IMU fault diagnosis and repair method, device, vehicle-mounted navigation equipment and storage medium

Abstract

The present invention discloses an IMU fault diagnosis and repair method, device, vehicle-mounted navigation equipment and storage medium. The IMU fault diagnosis and repair method includes: obtaining IMU time domain data, analyzing and processing to obtain initial measurement values; periodically analyzing the initial measurement values, calculating and statistically obtaining missing measurement points; diagnosing and analyzing the missing measurement points to obtain fault feature descriptions; establishing a fault data mapping relationship, and obtaining fault judgment parameters through feature decomposition; using the fault judgment parameters, calculating fault classification values; measuring data accuracy based on the fault classification values, and obtaining fault degree values ​​through threshold dynamic comparison; using the fault degree values, calculating correction parameters, and obtaining adjustment sequences through parameter conversion; performing extreme value decomposition, and obtaining steady-state quantities through waveform recombination; inputting quantized equalization equations, and frequency domain modulation to obtain the final repair results. The technical solution of the present invention improves the accuracy of fault diagnosis and the comprehensiveness of repair, and can accurately diagnose and repair faults under complex working conditions.

Description

Technical Field

[0001] The present invention relates to the field of vehicle-mounted navigation technology, and in particular to an IMU fault diagnosis and repair method, device, vehicle-mounted navigation equipment and storage medium. Background Art

[0002] IMU (Inertial Measurement Unit) sensors are widely used in intelligent driving, robot navigation, aerospace and other fields. However, in actual working process, IMU sensors are easily affected by factors such as temperature change, mechanical vibration, electromagnetic interference, etc., which may cause output data drift, mutation or failure. These faults not only affect the measurement accuracy, but also may cause the performance of the entire system to degrade. For example, in the vehicle navigation system, the abnormality of IMU sensor data will cause positioning deviation, which will seriously affect the navigation accuracy.

[0003] At present, the main methods used to diagnose and repair IMU sensor failures are Kalman filtering, threshold judgment, and other processing methods. Among them, the Kalman filtering method estimates and corrects sensor data by establishing a state space model, but it is highly dependent on the initial state and noise model; although the threshold judgment method is simple to implement, its fixed threshold setting makes it difficult to adapt to complex working conditions. Therefore, these methods generally have problems such as low fault diagnosis accuracy and single repair strategy in practical applications, and it is difficult to meet the working requirements of IMU sensors under complex working conditions. Summary of the invention

[0004] The present invention provides an IMU fault diagnosis and repair method, aiming to improve the fault diagnosis accuracy and repair comprehensiveness of IMU sensors, and to achieve accurate fault diagnosis and repair under complex working conditions.

[0005] To achieve the above object, the present invention proposes an IMU fault diagnosis and repair method, which is applied to a vehicle-mounted navigation device, comprising:

[0006] Acquire time domain data of the IMU sensor, sample and analyze the time domain data according to preset rules, and obtain initial measurement values ​​through integrity check and data preprocessing;

[0007] Perform periodic analysis on the initial measurement values, establish a reference benchmark based on the sensor sampling characteristics, and obtain complete information of the missing measurement points through benchmark comparison and regularity analysis;

[0008] The complete information of the missing measurement point is analyzed using bottom-level hardware diagnosis, middle-level data diagnosis and top-level system diagnosis to obtain a fault feature description;

[0009] According to the fault feature description, a fault data mapping relationship is established, and a fault judgment parameter is obtained through feature decomposition;

[0010] Using the fault judgment parameters, constructing a fault classification evaluation criterion, and calculating a fault classification value through the fault classification evaluation criterion;

[0011] Based on the fault classification value, data accuracy measurement is performed, and a fault degree value is obtained by dynamic comparison of threshold values;

[0012] Using the fault degree value, calculating the correction parameter, and obtaining the adjustment sequence through parameter conversion;

[0013] Performing extreme value decomposition on the regulation sequence, performing cross-spectral analysis on the decomposition result to eliminate the influence of noise, and obtaining the steady-state quantity by waveform reorganization;

[0014] The steady-state quantity is input into the quantized equalization equation, and the compensation factor is constructed in combination with the transient response characteristics of the sensor, and the final repair result is obtained through frequency domain modulation.

[0015] The present invention also proposes an IMU fault diagnosis and repair device, which is applied to a vehicle-mounted navigation device, and comprises:

[0016] An acquisition module is used to acquire time domain data of the IMU sensor, sample and analyze the time domain data according to preset rules, and obtain initial measurement values ​​through integrity check and data preprocessing;

[0017] A statistical module, used to perform periodic analysis on the initial measurement values, establish a reference benchmark according to the sensor sampling characteristics, and obtain complete information of missing measurement points through benchmark comparison and regularity analysis;

[0018] A diagnosis module, used for using bottom-level hardware diagnosis, middle-level data diagnosis and top-level system diagnosis on the complete information of the missing measurement point, and obtaining a fault feature description through diagnostic analysis;

[0019] A decomposition module, used to establish a fault data mapping relationship according to the fault feature description, and obtain a fault judgment parameter through feature decomposition;

[0020] A calculation module, used to construct a fault classification evaluation criterion by using the fault judgment parameter, and calculate a fault classification value by using the fault classification evaluation criterion;

[0021] A comparison module, used to measure data accuracy based on the fault classification value, and obtain a fault degree value through dynamic comparison of a threshold value;

[0022] A correction module, used to calculate correction parameters using the fault degree value, and obtain an adjustment sequence through parameter conversion;

[0023] A reorganization module is used to perform extreme value decomposition on the regulation sequence, perform cross-spectral analysis on the decomposition result to eliminate the influence of noise, and obtain a steady-state quantity by waveform reorganization;

[0024] The modulation module is used to input the steady-state quantity into the quantized equalization equation, construct a compensation factor in combination with the transient response characteristics of the sensor, and obtain the final repair result through frequency domain modulation.

[0025] The present invention also proposes a vehicle-mounted navigation device, including an IMU sensor, characterized in that the vehicle-mounted navigation device also includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program implements the steps of the above-mentioned IMU fault diagnosis and repair method when executed by the processor.

[0026] The present invention also proposes a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned IMU fault diagnosis and repair method are implemented.

[0027] The technical solution of the IMU fault diagnosis and repair method of the present invention solves the problem of incomplete fault feature extraction in traditional methods by establishing an adaptive data preprocessing and hierarchical diagnosis mechanism; the mechanism first performs dynamic sampling and analysis on the time domain data, then accurately calculates the lost data through periodic characteristics, and finally adopts a hierarchical diagnosis strategy to analyze layer by layer from hardware to system, breaking through the limitations of traditional single-level diagnosis, and can accurately identify and locate various types of faults under complex working conditions. Secondly, an innovative solution based on fault data mapping and quantitative evaluation is proposed to overcome the defect of fuzzy fault judgment; the solution achieves accurate classification and degree quantification of faults by establishing a mapping relationship between fault characteristics and judgment parameters, combined with dynamic threshold comparison technology, avoiding the subjectivity of traditional empirical judgment, and significantly improving the accuracy of fault diagnosis. In addition, an innovative repair strategy based on extreme value decomposition and frequency domain modulation is designed to solve the problem of poor adaptability of traditional repair methods. The strategy first obtains the adjustment sequence through parameter conversion, then uses extreme value decomposition to eliminate the influence of noise, and finally uses quantitative equalization to achieve accurate modulation in the frequency domain. This multi-level repair method breaks through the limitations of traditional single filtering or threshold correction, and can adaptively adjust the repair strategy according to fault characteristics, achieving more accurate data repair. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of a flow chart of an embodiment of an IMU fault diagnosis and repair method of the present invention;

[0029] Figure 2 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S20 in the embodiment;

[0030] Figure 3 for Figure 2 A schematic diagram of a flow chart of an embodiment of step S30 in the embodiment;

[0031] Figure 4 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S40 in the embodiment;

[0032] Figure 5 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S50 in the embodiment;

[0033] Figure 6 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S60 in the embodiment;

[0034] Figure 7 for Figure 1 A schematic diagram of a flow chart of an embodiment of step S70 in the embodiment;

[0035] Figure 8 A schematic diagram of the module structure of an embodiment of an IMU fault diagnosis and repair device of the present invention;

[0036] Fig. 9 It is a schematic diagram of the structure of an electronic device in a hardware operating environment involved in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] The present invention proposes an IMU fault diagnosis and repair method, which is applied to a vehicle-mounted navigation device. The implementation terminal of the IMU fault diagnosis and repair method of the present invention can be a vehicle-mounted navigation device or an electronic device, and the electronic device can be a controller of the vehicle-mounted navigation device, or a computing device such as a single-chip microcomputer, a desktop computer, a notebook, a palm computer, and a server.

[0039] like Figure 1 As shown, Figure 1 It is a flow chart of the first embodiment of the IMU fault diagnosis and repair method of the present invention.

[0040] In this embodiment, the IMU fault diagnosis and repair method includes:

[0041] Step S10, acquiring the time domain data of the IMU sensor, sampling and analyzing the time domain data according to preset rules, and obtaining the initial measurement value through integrity check and data preprocessing.

[0042] The time domain data output by the IMU (Inertial Measurement Unit) sensor mainly includes three-axis acceleration and three-axis angular velocity, which are transmitted through the data acquisition channel in the form of discrete digital signals. In the data acquisition stage, it is necessary to first establish a stable data acquisition channel. According to the system requirements, you can choose communication methods such as serial communication, CAN bus or SPI interface. Taking the application of vehicle-mounted IMU as an example, the CAN bus is usually used as the data transmission channel. This method has strong anti-interference ability and high transmission reliability. A key parameter in the data acquisition process is the sampling frequency, which needs to be reasonably set according to the motion characteristics of the carrier. For vehicle motion status monitoring, the sampling frequency is usually set between 100Hz-200Hz, which can accurately capture the vehicle's motion changes and avoid generating too much redundant data.

