Inertial navigation and mileage online correction method with multiple features fusion in pipeline

Abstract

The application relates to the technical field of pipeline detection navigation, in particular to a kind of inertial navigation and mileage online correction method of multi-feature fusion in pipeline, and the method comprises the following steps: collecting detection data; determining observation information based on the detection data; inputting the detection data into observation information prediction model, and outputting observation information prediction value; the observation information is input into observation information prediction model, and observation information prediction value is corrected to obtain observation information estimate value; based on the updated mileage wheel proportion factor in observation information estimate value, displacement pulse data in detection data is converted to obtain the target running mileage of detector. The application generates feature events and observation information during the running of the detector in the pipeline, and uses the extended Kalman filtering algorithm of state augmentation to jointly estimate and update the inertial device zero offset and the mileage wheel proportion factor in real time, thereby improving the positioning accuracy and mileage reliability of the detector in the pipeline.

Description

Technical Field

[0001] This application relates to the field of pipeline inspection and navigation technology, and in particular to an inertial navigation and online mileage correction method based on multi-feature fusion within a pipeline. Background Technology

[0002] In the field of pipeline inspection, accurately determining the position and mileage of the detector during its operation inside the pipeline is a crucial foundation for the geographical coordinate calibration of defect points and subsequent maintenance and repair work. The accuracy of the detector's positioning and mileage measurement directly determines the accuracy of defect location, thus affecting the cost and efficiency of the entire pipeline inspection project, as the detector operates over long distances inside buried pipelines completely shielded from external positioning signals.

[0003] To achieve accurate navigation and odometer correction, existing technologies employ a measurement scheme that combines inertial navigation and an odometer wheel. The detector is equipped with an inertial measurement unit and an odometer wheel. The inertial measurement unit collects motion data to calculate the trajectory, while the odometer wheel collects displacement pulse data to measure the distance traveled. After the detection is completed, a post-processing pipeline constraint algorithm is used to smooth and correct the odometer data.

[0004] However, existing technical solutions have fundamental flaws: in the combined inertial navigation and odometer wheel measurement scheme, the error of the inertial measurement unit accumulates over time, leading to a deterioration in positioning accuracy; the odometer wheel has an unavoidable systematic error when passing through bends, and cannot provide an effective heading reference in strong magnetic interference environments; although the post-processing pipeline constraint algorithm can smooth the odometer data, its correction process is entirely offline after the detection is completed, and it cannot provide accurate odometer information in real time during the detection process. Furthermore, this method indirectly corrects the final trajectory through geometric smoothing, without directly estimating and compensating for the scaling factor error of the odometer wheel, thus failing to address the root cause of the error. At the same time, it relies on a preset pipeline geometric model, and when the pipeline has actual deformation or a non-standard shape, its strong constraints may incorrectly correct the true odometer information, leading to a decrease in absolute accuracy. Summary of the Invention

[0005] This application provides a method for inertial navigation and online odometer correction in pipelines based on multi-feature fusion, in order to solve the technical problems of deteriorated inertial navigation positioning accuracy and unreliable odometer data caused by the accumulation of errors in inertial measurement units, the inherent system error of the odometer wheel, and the lack of heading reference under strong magnetic interference in existing pipeline detectors, which also makes online real-time correction impossible.

[0006] To achieve the above objectives, this application provides a multi-feature fusion inertial navigation and online odometer correction method for pipelines, applied to a detector within the pipeline. The detector includes multiple sensors, including:

[0007] Real-time acquisition of detection data; the detection data is data collected by sensors.

[0008] The observation information is determined based on the detection data; the observation information characterizes the characteristic events corresponding to the pipeline, including at least one of the following: circumferential weld events, elbow events, and zero-velocity events.

[0009] Input detection data into the observation information prediction model and output the observation information prediction value; the observation information prediction model includes the extended Kalman filter algorithm with state amplification; the observation information prediction value includes at least the mileage wheel scaling factor, which is the state quantity to be estimated incorporated by the extended Kalman filter algorithm after state amplification.

[0010] The observation information is input into the observation information prediction model, and the predicted values ​​of the observation information are corrected to obtain the estimated values ​​of the observation information.

[0011] Based on the updated mileage wheel scaling factor in the estimated value of the observation information, the displacement pulse data in the detection data is converted to obtain the target running mileage of the detector.

[0012] Optionally, the sensor includes an inertial measurement unit, an odometer wheel, and a magnetic flux leakage sensor; real-time acquisition of detection data includes:

[0013] The inertial measurement unit collects specific force and angular velocity information in real time along the three orthogonal axes of the detector.

[0014] Displacement pulse data is collected in real time using a mileage wheel; the displacement pulse data is pulse count data that characterizes the distance the detector travels.

[0015] Pipeline characteristic data is collected in real time using a magnetic flux leakage sensor. The pipeline characteristic data is the magnetic flux leakage signal data corresponding to the circumferential weld of the pipeline.

[0016] Optionally, the magnetic flux leakage sensor includes a first magnetic flux leakage probe and a second magnetic flux leakage probe arranged axially at a predetermined distance along the detector's forward direction.

[0017] Observational information is determined based on the detection data, including:

[0018] When the first and second magnetic flux leakage probes sequentially collect magnetic flux leakage signal data corresponding to the same circumferential weld, a circumferential weld event is generated.

[0019] Record the first timestamp corresponding to the peak value in the magnetic flux leakage signal data collected by the first magnetic flux leakage probe, and the second timestamp corresponding to the peak value in the magnetic flux leakage signal data collected by the second magnetic flux leakage probe.

[0020] Calculate the time difference between the first timestamp and the second timestamp.

[0021] Based on the time difference and the preset axial distance between the first and second magnetic flux leakage probes, the instantaneous velocity of the detector when passing through the circumferential weld is obtained; the instantaneous velocity is the observation information corresponding to the circumferential weld event.

[0022] Optionally, the detection data is input into the observation information prediction model, and the predicted observation information values ​​are output, including:

[0023] Construct a first observation function corresponding to the circumferential weld event; the first observation function is to project the three-dimensional velocity component of the state vector in the observation information prediction model onto the detector's forward direction.

[0024] The first observation function is executed by dot product operation, which is the dot product of the three-dimensional velocity components and the axial quantity of the body coordinate system, to obtain the velocity prediction value corresponding to the circumferential weld event.

[0025] Optionally, determining observation information based on detection data also includes:

[0026] Based on the angular velocity information in the detection data, if the angular velocity information along a single axis exhibits a smooth change characteristic and continues to exceed a preset time threshold, the angular velocity information exhibiting the smooth change characteristic is integrated to generate the total angular change.

[0027] If the difference between the total angle change and the preset bend angle is less than the preset angle threshold, a bend event is generated, and the start and end times of the bend event are recorded.

