CN102175266A - Fault diagnosis method for mobile gyroscope inertia subassembly - Google Patents

Abstract

A fault diagnosis method for a moving body gyro inertial component, specifically related to a health monitoring technology for a moving body inertial gyro component based on equivalence relations and empirical mode decomposition. The ability to isolate individual sensor faults is a relatively computationally intensive problem for signal processing-based health monitoring methods. The method of the present invention comprises: step 1, detects whether the inertial gyro assembly breaks down by the equivalence relation method; Step 2, when detecting a fault, collects N data points of each gyro sensor output signal in the inertial gyroscope assembly as its The fault data input signal is subjected to empirical mode decomposition, and the obtained first-order IM F component is used as its fault characteristic signal; Step 3, the accumulation and summation CUSUM processing of the statistical test is performed on the first-order IM F component to judge the gyro sensor Whether there is a fault, complete the health monitoring of the inertial gyro components.

Description

A Fault Diagnosis Method for Gyro Inertial Components of Moving Body

technical field

The invention relates to a fault diagnosis method for a moving body gyro inertial component, in particular to a health monitoring technology for a moving body inertial gyro component based on an equivalence relationship and empirical mode decomposition.

Background technique

The inertial gyro component consists of a number of gyro sensors, generally 2-4 according to different tasks. The gyro sensor is used to measure the angular velocity of moving bodies, especially space vehicles and marine vehicles relative to the inertial reference system, and is an important part of the attitude measurement of moving bodies. Components, their working performance and health status will directly affect the attitude measurement of the entire moving body and even the attitude control accuracy and reliability of the moving body.

Since the gyro sensor is a high-precision, relatively vulnerable part, it is prone to health problems such as failure and performance degradation after undergoing complex environments and strenuous exercise. At present, when monitoring the health status of inertial gyroscope components, in view of the fact that the traditional equivalence relationship method usually only has the ability to separate the failure of a single sensor, two approaches are usually adopted: one is to use the built-in test equipment to monitor some internal components of the gyroscope in real time. State variables, such as temperature, current, voltage, etc., are used to judge whether the gyro is faulty or not. This method does not use the output information of the gyro sensor and can only detect faults with large magnitudes. Another way is to use the redundant relationship of the attitude sensor to judge whether the gyroscope is faulty, but the application of this method is also subject to certain restrictions. On the one hand, this method is only valid under the condition that the attitude sensor system constitutes a redundant relationship; On the other hand, it is also necessary to consider the working conditions of redundant sensors.

In recent years, modern signal processing methods have been increasingly used for the extraction of sensor health feature information. For example, some researchers have applied wavelet transform to the fault diagnosis and classification of gyro sensors, and achieved good results. The advantage of sensor health monitoring based on signal processing methods is that it can directly extract health feature information from sensor output, which can overcome the limitation of redundant relationships. Moreover, in the case that multiple sensors fail at the same time, the detection and isolation of the failure can still be realized. But on the other hand, compared with methods based on hardware redundancy or analytical redundancy, health monitoring methods based on signal processing often require more calculations and processing, and the amount of calculation is relatively large.

Contents of the invention

The purpose of the present invention is to solve the problem that the traditional equivalence relation method usually only has the ability to isolate a single sensor fault, and the health monitoring method based on signal processing has a relatively large amount of calculation. Fault diagnosis method.

The fault diagnosis method of a kind of moving body gyro inertia assembly of the present invention comprises the following steps:

Step 1, utilizing the hardware redundancy of the inertial gyro assembly of the moving body, and detecting whether the inertial gyro assembly fails by an equivalence relation method;

The procedure for detecting failure of an inertial gyro component is:

Step 11, using the hardware redundancy configuration of the moving body inertial gyroscope component to construct an equivalence relationship, and calculate its equivalent vector;

Step 12, judging whether the norm of the equivalent vector is smaller than the fault detection threshold,

If the judgment result is yes, it is considered that the inertial gyro component has not failed; if the judgment result is no, it is considered that the inertial gyro component has failed.

