Health diagnosis method for inertial navigation system

Abstract

The application relates to a health diagnosis method of an inertial navigation system, which comprises the following steps: determining a phase angle deviation of a phase angle of each time sequence item in one-dimensional time sequence signals of the inertial navigation system under noise interference; generating a first matrix; generating a second matrix; generating a first image corresponding to the first matrix and a second image corresponding to the second matrix; and obtaining a health diagnosis result of the inertial navigation system according to the first image and the second image.

Description

Technical Field

[0001] This application belongs to the field of inertial navigation technology, and more specifically relates to a health diagnosis method for inertial navigation systems. Background Technology

[0002] Inertial navigation systems, such as those including gyroscopes, accelerometers, and inertial measurement units, can sometimes malfunction. For example, during complex and frequent missions, components of the inertial navigation system installed on an aircraft, such as inertial sensing elements and signal processing modules, may malfunction due to aging, vibration, or environmental factors, resulting in abnormal voltage or current waveforms. This can directly lead to a reduction in the control accuracy of the aircraft and even a decrease in its safety.

[0003] How to perform health diagnosis on inertial navigation systems, i.e., determine whether the inertial navigation system is abnormal, has become a technical problem that needs to be solved. Summary of the Invention

[0004] This application provides a health diagnosis method for inertial navigation systems to address the problem of how to perform health diagnoses on inertial navigation systems.

[0005] This application provides a health diagnosis method for an inertial navigation system, the method comprising:

[0006] Determine the phase angle offset of each time signal item in the one-dimensional time signal of the inertial navigation system under noise interference;

[0007] Generating a first matrix includes: generating each element of the first matrix, wherein the target element of the first matrix is ​​any element of the first matrix. Generating the target element of the first matrix includes: generating the target element of the first matrix based on the sum of the phase angles of the two first time-series signal terms corresponding to the target element of the first matrix, the phase angle offset of the two first time-series signal terms corresponding to the target element of the first matrix under noise interference, and the penalty term corresponding to the target element of the first matrix.

[0008] Generating the second matrix includes: generating each element of the second matrix, wherein the target second matrix element is any element of the second matrix. Generating the target second matrix element includes: generating the target second matrix element based on the difference in phase angles of the two second time-series signal terms corresponding to the target second matrix element, the phase angle offset of the two second time-series signal terms corresponding to the target second matrix element under noise interference, and the penalty term corresponding to the target second matrix element.

[0009] Generate a first image corresponding to the first matrix and a second image corresponding to the second matrix;

[0010] Based on the first image and the second image, the health diagnosis results of the inertial navigation system are obtained.

[0011] Beneficial effects:

[0012] The method provided in this application embodiment enables health diagnosis of an inertial navigation system, i.e., determining whether the inertial navigation system is abnormal.

[0013] The method provided in this application, when constructing the first and second matrices for health diagnosis of the inertial navigation system, considers that the inertial navigation system has long-term drift characteristics, making it unsuitable to use Gram matrices constructed solely using Euclidean inner products focused on local dynamics to predict anomalies, as is the case in related technologies. Therefore, the method provided in this application introduces a penalty term for the long-term drift characteristics of the inertial navigation system. This constructs the first and second matrices that conform to the long-term drift characteristics of the inertial navigation system. Using these first and second matrices to perform health diagnosis of the inertial navigation system—that is, to determine whether the inertial navigation system is abnormal—improves the accuracy of the health diagnosis.

[0014] The method provided in this application, when constructing the first and second matrices for health diagnosis of an inertial navigation system, considers non-Gaussian noise interference that can cause deviations between the acquired voltage or current and the actual voltage or current. Therefore, the method provided in this application applies an interference term to the phase angle of the timing signal term. This constructs the first and second matrices that consider non-Gaussian noise interference, comprehensively considering factors affecting the accuracy of health diagnosis of the inertial navigation system. Using the first and second matrices that consider non-Gaussian noise interference to perform health diagnosis of the inertial navigation system, i.e., determining whether the inertial navigation system is abnormal, improves the accuracy of health diagnosis of the inertial navigation system. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a flowchart illustrating the health diagnosis method for an inertial navigation system provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. It should be noted that, unless otherwise specified, the implementation methods and features in the implementation methods in this disclosure can be combined, separated, interchanged, and / or rearranged. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] The terminology used herein is for the purpose of describing particular embodiments and is not intended to be limiting. As used herein, unless the context clearly indicates otherwise, the singular forms “a” and “the” are intended to include the plural forms as well. Furthermore, when the terms “comprising” and / or “including” and variations thereof are used in this specification, it indicates the presence of the stated features, integrals, steps, operations, parts, components, and / or groups thereof, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, parts, components, and / or groups thereof. It should also be noted that, as used herein, the terms “substantially,” “about,” and other similar terms are used as approximate terms rather than as terms of degree, thus explaining the inherent biases in measurements, calculated values, and / or provided values ​​that would be recognized by one of ordinary skill in the art.