[0028] Based on the start and end times, the time interval for the bend to pass through is obtained.

[0029] Numerical integration is performed on the angular velocity information during the time interval of the bend passage to correct the three-dimensional motion trajectory of the detector's centroid during the bend passage time interval.

[0030] Based on the three-dimensional motion trajectory, the arc length corresponding to the three-dimensional motion trajectory is generated.

[0031] Obtain the number of pulses generated by the cumulative displacement pulse data within the time interval of passing through the bend, and calculate the mileage increment of the mileage wheel within the time interval of passing through the bend based on the number of pulses and the mileage wheel scaling factor.

[0032] Calculate the length difference between the arc length and the mileage increment to generate the arc length difference; the arc length difference is the observation information corresponding to the bend event.

[0033] Optionally, inputting detection data into the observation information prediction model and outputting predicted observation information values ​​also includes:

[0034] Construct a second observation function corresponding to the bend event. The second observation function is an integral function within the bend passage time interval.

[0035] Based on the second observation function, the cumulative number of pulses within the time interval of the bend is calculated, multiplied by the mileage wheel scaling factor of the mileage wheel, and then multiplied by the scaling factor error coefficient to be estimated in the state vector to obtain the predicted mileage increment corresponding to the bend event.

[0036] Optionally, determining observation information based on detection data also includes:

[0037] Based on specific force information and angular velocity information, the variance of specific force information and the variance of angular velocity information are monitored respectively;

[0038] Based on displacement pulse data and the online real-time estimated mileage wheel ratio factor, the instantaneous travel speed of the detector is calculated and generated.

[0039] If, within a consecutive preset time period, the variance of the force information and the variance of the angular velocity information are both lower than their respective preset variance thresholds, and the absolute value of the detector's instantaneous travel velocity is lower than a preset velocity threshold, then a zero-velocity event is determined to have occurred, and the theoretical zero-velocity vector is used as the observation information corresponding to the zero-velocity event.

[0040] Optionally, inputting detection data into the observation information prediction model and outputting predicted observation information values ​​also includes:

[0041] A third observation function corresponding to the zero-velocity event is constructed. The third observation function is used to extract the observation information and predict the three-dimensional velocity component of the state vector in the model.

[0042] The velocity prediction vector corresponding to the zero-velocity event is obtained by performing calculations using the third observation function.

[0043] Optionally, the observation information is input into the observation information prediction model to correct the predicted values ​​of the observation information, thereby obtaining estimated values ​​of the observation information, including:

[0044] The instantaneous velocity, arc length difference, and theoretical zero velocity vector are input into the observation information prediction model, and compared with the corresponding predicted velocity value, predicted mileage increment, and velocity prediction vector to generate the observation residual.

[0045] The inertial device zero bias in the predicted value of the observation information is weighted and corrected based on the observation residuals to obtain the estimated value of the observation information.

[0046] Optionally, based on the updated mileage wheel scaling factor in the estimated values ​​of the observation information, the displacement pulse data in the detection data are converted to obtain the target operating mileage of the detector, including:

[0047] Extract the mileage wheel scaling factor after correction for circumferential weld events and elbow events from the estimated values ​​of the observation information;

[0048] The cumulative number of pulses in the displacement pulse data is multiplied by the updated mileage wheel scale factor to obtain the distance increment for each time period.

[0049] The target operating mileage of the detector is obtained by summing the distance increments over all time periods.

[0050] As can be seen from the above technical solutions, this application provides an inertial navigation and online mileage correction method for multi-feature fusion in pipelines. The method includes: real-time acquisition of detection data; the detection data is data acquired by sensors; determination of observation information based on the detection data; the observation information characterizes feature events corresponding to the pipeline, including at least one of circumferential weld events, elbow events, and zero-velocity events; inputting the detection data into an observation information prediction model and outputting predicted observation information values; the observation information prediction model includes an extended Kalman filter algorithm with state amplification; the predicted observation information values ​​include at least a mileage wheel scaling factor, which is the state variable to be estimated incorporated after state amplification by the extended Kalman filter algorithm; inputting the observation information into the observation information prediction model to correct the predicted observation information values ​​and obtain estimated observation information values; and converting the displacement pulse data in the detection data based on the updated mileage wheel scaling factor in the estimated observation information values ​​to obtain the target operating mileage of the detector. This application generates information based on characteristic events and observations during the detector's operation within the pipeline, and utilizes a state-amplified extended Kalman filter algorithm to perform online real-time joint estimation and updating of the inertial device's zero bias and mileage wheel scaling factor, thereby improving the positioning accuracy and mileage reliability of the detector within the pipeline. Attached Figure Description

[0051] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 A flowchart illustrating the inertial navigation and online odometer correction method for multi-feature fusion within a pipeline, as provided in an embodiment of this application.

[0053] Figure 2 One of the schematic diagrams illustrating the principle of circumferential weld event recognition and instantaneous velocity generation in the inertial navigation and online mileage correction method for multi-feature fusion in pipelines provided in this application embodiment;

[0054] Figure 3 The second schematic diagram illustrates the principle of circumferential weld event recognition and instantaneous velocity generation in the inertial navigation and online mileage correction method for multi-feature fusion in pipelines provided in this application embodiment.

[0055] Figure 4A schematic diagram of the geometric principle of bend event recognition and arc length difference generation in the inertial navigation and online mileage correction method for multi-feature fusion in pipelines provided in the embodiments of this application;

[0056] Figure 5 One of the schematic diagrams illustrating the zero-speed event recognition principle of the inertial navigation and online odometer correction method for multi-feature fusion in pipelines provided in this application embodiment;

[0057] Figure 6 The second schematic diagram illustrates the zero-speed event recognition principle of the inertial navigation and online mileage correction method for multi-feature fusion in pipelines provided in this application embodiment.

[0058] Figure 7 This is the third schematic diagram illustrating the zero-speed event recognition principle of the inertial navigation and online mileage correction method for multi-feature fusion in pipelines provided in this application embodiment. Detailed Implementation

[0059] The embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described below do not represent all embodiments consistent with this application. They are merely examples of systems and methods consistent with some aspects of this application as detailed in the claims.

[0060] Pipeline detectors are complex systems operating inside pipelines with no satellite signals, strong magnetic interference, and complex geometries. Existing navigation and odometer correction schemes generally employ post-processing constrained smoothing models. These models force the initial trajectory of the inertial navigation and odometer wheel to a pre-defined pipeline geometry model for offline optimization after detection. However, post-processing constrained smoothing models have inherent limitations. They cannot provide accurate information online in real time and struggle to correct fundamental systematic errors in the odometer wheel at bends, resulting in insufficient absolute positioning accuracy and odometer reliability for the detector.