Step 2, when detecting that the inertial gyro assembly breaks down, collect N data points of each gyro sensor output signal in the inertial gyro assembly as the fault data input signal of the gyro sensor, and perform the fault data input signal on the fault data input signal Empirical mode decomposition, the obtained first-order IMF component is used as the gyro sensor fault characteristic signal;

The process of obtaining the first-order IMF component is:

Set the fault data input signal as x(t), t=1, 2,..., N,

Step 21, initialization of IMF decomposition process: n=1, and the relation r _n-1 (t)=x(t) is satisfied, where r _n-1 (t) is the trend function after the (n-1)th decomposition ;

Step 22: The screening process is initialized, k=1, and the relationship h _n (k-1)(t)=rn _-1 (t) is satisfied, where h _n (k-1)(t) is the nth time The residual function after the (k-1)th screening in the empirical mode decomposition;

Step 23, obtain the remaining function h _nk (t) after the kth screening according to the screening program;

The process of obtaining the residual function h _nk (t) is:

Step 231, using the cubic spline function to obtain the upper and lower values of the remaining function h n (k-1)(t) of the residual function h _n (k-1)(t) after the k-1th screening in the n-th empirical mode decomposition of the fault data input signal x(t). lower envelope,

Step 232, calculate the mean value of the upper and lower envelope curves of the residual function h _n (k-1)(t) at each t

Step 233, obtain the residual function of the fault data input signal x(t) after the kth screening in the nth empirical mode decomposition $h_{\mathrm{nk}}\left(t\right)=h_{\mathrm{no}(k-1)}\left(t\right)-{\stackrel{\‾}{m}}_{\mathrm{no}(k-1)}\left(t\right).$

Step 24, using the standard deviation criterion to judge whether the residual function h _nk (t) obtained in step 23 satisfies the condition of the intrinsic mode function IMF, namely Whether it is less than the threshold H _SD , 0.2≤H _SD ≤0.3;

If the judgment result is yes, execute step 25, if the judgment result is no, then k=k+1, then execute step 23,

Step 25, extracting the first-order IMF component: c ₁ (t)=h _1k (t).

Step 3, the first-order IMF component that each gyro sensor obtains in step 2 is carried out the accumulative summation CUSUM processing of statistical test, judges whether this gyro sensor has fault, and then separates out the gyro that has fault in the inertial gyro assembly The sensor completes the health monitoring of the inertial gyro components.

Step 31. Obtain the CUSUM calculation result W _i of a single data point in the first-order IMF component according to the following formula:

W _i =W _i-1 + |X _i |,

Wherein, i=1, 2,...N, W _i-1 is the CUSUM calculation result of the i-1th data point, and let W ₀ =0, _Xi is the first-order IMF value at the i-th data point,

Step 32. Determine whether the CUSUM calculation result W _N of the Nth data point is greater than the diagnostic threshold ε _I ,

If yes, there is a fault in the gyro sensor; otherwise, there is no fault in the gyro sensor.

Advantages of the present invention:

1) The health monitoring method proposed by the present invention monitors the fault of the inertial gyroscope assembly of the moving body through the equivalence relation method, and uses the empirical mode decomposition method to separate the faulty gyroscope at the same time. Compared with the fault diagnosis method that only uses signal processing, it can effectively reduce the Calculations.

2) The health monitoring method proposed by the present invention not only utilizes the hardware redundancy of the gyro components, but also effectively utilizes the output information of the gyro sensor itself, which enhances the fault diagnosis capability of the algorithm.

3) The health monitoring method proposed by the present invention only performs fault diagnosis on the basis of the output signal of the gyro component, without using information from other sensors, avoiding the introduction of other potential fault sources, and is conducive to improving the effectiveness of the fault diagnosis method.

4) The health monitoring method proposed by the present invention is not limited to the single-point fault assumption, and can conveniently realize the diagnosis of multiple faults, which breaks through the limitation that other methods can only diagnose single-point faults.

Description of drawings

Fig. 1 is the fault diagnosis method flow chart of a kind of moving body gyro inertia component based on equivalence relation and empirical mode decomposition;

Figure 2 is a flow chart of empirical mode decomposition;

Figure 3 is a schematic diagram of the experimental verification device;

Fig. 4 is the equivalent vector when the fault with constant value drift increases;

Fig. 5 is the first-order IMF signal when the fault with increasing constant value drift occurs;

Figure 6 is the equivalent vector when the noise level increases when the fault occurs;

Fig. 7 is the first-order IMF signal when the fault with increased noise level occurs;

Figure 8 is the equivalent vector when the X-axis gyro and the S-axis gyro have a sudden failure at the same time;

Figure 9 shows the first-order IMF signal when the X-axis gyro and the S-axis gyro have a sudden fault at the same time.

Detailed ways

Specific Embodiment 1: The present embodiment will be described below with reference to FIG. 1 and FIG. 2 .