[0019] refer to Figure 1 The diagram illustrates a flowchart of a health diagnosis method for an inertial navigation system provided in an embodiment of this application.

[0020] The health diagnosis method for an inertial navigation system provided in this application includes steps S101-S105.

[0021] In step S101, the phase angle offset of each time signal item in the one-dimensional time signal of the inertial navigation system under noise interference is determined;

[0022] In step S102, generating a first matrix includes: generating each element of the first matrix, wherein the target element of the first matrix is ​​any element of the first matrix. Generating the target element of the first matrix includes: generating the target element of the first matrix based on the sum of the phase angles of the two first time-series signal terms corresponding to the target element of the first matrix, the phase angle offset of the two first time-series signal terms corresponding to the target element of the first matrix under noise interference, and the penalty term corresponding to the target element of the first matrix.

[0023] In step S103, generating the second matrix includes: generating each element of the second matrix, wherein the target second matrix element is any element of the second matrix. Generating the target second matrix element includes: generating the target second matrix element based on the difference in phase angles of the two second time-series signal terms corresponding to the target second matrix element, the phase angle offset of the two second time-series signal terms corresponding to the target second matrix element under noise interference, and the penalty term corresponding to the target second matrix element.

[0024] In step S104, a first image corresponding to the first matrix and a second image corresponding to the second matrix are generated;

[0025] In step S105, the health diagnosis result of the inertial navigation system is obtained based on the first image and the second image.

[0026] In the embodiments of this application, the signal item in the one-dimensional time-series signal is called a time-series signal item.

[0027] In this embodiment, each timing signal item in the one-dimensional timing signal of the inertial navigation system has a different acquisition time. All timing signal items in the one-dimensional timing signal of the inertial navigation system can be arranged in order from front to back according to the acquisition time.

[0028] The matrix elements in the first matrix are called first matrix elements, and the matrix elements in the second matrix are called second matrix elements.

[0029] The timing signal item corresponding to the first matrix element in the first matrix is ​​called the first timing signal item, and the timing signal item corresponding to the second matrix element in the second matrix is ​​called the second timing signal item.

[0030] The timing signal term corresponding to each element of the first matrix in the first matrix is ​​the timing signal term in the one-dimensional timing signal. The timing signal term corresponding to each element of the second matrix in the second matrix is ​​the timing signal term in the one-dimensional timing signal.

[0031] The degree of difference between the two first time-series signal terms corresponding to the first matrix element is called the first degree of difference. The degree of difference between the two second time-series signal terms corresponding to the second matrix element is called the second degree of difference.

[0032] Before step S101, the phase angle and radius of each timing signal item in the one-dimensional timing signal of the inertial navigation system are determined.

[0033] Determining the phase angle and radius of information items in a timing signal is an existing technique. The following is a brief explanation of the process for determining the phase angle and radius of each timing signal item in a one-dimensional timing signal of an inertial navigation system:

[0034] One-dimensional time-series signal ,in n The total number of time points represents the number of time sequence items included in a one-dimensional time sequence signal.

[0035] First, the one-dimensional time-series signal in the rectangular coordinate system is expressed using formula (1). X ={ x 1, x 2…, x n The data is scaled to [–1, 1] to obtain a scaled one-dimensional time series signal:

[0036] (1)

[0037] The scaled one-dimensional time series signal includes: the scaled value of each time series signal item in the one-dimensional time series signal. Let i represent the time sequence term i in a one-dimensional time sequence signal. This represents the scaled value of timing signal term i;

[0038] Wherein, the timing signal term i can be any one of the one-dimensional timing signals of the inertial navigation system.