[0043] The acquired raw data needs to be analyzed and processed according to the preset sampling rules. The sampling rules first involve timing control, requiring the system to be able to stably collect data according to the set sampling period. This requires a high-precision clock source to control the sampling moment and control the jitter of the sampling interval within an acceptable range. In practical applications, sampling can be triggered by a hardware timer to ensure the uniformity of the sampling process. Another important aspect of the sampling rule is the normalization of the data format. The raw data is usually output in the form of analog-to-digital conversion values, which need to be converted into standard engineering units according to the range and resolution parameters of the IMU sensor. For acceleration data, convert it to m / s² units; for angular velocity data, convert it to rad / s units. This conversion ensures the consistency of the data in subsequent processing.

[0044] After acquiring data according to the sampling rules, a data integrity check is required. Data frame errors or losses may occur during data transmission. These anomalies can be identified by analyzing the sequence number and checksum of the data frame. When a data frame error is detected, it needs to be marked and recorded immediately to avoid inputting erroneous data into subsequent processing links. For detected data loss, different processing strategies need to be adopted according to the length of time the loss occurred. If the data loss is short-term, data repair can be performed through methods such as linear interpolation; if the data loss is longer, the loss interval needs to be recorded and specially handled in subsequent processing.

[0045] After the integrity check, the data needs to be preprocessed to improve the data quality. The first step of preprocessing is outlier detection, which uses a statistical characteristic-based method to identify possible outliers. The specific approach is to calculate the statistical characteristics of the data within a sliding time window, including the mean and standard deviation, and mark the data points that significantly deviate from the statistical characteristics as potential outliers. Outlier detection needs to be performed separately for acceleration and angular velocity data because they have different physical properties and change patterns. For the marked outliers, it is necessary to combine the physical characteristics of the IMU sensor and the actual application scenario to distinguish between real abnormal measurements and instantaneous interference.

[0046] The next step of preprocessing is data denoising, which is crucial to improving measurement accuracy. IMU sensor data is inevitably affected by various noises, including thermal noise of the IMU sensor itself, quantization noise, and external environmental interference. Noise reduction processing uses digital filtering technology to design a suitable filter based on the characteristics of IMU sensor data. A low-pass filter is usually used to suppress high-frequency noise, and the cutoff frequency of the filter needs to be determined according to actual application requirements. In vehicle-mounted applications, considering the frequency characteristics of vehicle motion, the filter cutoff frequency is usually set in the range of 20-30Hz, which can effectively suppress high-frequency interference while retaining valid signals.

[0047] The final step in data preprocessing is to assess the quality of the processed data. Quality assessment is performed from multiple dimensions, including data continuity, stability, and noise level. Data continuity is assessed by analyzing the integrity of the sampling sequence, recording any data loss or anomalies. Data stability is assessed by calculating the statistical characteristics of the measured values, including short-term stability and long-term stability. Noise level assessment focuses on the residual noise after filtering, which is quantified by calculating indicators such as signal-to-noise ratio.

[0048] Through the above series of processing steps, the normalized initial measurement values ​​of the IMU sensor are finally obtained. These initial measurement values ​​have a unified format and unit, have undergone integrity verification, outlier processing and noise reduction, and are accompanied by corresponding quality assessment results. Each measurement value contains timestamp information, three-axis acceleration data, and three-axis angular velocity data.

[0049] Step S20, periodically analyzing the initial measurement values, establishing a reference benchmark according to the sensor sampling characteristics, and obtaining complete information of the missing measurement points through benchmark comparison and regularity analysis.

[0050] The purpose of period analysis is to determine the distribution characteristics of data in time series. This distribution characteristic includes two key aspects: the density distribution of data points on the time axis and the time interval regularity between data points.

[0051] Combination Figure 2As shown, in some embodiments, step S20 may include: step S21, step S22, step S23, step S24, step S25 and step S26. Among them:

[0052] Step S21, determining a basic analysis period according to the sampling frequency of the IMU sensor;

[0053] Step S22, segmenting the initial measurement value into a number of analysis units according to the basic analysis cycle, counting the number of measurement values ​​actually received in each analysis unit, and calculating the time interval between adjacent measurement values ​​in each analysis unit to determine the data abnormality interval;

[0054] The analysis must first determine the selection value of the basic analysis cycle, which is directly related to the sampling frequency of the IMU sensor. For example, when the IMU uses a 100Hz sampling frequency, 1 second can be selected as the basic analysis cycle, because each cycle contains exactly 100 complete measurement points. When performing periodic analysis, the continuous data stream (i.e., the initial measurement value) needs to be segmented according to a 1-second cycle to be divided into several analysis units. In each analysis unit, the number of actual received measurement values ​​is counted, and the time interval between adjacent measurement values ​​is calculated. This interval value should theoretically be kept stable at 10 milliseconds. By comparing the deviation between the actual interval value and the theoretical interval value, the data abnormality interval can be accurately identified.

[0055] Taking the vehicle-mounted IMU as an example, when performing periodic analysis on a certain section of driving data, 10 consecutive seconds of data were selected for analysis. In these 10 seconds of data, by checking the data of each second, it was found that there was an obvious abnormality between the 3rd and 5th seconds: the amount of data in the 3rd second was reduced to 70, and many time intervals exceeded the normal range by 15 milliseconds; the situation was more serious in the 4th second, with only 20 measured values, and some time intervals expanded to 50 milliseconds; the number of measured values ​​in the 5th second was 60, and the time intervals were unevenly distributed. This three-second abnormal phenomenon reflects that there are persistent problems in the data collection process.

[0056] Step S23, based on the sampling frequency of the IMU sensor, establish a corresponding sampling time sequence as a reference benchmark;

[0057] In order to accurately calculate the amount of lost data, a standard sampling moment sequence is first established as a reference benchmark. At a sampling frequency of 100Hz, a standard sampling moment is set every 10 milliseconds. For example, the standard sequence of the 4th second contains 100 time points from 4.000 seconds to 4.990 seconds. The sampling characteristics of the IMU sensor stipulate that at a sampling frequency of 100Hz, a complete set of measurement data must be generated every 10 milliseconds, including three-axis acceleration (ax, ay, az) and three-axis angular velocity (ωx, ωy, ωz). This sampling specification provides a standard for calculating the amount of lost data.

[0058] Step S24, compare the actually received measurement value with the standard time in the sampling time sequence, and for the data abnormal interval, calculate the difference between the standard sampling quantity and the actually obtained valid measurement value quantity by the benchmark comparison method to determine the missing measurement point.

[0059] The actual measured values ​​are compared with these standard moments. Through the benchmark comparison method, the number of valid measured values ​​actually obtained is subtracted from the number of standard samples. For example, for the interval from 3 seconds to 5 seconds, 30 are lost in the 3rd second, 80 are lost in the 4th second, and 40 are lost in the 5th second, with a total of 150 data points lost. The most obvious is the loss of 20 consecutive sampling points from 350 milliseconds to 550 milliseconds in the 4th second.

[0060] Step S25: Analyze the distribution rules of the missing measurement points in terms of time dimension and measurement dimension.

[0061] Among them, the time dimension analysis includes calculating the density of missing points in different time windows, and the measurement dimension analysis includes calculating the difference in missing rates of horizontal and vertical parameters. For example, in the time dimension, the distribution density analysis found that the 200-millisecond time window in the middle of the 4th second concentrated 25% of the missing measurement points in that second; in the measurement dimension, the missing of the six basic parameters showed obvious differences: the missing of horizontal parameters was more serious, with the missing rate of x-axis acceleration data reaching 60%, while the missing rate of vertical parameters was only about 20%. This difference may be related to the vehicle's motion posture.

[0062] Step S26, combining the missing measurement points and the distribution law analysis results to obtain complete information of the missing measurement points.

[0063] Based on the missing measurement points determined by the benchmark comparison and the characteristic results of the distribution law analysis, the complete information of the missing measurement points in the observed data sequence was obtained, including the specific time position, distribution law and missing characteristics of the missing measurement points. In this way, the specific time position, distribution law and missing characteristics of the 150 missing measurement points between the 3rd and 5th seconds were clarified. In particular, the large-scale continuous missing phenomenon in the 4th second reflects that the system may have encountered serious data transmission problems during this period.

[0064] Step S30, using bottom-level hardware diagnosis, middle-level data diagnosis and top-level system diagnosis for the complete information of the missing measurement points, and obtaining a fault feature description through diagnostic analysis.

[0065] Combination Figure 3 As shown, in some embodiments, step S30 may include: step S31, step S32, step S33 and step S34. Among them:

[0066] Step S31, obtaining hardware diagnosis results by checking the basic electrical characteristics and physical status of the IMU sensor;

[0067] The underlying hardware diagnosis focuses on the physical state and basic electrical characteristics of the IMU sensor. Hardware diagnosis first checks the power supply status of the IMU sensor and evaluates the power quality by measuring the stability of the power supply voltage. Under normal working conditions, the power supply voltage fluctuation of the IMU sensor should be controlled within ±5%. For the IMU with a nominal 3.3V power supply, the voltage should be stable between 3.135V and 3.465V. The power supply check also includes the measurement of the power supply ripple, and the peak-to-peak ripple should not exceed 50mV. When power supply abnormalities are found, further analysis is required to determine whether there are problems such as poor contact, power module failure, or external interference.

[0068] On the basis of ensuring stable power supply, the next step of hardware diagnosis is communication interface inspection. Corresponding detection methods need to be adopted for IMUs with different interface types. Taking the I2C interface as an example, it is necessary to measure the level characteristics and timing relationship of the SCL and SDA signal lines to confirm the signal integrity; for the CAN bus interface, the waveform quality of the CAN_H and CAN_L differential signals is focused on, including key parameters such as signal amplitude, rise time and fall time.