In order to effectively use signal processing methods for sensor health monitoring, and to reduce the computational load of the algorithm to a certain extent, this patent proposes a sensor health monitoring method based on equivalence relations and Empirical Mode Decomposition (EMD) , for health monitoring of inertial gyro components.

The Empirical Mode Decomposition method was proposed by Dr. Huang E of the National Aeronautics and Space Administration (NASA) in 1998. It uses the change of the internal time scale of the signal to analyze the energy and frequency, and expands the signal into A limited number of Intrinsic Mode Functions (IMFs). Different from the traditional method that uses a fixed shape window as the boundary basis function, the basis function of EMD is extracted from the signal, that is, the IMF is used as the basis. The IMF must meet the following conditions:

1) In the whole function, the number of extremum points is equal to or differs by 1 from the number of crossing zero points;

2) At any moment, the local mean of the envelope defined by the local extremum envelope is zero. Among them, the first condition is similar to the narrow bandwidth requirement in the traditional Gaussian stationary process. The second condition is a new idea: change the overall requirement into a local requirement, so that the instantaneous frequency will not cause unnecessary shaking due to the existence of asymmetric waveforms. EMD built on the basis of these two conditions is considered to be a powerful adaptive method for solving nonlinear and non-stationary signals. It is a major breakthrough in traditional signal analysis methods such as Fourier transform in recent years, and has been widely used.

Considering that the failure of the gyro sensor will cause the change of the characteristics of the gyro output signal, this patent decomposes the gyro output signal into the signal superposition of IMF components from high frequency to low frequency through the EMD method, which provides a basis for feature selection in the health monitoring process. Reasonable way.

A fault diagnosis method for a moving body gyro inertial component described in this embodiment includes the following steps:

Step 1, utilizing the hardware redundancy of the inertial gyro assembly of the moving body, and detecting whether the inertial gyro assembly fails by an equivalence relation method;

Step 2, when detecting that the inertial gyro assembly breaks down, collect N data points of each gyro sensor output signal in the inertial gyro assembly as the fault data input signal of the gyro sensor, and perform the fault data input signal on the fault data input signal Empirical mode decomposition, the obtained first-order IMF component is used as the gyro sensor fault characteristic signal;

Step 3, the first-order IMF component that each gyro sensor obtains in step 2 is carried out the accumulative summation CUSUM processing of statistical test, judges whether this gyro sensor has fault, and then separates out the gyro that has fault in the inertial gyro assembly The sensor completes the health monitoring of the inertial gyro components.

In step 1, the process of detecting whether the inertial gyro component fails by the equivalence relation method is as follows:

Step 11, using the hardware redundancy configuration of the moving body inertial gyroscope component to construct an equivalence relationship, and calculate its equivalent vector;

Step 12, judging whether the norm of the equivalent vector is smaller than the fault detection threshold,

If the judgment result is yes, it is considered that the inertial gyro component has not failed; if the judgment result is no, it is considered that the inertial gyro component has failed.

The fault detection threshold ε _D is 8σ ~ 12σ, where σ is the standard deviation of the gyro noise. Theoretically, when ε _D is 4σ, the correct detection probability is very close to 1, but in order to reduce the false alarm rate, the fault detection threshold is increased to 8σ ~ 12σ.

The process of obtaining the first-order IMF component in step 2 is:

Set the fault data input signal as x(t), t=1, 2,..., N,

Step 21, initialization of IMF decomposition process: n=1, and the relation r _n-1 (t)=x(t) is satisfied, where r _n-1 (t) is the trend function after the (n-1)th decomposition ;

Step 22: The screening process is initialized, k=1, and the relation h _n(k-1) (t)=r _n-1 (t) is satisfied, where h _n(k-1) (t) is the nth time The residual function after the (k-1)th screening in the empirical mode decomposition;

Step 23, obtain the remaining function h _nk (t) after the kth screening according to the screening program;

是否小于阈值H_SD，0.2≤H_SD≤0.3；Step 24, using the standard deviation criterion to judge whether the residual function h _nk (t) obtained in step 23 satisfies the condition of the intrinsic mode function IMF, namely

Whether it is less than the threshold H _SD , 0.2≤H _SD ≤0.3;

If the judgment result is yes, execute step 25, if the judgment result is no, then k=k+1, then execute step 23,

Step 25, extracting the first-order IMF component: c ₁ (t)=h _1k (t).