[0039] Timing signal terms;

[0040] Formula (2) is used to transform the scaled one-dimensional time-series signal from the Cartesian coordinate system to the polar coordinate system, thereby obtaining the phase angle and radius of each time-series signal term in the one-dimensional time-series signal.

[0041] (2)

[0042] in, It is the timestamp of timing signal item i. N It is a constant. It is the phase angle. r It is the radius of the timing signal term i. The variable is monotonically decreasing within the interval [0, π], therefore, the above mapping relationship is a one-to-one correspondence.

[0043] In one possible implementation, the timing signal terms in the one-dimensional timing signal are the voltages or currents at the corresponding positions of the inertial navigation system.

[0044] The corresponding positions are preset.

[0045] The voltage and / or current at the corresponding position of the inertial navigation system are obtained by sampling the voltage signal and current signal at the corresponding position.

[0046] In step S104, the Gramian Angular Field (GAF) algorithm can be used to convert the Gramian Angular Field matrix corresponding to the time-series signal, which includes multiple signal items arranged in sequence, into an image. This converts the first matrix into a first image, thereby generating the first image corresponding to the first matrix. Similarly, the Gramian Angular Field algorithm can be used to convert the second matrix into a second image, also corresponding to the second matrix. This process involves converting the corresponding matrix elements in the Gramian Angular Field matrix into the pixel values ​​of the corresponding pixels in the image.

[0047] In one possible implementation, when the health diagnosis result of the inertial navigation system indicates that the inertial navigation system is abnormal, the inertial navigation system has at least one of the following abnormalities: abnormal on-state resistance, abnormal off-state resistance, and on / off delay.

[0048] Prolonged and frequent flights accelerate the aging of components in inertial navigation systems, altering characteristics such as resistance and delay. These changes may not cause malfunctions in the short term, but over time they can lead to irreversible failures, i.e., structural hard failures.

[0049] In this embodiment, for the problem that the voltage or current waveform may be abnormal due to defective states that may occur in the actual operation of the inertial navigation device, the abnormal voltage and current signals can be graphically represented and trained into a neural network, so as to realize the rapid diagnosis of abnormal voltage and current signals of the inertial navigation system.

[0050] In step S105, the first image and the second image are input into a neural network for obtaining health diagnosis results, and the neural network for obtaining health diagnosis results outputs the health diagnosis results of the inertial navigation system.

[0051] The neural network outputting the health diagnosis results of the inertial navigation system indicates whether the inertial navigation system is abnormal. If the health diagnosis results indicate that the inertial navigation system is abnormal, what types of abnormalities are included in the health diagnosis results?

[0052] In the embodiments of this application, the neural network used to obtain health diagnosis results can be any type of neural network capable of performing classification tasks.

[0053] The neural network used to obtain health diagnosis results predicts the probability of each anomaly in the inertial navigation system based on the first and second images.

[0054] As an example, if the probability of an inertial navigation system having such an anomaly is greater than a probability threshold, it can be determined that the inertial navigation system has such an anomaly.

[0055] If an inertial navigation system has at least one anomaly, then the inertial navigation system is determined to be abnormal. The health diagnosis result of the inertial navigation system indicates that the inertial navigation system is abnormal.

[0056] In one possible implementation, the neural network used to obtain the health diagnosis result includes a first convolutional neural network and a second convolutional neural network. In step S105, obtaining the health diagnosis result of the inertial navigation system based on the first image and the second image includes: inputting the first image into the first convolutional neural network to obtain a first feature output by the first convolutional neural network, and inputting the second image into the second convolutional neural network to obtain a second feature output by the second convolutional neural network; concatenating the first feature and the second feature to obtain a fused feature; obtaining the probability of the inertial navigation system having each type of abnormality based on the fused feature; and obtaining the health diagnosis result of the inertial navigation system based on the probability of the inertial navigation system having each type of abnormality.

[0057] As an example, the fused features are input into a fully connected layer in a neural network used to obtain health diagnosis results. The features output by the fully connected layer are then input into a Softmax classifier in the same neural network. The Softmax classifier outputs the probability of each abnormality in the inertial navigation system.

[0058] In this embodiment of the application, a neural network for obtaining health diagnosis results is trained in advance using a training dataset.

[0059] The labels on the training data indicate whether the inertial navigation system is abnormal. If the labels on the training data indicate that the inertial navigation system is abnormal, the labels on the training data also indicate a type of abnormality that the inertial navigation system possesses.