[0069] After completing the electrical characteristics check, hardware diagnosis also needs to monitor the operating temperature of the IMU sensor. The temperature data should be kept within the specified operating temperature range, usually -40°C to 85°C. Abnormal temperature may cause the performance of the IMU sensor to deteriorate, so it is necessary to analyze the cause of the temperature change in combination with the actual application environment.

[0070] Step S32, performing consistency checks on the raw data output by the IMU sensor in sequence, and analyzing the data noise on the basis of completing the consistency checks, so as to obtain a data diagnosis result;

[0071] After the hardware diagnosis is completed, the diagnosis work enters the middle-level data analysis stage. The middle-level data diagnosis mainly conducts a comprehensive analysis of the raw data output by the IMU sensor, focusing on the timing characteristics, measurement accuracy and noise level of the data. First, a data consistency check is performed, which includes a detailed analysis of the timestamps of consecutive data frames. At the standard 100Hz sampling frequency, the time interval between adjacent data frames should be stably maintained at around 10ms, and any time jitter exceeding ±1ms needs to be recorded and analyzed. The data consistency check also includes verification of the data frame format to ensure that each data frame contains complete acceleration and angular velocity information. For the acceleration data in the static test state, the Z-axis acceleration should be stable at about 9.81m / s², while the X and Y-axis accelerations should be close to zero, and the maximum allowable deviation is usually no more than 0.1m / s². Similarly, the standard deviation of the measured values ​​of the angular velocity data in the static state on the three axes should not exceed 0.1rad / s.

[0072] On the basis of completing the data consistency check, another key task of the middle-level diagnosis is to conduct an in-depth analysis of the data noise. By calculating the Allan variance of the measured value, various types of noise components in the data can be identified and quantified. For high-precision IMU, its angular random walk (ARW) index should normally be kept in the range of 0.1~0.5° / √h. In addition to random noise, it is also necessary to analyze possible periodic interference, which is usually related to external vibration sources or electromagnetic interference. By performing spectral analysis on the measured data, the characteristic frequency components can be identified, thereby tracking the source of interference. The results of the noise analysis directly reflect the performance status of the IMU sensor. If a significant increase in the noise index is observed, it indicates that the IMU sensor has performance degradation or failure.

[0073] Step S33, by analyzing the correlation between the IMU sensor data and other sensor information, and analyzing the response characteristics of the IMU sensor to known motion excitation, the dynamic performance of the IMU sensor is evaluated to obtain a system diagnosis result;

[0074] After completing the detailed diagnosis at the hardware and data levels, the diagnosis work enters the final top-level system analysis stage. The top-level system diagnosis mainly evaluates the working status of the IMU in the entire measurement system, focusing on the coordination of the IMU sensor output with the rest of the system. System diagnosis first analyzes the correlation between the IMU sensor data and other sensor information. This cross-validation can find system-level faults that are difficult to identify with a single sensor. Taking in-vehicle applications as an example, the acceleration data measured by the IMU can be compared with the vehicle speedometer signal to verify whether the acceleration integration result matches the speed change.

[0075] As an important part of system diagnosis, it is also necessary to evaluate the dynamic performance of the IMU sensor, including bandwidth characteristics and phase delay, by analyzing the response characteristics of the IMU to known motion excitation. For example, when the vehicle performs standard steering operations, the measurements of lateral acceleration and yaw rate should show accurate correlation and reasonable phase relationship. Any obvious response lag or signal distortion indicates that the system is abnormal.

[0076] Step S34, integrating the hardware diagnosis results, data diagnosis results and system diagnosis results to obtain a fault feature description.

[0077] Combining the diagnostic results of the above three levels, a complete fault feature description can be finally obtained. These features include the level at which the fault occurs (hardware layer, data layer or system layer), the specific manifestation of the fault (such as data loss, increased noise or abnormal response), and the timing characteristics of the fault (continuous or intermittent). Taking the diagnosis of the on-board IMU sensor as an example, when the system has data abnormalities during high-speed driving, the hierarchical diagnosis may obtain the following feature description: at the hardware level, the CAN bus communication quality is reduced, and the differential voltage swing of CAN_H and CAN_L is reduced to 2.0V (the standard value should be 3.0V); at the data level, the lateral acceleration data has periodic jumps, the amplitude exceeds the normal range to ±3g, and is accompanied by 50Hz high-frequency noise; at the system level, the correlation between the lateral acceleration and the yaw rate is significantly reduced, and the phase delay is increased to 150ms (the normal value should be within 50ms). This multi-level fault feature indicates that the fault may be caused by the loosening of the IMU sensor mounting seat due to the high-speed movement of the vehicle, which in turn causes the communication quality to deteriorate and the signal distortion.

[0078] Step S40: establishing a fault data mapping relationship according to the fault feature description, and obtaining fault judgment parameters through feature decomposition.

[0079] Combination Figure 4 As shown, in some embodiments, step S40 may include: step S41, step S42, step S43 and step S44. Among them:

[0080] Step S41, performing feature vectorization processing on the fault features in the fault feature description, and preliminarily mapping each fault feature into a corresponding initial feature vector, so as to obtain an initial hardware feature vector, an initial data feature vector and an initial system feature vector;

[0081] According to the fault features obtained in step S30, feature vectorization processing is first required to construct a standard feature vector space. This process preliminarily maps the original fault features at each level into corresponding feature vectors: hardware-level fault features (such as voltage anomalies, communication quality degradation, and temperature drift, etc.) are mapped to the initial hardware feature vector H'; data-level fault features (such as data frame loss, noise increase, and sampling instability, etc.) are mapped to the initial data feature vector D'; system-level fault features (such as dynamic response hysteresis, performance drift, and consistency deviation, etc.) are mapped to the initial system feature vector S'. These initial feature vectors contain the basic information of the original fault features, but because they come from different physical quantity measurements, their numerical ranges and units are different.

[0082] Step S42, respectively normalizing the initial hardware feature vector, the initial data feature vector, and the initial system feature vector to obtain a final hardware feature vector, a final data feature vector, and a final system feature vector;

[0083] In order to make these initial feature vectors comparable and analyzable, all the data in them need to be standardized. For the data in the initial hardware feature vector H', the voltage deviation rate ΔV is converted into a percentage relative to the nominal voltage. For example, when the operating voltage of an IMU sensor shifts from the nominal value of 3.3V to 3.5V, ΔV is calculated to be 6.06%. At the same time, the communication quality factor Qc is normalized by analyzing the signal integrity index, and the temperature drift coefficient αt is standardized based on the temperature change. These standardized parameters together constitute the final hardware feature vector H, and its values ​​are uniformly mapped to the [0, 1] interval.

[0084] Based on a similar standardization principle, the initial data feature vector D' is processed. The data validity index Ve is standardized by calculating the proportion of correct data frames to the total number of frames. For example, at a sampling frequency of 100Hz, if 95 valid data frames are successfully received within 1 second, the standardized value of Ve is 0.95. The noise level index Rn is normalized by calculating the logarithmic value of the signal-to-noise ratio, and the timing stability index Ts is expressed by the relative value of the standard deviation of the sampling interval. These standardized parameters constitute the final data feature vector D. In this standardization process, it is found that there is an obvious correlation between the various indicators, and their mutual influence relationship needs to be described by the correlation matrix R. For example, when the noise level Rn increases, it often leads to a decrease in the data validity Ve.

[0085] For the initial system eigenvector S', its standardization process needs to consider the dynamic characteristics of the performance parameters. The dynamic response coefficient Kd is standardized by the ratio of the actual response time to the nominal time. For example, if the nominal response time is 10ms and the actual measurement is 12ms, the standardized value of Kd is 1.2. The stability index Is is standardized by calculating the drift rate of the performance parameters, and the consistency parameter Cc is determined based on the data correlation analysis between IMU sensors. These standardized parameters form the final system eigenvector S. Since the system parameters have time-varying characteristics, the sliding window method is used for dynamic evaluation. The window size is determined according to the parameter change characteristics. Fast-changing parameters use minute-level windows, and slowly changing parameters use hour-level windows.

[0086] Step S43, using the principal component analysis method to analyze the feature vector space composed of the final hardware feature vector, the final data feature vector and the final system feature vector, by calculating the feature correlation coefficient matrix, the correlation strength between the features is obtained, and similar fault information is eliminated through feature selection to optimize the feature set;

[0087] After completing the standardization of the three vectors, the complete feature vector space {H, D, S} is obtained. On this basis, the principal component analysis method is used to analyze the correlation between features, and the correlation strength between each feature is obtained by calculating the feature correlation coefficient matrix C = [cij]. When the absolute value of the correlation coefficient between two features exceeds 0.8, it indicates that they convey similar fault information, and feature selection is needed to avoid information redundancy. After completing the feature correlation analysis and selection, the feature set is significantly optimized and simplified.

[0088] Step S44: Based on the optimized feature set, a singular value decomposition method is used to decompose the feature matrix to obtain fault judgment parameters.

[0089] Based on this optimized feature set, the singular value decomposition method is used to decompose the feature matrix F and obtain the standard form F = UΣVT. This decomposition process first constructs the sample covariance matrix Cov = FTF / (n-1), where F is the feature matrix, FT is the transposed matrix of F, n is the number of samples, and Cov represents the covariance matrix between features. Then solve the characteristic equation (Cov-λI)v=0 to obtain the eigenvalue and eigenvector, where λ is the eigenvalue, I is the unit matrix, and v is the eigenvector.

[0090] The feature matrix F is composed of 100 fault samples, each of which includes 9 features such as voltage deviation rate, communication quality factor, and temperature drift coefficient, forming a 100×9 matrix. The rows represent different fault samples, and the columns represent the characteristic parameters of each sample. The matrix U obtained after decomposition is composed of the left singular vectors of the original feature matrix F, forming a 100×100 matrix, and each column vector represents a new feature combination. Taking the IMU sensor as an example, for a left singular vector, if its components in the voltage deviation rate and temperature drift coefficient are 0.75 and 0.65 respectively, and the components in other features are less than 0.3, this indicates that the vector mainly reflects the feature combination at the hardware level. These vectors meet orthogonality, ensuring the independence of fault features.