The process of obtaining the remaining function h _nk (t) in step 23 is:

Step 231, using the cubic spline function to obtain the upper and lower values of the residual function h n(k-1) (t) of the residual function h _n(k-1) (t) after the k-1th screening in the n-th empirical mode decomposition of the fault data input signal x(t) lower envelope,

Step 232, calculate the mean value of the upper and lower envelope curves of the residual function h _n(k-1) (t) at each t

Step 233, obtain the residual function of the fault data input signal x(t) after the kth screening in the nth empirical mode decomposition $h_{\mathrm{nk}}\left(t\right)=h_{\mathrm{no}(k-1)}\left(t\right)-{\stackrel{\‾}{m}}_{\mathrm{no}(k-1)}\left(t\right).$

In the step 3, the accumulation and summation CUSUM processing of statistical inspection is carried out to the first-order IMF component, and the process of judging whether this gyro sensor has a fault is:

Step 31. Obtain the CUSUM calculation result W _i of a single data point in the first-order IMF component according to the following formula:

W _i =W _i-1 + |X _i |,

Wherein, i=1, 2,...N, W _i-1 is the CUSUM calculation result of the i-1th data point, and let W ₀ =0, _Xi is the first-order IMF value at the i-th data point,

Step 32. Determine whether the CUSUM calculation result W _N of the Nth data point is greater than the diagnostic threshold ε _I ,

If yes, there is a fault in the gyro sensor; otherwise, there is no fault in the gyro sensor.

If yes, there is a fault in the gyro sensor; otherwise, there is no fault in the gyro sensor.

The diagnostic threshold _εI is obtained according to the following formula:

ε _I =Nσ, where σ is the standard deviation of the gyro noise.

Embodiment 2: The difference between this embodiment and Embodiment 1 is that the fault detection threshold ε _D is 10σ, and other aspects are the same as Embodiment 1.

Embodiment 3: The difference between this embodiment and Embodiment 1 is that in step 24 H _SD =0.25, and the others are the same as Embodiment 1.

Specific Embodiment 4: The present embodiment will be described below with reference to FIG. 3 to FIG. 9 . This embodiment provides a specific example: the schematic diagram of the test verification device is shown in FIG. 3 . The monitoring object is a gyro inertial assembly composed of four single-axis fiber optic gyroscopes VG951D. The motion of the moving body can be simulated by a three-axis turntable. The inertial gyro assembly is loaded on the three-axis turntable. The measurement output of the inertial gyro assembly is Rotational angular velocity. The health monitoring processor uses LPC2478 as the main controller. LPC2478 is a microcontroller designed by NXP Semiconductors with extremely high integration and ARM7TDMI-S as the core.

Execution step 1: Utilize the hardware redundancy of the inertial gyroscope component of the moving body, and detect whether the inertial gyroscope component fails by using an equivalence relation method.

In this step, the equivalence relation method is used to detect the fault of the gyro component. In order to ensure the safety and reliability of the system, most of the gyro sensor systems of the moving body inertial gyro components adopt redundant configuration, so the equivalence relationship can be constructed and the equivalent vector can be calculated for sensor fault detection and separation. Suppose the measurement equation of the system is:

m＝Hx+w+f (1)

为l个传感器的输出信号；

为传感器安装矩阵；为n维被测量信号；

分别为测量噪声和附加故障信号。in,

is the output signal of l sensors;

Install the matrix for the sensor; is the n-dimensional measured signal;

are the measurement noise and the additional fault signal, respectively.

If choose matrix satisfy:

VH＝0 _(ln)×n (2)

VV ^T = I _ln (3)

That is, the column vectors of H form an n-dimensional space, and the row vectors of V form the (l-n)-dimensional orthogonal space of this space.

Equivalent vectors are defined as:

p＝Vm＝V(w+f) (4)

Then the equivalent vector p has nothing to do with the measured signal, it is only a function of noise w and fault f. If the influence of noise is not considered, the equivalent vector p is the component of the fault vector f in the (l-n) dimensional subspace spanned by V. Therefore, a certain threshold can be set to detect the fault of the sensor system according to the norm of the equivalent vector p.

The inertial gyroscope assembly in the moving body usually adopts a three-orthogonal-inclined configuration. This patent illustrates the effectiveness of the proposed fault diagnosis method for this gyroscope configuration.