[0060] The labels of training data used as negative samples indicate anomalies in the inertial navigation system, and the labels of training data used as negative samples also indicate anomalies in the inertial navigation system.

[0061] The training data serving as negative samples consists of the third and fourth images representing the time-series signal under the anomaly. This time-series signal is either a voltage sequence or a current sequence under the anomaly, and the type of signal terms in the time-series signal under the anomaly is the same as the type of signal terms in a one-dimensional time-series signal sequence. The voltage sequence under the anomaly includes multiple voltages used for training, ordered sequentially from the time the voltage was acquired. The current sequence under the anomaly includes multiple currents used for training, ordered sequentially from the time the current was acquired.

[0062] For an anomaly, the voltage or current used for training under that anomaly is the voltage or current at the corresponding position of the inertial navigation system when the anomaly occurs.

[0063] In this embodiment, a simulation model can be established, which includes each electronic component in the inertial navigation system and the connection relationships between them. The simulation model can be used to simulate anomalies in the inertial navigation system. Simultaneously, the simulation model is used to calculate the voltage or current at each moment of the corresponding position of the inertial navigation system under a specific anomaly. These voltages or currents at each moment of the corresponding position under the specific anomaly are used as multiple training voltages or currents under that specific anomaly, resulting in a timing signal including multiple training voltages or currents under that specific anomaly.

[0064] As an example, a simulation model is used to repeatedly change the values ​​of the series and parallel resistors and the parallel capacitor in the inertial navigation system to simulate an anomaly in the on-state resistance. The simulation model is then used to calculate the voltage or current at each moment of the corresponding position of the inertial navigation system under the on-state resistance anomaly, obtaining timing signals for multiple voltages or currents used for training under the on-state resistance anomaly. Similarly, the simulation model is used to repeatedly change the values ​​of the series and parallel resistors and the parallel capacitor in the inertial navigation system to simulate an anomaly in the off-state resistance. The simulation model is used to calculate the voltage or current at each moment of the corresponding position of the inertial navigation system under the off-state resistance anomaly, obtaining timing signals for multiple voltages or currents used for training under the off-state resistance anomaly. Finally, the simulation model is used to repeatedly change the values ​​of the series and parallel resistors and the parallel capacitor in the inertial navigation system to simulate an anomaly in the on-off delay. The voltage or current of the inertial navigation system at each time point in multiple moments is calculated using a simulation model when the inertial navigation system has an on / off delay anomaly. This yields timing signals of multiple voltages or currents used for training under the on / off delay anomaly.

[0065] The training data used as positive samples consists of the third and fourth images corresponding to the time-series signals under normal conditions. The time-series signals under normal conditions are voltage or current sequences, which include multiple voltage or current samples used for training, ordered sequentially from the time the voltage or current was acquired. The voltage or current used for training under normal conditions represents the voltage or current at the corresponding position of the inertial navigation system when it is functioning normally.

[0066] The labels of the training data, which serve as positive samples, indicate that the inertial navigation system is normal, i.e., not abnormal.

[0067] During the training of the neural network used to obtain health diagnosis results, training data is input into the neural network, and the network outputs a prediction result corresponding to the training data. This prediction result includes the predicted probability of the inertial navigation system having each type of anomaly. Based on the prediction result and the label of the training data, it is determined whether to generate a loss corresponding to the training data. If it is determined to generate a loss corresponding to the training data, then the loss is generated and used to update the parameters of the neural network used to obtain health diagnosis results.

[0068] In one possible implementation, steps S106 and S107 are also included.

[0069] In step S106, a difference measure term is generated for the two first time-series signal terms corresponding to the first element of the target matrix. Generating the difference measure term for the first time-series signal terms corresponding to the first element of the target matrix includes: based on the phase angle and radius of the first time-series signal term. ， Generate a difference measure for the first time-series signal item of the target, wherein the first time-series signal item of the target is any first time-series signal item corresponding to the first matrix element of the target; generate a penalty term corresponding to the first matrix element of the target based on the first difference measure between the two first time-series signal items corresponding to the first matrix element of the target;

[0070] In step S107, a difference measure term is generated for the two second time-series signal terms corresponding to the target second matrix element. Generating the difference measure term for the target second time-series signal term corresponding to the target second matrix element includes: based on the phase angle and radius of the target second time-series signal term. ， Generate a difference measure term for the target second time-series signal item, wherein the target second time-series signal item is any second time-series signal item corresponding to the target second matrix element; generate a penalty term corresponding to the target second matrix element based on the second difference measure term between the two second time-series signal items corresponding to the target second matrix element.