[0061] To solve the above problems, see [link to relevant documentation]. Figure 1 This application provides a method for online inertial navigation and odometer correction based on multi-feature fusion within a pipeline, comprising:

[0062] S101. Real-time acquisition of detection data; the detection data is the data collected by the sensor.

[0063] The detection data refers to the raw measurement data synchronously collected by different functional sensors integrated on the detector during its operation inside the pipeline.

[0064] Specifically, the sensors include an inertial measurement unit, an odometer wheel, and a magnetic flux leakage sensor; real-time acquisition of detection data includes:

[0065] The inertial measurement unit collects specific force and angular velocity information in real time along the three orthogonal axes of the detector.

[0066] The inertial measurement unit (IMU) directly measures the inertial physical quantities generated by the detector as it moves within the pipe. These inertial physical quantities include specific force and angular velocity information along the detector's three orthogonal axes. Specific force information, measured by the accelerometer unit within the IMU, reflects the acceleration effect caused by all external forces acting on the detector except gravity. Angular velocity information, measured by the gyroscope unit within the IMU, reflects the angular rate of the detector's rotation around its three orthogonal axes.

[0067] Displacement pulse data is collected in real time using a mileage wheel; the displacement pulse data is pulse count data that characterizes the distance the detector travels.

[0068] The displacement pulse data is collected by an integrated odometer wheel on the detector. The odometer wheel is a contact or non-contact displacement measurement device that operates based on the detector's movement relative to the pipe's inner wall. The displacement pulse data is a pulse count representing the detector's travel distance. The odometer wheel generates a pulse signal each time it senses the detector's travel of a fixed unit distance, and the relative displacement change is represented by accumulating the pulse count.

[0069] Pipeline characteristic data is collected in real time using a magnetic flux leakage sensor. The pipeline characteristic data is the magnetic flux leakage signal data corresponding to the circumferential weld of the pipeline.

[0070] The detector integrates a magnetic flux leakage sensor to collect pipeline characteristic data. This sensor operates based on magnetic principles and is used to detect the physical properties of the pipe wall. The pipeline characteristic data consists of magnetic flux leakage signal data corresponding to the circumferential weld seam. When the detector passes the circumferential weld seam, the continuity of the pipe wall material changes, and the magnetic flux leakage sensor can capture the resulting magnetic field disturbance signal. This magnetic flux leakage signal data forms the direct basis for identifying circumferential weld seam events.

[0071] After acquiring the specific force and angular velocity information generated by the inertial measurement unit (IMU) detector moving within the pipeline, the displacement pulse data collected by the odometer wheel, and the pipeline characteristic data acquired in real time by the magnetic flux leakage sensor, it is necessary to perform time synchronization processing on the force and angular velocity information, displacement pulse data, and pipeline characteristic data. This time synchronization processing involves timing calibration of the force and angular velocity information, displacement pulse data, and pipeline characteristic data. Differences in the data acquisition clocks within different sensors lead to inconsistencies in the acquisition time references of the motion data, displacement pulse data, and pipeline characteristic data. Time synchronization processing establishes a unified time reference, strictly aligning the acquisition times of the motion data, displacement pulse data, and pipeline characteristic data. This ensures that data points reflecting the detector's state at the same moment in the motion data, displacement pulse data, and pipeline characteristic data have synchronized time markers, eliminating potential errors in subsequent calculations caused by timing misalignment between the motion data, displacement pulse data, and pipeline characteristic data.

[0072] S102. Determine observation information based on detection data.

[0073] Among them, the observation information characterizes the characteristic events corresponding to the pipeline, and the characteristic events include at least one of the following: circumferential weld events, elbow events, and zero-speed events.

[0074] Specifically, based on the time-synchronized observation information, characteristic events during the detector's operation within the pipeline are identified. The identification of circumferential weld events is based on pipeline characteristic data, while the identification of elbow and zero-velocity events is based on specific force and angular velocity information. According to the identified characteristic event type, observation information corresponding to the circumferential weld event, elbow event, or zero-velocity event is generated. This observation information serves as input data for the extended Kalman filter algorithm used in state augmentation to perform measurement updates, providing observational constraints for subsequent estimation updates of the inertial device's zero bias and odometer wheel scaling factor, as well as corrections to navigation state quantities.

[0075] Among them, see Figure 2 and Figure 3 The magnetic flux leakage sensor includes a first magnetic flux leakage probe and a second magnetic flux leakage probe arranged axially at a preset distance along the detector's forward direction.

[0076] Observational information is determined based on the detection data, including:

[0077] When the first and second magnetic flux leakage probes sequentially collect magnetic flux leakage signal data corresponding to the same circumferential weld, a circumferential weld event is generated.

[0078] Record the first timestamp corresponding to the peak value in the magnetic flux leakage signal data collected by the first magnetic flux leakage probe, and the second timestamp corresponding to the peak value in the magnetic flux leakage signal data collected by the second magnetic flux leakage probe.

[0079] Calculate the time difference between the first timestamp and the second timestamp.

[0080] Based on the time difference and the preset axial distance between the first and second magnetic flux leakage probes, the instantaneous velocity of the detector when passing through the circumferential weld is obtained; the instantaneous velocity is the observation information corresponding to the circumferential weld event.

[0081] Specifically, the first timestamp corresponding to the peak value in the magnetic flux leakage signal data acquired by the first magnetic flux leakage probe and the second timestamp corresponding to the peak value in the magnetic flux leakage signal data acquired by the second magnetic flux leakage probe are recorded. The first and second timestamps are recorded based on a high-precision time reference provided by a time synchronization system.

[0082] Calculate the time difference between the first timestamp and the second timestamp.

[0083] Based on the time difference and the preset axial distance between the first and second magnetic flux leakage probes, the instantaneous velocity of the detector when passing through the circumferential weld is calculated, and the instantaneous velocity is the observation information corresponding to the circumferential weld event.

[0084] Specifically, when the first and second magnetic leakage probes, which are axially arranged at a preset distance along the detector's forward direction, sequentially acquire magnetic leakage signal data corresponding to the same circumferential weld, the system determines that a circumferential weld event has been generated. Subsequently, based on the high-precision time reference provided by the time synchronization system, the system records the first timestamp of the peak value in the magnetic leakage signal data captured by the first magnetic leakage probe. And the second timestamp of the peak value in the same magnetic flux leakage signal data captured by the second magnetic flux leakage probe. The time difference is obtained by calculating the difference between two timestamps. Combined with the pre-set fixed axial distance between the first and second magnetic flux leakage probes According to the kinematic formula: Calculate the instantaneous velocity of the detector as it passes through the circumferential weld. This instantaneous velocity is independent of the measurement data from the inertial measurement unit (IMU) and the odometer wheel. It is a high-precision absolute velocity reference and serves as the core observation information corresponding to the circumferential weld event, providing key input for the correction of the subsequent observation information prediction model.