The measurement matrix of the gyroscope component in the three-orthogonal-inclined configuration is:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 0.5774</span>}& \mathrm{<span\; class="notranslate">\; 0.5774</span>}& \mathrm{<span\; class="notranslate">\; 0.5774</span>}\end{array}\right]<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

The projection matrix that satisfies formulas (2) and (3) is obtained as

V＝[0.4082 0.4082 0.4082 -0.707] (6)

The equivalent vector is:

p=Vm (7)

The fault detection law is:

Among them, ε _D is the fault detection threshold, which can be selected as 10σ, and σ is the standard deviation of the gyro noise. Whether the gyro sensor group fails is detected by the equivalence relation method.

Execute step 2: when detecting that the inertial gyro assembly breaks down, collect N data points of each gyro sensor output signal in the inertial gyro assembly as the fault data input signal of the gyro sensor, and input the fault data to the input signal The empirical mode decomposition is carried out, and the obtained first-order IMF component is used as the fault characteristic signal of the gyro sensor.

After the fault of the gyroscope component is detected, N=128 data points are collected around the time of fault detection, and an empirical mode decomposition is performed. Since the first-order IMF contains sufficient feature information, EMD can stop after decomposing the first IMF component c ₁ (t), which can effectively improve the processing speed and reduce the amount of calculation.

Step 3: Carry out a CUSUM operation on the obtained first-order IMF, and diagnose whether the gyro of each axis is faulty or not by passing the threshold value ε _I =Nσ.

The validity of the fault diagnosis method proposed by the present invention will be verified by adopting a typical inertial gyro assembly failure below. The time interval adopted in the simulation is 0.025s, and the standard deviation of the gyro noise is 1×10 ^-6 rad/s (ie ε _D =10σ=1×10 ^-5 rad/s, ε _I =Nσ=1.28×10 ^-4 rad/s). The three faults are:

1) At t=6s, the X-axis gyro sensor has a constant value drift increase fault, and the mutation amplitude is -1×10 ^-4 rad/s, the equivalent vector is shown in Figure 4, and the first-order IMF signal is shown in Figure 5 shown. The calculated CUSUM calculation results W _N of the Nth data point in the first-order IMF component of each axis gyroscope are: X axis: 3.354×10 ^-4 rad/s, Y axis: 9.5566×10 ^-5 rad/s, Z Axis: 9.5738×10 ^-5 rad/s, S axis: 8.7282×10 ^-5 rad/s. Wherein, the W _N of the X-axis is greater than the diagnostic threshold ε _I (1.28×10 ^-4 rad/s), therefore, it can be determined that the X-axis gyro is faulty.

2) At t=11s, the Z-axis gyro fails, causing the noise variance to increase to 1×10 ^-9 rad/s. The equivalent vector is shown in Figure 6, and the first-order IMF signal is shown in Figure 7. The calculated CUSUM calculation results W _N of the Nth data point in the first-order IMF component of each axis gyroscope are: X axis: 8.8193×10 ^-5 rad/s, Y axis: 9.8038×10 ^-5 rad/s, Z Axis: 1.7×10 ^-3 rad/s, S axis: 9.2548×10 ^-5 rad/s. Wherein, the W _N of the Z-axis is greater than the diagnostic threshold ε _I (1.28×10 ^-4 rad/s), therefore, it can be determined that the Z-axis gyro is faulty.

3) At t=15s, the X-axis gyro and the S-axis gyro have a sudden fault at the same time, and the fault size is 6×10 ^-5 rad/s. The equivalent vector is shown in Figure 8, and the first-order IMF signals are shown in Figure 9 Show. The calculated CUSUM calculation results W _N of the Nth data point in the first-order IMF component of each axis gyroscope are: X axis: 2.1543×10 ^-4 rad/s, Y axis: 9.1255×10 ^-5 rad/s, Z Axis: 8.149×10 ^-5 rad/s, S axis: 1.7414×10 ^-4 rad/s, where W _N of both X and S axes is greater than the diagnostic threshold ε _I (1.28×10 ^-4 rad/s), From this, it can be determined that the X-axis gyro and the S-axis gyro are faulty.

The above shows that this method does not need other sensors to provide information, and can effectively diagnose the fault when multiple faults occur at the same time.