[0091] The Σ matrix in the middle is a 100×9 diagonal matrix. The elements σi on the diagonal are singular values, which are arranged in descending order to reflect the importance of each feature combination. The VT matrix is ​​a 9×9 matrix whose elements provide the conversion coefficients from the original features to the new feature combinations. After obtaining the singular value decomposition results, it is necessary to determine the number of eigenvectors to be retained by analyzing the distribution of singular values ​​in the Σ matrix. The cumulative contribution rate CR(k)=∑(i=1 to k)σi / ∑(i=1 to n)σi is used as the judgment basis. When the cumulative contribution rate exceeds the preset 85% threshold, the corresponding k value is the number of eigenvectors that need to be retained.

[0092] Specifically, in the test data of a certain IMU sensor, assuming that the singular values ​​obtained are σ1=5.2, σ2=2.8, σ3=1.5, σ4=0.8, σ5=0.4, etc., the cumulative contribution rate of the first four singular values ​​is calculated to be (5.2+2.8+1.5+0.8) / (5.2+2.8+1.5+0.8+0.4)=0.96, which significantly exceeds the threshold of 85%. This shows that the first four eigenvectors already contain the main information in the original feature space, which is sufficient for constructing fault judgment parameters. Each eigenvector represents a combination pattern of fault features, and the size of each component directly reflects the weight contribution of the original features in the combination.

[0093] Based on these four main eigenvectors, the corresponding fault judgment parameters are constructed. The first eigenvector presents high component values ​​of 0.75 and 0.65 in the voltage deviation rate and temperature drift coefficient, respectively, while the components are smaller in other features. This distribution feature shows that it mainly reflects the fault performance at the hardware level. Therefore, the zero bias stability parameter Pb is constructed based on it, which is obtained by the linear combination of these two main components. The second eigenvector shows significant components of 0.82 and 0.78 in the data validity index and noise level index, and the gain consistency parameter Pg is constructed based on this, reflecting the fault characteristics at the data quality level. The third eigenvector presents characteristic components of 0.70 and 0.68 in the timing stability index and system consistency parameter, respectively. Based on this distribution feature, the nonlinear error parameter Pn is constructed. The fourth eigenvector has components of 0.85 and 0.60 in the dynamic response coefficient and stability index, which are used to construct the bandwidth characteristic parameter Pw. The calculation of these judgment parameters uses the component values ​​of the corresponding eigenvectors as the combination weights, realizing the accurate mapping conversion from the original features to the fault judgment parameters. For example, in a certain IMU sensor fault test, when the voltage deviation rate abnormally rises to 0.8 and the temperature drift coefficient reaches 0.6, the zero bias stability parameter Pb calculated through the weight combination of the first eigenvector will significantly deviate from the normal range, directly indicating the occurrence of a hardware-level fault.

[0094] Step S50: construct a fault classification evaluation criterion using the fault judgment parameter, and calculate the fault classification value through the fault classification evaluation criterion.

[0095] Based on the fault judgment parameters obtained in step S40 (including zero bias stability parameter Pb, gain consistency parameter Pg, nonlinear error parameter Pn and bandwidth characteristic parameter Pw), a fault classification evaluation criterion is constructed to determine the fault type. The evaluation criterion needs to comprehensively analyze the dynamic change characteristics of each fault judgment parameter, including instantaneous changes and long-term trends, to determine the fault classification value.

[0096] Combined with reference Figure 5 In some embodiments, step S50 may include step S51, step S52, step S53 and step S54.

[0097] Step S51, establish the concept of fault classification value, use integers from 0 to 3 to represent the four basic states of normal, mutation, drift and failure, and set the mapping relationship between the fault classification value and the standardized score of each fault judgment parameter;

[0098] Step S52, calculating the change rate and long-term change trend of each fault determination parameter;

[0099] First, the concept of fault classification value F is established, where F=0 represents the normal state, F=1 represents the mutation state, F=2 represents the drift state, and F=4 represents the failure state.

[0100] The core of fault classification evaluation is to analyze the time domain characteristics of fault judgment parameters. For each fault judgment parameter, its change rate λi = |(Pi(t) - Pi(t-Δt))| / (Pi(t-Δt)×Δt) must be calculated first, where i represents the fault judgment parameter type (b, g, n, w), corresponding to the four parameters of zero bias stability, gain consistency, nonlinear error and bandwidth characteristics, Pi(t) represents the parameter value at the current moment t, Pi(t-Δt) represents the parameter value at the previous moment, and Δt is the sampling time interval (usually 0.01s). This change rate λi reflects the instantaneous change degree of the fault judgment parameter and is an important indicator for identifying sudden faults. At the same time, in order to capture the slow change of the parameters, it is also necessary to calculate the long-term change trend θi = (Pi(t) - Pi(t0)) / (t - t0), where t0 is the start time of observation, Pi(t0) is the parameter value at the start time, and (t - t0) is the total observation time. The value of θi reflects the average change rate of the fault judgment parameter on a longer time scale, which is particularly important for identifying drift faults.

[0101] Under normal working conditions, the various fault judgment parameters of the IMU sensor should remain relatively stable. Taking a certain model of vehicle-mounted IMU as an example, the change rate λb of its zero bias stability parameter Pb is usually kept below 0.05% / s, indicating that the zero point position is stable; the long-term trend θb does not exceed 0.008% / s, indicating that there is no obvious drift. The gain consistency parameter Pg should be less than λg 0.03% / s when working stably, and θg usually does not exceed 0.005% / s. The nonlinear error parameter Pn and the bandwidth characteristic parameter Pw should also show similar stable characteristics, and their change rate and trend values ​​are kept below 0.04% / s and 0.006% / s respectively. When all these fault judgment parameters meet the stability requirements, it can be confirmed that the IMU sensor is in normal working condition.

[0102] When the IMU sensor is subject to external interference, mutation characteristics of the fault judgment parameters are usually observed. For example, when the vehicle passes through a bumpy road, the change rate of the zero bias stability parameter λb may suddenly rise to 1.5% / s, far exceeding the normal level, while its long-term trend θb remains at around 0.01% / s. This feature clearly points to a mutation fault. At the same time, the gain consistency parameter may also undergo a synchronous mutation, with λg reaching 1.2% / s, but θg remains stable. This synchronous mutation of multiple parameters is usually related to mechanical vibration or electromagnetic interference. The typical feature of a mutation fault is a significant increase in the parameter change rate while the long-term trend remains stable.

[0103] Drift-type faults exhibit different characteristic patterns. When the IMU sensor is in an environment where the temperature is constantly changing, a slow but continuous shift in the fault judgment parameters may be observed. For example, although the rate of change of the zero-bias stability parameter λb is always less than 0.3% / s, indicating that there is no drastic change, its long-term trend θb continues to remain at 0.06% / s, which is significantly higher than normal levels. The gain consistency parameter may show similar drift characteristics, with θg remaining around 0.04% / s. This slow but continuous parameter shift is a typical manifestation of drift-type faults, which are usually related to environmental factors or performance degradation of the IMU sensor itself.

[0104] The most serious case is a failure type fault, which is manifested by multiple parameters being severely abnormal at the same time. For example, when there is a problem with the IMU sensor power supply system, severe fluctuations in the zero bias stability and gain consistency parameters may be observed at the same time, with both λb and λg exceeding 1.5% / s. Nonlinear error parameters and bandwidth characteristic parameters may also deviate severely, with their long-term trends θn and θw exceeding 0.15% / s at the same time. This situation of multiple parameters being abnormal at the same time usually indicates that the IMU sensor has suffered a serious functional failure.

[0105] Step S53, calculating the standardized score of each fault determination parameter based on the change rate and long-term change trend of the fault determination parameter;

[0106] Step S54, combining the standardized scores of various fault judgment parameters to determine the fault classification value of the IMU sensor.

[0107] The calculation of the fault classification value F adopts a multi-level judgment method, which needs to consider the numerical characteristics of the change rate λi and the long-term trend θi of the fault judgment parameter at the same time. First, the standardized score Si of all fault judgment parameters is calculated:

[0108] Si = 100×(1 -λi / λi_max)×(1 - |θi| / θi_max)

[0109] Among them, λi_max is the maximum allowable change rate of each fault judgment parameter (the zero bias stability parameter is 2% / s, the gain consistency parameter is 1.5% / s, the nonlinear error parameter is 1.2% / s, and the bandwidth characteristic parameter is 1% / s), and θi_max is the maximum allowable trend value (0.2% / s, 0.15% / s, 0.12% / s and 0.1% / s respectively).

[0110] Then, according to the mapping relationship between the fault classification value F and the standardized scores of each fault determination parameter, the fault classification value F is determined based on the calculated standardized scores of each fault determination parameter. The mapping relationship between the fault classification value F and the standardized scores of each fault determination parameter can be as follows:

[0111] When the Si of all fault judgment parameters is greater than 90 points, it indicates that the IMU sensor performance is stable, and F = 0. Taking a certain IMU sensor as an example, when λb = 0.03% / s and θb = 0.005% / s are measured, Sb = 97.5 points are calculated; at the same time, other fault judgment parameters also obtained high scores (Sg = 96.8 points, Sn = 95.4 points, Sw = 94.9 points), so it is judged to be in normal state.

[0112] When the Si of any fault judgment parameter is lower than 90 points but all Si are higher than 70 points, the ratio of λi to θi is further analyzed. If max(λi) / λi_max>0.8 and the corresponding |θi| / θi_max<0.2, F=1 is judged, which reflects the typical mutation characteristics.

[0113] When the fault judgment parameter Si is lower than 70 points and max(|θi| / θi_max)>0.8 and all λi / λi_max<0.3, it is judged that F=2, indicating that a significant drift phenomenon has occurred.