[0071] The two timing signal terms corresponding to the matrix element in the j-th row and k-th column of matrix i are: the j-th timing signal term and the k-th timing signal term in the one-dimensional timing signal of the inertial navigation system.

[0072] Where matrix i is either the first matrix or the second matrix.

[0073] For the element in the j-th row and k-th column of matrix i, if j equals k, then the two timing signal terms corresponding to the element in the j-th row and k-th column of matrix i are the same. If j does not equal k, then each of the two timing signal terms corresponding to the element in the j-th row and k-th column of matrix i has a different acquisition time, and the two timing signal terms corresponding to the element in the j-th row and k-th column of matrix i are different.

[0074] In one possible implementation, the difference measure term z of the timing signal term i i It can be represented as:

[0075]

[0076] Wherein, the timing signal term i can be any one of the one-dimensional timing signals of the inertial navigation system.

[0077] Timing signal items, This represents the phase angle of timing signal term i. Represents the radius of the timing signal term i.

[0078] In one possible implementation, the degree of difference between the two time-series signal terms corresponding to the matrix element in the j-th row and k-th column of matrix i is: the Kullback-Leibler divergence between the two time-series signal terms corresponding to the matrix element in the j-th row and k-th column of matrix i, z j z represents the measure of the degree of difference of the j-th time-series signal item. k The Kullback-Leibler divergence between the difference measure terms of the two time series signal terms corresponding to the matrix element in the j-th row and k-th column of matrix i is denoted as . .

[0079] In this embodiment, the first matrix element in the j-th row and k-th column of the first matrix can be: the product of cos(the sum of the phase angles of the two first timing signal terms corresponding to the first matrix element in the j-th row and k-th column of the first matrix + the sum of the phase angle offsets of the two first timing signal terms corresponding to the first matrix element in the j-th row and k-th column of the first matrix under noise interference) and the penalty term corresponding to the first matrix element in the j-th row and k-th column of the first matrix.

[0080] Where cos represents the cosine function.

[0081] In one possible implementation, the penalty term corresponding to the first matrix element in the j-th row and k-th column of the first matrix can be: e raised to the power of -x1, where x1 is the first difference between the two first time-series signal terms corresponding to the first matrix element in the j-th row and k-th column of the first matrix.

[0082] As an example, the degree of difference between the two first time-series signal terms corresponding to the first matrix element in the j-th row and k-th column of the first matrix can be: the Kullback-Leibler divergence between the two first time-series signal terms corresponding to the first matrix element in the j-th row and k-th column of the first matrix.

[0083] In this embodiment, the matrix element in the j-th row and k-th column of the second matrix can be the product of sin(the difference in phase angles of the two second timing signal terms corresponding to the second matrix element in the j-th row and k-th column of the second matrix + the difference in phase angle offsets of the two second timing signal terms corresponding to the matrix element in the j-th row and k-th column of the first matrix under noise interference) and the penalty term corresponding to the matrix element in the j-th row and k-th column of the second matrix.

[0084] Here, sin represents the sine function.

[0085] In one possible implementation, the penalty term corresponding to the second matrix element in the j-th row and k-th column of the second matrix can be: e raised to the power of -x2, where x2 is the second degree of difference between the two second time-series signal terms corresponding to the matrix element in the j-th row and k-th column of the second matrix.

[0086] As an example, the degree of difference between the two second time series signal terms corresponding to the second matrix element in the j-th row and k-th column of the second matrix can be: the Kullback-Leibler divergence between the two second time series signal terms corresponding to the matrix element in the j-th row and k-th column of the second matrix.

[0087] In one possible implementation, in step S106, generating a penalty term corresponding to the target first matrix element based on the first difference degree between the two first time-series signal terms corresponding to the target first matrix element includes: generating a penalty term corresponding to the target first matrix element based on the first difference degree and the penalty intensity.

[0088] In one possible implementation, in step S107, generating a penalty term corresponding to the target second matrix element based on the second difference degree between the two second time-series signal terms corresponding to the target second matrix element includes: generating a penalty term corresponding to the target second matrix element based on the second difference degree and the penalty intensity.