Claims (9)

1. a method for fault diagnosis of a moving body gyro inertial assembly, characterized in that the method may further comprise the steps: Step 1, utilizing the hardware redundancy of the inertial gyro assembly of the moving body, and detecting whether the inertial gyro assembly fails by an equivalence relation method; Step 2, when detecting that the inertial gyro assembly breaks down, collect N data points of each gyro sensor output signal in the inertial gyro assembly as the fault data input signal of the gyro sensor, and perform the fault data input signal on the fault data input signal Empirical mode decomposition, the obtained first-order IMF component is used as the gyro sensor fault characteristic signal; Step 3, the first-order IMF component that each gyro sensor obtains in step 2 is carried out the accumulative summation CUSUM processing of statistical test, judges whether this gyro sensor has fault, and then separates out the gyro that has fault in the inertial gyro assembly The sensor completes the health monitoring of the inertial gyro components. 2. the fault diagnosis method of a kind of moving body gyroscope inertial assembly according to claim 1, is characterized in that, in the step 1, the process of detecting whether the inertial gyroscope assembly breaks down by the equivalence relation method is: Step 11, using the hardware redundancy configuration of the moving body inertial gyroscope component to construct an equivalence relationship, and calculate its equivalent vector; Step 12, judging whether the norm of the equivalent vector is smaller than the fault detection threshold, If the judgment result is yes, it is considered that the inertial gyro component has not failed; if the judgment result is no, it is considered that the inertial gyro component has failed. 3. A method for fault diagnosis of a moving body gyro-inertial component according to claim 2, wherein the fault detection threshold ε _D is 8σ˜12σ, where σ is the standard deviation of gyro noise. 4. A fault diagnosis method for a moving body gyro-inertial component according to claim 2, wherein the fault detection threshold ε _D is 10σ, where σ is the standard deviation of the gyro noise. 5. the fault diagnosis method of a kind of moving body gyro inertia assembly according to claim 1, is characterized in that, the process of obtaining first-order IMF component in step 2 is: Set the fault data input signal as x(t), t=1, 2,..., N, Step 21, initialization of IMF decomposition process: n=1, and the relation r _n-1 (t)=x(t) is satisfied, where r _n-1 (t) is the trend function after the (n-1)th decomposition ; Step 22: The screening process is initialized, k=1, and the relation h _n(k-1) (t)=r _n-1 (t) is satisfied, where h _n(k-1) (t) is the nth time The residual function after the (k-1)th screening in the empirical mode decomposition; Step 23, obtain the remaining function h _nk (t) after the kth screening according to the screening program; Step 24, using the standard deviation criterion to judge whether the residual function h _nk (t) obtained in step 23 satisfies the condition of the intrinsic mode function IMF, namely

Whether it is less than the threshold H _SD , 0.2≤H _SD ≤0.3; If the judgment result is yes, execute step 25, if the judgment result is no, then k=k+1, then execute step 23, Step 25, extracting the first-order IMF component: c ₁ (t)=h _1k (t). 6. the fault diagnosis method of a kind of moving body gyro inertia assembly according to claim 5, is characterized in that, the process of obtaining remaining function h _nk (t) in step 23 is: Step 231, using the cubic spline function to obtain the upper and lower values of the residual function h n(k-1) (t) of the residual function h _n(k-1) (t) after the k-1th screening in the n-th empirical mode decomposition of the fault data input signal x(t) lower envelope, Step 232, calculate the mean value of the upper and lower envelope curves of the residual function h _n(k-1) (t) at each t

Step 233, obtain the residual function of the fault data input signal x(t) after the kth screening in the nth empirical mode decomposition

7 . The fault diagnosis method for the gyro-inertial assembly of the moving body according to claim 5 , wherein H _SD =0.25 in step 24 . 8. the fault diagnosis method of a kind of moving body gyro inertial assembly according to claim 1, is characterized in that, in step 3, the accumulative summation CUSUM processing that statistical inspection is carried out to first-order IMF component, judges this gyro sensor The process of whether there is a fault is: Step 31, obtain the CUSUM calculation result Wi of a single data point in the first-order IMF component according to the following formula: W _i =W _i-1 + |X _i |, Wherein, i=1, 2,...N, W _i-1 is the CUSUM calculation result of the i-1th data point, and let W ₀ =0, _Xi is the first-order IMF value at the i-th data point, Step 32. Determine whether the CUSUM calculation result W _N of the Nth data point is greater than the diagnostic threshold ε _I , If yes, the gyro sensor is faulty; if not, the gyro sensor is not faulty. 9. the fault diagnosis method of a kind of moving body gyro inertia assembly according to claim 8, is characterized in that, diagnosis threshold ε ₁ obtains by following formula: ε _I =Nσ, where σ is the standard deviation of the gyro noise.

Images