[0114] When the Si of multiple fault judgment parameters is lower than 60 points at the same time, or when multiple parameters max(λi) / λi_max>0.8 and max(|θi| / θi_max)>0.8 occur, F=3 is judged, indicating that the IMU sensor has a serious fault.

[0115] Through this fault classification method based on quantitative analysis, combined with standardized scoring of parameter change characteristics and multi-level criteria, accurate identification of IMU sensor fault types can be effectively achieved.

[0116] Step S60, based on the fault classification value, data accuracy measurement is performed, and the fault degree value is obtained by dynamic comparison of the threshold value.

[0117] Based on the fault classification value F (0-3 corresponds to normal, mutation, drift and failure faults respectively) obtained in step S50, data accuracy measurement is required to determine the specific fault severity value. By dynamically comparing the threshold, the IMU sensor data is compared with the standard state to quantitatively evaluate the severity of the fault. The core of the dynamic threshold comparison method is to adjust the evaluation criteria in real time according to the system operating status. This adaptive characteristic makes fault assessment more accurate.

[0118] Combination Figure 6 As shown, in some embodiments, step S60 may include: step S61, step S62 and step S63. Among them:

[0119] Step S61, for different fault classification values, corresponding threshold reference models are established respectively, wherein when the fault classification values ​​are 0, 1, 2, and 3, they correspond to the normal state threshold model, the mutation state threshold model, the drift state threshold model, and the failure state threshold model respectively;

[0120] In the normal state (F=0), the measured data x(t) should meet the standard threshold range, and the normal state threshold model is:

[0121] x(t) = x_base +μ(t)

[0122] Among them, t represents the sampling time, x(t) is the output value of the IMU sensor at time t, x_base is the reference value under standard working conditions, and μ(t) is the allowable fluctuation range. For example, when the IMU sensor is stationary, its Z-axis acceleration x_base should be maintained at 9.8m / s², and the allowable fluctuation range μ(t) is ±0.1m / s². This benchmark model provides a basis for fault identification by defining the parameter range under normal working conditions.

[0123] When a mutation fault (F=1) is detected, the threshold model needs to be extended to a mutation state threshold model:

[0124] x(t) = x_base + A(t) +μ(t)

[0125] Among them, A(t) represents the mutation amplitude component, which is used to characterize the instantaneous jump characteristics of the data. The introduction of the mutation component A(t) enables the model to capture the rapid changes of the data. Taking the on-board IMU as an example, when the vehicle passes over a speed bump, if the Z-axis acceleration suddenly fluctuates abnormally by more than ±5g, A(t) will obviously exceed the fluctuation range under normal driving conditions. This mutation feature is often related to external interference or abnormal transient response of the IMU sensor.

[0126] For drift faults (F=2), the threshold model further extends the drift state threshold model:

[0127] x(t) = x_base + D(t) +μ(t)

[0128] Among them, D(t) is the drift trend component, which indicates the slow deviation of the data. Drift faults usually manifest as gradual changes in data. For example, under the influence of temperature changes, the gyroscope output may produce continuous zero bias changes. By recording and analyzing the drift component D(t), the long-term change trend of IMU sensor performance can be tracked. For vehicle-mounted IMUs, the drift caused by temperature is particularly obvious at the beginning of startup. As the system temperature stabilizes, the drift degree will gradually decrease.

[0129] For failure type faults (F=3), the model is extended to require a more complex form, namely the failure state threshold model:

[0130] x(t) = x_base + E(t) +μ(t)

[0131] Among them, E(t) represents the failure characteristic component, reflecting the serious degradation of the IMU sensor performance. This kind of fault often manifests as a complete abnormality of the data, such as the unstable power supply of the IMU sensor causing random jumps in the output value. The failure characteristic component E(t) usually contains the superposition of multiple abnormalities, and its manifestation needs to be comprehensively analyzed.

[0132] Step S62, obtaining a measurement result by calculating the deviation between the measured data and the model prediction value;

[0133] Step S63, updating the threshold parameters in real time according to the actual operating status of the IMU, and introducing the deviation mean, maximum deviation and fluctuation coefficient to determine the fault degree value based on the measurement results.

[0134] The accuracy measurement process uses a dynamic threshold comparison method. The measurement result is obtained by calculating the deviation between the measured data and the model prediction value:

[0135] δ(t) = w(t)|x(t) - x_model(t)|

[0136] Among them, δ(t) represents the weighted deviation value at time t, w(t) is the time weight factor, which reflects the timeliness of the data, x(t) is the actual measurement value, x_model(t) is the predicted value of the corresponding threshold model, and |·| represents the absolute value. The design of the time weight w(t) is particularly important. It not only considers the timeliness of the data, but also takes into account the dynamic characteristics of the fault development. For example, for mutation faults, a larger time weight is required to quickly respond to data changes; for drift faults, a more gentle weight change can be used to suppress the impact of short-term fluctuations.

[0137] The dynamic threshold adjustment mechanism is the key to ensure measurement accuracy. The electronic device will update the threshold parameters in real time according to the operating status. For example, when the vehicle is driving at high speed, considering the influence of road vibration and air disturbance, the allowable fluctuation range μ(t) of the IMU sensor data will be appropriately increased; at low speed or stationary, due to less external interference, the evaluation criteria will be more stringent. This adaptive mechanism ensures the reliability of fault assessment. For example, when the vehicle speed reaches 100km / h, the allowable fluctuation range of the angular velocity sensor may need to be expanded to ±50° / s, while at 30km / h, the range can be tightened to ±20° / s.

[0138] Based on the measurement deviation, three evaluation indicators are introduced to comprehensively reflect the degree of fault:

[0139] P = k1×M + k2×Ma + k3×V

[0140] Among them, P is the fault degree value (0-100%), M is the deviation mean, reflecting the overall deviation level, Ma is the maximum deviation, reflecting the peak deviation degree, V is the fluctuation coefficient, reflecting the discrete degree of the data, k1, k2, k3 are weight coefficients, satisfying k1+k2+k3=1. This multi-index evaluation method can comprehensively reflect the various characteristics of the fault. Different weight configurations are used for different types of faults: for mutation faults, the maximum deviation Ma is focused on, and the value of k2 can reach more than 0.5; for drift faults, more attention is paid to the mean M and the fluctuation coefficient V, and the values ​​of k1 and k3 are usually between 0.3-0.4; for failure faults, the weights of the three indicators are similar, each about 0.33.

[0141] The final fault severity value P is in the range of 0-100%, reflecting the severity of the fault through graded evaluation:

[0142] P<30% is a minor fault, in which case the data still has good availability and only requires appropriate compensation processing;

[0143] 30%≤P<70% is a moderate fault, data reliability is significantly reduced, and corrective measures need to be taken;

[0144] P ≥ 70% is a severe fault, indicating that the data has been severely distorted and the IMU sensor may need to be stopped.

[0145] This multi-indicator-based fault severity assessment method not only takes into account multiple characteristics of data deviation, but also realizes targeted assessment of different types of faults through dynamic weights, providing a reliable basis for subsequent fault repair.

[0146] Step S70, using the fault degree value, calculates the correction parameter, and obtains the adjustment sequence through parameter conversion.

[0147] The calculation principle of the correction parameters is different for different faults. For mild faults, linear correction is mainly considered, and data repair can be achieved by adjusting gain and zero bias; for moderate faults, nonlinear characteristics need to be considered, and quadratic correction is introduced; and for severe faults, dynamic response compensation needs to be added, and the change characteristics of the data need to be comprehensively considered.

[0148] Combination Figure 7 As shown, in some implementations, step S70 may include: step S71, step S72, step S73 and step S74. Among them:

[0149] Step S71, establishing compensation models corresponding to slight faults, moderate faults, and severe faults;

[0150] Step S72, according to the fault degree value, select a corresponding compensation model to perform correction calculation to obtain correction parameters;

[0151] The calculation of the correction parameters must first establish a compensation model. For slight faults, moderate faults, and severe faults, a slight compensation model, a moderate compensation model, and a severe compensation model are established respectively.

[0152] For minor faults (P<30%), a minor compensation model is used. For example, the minor compensation model uses a linear correction model:

[0153] y(t) =α(t)x(t) +β(t)

[0154] Among them, t represents the sampling time, x(t) is the output value of the IMU sensor, y(t) is the value after compensation, α(t) is the gain correction coefficient, and β(t) is the zero bias correction coefficient. The parameters of this linear model are determined using the least squares method:

[0155] min J =Σ[y(t) - y_ref(t)]²

[0156] Among them, y_ref(t) is the reference value, and the optimal α(t) and β(t) are determined by minimizing the objective function J. Taking an IMU accelerometer as an example, when a slight zero bias occurs, by solving the least squares problem, α(t)≈1.02 and β(t)≈-0.15g are obtained, achieving effective compensation for the zero bias.

[0157] For moderate faults (30%≤P<70%), a moderate compensation model is used:

[0158] y(t) =α(t)x(t) +β(t) +γ(t)x²(t)

[0159] A nonlinear correction term γ(t)x²(t) is added, where γ(t) is the nonlinear correction coefficient. The parameter determination adopts an iterative optimization method: first, the initial solution is obtained using the linear term, and then the nonlinear coefficient is gradually optimized by the gradient descent method. The objective function is expanded to:

[0160] J =Σ[y(t) - y_ref(t)]² +λ|γ(t)|

[0161] Among them, λ is the regularization coefficient, which is used to control the influence of nonlinear terms. The introduction of this quadratic term can effectively compensate for the nonlinear distortion of the IMU sensor, such as the characteristic bending caused by temperature changes.

[0162] In case of severe failure (P ≥ 70%), the severe compensation model is used:

[0163] y(t) =α(t)x(t) +β(t) +γ(t)x²(t) +η(t)dx / dt

[0164] The dynamic response correction term η(t)dx / dt is added, where dx / dt represents the data change rate and η(t) is the dynamic response coefficient. The parameter optimization of this complex model adopts a step-by-step iterative strategy: first fix the dynamic term and optimize the static parameter; then fix the static parameter and optimize the dynamic coefficient. The objective function becomes:

[0165] J =Σ[y(t) - y_ref(t)]² +λ1|γ(t)| +λ2|η(t)|

[0166] Among them, λ1 and λ2 control the weights of the nonlinear term and the dynamic term respectively.