[0085] In addition, determining observation information based on detection data also includes:

[0086] Based on the angular velocity information in the detection data, if the angular velocity information along a single axis exhibits a smooth change characteristic and continues to exceed a preset time threshold, the angular velocity information exhibiting the smooth change characteristic is integrated to generate the total angular change.

[0087] Specifically, based on the angular velocity information in the detection data after time synchronization processing, the angular velocity information includes the angular velocity value of a time series. The changes in the angular velocity value along the three orthogonal axes of the detector are monitored in real time. If the angular velocity value of a certain single axis is found to have a continuous, smooth and stable amplitude change characteristic, that is, the angular velocity value continuously increases or decreases monotonically, and the absolute value of the angular velocity value remains within a preset range, and the duration of the smooth change characteristic exceeds a preset time threshold, then the angular velocity information in the time period of the smooth change characteristic is numerically integrated along the time dimension, and the total angle change of the detector around the single axis in the time period is obtained by accumulating the integral.

[0088] If the difference between the total angle change and the preset bend angle is less than the preset angle threshold, a bend event is generated, and the start and end times of the bend event are recorded.

[0089] Specifically, the calculated total angle change is compared with a preset bend angle. The preset bend angle is the standard bend angle of the pipeline, such as 90 degrees, 45 degrees, or 30 degrees. If the difference between the total angle change and the preset bend angle is less than a preset angle threshold, a bend event is determined to have occurred. The preset angle threshold is, for example, 5 degrees, as an allowable recognition tolerance. The times corresponding to the start and end points of this smooth angular velocity change characteristic are recorded, serving as the start and end times of the bend event, respectively. The time interval between the start and end times is defined as the bend passage time interval.

[0090] Based on the start and end times, the time interval for the bend to pass through is obtained.

[0091] Numerical integration is performed on the angular velocity information during the time interval of the bend passage to correct the three-dimensional motion trajectory of the detector's centroid during the bend passage time interval.

[0092] Specifically, the angular velocity information of the bend during the time interval is numerically integrated. The integration process, based on rigid body rotational kinematics, reconstructs the continuous three-dimensional spatial trajectory of the detector's center of mass as it passes through the bend within that time interval. Subsequently, the arc length of this three-dimensional trajectory is calculated.

[0093] Based on the start and end times recorded in the bend event recognition process, the bend passage time interval [t_start, t_end] is defined for the detector to pass through the bend. The bend passage time interval includes the entire process in which the angular velocity exhibits a smooth change characteristic, providing a time range constraint for subsequent trajectory reconstruction.

[0094] Subsequently, the angular velocity information processed by time synchronization within the time interval of the bend is integrated according to the principle of rigid body rotation kinematics and the differential equation of inertial navigation. The attitude change of the detector is calculated step by step through the integration process, and then the continuous three-dimensional motion trajectory of the detector's center of mass during the bend is reconstructed in reverse. This corrects the trajectory deviation caused by the limitations of the odometer wheel measurement and ensures that the trajectory can truly reflect the actual motion path of the detector's center of mass along the center line of the pipeline.

[0095] Based on the three-dimensional motion trajectory, the arc length corresponding to the three-dimensional motion trajectory is generated.

[0096] Obtain the number of pulses generated by the cumulative displacement pulse data within the time interval of passing through the bend, and calculate the mileage increment of the mileage wheel within the time interval of passing through the bend based on the number of pulses and the mileage wheel scaling factor.

[0097] Calculate the length difference between the arc length and the mileage increment to generate the arc length difference; the arc length difference is the observation information corresponding to the bend event.

[0098] Based on the detector's centroid's three-dimensional motion trajectory corrected by numerical integration of angular velocity, continuous numerical integration is performed on the trajectory according to the inertial navigation differential equation. By progressively accumulating the lengths of trajectory micro-elements, the trajectory arc length Simu corresponding to the three-dimensional motion trajectory is generated. Simultaneously, displacement pulse data collected by the odometer wheel during the bend passage time interval is extracted, and the total number of pulses N during this period is accumulated. The currently estimated odometer wheel scaling factor (unit: meters / pulse) in the observation information prediction model is retrieved, and the odometer wheel mileage increment during the bend passage time interval is obtained according to the calculation logic of the odometer wheel scaling factor. The mileage increment reflects the distance traveled measured along the pipe wall by the mileage wheel; finally, the arc length of the three-dimensional motion trajectory is calculated. Mileage increment The length difference between them generates the arc length difference. The arc length difference directly reflects the fundamental systematic error caused by the difference in geometric radii between the detector centroid trajectory and the mileage wheel trajectory at the bend. As the core observation information corresponding to the bend event, it provides a key input for the subsequent observation information prediction model to correct the mileage wheel scaling factor.

[0099] The formula for calculating the mileage increment of the odometer wheel during the time interval of passing the bend is as follows:

[0100] ;

[0101] in, This represents the total number of pulses. Scale_odom is the mileage increment, and Scale_odom is the mileage wheel scaling factor.

[0102] In this embodiment, since the observation information corresponding to the bend event is the arc length difference, and the observation function of this arc length difference is directly related to the mileage wheel scaling factor in the state vector, the resulting observation residual can directly and significantly update the mileage wheel scaling factor. This process achieves source compensation for the fundamental systematic error in mileage measurement at bends.

[0103] In addition, determining observation information based on detection data also includes:

[0104] Based on specific force information and angular velocity information, the variance of specific force information and the variance of angular velocity information are monitored respectively;

[0105] Based on displacement pulse data and the online real-time estimated mileage wheel ratio factor, the instantaneous travel speed of the detector is calculated and generated.

[0106] If, within a consecutive preset time period, the variance of the force information and the variance of the angular velocity information are both lower than their respective preset variance thresholds, and the absolute value of the detector's instantaneous travel velocity is lower than a preset velocity threshold, then a zero-velocity event is determined to have occurred, and the theoretical zero-velocity vector is used as the observation information corresponding to the zero-velocity event.

[0107] Based on the specific force information collected by the accelerometer unit of the inertial measurement unit and the angular velocity information collected by the gyroscope unit of the inertial measurement unit after time synchronization processing, the variance of the two is calculated and monitored within the sliding time window. The variance is used to quantify the degree of fluctuation of the sensor signal around the mean in a short period of time. A low variance value indicates that the detector's motion state is stable or stationary, and the preset variance threshold is calibrated according to the noise characteristics of the inertial measurement unit in a stationary state.