[0089] In one possible implementation, the penalty term corresponding to the first matrix element in the j-th row and k-th column of the first matrix can be: (e^(-x³)), where x³ is the first difference measure between the two first time-series signal terms corresponding to the first matrix element in the j-th row and k-th column of the first matrix. α The product of α The severity of the punishment.

[0090] In one possible implementation, the penalty term corresponding to the second matrix element in the j-th row and k-th column of the second matrix can be: (e^(-x4)), where x4 is the second difference measure between the two second time-series signal terms corresponding to the second matrix element in the j-th row and k-th column of the second matrix. α The product of α The severity of the punishment.

[0091] In this embodiment, the following considerations are taken into account: Inertial navigation systems exhibit long-term drift characteristics, making it unsuitable to use Gram matrices constructed solely using Euclidean inner products of focused local dynamics, as employed in related technologies, to predict anomalies. Therefore, this embodiment introduces a penalty term to address the long-term drift characteristics of the inertial navigation system. Furthermore, this embodiment considers non-Gaussian noise interference that can cause deviations between the acquired voltage or current and the actual voltage or current. Therefore, an interference term is applied to the phase angle of the timing signal term.

[0092] In one possible implementation, the first matrix is:

[0093]

[0094] in, The phase angle of the i-th timing signal term Phase angle offset under noise interference. H Indicates the Kullback-Leibler divergence. α To determine the severity of the punishment, α The larger the value, the more the amplitude difference is penalized.

[0095] The matrix element in the first row and first column of the first matrix is In the first matrix, the two first time-series signal terms corresponding to the matrix element in the first row and first column are the same. Each of the two first time-series signal terms corresponding to the matrix element in the first row and first column of the first matrix is ​​the first time-series signal term in the one-dimensional time-series signal. The phase angle of the first timing signal term. This represents the phase angle offset of the first timing signal term under noise interference. This is the penalty term corresponding to the first matrix element in the first row and first column of the first matrix. In H (z1||z1) is the Kullback-Leibler divergence between the two first time-series signal terms corresponding to the first matrix element in the first row and first column of the first matrix.

[0096] The matrix element in the 2nd row and 1st column of the first matrix is The two first time-series signal terms corresponding to the matrix element in the second row and first column of the first matrix are different. The two first time-series signal terms corresponding to the matrix element in the second row and first column of the first matrix are the first and second time-series signal terms in this one-dimensional time-series signal. The phase angle of the first timing signal term. The phase angle of the second timing signal term. This represents the phase angle offset of the first timing signal term under noise interference. This represents the phase angle offset of the second timing signal term under noise interference. This is the penalty term corresponding to the first matrix element in the second row and first column of the first matrix. H (z2||z1) is the Kullback-Leibler divergence between the two first time-series signal terms corresponding to the first matrix element in the second row and first column of the first matrix.

[0097] In one possible implementation, the second matrix is:

[0098]

[0099] The element of the second matrix in the first row and first column is: The two second time-series signal terms corresponding to the second matrix element in the first row and first column of the second matrix are the same, and each of the two second time-series signal terms corresponding to the second matrix element in the first row and first column of the second matrix is ​​the first time-series signal term in the one-dimensional time-series signal.

[0100] The second matrix element in the second row and first column of the second matrix is: The two second time-series signal terms corresponding to the matrix element in the second row and first column of the second matrix are different. The two second time-series signal terms corresponding to the matrix element in the second row and first column of the second matrix are the first time-series signal term and the second time-series signal term in the one-dimensional time-series signal.

[0101] In one possible implementation, in step S101, determining the phase angle offset of the target timing signal item in the one-dimensional timing signal of the inertial navigation system under noise interference includes: determining the phase angle offset of the target timing signal item under noise interference as the phase angle offset of the target timing signal item under noise interference according to a pre-determined correspondence between the phase angle of the timing signal item and the phase angle offset of the timing signal item under noise interference, wherein the target timing signal item is any one of the timing signal items in the one-dimensional timing signal.

[0102] In one possible implementation, in step S101, determining the phase angle offset of the target timing signal item in the one-dimensional timing signal of the inertial navigation system under noise interference includes: sampling the phase angle offset of the target timing signal item under noise interference from a predetermined distribution of phase angle offsets under noise interference.