[0167] The value range of each correction parameter is directly related to the fault degree value P:

[0168] For minor faults: the gain correction coefficient α(t) varies from 0.9 to 1.1, and the zero bias correction coefficient β(t) does not exceed ±0.5% of the full scale. For example, when the full scale of a certain model of IMU is ±16g, the adjustment range of β(t) is about ±0.08g.

[0169] For moderate faults: the range of α(t) is expanded to 0.8-1.3, β(t) can reach ±1% of the full scale, and the range of the nonlinear coefficient γ(t) is ±0.05.

[0170] For severe faults: α(t) can reach 0.5-1.5, β(t) can reach ±2% of the full scale, γ(t) can reach ±0.1, and the range of the dynamic response coefficient η(t) is determined according to the system response time.

[0171] Step S73, performing time series expansion on the correction parameters to make them into continuous time functions, wherein the time series expansion adopts a piecewise linear interpolation method to interpolate each correction parameter between adjacent calculation moments;

[0172] After obtaining the correction parameters, it is necessary to generate the adjustment sequence through parameter conversion. This process (1) first requires the time series expansion of the correction parameters to make them a continuous time function. The time series expansion uses the piecewise linear interpolation method to interpolate each correction parameter between adjacent calculation times. The interpolation formula is: p(t) = p(t1) + [p(t2) - p(t1)](t - t1) / (t2- t1), where p(t) can represent any correction parameter, and t1 and t2 are adjacent calculation times. This piecewise linear interpolation can ensure the smoothness of parameter changes over time and avoid sudden changes during the correction process.

[0173] Step S74, when all correction parameters have completed the time series expansion, these parameters are substituted into the corresponding compensation model, and the sliding window method is used to organize and obtain the adjustment sequence.

[0174] When all correction parameters have completed the time series expansion, they need to be substituted into the corresponding compensation model. For mild faults, the expanded gain correction coefficient α(t) and zero bias correction coefficient β(t) are substituted into the linear compensation model y(t) = α(t)x(t) + β(t); for moderate faults, the nonlinear correction coefficient γ(t) needs to be substituted into the extended model y(t) = α(t)x(t) + β(t) + γ(t)x²(t); and for severe faults, the dynamic response coefficient η(t) needs to be substituted into the complete model y(t) = α(t)x(t) + β(t) + γ(t)x²(t) + η(t)dx / dt. Through this substitution calculation, the correction value y(t) at each moment can be obtained.

[0175] In order to ensure the stability and continuity of the correction process, it is necessary to use a sliding window method to organize the adjustment sequence. With the current moment k as the center, take n points forward to form a sequence S(k) = [y(k), y(k-1), ..., y(k-n+1)]. The choice of window length n is directly related to the time characteristics of the correction process. For example, for an IMU sensor with a sampling frequency of 100Hz, if the correction process is required to be completed within 0.1 seconds, n=10 sampling points are required. The key to this sliding window mechanism lies in the overlapping design between windows. When time advances to moment k+1, the new window S(k+1) = [y(k+1), y(k), ..., y(k-8)] overlaps with the window of the previous moment by 9 points. This highly overlapping structure ensures a smooth transition of the correction process.

[0176] Through this parameter conversion and sequence generation method, the final adjustment sequence S(k) contains complete correction information. Each element in the sequence corresponds to the optimal correction value at a certain moment. The length of the sequence reflects the time span of the correction, and the overlapping structure between sequences ensures the continuity of the entire correction process. This hierarchical correction method based on the degree of fault achieves accurate correction of IMU sensor data by establishing a progressive compensation model, strict parameter optimization and smooth sequence generation. The advantages of this method are that the compensation model structure is flexible and can adapt to different degrees of faults; the parameter optimization theory is rigorous, which ensures the accuracy of the correction; the sequence generation strategy is reasonable, which ensures the stability of the correction process.

[0177] Step S80, performing extreme value decomposition on the adjustment sequence, performing cross-spectral analysis on the decomposition result to eliminate the influence of noise, and obtaining the steady-state quantity through waveform reorganization.

[0178] The purpose of extreme value decomposition is to decompose the adjustment sequence into components with different frequency characteristics, and cross-spectral analysis is used to identify and eliminate the noise components in these components, and obtain a more stable repair effect through recombination. In view of the characteristics of IMU sensors, this decomposition and recombination method can not only effectively process random noise in the data, but also identify periodic interference introduced by factors such as mechanical vibration and temperature change. For example, the vehicle-mounted IMU may be interfered by multiple frequencies such as engine vibration (20-30Hz), road bumps (5-10Hz), and steering movements (1-2Hz) during driving. These interferences of different frequencies need to be processed separately through decomposition and analysis.

[0179] The extreme value decomposition first performs a characteristic scale analysis on the regulation sequence S[k]. This analysis process starts by finding the local extreme points of the sequence, with the goal of extracting each intrinsic mode component (IMF). The identification of local extreme points uses a three-point comparison method: by comparing each data point with its adjacent points, when a point is greater than the adjacent points on both sides, it is a local maximum point, and when it is less than the adjacent points on both sides, it is a local minimum point.

[0180] Each IMF component needs to meet two basic conditions: first, the number of extreme points in the entire sequence does not differ from the number of zero-crossing points by more than 1, which ensures that the component has good oscillation characteristics; second, the local mean is zero at any time, which ensures the central symmetry of the component. The setting of these two conditions ensures that each IMF component has a clear physical meaning.

[0181] In order to extract the IMF components that meet the conditions, an iterative "screening" process is required. First, the identified local extreme points are processed: all local maximum points are found, and these discrete points are connected with a cubic spline function to form a smooth upper envelope, which will wrap around the upper peak of the signal like an envelope; similarly, all local minimum points are found, and they are connected with a cubic spline function to form a smooth lower envelope, which will wrap around the lower trough of the signal. Using cubic spline interpolation ensures that the two envelopes transition smoothly between the extreme points without mutations. Then the arithmetic mean of the two envelopes at each sampling moment is calculated to obtain a reference center line located between the upper and lower envelopes. The new sequence h1[k] is obtained by subtracting this reference center line from the original sequence.

[0182] Next, we need to check whether h1[k] meets the two conditions of IMF. The checking process includes: counting the number of extreme points and zero-crossing points of the sequence and comparing their differences; calculating the mean in the local time window and checking whether it is close to zero. If either of these two conditions is not met, the screening process continues with h1[k] as the new input data. This iterative screening process can be expressed by the formula: h1, n[k] = h1, n-1[k] - m1, n-1[k], where n represents the number of iterations and m1, n-1[k] is the mean sequence obtained at the n-1th iteration.

[0183] When the sequence after a certain iteration meets the IMF condition, the first IMF component c1[k] is obtained. Subtract this component from the original sequence: r1[k] = S[k]- c1[k] to obtain the residual sequence r1[k]. Using this residual sequence as the new sequence to be processed, repeat the above decomposition process to obtain the second IMF component. This step-by-step decomposition is performed until the residue becomes a monotonic sequence or meets the preset termination condition.

[0184] The entire decomposition process can be expressed as: S[k] =Σci[k]+rn[k], where ci[k] is the i-th IMF component and rn[k] is the final residual term. Each IMF component represents an inherent oscillation mode in the sequence, and is arranged in descending order of frequency. This order reflects the multi-scale characteristics of the signal. For example, when processing IMU sensor data, higher-frequency IMF components usually correspond to electrical noise (>100Hz), mid-frequency components reflect mechanical vibration characteristics (50-100Hz), and low-frequency components are related to the main motion characteristics of the system (10-50Hz).

[0185] In the IMU sensor data processing, the obtained IMF components need to be deeply analyzed. These components have clear physical meanings: the high-frequency first component (greater than 100Hz) mainly corresponds to electrical noise, the second component (50-100Hz) mainly reflects the mechanical vibration characteristics, and the third component (10-50Hz) is related to the main motion characteristics of the system. In order to accurately distinguish useful signals from interference, it is necessary to analyze the relationship between these components.

[0186] This analysis first calculates the correlation between different components through the cross spectrum: Pij(f) = Xi(f)×Xj*(f). Here Xi(f) and Xj(f) are the discrete Fourier transforms of the i-th and j-th IMF components, respectively, and Xj*(f) is the conjugate complex number of Xj(f). The amplitude of the cross spectrum obtained by this calculation reflects the strength of the correlation between the components, and the phase represents the time delay relationship between them.

[0187] Based on this analysis, the correlation threshold is usually set to 0.7 to classify IMF components into two categories: signal body and noise. When judging, not only the correlation between components should be considered, but also the energy distribution characteristics of each component should be analyzed. Taking gyroscope data processing as an example, when the correlation between an IMF component and other components is lower than the threshold, and its energy distribution shows typical white noise characteristics (i.e., the energy spectrum is flat), it can be judged as a noise component. This dual judgment method combining correlation and energy characteristics significantly improves the accuracy of noise identification.

[0188] After determining the properties of each IMF component, waveform reorganization is required to obtain the steady-state quantity. First, the energy value of each IMF component is calculated: Ei = Σ|ci[k]|², where ci[k] is the component sequence value. The energy value obtained by this calculation directly reflects the importance of the component in the original signal. Based on the energy value, the weight coefficient of each IMF component is further determined: wi = Ei / ΣEi, where ΣEi represents the sum of the energy of all components. This weight calculation method based on energy ratio realizes the scientific quantification of component contribution.