[0108] Simultaneously, based on the displacement pulse data after time synchronization processing, and combined with the latest mileage wheel ratio factor estimated online in real time by the extended Kalman filter algorithm with state amplification, the displacement pulse data is converted into the instantaneous travel speed of the detector in real time. If, within a continuous preset time period, the variance of the force information and the variance of the angular velocity information are both continuously lower than their respective preset variance thresholds, and the absolute value of the instantaneous travel speed of the detector is continuously lower than the preset speed threshold, then a zero-speed event is determined to have occurred. The theoretical zero-speed vector [0, 0, 0]^T, in which the three axial velocity components in the navigation coordinate system are all zero, is taken as the observation information corresponding to the zero-speed event. This observation information provides constraints for the subsequent observation information prediction model to correct the zero bias of the inertial device and suppress error divergence.

[0109] By using the physical constraint of absolute zero velocity provided by zero-velocity events, the velocity error accumulated by the inertial navigation system during the stationary period can be corrected, and the drift of zero bias of the inertial device can be suppressed.

[0110] S103. Input the detection data into the observation information prediction model and output the predicted value of the observation information.

[0111] The observation information prediction model includes an extended Kalman filter algorithm with state amplification; the predicted observation information value includes at least a mileage wheel scaling factor, which is the estimated state quantity incorporated by the extended Kalman filter algorithm after state amplification.

[0112] Specifically, the detection data is used as input, and the extended Kalman filter algorithm for state amplification performs recursive calculations based on the detection data and the dynamic model describing the evolution of the system state to generate predicted values ​​of observation information. The predicted values ​​of observation information include the inertial device zero bias, the odometer wheel scaling factor, the velocity component, the position component, and the attitude component.

[0113] Among them, the extended Kalman filter algorithm with state augmentation can incorporate sensor error parameters and navigation parameters into the same high-dimensional state vector for real-time dynamic joint estimation, rather than only estimating basic navigation parameters such as position, velocity, and attitude. It compensates for sensor errors at their source by leveraging observation constraints provided by inherent pipeline characteristic events to avoid deterioration in positioning accuracy. The state vector is a 16-dimensional vector, defined as follows:

[0114] X = [position component, velocity component, attitude component, inertial device zero bias, odometer wheel scaling factor]^T.

[0115] in:

[0116] The position components (P_x, P_y, P_z) represent the three-dimensional coordinates of the detector in the navigation coordinate system.

[0117] The velocity components (V_x, V_y, V_z) represent the three-dimensional velocity of the detector in the navigation coordinate system.

[0118] The attitude components (φ, θ, ψ) represent the three-dimensional attitude of the detector (roll angle, pitch angle, yaw angle).

[0119] Inertial device zero bias includes accelerometer zero bias (b_ax, b_ay, b_az) and gyroscope zero bias (b_gx, b_gy, b_gz).

[0120] The odometer wheel scale factor (Scale_odom) is the scaling factor that converts the odometer wheel pulse count into actual displacement.

[0121] Incorporating the mileage wheel scaling factor into the predicted value of the observation information as a state variable to be estimated is the key to achieving online real-time calibration, breaking the limitation of the traditional method of treating it as a fixed constant.

[0122] The state equation describes the evolution of the state vector over time, enabling state prediction from time k-1 to time k. The formula is as follows:

[0123] ;

[0124] in, It is a nonlinear state transition function based on the inertial navigation mechanics arrangement equations. For acceleration and angular velocity measurements, To characterize the uncertainty of the state transition model, we need to represent the process noise that follows a Gaussian distribution. Let k be the state vector at time k. The state transition process includes, in sequence, attitude update based on gyroscope measurements, velocity update after conversion from body coordinate system to navigation coordinate system, position update after velocity integration, and error state update based on IMU bias and odometer wheel scaling factor, ensuring full coverage of the temporal evolution of navigation parameters and error parameters.

[0125] The observation equation is used to establish the mapping relationship between the state vector and external observation information. The formula is:

[0126] ;

[0127] in, The observation vector is dynamically changed with the characteristic event. To adapt to the observation functions of different events, To mitigate observation noise that conforms to a Gaussian distribution; for circumferential weld events, the observation vector is the instantaneous precise velocity, and the observation function projects the three-dimensional velocity onto the detector's forward direction through dot product operations to correct random errors in the mileage wheel; for bend events, the observation vector is the arc length difference, and the observation function obtains the predicted mileage increment by multiplying the accumulated pulse count, the nominal scaling factor, and the error coefficient of the scaling factor to be estimated, specifically compensating for the fundamental systematic errors at bends; for zero-speed events, the observation vector is the theoretical zero-speed vector, and the observation function directly extracts the three-dimensional velocity component from the state vector to suppress IMU drift. These three observation models form a multi-level correction system.

[0128] The operational logic of the extended Kalman filter algorithm for state augmentation is as follows: Continuous optimization of the state vector is achieved through an iterative cycle of prediction-update-feedback. Firstly, in the prediction step, the state vector is... The state estimate at time k and the raw IMU measurement data are used to derive the prior state prediction and prior covariance matrix at time k through recursion of the state equation, thus quantifying the prediction uncertainty. Subsequently, a feature event recognition algorithm is run in parallel to monitor the occurrence of circumferential welds, elbows, and zero-speed events in real time and extract the corresponding observation vectors.

[0129] After entering the update step, an observation function is constructed based on the event type and the predicted observation value is calculated. The observation residual is generated by comparing the predicted observation value with the actual observation vector. Then, the Kalman gain is calculated. The prior state prediction value is corrected using the gain and residual, the posterior state estimate is obtained, and the covariance matrix is ​​updated.

[0130] Finally, through closed-loop feedback, the real-time feedback of IMU zero bias in the posterior state is used to compensate for the original IMU data, and the dynamically updated mileage wheel scaling factor feedback is used for mileage increment calculation, providing optimized input for the next iteration cycle, thus forming an adaptive correction closed loop.

[0131] Specifically, the detection data is input into the observation information prediction model, and the predicted observation information values ​​are output, including:

[0132] Construct a first observation function corresponding to the circumferential weld event; the first observation function is to project the three-dimensional velocity component of the state vector in the observation information prediction model onto the detector's forward direction.

[0133] The first observation function is executed by dot product operation, which is the dot product of the three-dimensional velocity components and the axial quantity of the body coordinate system, to obtain the velocity prediction value corresponding to the circumferential weld event.

[0134] Specifically, when inputting detection data into the observation information prediction model to output the predicted observation information value corresponding to the circumferential weld event, it is first necessary to construct a first observation function corresponding to the circumferential weld event. The first observation function takes the state vector in the observation information prediction model as the core basis, first extracts the three-dimensional velocity components (V_x, V_y, V_z) from the state vector, and at the same time clarifies that the axis vector of the body coordinate system corresponding to the detector's forward direction is the X-axis vector of the body coordinate system. Then, a first observation function is established that projects the three-dimensional velocity components of the state vector onto the detector's forward direction. The first observation function derives the velocity prediction benchmark that matches the circumferential weld event through the velocity parameters in the state vector.

