A combined navigation multi-time scale fault detection and isolation method

Abstract

This invention discloses a multi-timescale fault detection and isolation method for integrated navigation systems. Targeting integrated navigation systems employing a federated filtering structure, the method acquires state estimation information for each local filtering branch, constructs a residual quantity reflecting changes in consistency between branches, and normalizes this residual by combining it with corresponding uncertainty information to obtain a standardized deviation. Normal fluctuations are suppressed through soft thresholding, and non-negative effective deviations are extracted. Based on the effective deviations, fast-timescale and slow-timescale statistics are constructed respectively, and the impact of operational condition changes on the detection results is mitigated through baseline updates and platform removal processing. Finally, fault detection and isolation conclusions are output based on the joint decision results of the fast and slow-timescale statistics. This invention can balance rapid response to abrupt faults with cumulative perception of gradual faults, and is applicable to local branch fault detection and isolation in multi-source integrated navigation systems.

Description

Technical Field

[0001] This invention belongs to the field of integrated navigation fault diagnosis technology, and in particular relates to a multi-timescale fault detection and isolation method applicable to federated filter integrated navigation systems. Background Technology

[0002] Integrated navigation systems improve positioning accuracy, velocity measurement accuracy, and operational reliability by fusing information from multiple sensors, including inertial devices, satellite navigation equipment, and Doppler velocimetry. These systems are widely used in unmanned aerial vehicles, intelligent vehicles, and underwater vehicles. However, in practical applications, factors such as device performance degradation, environmental interference, measurement anomalies, and model uncertainties can cause sudden faults (e.g., instantaneous bias anomalies) or gradual degradation faults (e.g., slow drift) in local sensor branches. If these anomalies are not detected and isolated in a timely manner, they may enter the integrated navigation system through the filtering and update process, thus affecting the overall navigation accuracy and stability of the system.

[0003] Existing fault detection methods for integrated navigation systems mostly rely on single-time residuals or statistical features at a single time scale for anomaly identification. These methods typically have a good response speed to abrupt faults, but they often suffer from detection lag and high false negative rates for gradually accumulating faults. Furthermore, single statistical quantities are easily affected by platform drift or baseline offset, leading to an increased risk of false alarms.

[0004] Therefore, there is a need for a combined navigation fault detection and isolation method that can simultaneously take into account the ability to quickly detect sudden faults and the ability to perceive the accumulation of gradual faults, while also suppressing the impact of changes in operating conditions. Summary of the Invention

[0005] Purpose of the invention: This invention provides a method for multi-timescale fault detection and isolation in integrated navigation, in order to solve the problems of insufficient ability to perceive gradual faults, sensitivity to changes in operating conditions, and frequent switching of fault states in the prior art.

[0006] Summary of the Invention: This invention discloses a multi-timescale fault detection and isolation method for integrated navigation systems, applied to integrated navigation systems employing a federated filtering structure; the integrated navigation system includes an inertial navigation system and at least two local filtering branches combined with it; the method includes the following steps:

[0007] (1) Obtain the state estimate of each local filter branch at the current time and the covariance information representing the corresponding estimation uncertainty;

[0008] (2) Based on the estimation difference of the corresponding state components between each local filtering branch, construct the consistency relationship graph between branches; take each local filtering branch as the graph node, take the state estimation difference between any two local filtering branches as the graph edge, and determine the graph edge weight by combining the corresponding covariance information.

[0009] (3) According to the consistency relationship diagram, the weighted deviation degree between each local filtering branch and the other local filtering branches is aggregated to obtain the consistency deviation index of the corresponding local filtering branch. The consistency deviation index represents the abnormal contribution of the local filtering branch in the overall consistency mismatch. The index is then subjected to adaptive standardization and soft threshold mapping to obtain the effective deviation of the corresponding local filtering branch.

[0010] (4) Construct fast timescale statistics to characterize instantaneous anomalies and slow timescale statistics to characterize the accumulation of persistent anomalies based on effective deviations; and perform baseline updates and deplatformization on the fast timescale statistics and slow timescale statistics to obtain deplatformized statistics.

[0011] (5) Make a joint decision based on the deplatform statistics and abnormal contribution of each local filter branch; when a fault is determined, output the fault detection result and isolation command of the corresponding local filter branch, and control the local filter branch to exit the federated main filter fusion or reduce its weight in the federated main filter fusion; when no fault is determined, maintain the normal operation of the corresponding local filter branch.