[0189] Finally, the waveform is reorganized through weighted combination to obtain the steady-state quantity: Y[k] =Σwici[k]. This reorganization process is not a simple waveform superposition, but a selective combination of IMF components with different frequency characteristics through weight coefficients. For example, when processing the data of the vehicle-mounted IMU, the medium-frequency IMF component (10-50Hz) that reflects the main motion characteristics of the vehicle usually has a larger energy value and thus obtains a higher weight, while the high-frequency noise component is appropriately suppressed due to its smaller energy.

[0190] This processing method based on extreme value decomposition and cross-spectral analysis achieves deep optimization of the adjustment sequence through scientific energy calculation, reasonable weight distribution and selective waveform reorganization. It can not only adaptively process nonlinear and non-stationary signals, which is particularly suitable for processing complex sensor data such as IMU, but also provides an objective noise identification standard through cross-spectral analysis, avoiding the subjectivity of empirical judgment. At the same time, the waveform reorganization strategy based on energy characteristics also ensures that the steady-state quantity has a clear physical meaning.

[0191] Step S90, input the steady-state quantity into the quantized equalization equation, construct the compensation factor in combination with the transient response characteristics of the sensor, and obtain the final repair result through frequency domain modulation.

[0192] Based on the steady-state quantity Y[k] obtained in step S80, it is necessary to obtain the final repair result through quantization equalization processing and frequency domain modulation. First, the discrete time series Y[k] needs to be converted to the frequency domain through discrete Fourier transform (DFT) for processing to obtain its spectrum representation Y(f). By designing appropriate equalization coefficients and compensation mechanisms, different frequency components are optimized and modulated, and then the repair result is obtained through inverse transformation. The whole process must consider both the spectral distribution characteristics of the signal and the transient response characteristics of the IMU sensor itself. For example, when processing the on-board IMU sensor data, it is necessary to consider the combined impact of vehicle motion and road environment on the IMU sensor output.

[0193] Quantization equalization modulates the signal in the frequency domain. The basic quantization equalization equation can be expressed as: Z(f) = Y(f)×H(f)×C(f), where Z(f) is the repaired spectrum, Y(f) is the spectrum of the steady-state quantity, H(f) is the equalization coefficient, C(f) is the compensation factor, and f is the frequency variable. This equation implements the core function of frequency domain modulation: adjusting the spectrum distribution through the equalization coefficient and correcting the IMU sensor response characteristics using the compensation factor.

[0194] The design of the equalization coefficient H(f) is the key to achieve spectrum optimization. According to the characteristics of IMU sensors, different modulation strategies need to be adopted for different frequency bands. The low frequency band (less than 10Hz) mainly considers the drift caused by environmental factors such as temperature changes, and the equalization coefficient is designed to be a gradual characteristic to ensure a smooth transition; the mid-frequency band (10-100Hz) contains the main motion information measured by the IMU sensor, and the equalization coefficient of this part of the frequency band needs to maintain a relatively flat characteristic to ensure that the effective signal is not over-modulated; the high frequency band (greater than 100Hz) mainly contains electrical noise and mechanical vibration, and the equalization coefficient presents a decreasing characteristic to achieve selective filtering. For example, when a vehicle is driving on a bumpy road, the mid-frequency signal contains the main motion information, and the equalization coefficient of this part should be kept between 0.9-1.1, while the equalization coefficient of the high frequency band can be reduced to below 0.5.

[0195] The transient response characteristics of the IMU sensor are corrected by the compensation factor C(f). Based on the frequency response function G(f) of the IMU sensor, the compensation factor is constructed: C(f) = 1 / |G(f)|×exp(-jφ(f)), where |G(f)| represents the amplitude-frequency characteristics of the IMU sensor, φ(f) is the phase-frequency characteristic, exp(-jφ(f)) is the phase compensation term, and j is the imaginary unit. Through this compensation mechanism, the response differences of the IMU sensor at different frequencies can be effectively corrected. For example, when the vehicle is traveling at high speed, the acceleration sensor may have insufficient dynamic response due to limited bandwidth, and the high-frequency response characteristics can be improved by the compensation factor.

[0196] The frequency domain modulation process needs to consider the effects of the equalization coefficient and the compensation factor at the same time. Different modulation strategies are used for the three common working states of IMU: static, low-speed movement and high-speed movement. In the static state, the main focus is on suppressing low-frequency drift, and the focus is on adjusting the equalization coefficient of the low-frequency band; in low-speed movement, it is necessary to protect the motion information of 10-50Hz, while appropriately suppressing high-frequency noise; in high-speed movement, it is necessary to use compensation factors to enhance the dynamic response capability of the IMU sensor.

[0197] The modulated spectrum Z(f) is transformed into a time domain sequence through the inverse discrete Fourier transform (IDFT): z[k] = IDFT{Z(f)}, where z[k] is the final repair result sequence. This conversion process needs to ensure that the frequency domain correction does not introduce additional phase distortion. For IMU sensor data, the repaired signal should be able to accurately reflect the changes in the physical quantity measured by the IMU sensor. For example, when measuring the angular velocity of a vehicle during steering, the repair result should maintain the transient characteristics of the angular velocity signal and effectively suppress the interference introduced by the body vibration.

[0198] This repair method based on quantitative equalization and frequency domain modulation realizes the conversion from steady-state quantity to final repair result through scientific spectrum analysis and carefully designed modulation strategy. The equalization coefficient ensures the reasonable distribution of spectrum components, the compensation factor realizes the correction of IMU sensor characteristics, and the application of discrete Fourier transform ensures the digital realization of the entire processing process. The final repair sequence z[k] not only retains the effective information of IMU sensor data, but also realizes the targeted suppression of various interferences, marking the successful completion of the entire fault diagnosis and repair process. The innovation of this process lies in: the precise modulation of frequency domain characteristics is achieved through quantitative equalization, the compensation factor is used to ensure the accurate restoration of dynamic characteristics, and the segmented modulation strategy is adopted to adapt to the requirements of different working states, and finally a complete signal repair solution is established.

[0199] The technical solution of the IMU fault diagnosis and repair method of the present invention, firstly, solves the problem of incomplete fault feature extraction in the traditional method by establishing an adaptive data preprocessing and hierarchical diagnosis mechanism. The mechanism first performs dynamic sampling and analysis on the time domain data, then accurately calculates the lost data through periodic characteristics, and finally adopts a hierarchical diagnosis strategy to analyze layer by layer from hardware to system, breaking through the limitations of traditional single-level diagnosis, and can accurately identify and locate various types of faults under complex working conditions. Secondly, an innovative scheme based on fault data mapping and quantitative evaluation is proposed to overcome the defect of fuzzy fault judgment. The scheme realizes accurate classification and degree quantification of faults by establishing a mapping relationship between fault characteristics and judgment parameters, combined with dynamic threshold comparison technology, avoiding the subjectivity of traditional empirical judgment, and significantly improving the accuracy of fault diagnosis. In addition, an innovative repair strategy based on extreme value decomposition and frequency domain modulation is designed to solve the problem of poor adaptability of traditional repair methods. The strategy first obtains the adjustment sequence through parameter conversion, then uses extreme value decomposition to eliminate the influence of noise, and finally uses quantitative equalization to achieve accurate modulation in the frequency domain. This multi-level repair method breaks through the limitations of traditional single filtering or threshold correction, and can adaptively adjust the repair strategy according to fault characteristics, achieving more accurate data repair.

[0200] It should be noted that all the above-mentioned embodiments of the IMU fault diagnosis and repair method of the present invention can be arbitrarily combined or combined to form a new embodiment when there is no contradiction or conflict between them.

[0201] Reference Figure 8 The present invention also proposes an IMU fault diagnosis and repair device 100, which is applied to a vehicle-mounted navigation device.

[0202] In this embodiment, the IMU fault diagnosis and repair device 100 includes:

[0203] The acquisition module 101 is used to acquire the time domain data of the IMU sensor, sample and analyze the time domain data according to a preset rule, and obtain an initial measurement value through integrity check and data preprocessing;

[0204] A statistics module 102 is used to perform periodic analysis on the initial measurement values, calculate the amount of lost data according to the sensor sampling characteristics, and obtain missing measurement points through data statistics;

[0205] Diagnostic module 103, used for using bottom-level hardware diagnosis, middle-level data diagnosis and top-level system diagnosis on the obtained missing measurement points, and obtaining a fault feature description through diagnostic analysis;

[0206] A decomposition module 104 is used to establish a fault data mapping relationship according to the fault feature description, and obtain a fault judgment parameter through feature decomposition;

[0207] A calculation module 105 is used to construct a fault classification evaluation criterion using the fault judgment parameter, and calculate a fault classification value using the fault classification evaluation criterion;

[0208] A comparison module 106, configured to measure data accuracy based on the fault classification value and obtain a fault degree value by dynamic comparison of a threshold value;

[0209] A correction module 107, used to calculate correction parameters using the fault degree value, and obtain an adjustment sequence through parameter conversion;

[0210] The reorganization module 108 is used to perform extreme value decomposition on the adjustment sequence, perform cross-spectral analysis on the decomposition result to eliminate the influence of noise, and obtain the steady-state quantity by waveform reorganization;

[0211] The modulation module 109 is used to input the steady-state quantity into the quantized equalization equation, construct a compensation factor in combination with the transient response characteristics of the sensor, and obtain the final repair result through frequency domain modulation.

[0212] The IMU fault diagnosis and repair device 100 of the present invention can implement the steps of the above-mentioned IMU fault diagnosis and repair method, and therefore has at least all the beneficial effects brought about by the technical solutions of the above-mentioned IMU fault diagnosis and repair method embodiments, which will not be described one by one here.

[0213] The present invention also provides a vehicle-mounted navigation device, which includes an IMU sensor and an electronic device, referring to Fig. 9 , Fig. 9 It is a schematic diagram of the structure of an electronic device in a hardware operating environment involved in an embodiment of the present invention.