[0135] Subsequently, the first observation function is executed through dot product operation. Specifically, the extracted three-dimensional velocity components (V_x, V_y, V_z) are multiplied by the preset body coordinate system axis vectors. The dot product operation is used to project the three-dimensional velocity vectors onto the detector's forward direction, ultimately obtaining the velocity prediction value corresponding to the circumferential weld event. The velocity prediction value will be used to compare with the accurate instantaneous velocity Vw calculated by the leakage magnetic field sensor in the weld event, providing observation residual information for the random error correction of the mileage wheel scaling factor in the extended Kalman filter.

[0136] In addition, the process of inputting detection data into the observation information prediction model and outputting predicted observation information values ​​also includes:

[0137] Construct a second observation function corresponding to the bend event. The second observation function is an integral function within the bend passage time interval.

[0138] Based on the second observation function, the cumulative number of pulses within the time interval of the bend is calculated, multiplied by the mileage wheel scaling factor of the mileage wheel, and then multiplied by the scaling factor error coefficient to be estimated in the state vector to obtain the predicted mileage increment corresponding to the bend event.

[0139] Specifically, when inputting detection data into the observation information prediction model to output the predicted observation information value corresponding to the bend event, it is first necessary to construct a second observation function corresponding to the bend event. The second observation function is an integral function within the bend passage time interval [t_start, t_end], where the time interval [t_start, t_end] comes from the start and end times of bend passage recorded in the bend event identification stage. The role of the second observation function is to establish the correlation between the mileage wheel measurement data and the state vector of the observation information prediction model.

[0140] Subsequently, calculations are performed based on the second observation function. First, the number of pulses N accumulated by the odometer encoder within the aforementioned bend passage time interval is extracted. Then, a preset nominal scale factor of the odometer (unit: meters / pulse) is retrieved. The accumulated pulse number N is multiplied by this nominal scale factor. The result is then further multiplied by the scale factor error coefficient scale_odom_k to be estimated in the state vector of the observation information prediction model. Finally, the predicted mileage increment corresponding to the bend event is obtained. This predicted mileage increment will be used as the basis for subsequent calculation of the arc length difference ΔS (IMU integral true arc length). The difference between the mileage increment and the predicted mileage increment is a key basis for providing observational input for the extended Kalman filter to correct the fundamental systematic error at the mileage wheel bend.

[0141] In addition, the process of inputting detection data into the observation information prediction model and outputting predicted observation information values ​​also includes:

[0142] A third observation function corresponding to the zero-velocity event is constructed. The third observation function is used to extract the observation information and predict the three-dimensional velocity component of the state vector in the model.

[0143] The velocity prediction vector corresponding to the zero-velocity event is obtained by performing calculations using the third observation function.

[0144] Specifically, when inputting detection data into the observation information prediction model to output the predicted observation information value corresponding to the zero-velocity event, a third observation function corresponding to the zero-velocity event is first constructed. The third observation function is based on the 16-dimensional state vector of the observation information prediction model. It is defined that the third observation function is to extract the three-dimensional velocity components (V_x, V_y, V_z) representing the motion state of the detector in the state vector, ensuring that the third observation function directly matches the velocity constraint requirements of the zero-velocity event. Subsequently, the operation is performed through the third observation function to directly extract the real-time estimated value of the three-dimensional velocity components from the state vector updated after the prediction step, forming the velocity prediction vector [V_x, V_y, V_z]^T corresponding to the zero-velocity event. The velocity prediction vector [V_x, V_y, V_z]^T will be used to compare with the theoretical observation vector [0, 0, 0]^T of the zero-velocity event. The resulting observation residual will be used as the basis for the extended Kalman filter to correct the IMU accelerometer bias, gyroscope bias and velocity error, thereby effectively suppressing the IMU error divergence.

[0145] S104. Input the observation information into the observation information prediction model, correct the predicted value of the observation information, and obtain the estimated value of the observation information.

[0146] Specifically, the measurement update step uses an extended Kalman filter algorithm to amplify the input state of the observation information. The extended Kalman filter algorithm converts the predicted values ​​of the observation information into corresponding predicted observation values ​​using a pre-defined observation information prediction model. The predicted observation values ​​are then compared with the input observation information to obtain the observation residuals. Based on these residuals, the predicted observation information values ​​are weighted and corrected to obtain estimated observation information values, thereby enabling online real-time updates of the inertial device's zero bias and the odometer wheel's scaling factor.

[0147] Specifically, the observation information is input into the observation information prediction model, the predicted values ​​of the observation information are corrected, and the estimated values ​​of the observation information are obtained, including:

[0148] The instantaneous velocity, arc length difference, and theoretical zero velocity vector are input into the observation information prediction model, and compared with the corresponding predicted velocity value, predicted mileage increment, and velocity prediction vector to generate the observation residual.

[0149] The inertial device zero bias in the predicted value of the observation information is weighted and corrected based on the observation residuals to obtain the estimated value of the observation information.

[0150] Specifically, see Figure 2 and Figure 3 Once a circumferential weld event is identified and corresponding instantaneous velocity observation information is generated, the instantaneous velocity is input to the extended Kalman filter algorithm amplified by the state to perform measurement updates.

[0151] The extended Kalman filter (EDF) algorithm with state augmentation performs a measurement update step. In this step, the EDF algorithm first calculates a predicted velocity based on the observed information prediction value using a pre-defined observation information prediction model. The predicted velocity is the detector velocity deduced by the EDF algorithm based on the observed information prediction value.

[0152] Subsequently, the extended Kalman filter algorithm with state amplification compares the instantaneous velocity, which serves as the input observation information, with the calculated velocity prediction value to obtain the observation residual. The observation residual simultaneously reflects the error between the predicted velocity and the predicted value of the mileage wheel scaling factor.

[0153] The extended Kalman filter algorithm with state augmentation performs a weighted correction on the predicted values ​​of the observation information based on the observation residuals. Since the mileage wheel scaling factor is a component of the predicted values ​​of the observation information, this correction process directly realizes the online real-time update of the mileage wheel scaling factor.

[0154] Updating the mileage wheel scale factor using instantaneous speed observation information provided by circumferential weld events can effectively compensate for the instantaneous error of the mileage wheel scale factor caused by random factors such as mileage wheel slippage and freewheeling, providing more accurate parameters for subsequent mileage conversion based on the updated mileage wheel scale factor.