[0012] Furthermore, the corresponding state component mentioned in step (2) includes one or more of the following: velocity component, position component, or attitude component.

[0013] Furthermore, the state estimation difference between any two local filter branches in step (2) is as follows:

[0014]

[0015]

[0016] in, Used to characterize the degree of consistency deviation between local filter branch i and local filter branch j.

[0017] Furthermore, the adaptive standardization process in step (3) incorporates at least one or more of the following information: the covariance level at the current moment, the intensity of historical baseline fluctuations, and the degree of consistency dispersion of multiple branches; the soft threshold mapping is used to suppress normal fluctuations below the initial threshold and to map deviations above the initial threshold as non-negative effective deviations.

[0018] Furthermore, the fast timescale statistic mentioned in step (4) is:

[0019]

[0020] Among them, e i (t) represents the effective deviation of the i-th local filter branch at time t. The rate of change of deviation. Used for burst enhancement to improve the response to sudden failures; α1, α2, and α3 are preset weighting coefficients.

[0021] The slow timescale statistic is:

[0022]

[0023] Among them, C i (t) represents the continuous count of out-of-baseline events, used to accumulate gradual anomalies, and β1, β2, and β3 are preset weighting coefficients.

[0024] Furthermore, in step (4), the baseline update and de-platform processing update the corresponding statistical baseline only when the target local filtering branch meets the preset health constraints; when the health constraints are not met, the update of the corresponding statistical baseline is suspended; the statistical baseline that is suspended from updating is only resumed when the recovery threshold and consistency recovery conditions are met simultaneously at multiple consecutive times.

[0025] Further, the joint decision implementation process in step (5) is as follows: when the slow time scale deplatformation statistic exceeds the slow statistics threshold, it is determined that the corresponding local filtering branch has a gradual abnormal trend; when the fast time scale deplatformation statistic exceeds the fast statistics threshold, it is determined that the corresponding local filtering branch has a sudden abnormal trend; and the fault status of the corresponding local filtering branch is output.

[0026] Furthermore, the joint decision in step (5) adopts a hysteresis decision mechanism, setting fault entry threshold and fault recovery threshold respectively to reduce the frequent switching of fault state within the critical interval.

[0027] Beneficial Effects: Compared with the prior art, the beneficial effects of this invention are as follows: By constructing a consistency relationship graph between branches, this invention can fully utilize the mutual constraints between multiple branches and improve the information utilization rate of fault detection; by simultaneously constructing fast-time-scale statistics and slow-time-scale statistics, this invention can balance the rapid response capability to sudden faults and the cumulative detection capability to gradual faults; by suppressing normal fluctuations through soft threshold mapping, this invention helps to reduce the impact of noise disturbances and small random fluctuations on fault detection results; by baseline updating and plateau removal processing, this invention weakens the adverse effects of operating condition changes and statistical plateau drift, which helps to improve detection stability. Attached Figure Description

[0028] Figure 1 This is a flowchart of the present invention;

[0029] Figure 2 The simulation curve of the eastward velocity error using the present invention and other methods is shown.

[0030] Figure 3 The figure shows the simulation curve of the northbound velocity error using the present invention and other methods. Detailed Implementation

[0031] The invention will now be further described with reference to the accompanying drawings.

[0032] like Figure 1 As shown, this invention proposes a multi-timescale fault detection and isolation method for integrated navigation systems, applicable to integrated navigation systems employing a federated filtering structure. The integrated navigation system includes an inertial navigation system and at least two combined local filtering branches. These local filtering branches may include inertial navigation and Doppler velocimeter local filtering branches, inertial navigation and GPS local filtering branches, or other combined branches containing external measurement information. Each local filtering branch outputs its corresponding state estimate and covariance information. The federated master filter fuses the information output by each local filtering branch to form the integrated navigation result. The specific implementation process is as follows:

[0033] Step 1: Obtain the state estimate X of each local filter branch at the current time. i The covariance information P that characterizes the uncertainty of the corresponding estimate i , where i = 1, 2, ..., M, and M is the total number of local filter branches. State estimates may include velocity, position, or attitude components. Data may come from inertial navigation systems, Doppler velocimeters, or GPS, and requires time alignment and preprocessing.