[0214] The electronic device of the embodiment of the present invention may be a controller of a vehicle-mounted navigation device, or a computing device such as a single-chip microcomputer, a notebook, a palm computer, and a server. Fig. 9 As shown, the electronic device may include: a processor 1001 (such as a CPU), a network interface 1004, a user interface 1003, a memory 1005 and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit, such as a keyboard (Keyboard), and the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0215] Those skilled in the art will understand that Fig. 9 The electronic device structure shown in the figure does not constitute a limitation of the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0216] like Fig. 9 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module and a computer program.

[0217] exist Fig. 9 In the electronic device shown, the network interface 1004 is mainly used to connect to the background server and communicate data with the background server; the user interface 1003 is mainly used to connect to the client (user end) and communicate data with the client; and the processor 1001 can be used to call the computer program stored in the memory 1005. When the computer program is called and executed by the processor 1001, the steps of the above-mentioned IMU fault diagnosis and repair method are implemented.

[0218] Based on the computer program proposed in the aforementioned embodiment, the present invention further proposes a storage medium, which stores a computer program. When the computer program is executed by a controller, the IMU fault diagnosis and repair method recorded in the aforementioned embodiment is implemented.

[0219] Since the vehicle-mounted navigation device and storage medium of the present invention can implement the steps of the above-mentioned IMU fault diagnosis and repair method, they at least have all the beneficial effects brought by the technical solutions of the above-mentioned IMU fault diagnosis and repair method embodiments, which will not be described one by one here.

[0220] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0221] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0222] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0223] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0224] The above description is only a partial or preferred embodiment of the present invention. Neither the text nor the drawings can limit the scope of protection of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the overall concept of the present invention, or direct / indirect application in other related technical fields are included in the scope of protection of the present invention.

Claims

1. An IMU fault diagnosis and repair method, applied to a vehicle-mounted navigation device, characterized in that: include: Acquire time domain data of the IMU sensor, sample and analyze the time domain data according to preset rules, and obtain initial measurement values ​​through integrity check and data preprocessing; Perform periodic analysis on the initial measurement values, establish a reference benchmark based on the sensor sampling characteristics, and obtain complete information of the missing measurement points through benchmark comparison and regularity analysis; The complete information of the missing measurement point is analyzed using bottom-level hardware diagnosis, middle-level data diagnosis and top-level system diagnosis to obtain a fault feature description; According to the fault feature description, a fault data mapping relationship is established, and a fault judgment parameter is obtained through feature decomposition; Using the fault judgment parameters, constructing a fault classification evaluation criterion, and calculating a fault classification value through the fault classification evaluation criterion; Based on the fault classification value, data accuracy measurement is performed, and a fault degree value is obtained by dynamic comparison of threshold values; Using the fault degree value, calculating the correction parameter, and obtaining the adjustment sequence through parameter conversion; Performing extreme value decomposition on the regulation sequence, performing cross-spectral analysis on the decomposition result to eliminate the influence of noise, and obtaining the steady-state quantity by waveform reorganization; The steady-state quantity is input into the quantized equalization equation, and the compensation factor is constructed in combination with the transient response characteristics of the sensor, and the final repair result is obtained through frequency domain modulation.

2. The IMU fault diagnosis and repair method according to claim 1, characterized in that: The initial measurement values ​​are periodically analyzed, a reference benchmark is established according to the sensor sampling characteristics, and complete information of the missing measurement points is obtained through benchmark comparison and regularity analysis, including: Determine a basic analysis period according to the sampling frequency of the IMU sensor; The initial measurement value is segmented into a plurality of analysis units according to the basic analysis period, the number of measurement values ​​actually received in each analysis unit is counted, and the time interval between adjacent measurement values ​​in each analysis unit is calculated to determine the data abnormality interval; Based on the sampling frequency of the IMU sensor, a corresponding sampling time sequence is established as a reference benchmark; Compare the actually received measurement values ​​with the standard time in the sampling time sequence, and for the data abnormal interval, calculate the difference between the standard sampling quantity and the actually obtained valid measurement value quantity by the benchmark comparison method to determine all missing measurement points; Performing distribution law analysis of the missing measurement points in terms of time dimension and measurement dimension, wherein the time dimension analysis includes calculating the density of missing points in different time windows, and the measurement dimension analysis includes calculating the difference in missing rate of parameters in the horizontal direction and the vertical direction; By combining the missing measurement points and the distribution law analysis results, complete information of the missing measurement points is obtained, including specific time locations, distribution laws and missing characteristics.

3. The IMU fault diagnosis and repair method according to claim 1, characterized in that: The complete information of the missing measurement point uses bottom-level hardware diagnosis, middle-level data diagnosis and top-level system diagnosis to obtain a fault feature description through diagnostic analysis, including: Obtaining hardware diagnostic results by performing basic electrical characteristics and physical status checks on the IMU sensor; By sequentially performing consistency checks on the raw data output by the IMU sensor, and analyzing the data noise on the basis of completing the consistency checks, a data diagnosis result is obtained; By analyzing the correlation between the data of the IMU sensor and other sensor information, and analyzing the response characteristics of the IMU sensor to known motion excitation, the dynamic performance of the IMU sensor is evaluated to obtain a system diagnosis result; The fault feature description is obtained by combining the hardware diagnosis result, the data diagnosis result and the system diagnosis result.

4. The IMU fault diagnosis and repair method according to claim 1, characterized in that: The step of establishing a fault data mapping relationship according to the fault feature description and obtaining a fault determination parameter by feature decomposition includes: Performing feature vectorization processing on the fault features in the fault feature description, and preliminarily mapping each fault feature into a corresponding initial feature vector, so as to obtain an initial hardware feature vector, an initial data feature vector, and an initial system feature vector; The initial hardware feature vector, the initial data feature vector and the initial system feature vector are respectively standardized to obtain a final hardware feature vector, a final data feature vector and a final system feature vector; The principal component analysis method is used to analyze the feature vector space composed of the final hardware feature vector, the final data feature vector and the final system feature vector, and the correlation strength between the features is obtained by calculating the feature correlation coefficient matrix, and similar fault information is eliminated through feature selection to optimize the feature set; Based on the optimized feature set, the feature matrix is ​​decomposed by using a singular value decomposition method to obtain the fault judgment parameter.

5. The IMU fault diagnosis and repair method according to claim 1, characterized in that: The method of using the fault judgment parameter to construct a fault classification evaluation criterion, and calculating a fault classification value by using the fault classification evaluation criterion, includes: Establish the concept of fault classification value, use integers from 0 to 3 to represent the four basic states of normal, mutation, drift and failure, and set the mapping relationship between the fault classification value and the standardized score of each fault judgment parameter; Calculating the change rate and long-term change trend of each of the fault determination parameters; Calculating a standardized score of each fault determination parameter based on the change rate and long-term change trend of the fault determination parameter; The standardized scores of the fault judgment parameters are comprehensively considered to determine the fault classification value of the IMU sensor.

6. The IMU fault diagnosis and repair method according to claim 1, characterized in that: The method of measuring data accuracy based on the fault classification value and obtaining a fault degree value by dynamically comparing a threshold value includes: For different fault classification values, corresponding threshold reference models are established respectively, wherein when the fault classification values ​​are 0, 1, 2, and 3, they correspond to the normal state threshold model, the mutation state threshold model, the drift state threshold model, and the failure state threshold model respectively; The measurement result is obtained by calculating the deviation between the measured data and the model prediction value; The threshold parameters are updated in real time according to the actual operating status of the IMU, and based on the measurement results, the deviation mean, maximum deviation and fluctuation coefficient are introduced to determine the fault degree value.

7. The IMU fault diagnosis and repair method according to any one of claims 1 to 6, characterized in that: The method of using the fault degree value to calculate the correction parameter and obtaining the adjustment sequence through parameter conversion includes: Establish compensation models corresponding to minor faults, moderate faults, and severe faults; According to the fault degree value, a corresponding compensation model is selected to perform correction calculation to obtain correction parameters; Expanding the correction parameters in time series to make them into continuous time functions, wherein the time series expansion adopts a piecewise linear interpolation method to interpolate each correction parameter between adjacent calculation moments; After all correction parameters have completed time series expansion, these parameters are substituted into the corresponding compensation model, and the sliding window method is used to organize and obtain the adjustment sequence.

8. An IMU fault diagnosis and repair device, applied to a vehicle-mounted navigation device, characterized in that: include: An acquisition module is used to acquire time domain data of the IMU sensor, sample and analyze the time domain data according to preset rules, and obtain initial measurement values ​​through integrity check and data preprocessing; A statistics module, used to perform periodic analysis on the initial measurement values, calculate the amount of lost data according to the sampling characteristics of the sensor, and obtain missing measurement points through data statistics; A diagnostic module is used to use low-level hardware diagnosis, middle-level data diagnosis, and top-level system diagnosis on the missing measurement points, and obtain a fault feature description through diagnostic analysis; A decomposition module, used to establish a fault data mapping relationship according to the fault feature description, and obtain a fault judgment parameter through feature decomposition; A calculation module, used to construct a fault classification evaluation criterion by using the fault judgment parameter, and calculate a fault classification value by using the fault classification evaluation criterion; A comparison module, used to measure data accuracy based on the fault classification value, and obtain a fault degree value through dynamic comparison of a threshold value; A correction module, used to calculate correction parameters using the fault degree value, and obtain an adjustment sequence through parameter conversion; A reorganization module is used to perform extreme value decomposition on the regulation sequence, perform cross-spectral analysis on the decomposition result to eliminate the influence of noise, and obtain a steady-state quantity by waveform reorganization; The modulation module is used to input the steady-state quantity into the quantized equalization equation, construct a compensation factor in combination with the transient response characteristics of the sensor, and obtain the final repair result through frequency domain modulation.

9. A vehicle-mounted navigation device, comprising an IMU sensor, characterized in that: The vehicle-mounted navigation device also includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the IMU fault diagnosis and repair method as described in any one of claims 1 to 7 are implemented.

10. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the steps of the IMU fault diagnosis and repair method according to any one of claims 1 to 7.

Images