[0155] In this embodiment, the observation information provided by the circumferential weld event is instantaneous velocity, which serves as the absolute velocity reference for the detector. By comparing the instantaneous velocity with the velocity predicted by the extended Kalman filter algorithm with state amplification, the resulting velocity observation residual directly reflects the instantaneous error of the odometer wheel scaling factor. Therefore, this observation residual can quickly and sensitively update the odometer wheel scaling factor. This process effectively suppresses instantaneous errors caused by random factors such as odometer wheel slippage and idling.

[0156] In addition, an extended Kalman filter algorithm is needed to amplify the state by using the arc length difference corresponding to the bend event as the input of the observation information, so as to estimate and update the odometer scale factor online in real time. When the observation information is the theoretical zero velocity vector corresponding to the zero velocity event, the theoretical zero velocity vector is input into the extended Kalman filter algorithm to update the zero bias of the inertial device online in real time.

[0157] Specifically, in some embodiments, see [link to relevant documentation]. Figure 4 To update the zero bias of the inertial device and the odometer scale factor in real time online, including:

[0158] The extended Kalman filter algorithm, which uses the arc length difference corresponding to the bend event as the input observation information, is used to estimate and update the mileage wheel scale factor online in real time.

[0159] Specifically, see Figure 4Once a bend event is identified and its corresponding arc length difference is generated, this arc length difference is used as the observation information input to the extended Kalman filter algorithm for state amplification.

[0160] The extended Kalman filter algorithm with state augmentation performs a measurement update step. In this step, the extended Kalman filter algorithm first calculates a predicted arc length difference based on the predicted value from the observation information using a preset observation model. This predicted arc length difference is calculated by the extended Kalman filter algorithm based on the cumulative number of pulses within the bend passage time interval and the mileage wheel scaling factor in the current observation information prediction value.

[0161] Subsequently, the arc length difference, which serves as the input of observational information, is compared with the calculated predicted mileage increment to obtain the observation residual. This observation residual directly reflects the error of the predicted value of the wheel scale factor for the current mileage.

[0162] The extended Kalman filter algorithm with state augmentation performs a weighted correction on the predicted values ​​of the observation information based on the observation residuals. Since the mileage wheel scaling factor is a component of the predicted values ​​of the observation information, this correction process directly realizes the online real-time update of the mileage wheel scaling factor.

[0163] Updating the mileage wheel scaling factor using arc length difference observations from bend events can directly compensate for the fundamental systematic error caused by the difference in geometric radii between the detector centroid trajectory and the mileage wheel's travel trajectory at bends. This is a strong geometric constraint correction to the mileage wheel scaling factor, providing more accurate conversion parameters for subsequent mileage calculations.

[0164] See Figure 5 , Figure 6 and Figure 7 Online real-time updates of inertial device zero bias and odometer scale factor, including:

[0165] When the observed information is the theoretical zero velocity vector corresponding to the zero velocity event, the theoretical zero velocity vector is input into the extended Kalman filter algorithm of state amplification to update the zero bias of the inertial device online in real time.

[0166] Specifically, after the zero-velocity event is determined to have occurred and the theoretical zero-velocity vector is generated, the theoretical zero-velocity vector is used as the observation input state amplification extended Kalman filter algorithm.

[0167] The extended Kalman filter (EDF) algorithm with state augmentation performs a measurement update step. In this step, the EDF algorithm first predicts values ​​based on observation information and calculates a predicted velocity vector using an observation information prediction model. This predicted velocity vector represents the detector's three-dimensional velocity, calculated by the EDF algorithm based on the predicted values.

[0168] Subsequently, the extended Kalman filter algorithm for state amplification compares the theoretical zero-velocity vector, which serves as the input observation information, with the calculated predicted velocity vector to obtain the observation residual. This observation residual directly reflects the deviation of the predicted velocity from the true zero velocity. Under zero-velocity events, this deviation is related to the inertial device zero-bias error in the state vector.

[0169] The extended Kalman filter algorithm for state augmentation applies a weighted correction to the predicted values ​​of the observation information based on the observation residuals. Since the inertial device's zero bias is a component of the predicted values ​​of the observation information, this correction process enables online real-time updates of the inertial device's zero bias.

[0170] By updating the zero bias of the inertial device using the theoretical zero velocity vector provided by the zero velocity event, the accumulation of velocity and position errors caused by zero bias drift in the inertial navigation system can be effectively suppressed during the detector's stationary period, providing more accurate sensor error compensation for navigation calculations during subsequent motion.

[0171] S105. Based on the updated mileage wheel ratio factor in the estimated value of the observation information, the displacement pulse data in the detection data is converted to obtain the target running mileage of the detector.

[0172] Specifically, based on the updated mileage wheel scaling factor in the estimated values ​​of the observation information, the displacement pulse data after time synchronization processing is converted. The conversion step uses the updated mileage wheel scaling factor to convert the pulse counts in the displacement pulse data into the corresponding actual distance increments. By accumulating the converted distance increments, the corrected operating mileage of the detector is obtained.

[0173] Specifically, based on the updated mileage scale factor in the estimated values ​​of the observation information, the displacement pulse data in the detection data are converted to obtain the target operating mileage of the detector, including:

[0174] Extract the mileage wheel scaling factor after correction for circumferential weld events and elbow events from the estimated values ​​of the observation information;

[0175] The cumulative number of pulses in the displacement pulse data is multiplied by the updated mileage wheel scale factor to obtain the distance increment for each time period.

[0176] The target operating mileage of the detector is obtained by summing the distance increments over all time periods.

[0177] Specifically, the mileage wheel scaling factor, which has been dynamically corrected twice for circumferential weld events and elbow events, is extracted from the estimated values ​​of the observation information output by the extended Kalman filter. The mileage wheel scaling factor is a dynamic parameter obtained by the system in real time through a hierarchical correction mechanism, rather than a fixed constant, which has fully offset the measurement deviations caused by factors such as slippage, idling, and the difference between the inner and outer diameters of elbows.

[0178] Subsequently, displacement pulse data is extracted, and time periods are divided according to the time dimension. The cumulative number of pulses in each time period is multiplied by the updated mileage wheel scaling factor for the corresponding time period. Based on the conversion logic of distance increment = cumulative number of pulses × dynamically corrected scaling factor, the travel distance increment of the detector in each time period is obtained, ensuring that the distance calculation in each time period is based on the current optimal scaling factor.

[0179] Finally, the distance increments for all consecutive time periods are summed in chronological order. Through continuous incremental accumulation throughout the entire process, the target running mileage of the detector in the pipeline is finally obtained. The target running mileage has incorporated the correction information of the inherent characteristics of the pipeline, realizing long-distance, high-precision online mileage output, which meets the core requirements of pipeline defect location and GIS system integration.