[0103] In one possible implementation, the method further includes: determining the true probability density function of the Gaussian distribution that the noise sampled in the inertial navigation system conforms to; constructing an optimization function to measure the difference between the true probability density function and the Gaussian probability density function, wherein the Gaussian probability density function is a weighted sum of the probability density functions of multiple Gaussian distributions; solving for the Gaussian probability density function with the objective of minimizing the function value of the optimization function; and determining the phase angle offset of the target time-series signal term in the one-dimensional time-series signal of the inertial navigation system under noise interference by: generating random numbers based on the Gaussian probability density function, and using the generated random numbers as the phase angle offset of the target time-series signal term under noise interference, wherein the target time-series signal term is any one of the time-series signal terms in the one-dimensional time-series signal. Alternatively, it can be described as sampling random numbers from the Gaussian probability density function as the phase angle offset of the target time-series signal term under noise interference.

[0104] In this method, random numbers are generated based on the Gaussian sum probability density function. These random numbers are then used as the phase angle offset of the target time-series signal term under noise interference. It is considered that the normal distribution is commonly used to describe Gaussian noise but cannot be used for non-Gaussian noise. Signal noise generated in inertial navigation systems is typically non-Gaussian noise. Related technologies cannot analytically provide propagation models for uncertainties in nonlinear, non-Gaussian systems; they can only provide approximate solutions. Therefore, in this embodiment, based on Bayesian estimation theory, a "Gaussian sum" state estimation method is proposed to approximate non-Gaussian noise by continuously approximating the state probability density function. According to Gaussian sum theory, any probability density function can be approximated as a weighted sum of multiple Gaussian noise terms, and the approximation error can be made arbitrarily small by increasing the number of Gaussian noise terms.

[0105] In this embodiment, Gaussian and probability density functions are used to generate the phase angle offset of the time sequence signal term in a one-dimensional time sequence signal under noise interference more accurately, thereby improving the accuracy of the generated matrix reflecting the characteristics of the signal term.

[0106] Gaussian and probability density functions can be expressed as:

[0107]

[0108] The optimization function can be expressed as:

[0109]

[0110] The optimization objective, which is to minimize the function value of the optimization function, can be expressed as:

[0111]

[0112] in, Let M be the approximate probability density function of the Gaussian sum, i.e., the Gaussian sum probability density function, where M is the number of terms in the Gaussian sum. To meet The weighting coefficients. It is a Gaussian probability density function with a mean. s ,variance e . This is the true probability density function, i.e., the true probability density function of the Gaussian distribution that the noise sampled in the inertial navigation system follows. This represents the optimization index, which is the function value of the optimization function.

[0113] In the embodiments of this application, any algorithm for optimizing multiple variables under a given optimization objective, such as stochastic gradient descent or the Grey Wolf algorithm, can be used to minimize the function value of the optimization function and solve for the Gaussian and probability density functions.

[0114] In one possible implementation, multiple Gaussian distributions are converted into four Gaussian distributions.

[0115] In one possible implementation, the probability density functions of the four Gaussian distributions all have a mean of zero, and the variances of the probability density functions of the four Gaussian distributions are 0.03, 0.06, 0.09, and 0.12, respectively.

[0116] In one possible implementation, the weights of the probability density functions of the four Gaussian distributions are all 0.25.

[0117] In this embodiment of the application, considering the operating environment of the inertial navigation element under low-altitude flight conditions, the following is taken: M =4. The four types of Gaussian distributions are independent, with all having a mean of zero and a variance of 0. e 1 = 0.03, e 2 = 0.06, e 3 = 0.09, e 4 = 0.12, weight β 1= β 2= β 3= β 4 = 0.25.

[0118] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0119] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0120] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A health diagnosis method for an inertial navigation system, characterized in that: The method includes: Determine the phase angle offset of each time signal item in the one-dimensional time signal of the inertial navigation system under noise interference; Generating a first matrix includes: generating each element of the first matrix, wherein the target element of the first matrix is ​​any element of the first matrix. Generating the target element of the first matrix includes: generating the target element of the first matrix based on the sum of the phase angles of the two first time-series signal terms corresponding to the target element of the first matrix, the phase angle offset of the two first time-series signal terms corresponding to the target element of the first matrix under noise interference, and the penalty term corresponding to the target element of the first matrix. Generating the second matrix includes: generating each element of the second matrix, wherein the target second matrix element is any element of the second matrix. Generating the target second matrix element includes: generating the target second matrix element based on the difference in phase angles of the two second time-series signal terms corresponding to the target second matrix element, the phase angle offset of the two second time-series signal terms corresponding to the target second matrix element under noise interference, and the penalty term corresponding to the target second matrix element. Generate a first image corresponding to the first matrix and a second image corresponding to the second matrix; Based on the first image and the second image, the health diagnosis results of the inertial navigation system are obtained.

