. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 8, characterized in that: Step S4) also includes: S43), after the degradation state U(k+1|k) is estimated a priori at time k+1, the reliability at time k+1 can be predicted; Where a0 is the initial value of the reaction wheel bearing temperature telemetry, a1 is the drift parameter of the reaction wheel bearing temperature telemetry, σ B is the diffusion coefficient; Xm is the failure threshold; U(k+1|k) is the prior estimate of the state, which represents the prior estimate of the temperature degradation in this Kalman filter.

10. The method for predicting the reliability of a reaction wheel during a degradation process according to claim 9, characterized in that: In step S5), the reliability of the reaction wheel is predicted using the prediction model, which specifically includes the following steps: S51), initialization parameters, including failure threshold Xm, initial value a0 of reaction wheel bearing temperature telemetry, drift parameter a1 of reaction wheel bearing temperature telemetry, diffusion coefficient σ B , the initial state values ​​U(0|0) and G(0|0) are the initial covariance values; S52) Use the state prediction equation U(k+1|k)=A(k)U(k|k)+δ(k)+E(k) to predict the mean value at the next moment, and obtain the predicted value at the k+1 moment based on the optimal estimate at the k moment; S53) Utilization Calculate the probability density function at this time, and then use The calculation expression of reliability is used to calculate the predictable reliability P(X(t)<Xm) at the k+1th moment; S54), collecting n samples within the k+1th Δt moment, and calculating the moment estimate as the observation value Z(k+1) at the k+1th moment; S55), prediction covariance G(k+1|k)=A(k)G(k|k)+E(k); S56), update the gain K(k+1), the optimal state estimate U(k+1|k+1) and the covariance G(k+1|k+1); S56), k=k+1, return to step S52).

Images