[0034] Step 2: Construct a consistency graph between branches based on the estimation differences of the corresponding state components between each local filtering branch; use each local filtering branch as a graph node, the state estimation difference between any two local filtering branches as a graph edge, and determine the graph edge weights by combining the corresponding covariance information.

[0035] The difference in state estimates between any two local filter branches i and j is:

[0036]

[0037] Determine the graph edge weights based on the corresponding covariance information:

[0038]

[0039] in, Used to characterize the degree of consistency deviation between local filter branch i and local filter branch j.

[0040] Step 3: Based on the consistency graph, aggregate the weighted deviation between each local filter branch i and the other local filter branches to obtain the consistency deviation index of the corresponding local filter branch.

[0041] The consistency deviation index characterizes the abnormal contribution of a local filter branch to the overall consistency mismatch. Adaptive standardization and soft thresholding are applied to the consistency deviation index to obtain the effective deviation e of the corresponding local filter branch. i (t); where adaptive standardization can combine one or more of the current covariance level, historical baseline volatility intensity, and multi-branch consistency dispersion. Soft thresholding is used to suppress normal fluctuations below the initial threshold and map deviations above the initial threshold to non-negative effective deviations.

[0042] Step 4: Construct fast timescale statistics to represent instantaneous anomalies and slow timescale statistics to represent the accumulation of persistent anomalies based on effective deviations; and perform baseline updates and deplatformization on the fast and slow timescale statistics to obtain deplatformized statistics.

[0043] The fast time-scale statistic for the i-th local filter branch is:

[0044]

[0045] Among them, e i (t) represents the effective deviation of the i-th local filter branch at time t. The rate of change of deviation. Used for burst enhancement to improve the response to sudden failures, α1, α2, and α3 are preset weighting coefficients.

[0046] The slow-time-scale statistic for the i-th local filtering branch is:

[0047]

[0048] Among them, C i (t) represents the continuous count of out-of-baseline events, used to accumulate gradual anomalies, and β1, β2, and β3 are preset weighting coefficients.

[0049] The corresponding statistical baseline is updated only when the target local filtering branch meets the preset health constraints; when the health constraints are not met, the update of the corresponding statistical baseline is paused; the statistical baseline that has been paused is only resumed when the recovery threshold and consistency recovery conditions are met simultaneously at multiple consecutive time points.

[0050] Step 5: Perform a joint decision based on the de-platform statistics and outlier contributions of each local filter branch: when the slow timescale statistics... When the slow statistical threshold is exceeded, the branch is determined to have a gradual abnormal trend; when the fast timescale statistics... When the fast statistical threshold is exceeded, the branch is determined to have a sudden abnormal trend. Further, a hysteresis decision mechanism can be adopted, setting fault entry and recovery thresholds to reduce frequent switching near these thresholds. Local filter branches determined to be faulty are removed from federated main filter fusion or have their weight in federated main filter fusion reduced to ensure the overall accuracy and stability of the integrated navigation system. Local filter branches determined to be normal maintain normal operation.

[0051] In this embodiment, the integrated navigation system consists of an inertial navigation system, a Doppler velocimeter, and a global positioning system, and employs a federated filtering structure to achieve information fusion. Specifically, the local filtering branches for the inertial navigation and Doppler velocimeter, and the local filtering branches for the inertial navigation and global positioning system, respectively output their corresponding state estimation results and covariance information. The main filter fuses the information from the two local filtering branches. In this embodiment, the velocity component is preferably selected as the state component for consistency deviation calculation to improve detection accuracy.

[0052] In this embodiment, the GPS positioning accuracy is 1.4 cm, the velocity accuracy is 0.002 m / s, and the long-term velocity measurement accuracy of DVL is 0.2% ± 0.3 cm. To verify the detection and isolation capabilities of the method of the present invention for abrupt and gradual faults, fault injection is performed in the forward velocity channel of DVL: a constant bias with an amplitude of 0.8 m / s is injected in the interval of 750 s to 1050 s to simulate abrupt faults; and a drift coefficient of 0.003 m / s is injected in the interval of 3250 s to 3750 s. 2 The linear drift is used to simulate slowly varying faults.