2. The method according to claim 1, characterized in that: The method further includes: Generate a difference measure for the two first time-series signal terms corresponding to the first element of the target matrix, wherein generating the difference measure for the first time-series signal terms corresponding to the first element of the target matrix includes: based on the phase angle and radius of the first time-series signal terms. ， Generate a difference measure for the first time-series signal item of the target, wherein the first time-series signal item of the target is any one of the first time-series signal items corresponding to the elements of the first matrix of the target; The method further includes generating a penalty term corresponding to the first matrix element based on the first difference degree between the two first time-series signal terms corresponding to the first matrix element; and the method further includes: Generate a difference measure for the two second time-series signal terms corresponding to the target second matrix element, wherein generating the difference measure for the target second time-series signal terms corresponding to the target second matrix element includes: based on the phase angle and radius of the target second time-series signal terms. ， Generate a difference measure for the target second time-series signal item, wherein the target second time-series signal item is any second time-series signal item corresponding to the element of the target second matrix; Based on the second degree of difference between the two second time-series signal terms corresponding to the second matrix element of the target, a penalty term corresponding to the second matrix element of the target is generated.

3. The method according to claim 2, characterized in that: Based on the first difference degree between the two first time-series signal terms corresponding to the first element of the target matrix, the penalty term corresponding to the first element of the target matrix is ​​generated, including: Based on the first degree of difference and the penalty intensity, a penalty term corresponding to the target first matrix element is generated; and based on the second degree of difference between the two second time-series signal terms corresponding to the target second matrix element, a penalty term corresponding to the target second matrix element is generated, including: Based on the second degree of difference and the penalty intensity, the penalty term corresponding to the second matrix element of the target is generated.

4. The method according to claim 1, characterized in that: The method further includes: Determine the true probability density function of the Gaussian distribution that the noise sampled in the inertial navigation system conforms to; Construct an optimization function to measure the difference between the true probability density function and the Gaussian probability density function, where the Gaussian probability density function is a weighted sum of the probability density functions of multiple Gaussian distributions; With the objective of minimizing the function value of the optimization function, the Gaussian and probability density functions are solved; and The phase angle offset of the target time sequence signal term in the one-dimensional time sequence signal of the inertial navigation system under noise interference is determined by: Based on the Gaussian and probability density functions, random numbers are generated, and the generated random numbers are used as the phase angle offset of the target time-series signal term under noise interference. The target time-series signal term is any one of the time-series signal terms in the one-dimensional time-series signal.

5. The method according to claim 4, characterized in that: The multiple Gaussian distributions are four Gaussian distributions.

6. The method according to claim 5, characterized in that: The mean of the probability density functions of the four Gaussian distributions is zero, and the variances of the probability density functions of the four Gaussian distributions are 0.03, 0.06, 0.09, and 0.12, respectively.

7. The method according to claim 5, characterized in that: The weights of the probability density functions of the four Gaussian distributions are all 0.

25.

8. The method according to claim 1, characterized in that: The timing signal item in the one-dimensional timing signal is the voltage or current at the corresponding position of the inertial navigation system.

9. The method according to claim 1, characterized in that: When the health diagnosis results of the inertial navigation system indicate that the inertial navigation system is abnormal, the inertial navigation system has at least one of the following abnormalities: abnormal on-state resistance, abnormal off-state resistance, and on / off delay.

10. The method according to claim 1, characterized in that: Based on the first image and the second image, the health diagnosis results of the inertial navigation system include: The first image is input into a first convolutional neural network to obtain a first feature, and the second image is input into a second convolutional neural network to obtain a second feature; The first feature and the second feature are concatenated to obtain the fused feature; Based on the fusion features, the probability of each anomaly in the inertial navigation system is obtained; Based on the probability of each type of anomaly in the inertial navigation system, the health diagnosis result of the inertial navigation system is obtained.

Images