[0180] Similar parts between the embodiments provided in this application can be referred to mutually. The specific implementation methods provided above are only a few examples under the overall concept of this application and do not constitute a limitation on the scope of protection of this application. For those skilled in the art, any other implementation methods extended from the solution of this application without creative effort shall fall within the scope of protection of this application.

Claims

1. A method for online inertial navigation and odometer correction based on multi-feature fusion within a pipeline, characterized in that, A detector used in pipelines, the detector comprising multiple sensors, including: Real-time acquisition of detection data; the detection data is the data acquired by the sensor, wherein the sensor includes a magnetic flux leakage sensor, and the magnetic flux leakage sensor is used to acquire pipeline characteristic data in real time, wherein the pipeline characteristic data is the magnetic flux leakage signal data corresponding to the circumferential weld of the pipeline; The observation information is determined based on the detection data; the observation information characterizes the characteristic events corresponding to the pipeline, and the characteristic events include at least one of the circumferential weld event, elbow event and zero speed event, wherein the magnetic flux leakage sensor includes a first magnetic flux leakage probe and a second magnetic flux leakage probe arranged axially at a preset distance along the advancing direction of the detector; The determination of observation information based on the detection data includes: When the first magnetic flux leakage probe and the second magnetic flux leakage probe sequentially collect magnetic flux leakage signal data corresponding to the same circumferential weld, a circumferential weld event is generated. Record the first timestamp corresponding to the peak value in the magnetic flux leakage signal data collected by the first magnetic flux leakage probe, and the second timestamp corresponding to the peak value in the magnetic flux leakage signal data collected by the second magnetic flux leakage probe; Calculate the time difference between the first timestamp and the second timestamp; Based on the time difference and the preset axial distance between the first and second magnetic flux leakage probes, the instantaneous velocity of the detector when it passes through the circumferential weld is obtained; the instantaneous velocity is the observation information corresponding to the circumferential weld event. The detection data is input into the observation information prediction model, and the predicted observation information value is output. The observation information prediction model includes an extended Kalman filter algorithm with state amplification. The predicted observation information value includes at least a mileage wheel scaling factor, which is the estimated state quantity incorporated by the extended Kalman filter algorithm after state amplification. A first observation function corresponding to the circumferential weld event is constructed. The first observation function projects the three-dimensional velocity components of the state vector in the observation information prediction model onto the detector's forward direction. The first observation function is executed by dot product operation, which is the dot product of the three-dimensional velocity component and the axial quantity of the body coordinate system, to obtain the velocity prediction value corresponding to the circumferential weld event. The observation information is input into the observation information prediction model, and the predicted value of the observation information is corrected to obtain the estimated value of the observation information. Based on the updated mileage wheel scaling factor in the estimated value of the observation information, the displacement pulse data in the detection data is converted to obtain the target operating mileage of the detector.

2. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 1, characterized in that, The sensor includes an inertial measurement unit and a mileage wheel; The real-time acquired detection data includes: The inertial measurement unit collects specific force and angular velocity information in real time along the three orthogonal axes of the detector. Displacement pulse data is collected in real time through the mileage wheel; the displacement pulse data is pulse count data that characterizes the running distance of the detector.

3. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 2, characterized in that, The step of determining observation information based on the detection data further includes: Based on the angular velocity information in the detection data, if the angular velocity information along a single axis is detected to exhibit a smooth change characteristic and continues to exceed a preset time threshold, the angular velocity information exhibiting the smooth change characteristic is integrated to generate the total angular change. If the difference between the total angle change and the preset bend angle is less than the preset angle threshold, a bend event is generated, and the start and end times of the bend event are recorded. Based on the start time and the end time, the time interval for the bend to pass is obtained; The angular velocity information of the bend during the time interval is numerically integrated to correct the three-dimensional motion trajectory of the detector's centroid during the bend's time interval. Based on the three-dimensional motion trajectory, generate the arc length corresponding to the three-dimensional motion trajectory; The number of pulses generated by the cumulative displacement pulse data within the time interval of the bend passage is obtained, and the mileage increment of the mileage wheel within the time interval of the bend is calculated based on the number of pulses and the mileage wheel scaling factor. Calculate the length difference between the arc length and the mileage increment to generate the arc length difference; the arc length difference is the observation information corresponding to the bend event.

4. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 3, characterized in that, The method further includes inputting the detection data into the observation information prediction model and outputting the predicted observation information value, and also includes: Construct a second observation function corresponding to the bend event, wherein the second observation function is an integral function over the time interval of the bend passage; Based on the second observation function, the cumulative number of pulses within the time interval of the bend is calculated, multiplied by the mileage wheel scaling factor of the mileage wheel, and then multiplied by the scaling factor error coefficient to be estimated in the state vector to obtain the predicted mileage increment corresponding to the bend event.

5. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 4, characterized in that, The step of determining observation information based on the detection data further includes: Based on the specific force information and the angular velocity information, monitor the variance of the specific force information and the variance of the angular velocity information respectively; Based on the displacement pulse data and the online real-time estimated mileage wheel ratio factor, the instantaneous travel speed of the detector is calculated and generated. If, within a consecutive preset time period, the variance of the force information and the variance of the angular velocity information are both lower than their respective preset variance thresholds, and the absolute value of the instantaneous travel speed of the detector is lower than a preset speed threshold, then a zero-speed event is determined to have occurred, and the theoretical zero-speed vector is used as the observation information corresponding to the zero-speed event.

6. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 5, characterized in that, The method further includes inputting the detection data into the observation information prediction model and outputting the predicted observation information value, and also includes: Construct a third observation function corresponding to the zero-velocity event, wherein the third observation function is to extract the three-dimensional velocity component of the state vector in the prediction model based on the observation information; The velocity prediction vector corresponding to the zero-speed event is obtained by performing calculations using the third observation function.

7. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 6, characterized in that, The step of inputting the observation information into the observation information prediction model, correcting the predicted value of the observation information, and obtaining the estimated value of the observation information includes: The instantaneous velocity, the arc length difference, and the theoretical zero velocity vector are input into the observation information prediction model, and compared with the corresponding predicted velocity value, predicted mileage increment, and the predicted velocity vector to generate the observation residual. The inertial device zero bias in the predicted observation information is weighted and corrected based on the observation residual to obtain the estimated observation information.

8. The inertial navigation and online odometer correction method for multi-feature fusion in pipelines according to claim 7, characterized in that, Based on the updated mileage wheel scaling factor in the estimated value of the observed information, the displacement pulse data in the detection data are converted to obtain the target operating mileage of the detector, including: Extract the mileage wheel scaling factor from the estimated value of the observed information after correction for circumferential weld events and elbow events; The cumulative number of pulses in the displacement pulse data is multiplied by the updated mileage wheel ratio factor to obtain the distance increment for each time period; The target operating mileage of the detector is obtained by summing the distance increments over all time periods.

Images