[0053] Through computer simulation, by Figure 2 , Figure 3 As can be seen from the comparison curves, in the two abnormal intervals, the velocity error of the system without fault-tolerance measures accumulates rapidly; the velocity error level of the system using the traditional chi-square detection method is suppressed to a certain extent, but there is still obvious residual error in the gradual failure stage; while the velocity error of the system using the multi-timescale integrated navigation system fault detection and isolation method proposed in this invention is smaller, and the curve fluctuation amplitude is significantly smaller than that of the comparison method. This invention effectively suppresses the spread of velocity error during the fault occurrence, thereby ensuring the overall positioning accuracy of the integrated navigation system.

[0054] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for multi-timescale fault detection and isolation in integrated navigation systems, characterized in that, The method is applied to a combined navigation system employing a federated filtering structure; the combined navigation system includes an inertial navigation system and at least two local filtering branches combined with it; the method includes the following steps: (1) Obtain the state estimate of each local filter branch at the current time and the covariance information representing the corresponding estimation uncertainty; (2) Based on the estimation difference of the corresponding state components between each local filtering branch, construct the consistency relationship graph between branches; take each local filtering branch as the graph node, take the state estimation difference between any two local filtering branches as the graph edge, and determine the graph edge weight by combining the corresponding covariance information. (3) According to the consistency relationship diagram, the weighted deviation degree between each local filtering branch and the other local filtering branches is aggregated to obtain the consistency deviation index of the corresponding local filtering branch. The consistency deviation index represents the abnormal contribution of the local filtering branch in the overall consistency mismatch. The index is then subjected to adaptive standardization and soft threshold mapping to obtain the effective deviation of the corresponding local filtering branch. (4) Construct fast timescale statistics to characterize instantaneous anomalies and slow timescale statistics to characterize the accumulation of persistent anomalies based on effective deviations; and perform baseline updates and deplatformization on the fast timescale statistics and slow timescale statistics to obtain deplatformized statistics. (5) Make a joint decision based on the deplatform statistics and abnormal contribution of each local filter branch; when a fault is determined, output the fault detection result and isolation command of the corresponding local filter branch, and control the local filter branch to exit the federated main filter fusion or reduce its weight in the federated main filter fusion; when no fault is determined, maintain the normal operation of the corresponding local filter branch.

2. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, The corresponding state components mentioned in step (2) include one or more of the velocity component, position component, or attitude component.

3. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, The state estimation difference between any two local filter branches mentioned in step (2) is: ； ； in, Used to characterize the degree of consistency deviation between local filter branch i and local filter branch j.

4. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, The adaptive standardization process described in step (3) incorporates at least one or more of the following information: the covariance level at the current time, the intensity of historical baseline fluctuations, and the degree of consistency dispersion of multiple branches; The soft threshold mapping is used to suppress normal fluctuations below the initial threshold and to map deviations above the initial threshold as non-negative effective deviations.

5. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, The fast timescale statistic mentioned in step (4) is: ； Among them, e i (t) represents the effective deviation of the i-th local filter branch at time t. The rate of change of deviation. Used for burst enhancement to improve the response to sudden failures; α1, α2, and α3 are preset weighting coefficients. The slow timescale statistic is: ； Among them, C i (t) represents the continuous count of out-of-baseline events, used to accumulate gradual anomalies, and β1, β2, and β3 are preset weighting coefficients.

6. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, In step (4), the baseline update and de-platformization process only updates the corresponding statistical baseline when the target local filter branch meets the preset health constraints; when the health constraints are not met, the update of the corresponding statistical baseline is paused; the statistical baseline that is paused is only resumed when the recovery threshold and consistency recovery conditions are met simultaneously at multiple consecutive times.

7. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, The joint decision implementation process described in step (5) is as follows: when the slow time scale deplatformation statistic exceeds the slow statistics threshold, it is determined that the corresponding local filter branch has a gradual abnormal trend; when the fast time scale deplatformation statistic exceeds the fast statistics threshold, it is determined that the corresponding local filter branch has a sudden abnormal trend; and the fault status of the corresponding local filter branch is output.

8. The method for multi-timescale fault detection and isolation in integrated navigation according to claim 1, characterized in that, The joint decision in step (5) adopts a hysteresis decision mechanism, setting fault entry threshold and fault recovery threshold respectively to reduce the frequent switching of fault state within the critical interval.

Images