A high-precision dynamic measurement module while-drilling real-time measurement system

Abstract

The application relates to the technical field of downhole measurement, and discloses a high-precision real-time measurement system of a dynamic measurement module while drilling. The method comprises the following steps: a collection module synchronously collects sensor data and classifies environment states; an embedding module extracts an error weight vector from a mapping matrix according to the environment states and embeds the error weight vector into a covariance matrix; a fusion module substitutes an adaptive covariance matrix into Kalman filtering to perform data fusion; and an updating module calculates a residual error to trigger hardware calibration and output measurement parameters. The application improves the measurement accuracy under different working conditions, the pertinence and real-time performance of algorithm compensation, and the reliability of long-term continuous operation of the system.

Description

Technical Field

[0001] This application relates to the field of downhole measurement technology, and in particular to a real-time measurement system for a high-precision dynamic measurement module while drilling. Background Technology

[0002] Measurement while drilling (MWD) is a crucial measurement method in coal mine downhole and oil drilling operations. Existing high-precision MWD systems typically employ a combination of triaxial MEMS accelerometers, magnetometers, and gyroscopes to measure key parameters such as well inclination angle, azimuth angle, and drill string rotation speed. The basic workflow involves acquiring raw data from sensors, fusing the multi-sensor data using a Kalman filter algorithm, and then obtaining attitude parameters such as well inclination angle and tool face angle through geometric calculations. Finally, the measurement results are transmitted to the ground control system via a communication interface for drilling trajectory control. Existing Kalman filters typically use fixed process noise covariance matrices and observation noise covariance matrices. These parameters are set empirically during system initialization or obtained through offline calibration and remain constant throughout the measurement process. Error compensation methods primarily target static sensor errors such as zero bias and scaling factor through one-time calibration correction. Some systems employ temperature compensation models to correct the sensor's temperature drift characteristics.

[0003] However, existing technologies have the following shortcomings: First, the underground environment of coal mines is complex and changeable. During the drilling process, various working conditions such as static drilling, dynamic impact, electromagnetic interference, and drastic temperature changes will be experienced. The nature and intensity of sensor error sources differ significantly under different working conditions. For example, during static drilling, the long-term zero-bias drift of the gyroscope is the main error source, while during dynamic impact, the high-frequency noise of the accelerometer becomes the dominant error. However, existing technologies use Kalman filters with fixed parameters, which cannot dynamically adjust the suppression intensity of the filter for different error sources according to environmental changes. This leads to a serious mismatch between the filter parameters and the actual noise characteristics under certain working conditions, and the measurement error of the well inclination angle can reach more than 0.5 degrees. Secondly, existing error compensation methods typically use a single algorithm to compensate for all types of errors, or only perform pre-calibration corrections for static errors. They lack targeted layered processing mechanisms for dynamic errors such as transient disturbances caused by vibration and long-term drift caused by temperature changes. When the drill bit encounters a severe impact, the compensation algorithm used to suppress long-term drift cannot respond quickly to transient disturbances, while the algorithm used for transient suppression will introduce unnecessary computational overhead and compensation errors in a static environment, resulting in poor compensation effect and poor real-time performance. Summary of the Invention

[0004] This application provides a high-precision real-time measurement system for a dynamic measurement while drilling module, which solves the problems of low measurement accuracy, poor environmental adaptability, and insufficient long-term stability in complex and variable downhole environments caused by the use of fixed filtering parameters and a single compensation algorithm in existing measurement while drilling systems. It improves the measurement accuracy under different working conditions, the pertinence and real-time performance of algorithm compensation, and the reliability of the system for long-term continuous operation.

[0005] This application provides a high-precision real-time measurement system for a dynamic measurement while drilling module, the high-precision real-time measurement system for a dynamic measurement while drilling module includes: The data acquisition module is used to simultaneously acquire triaxial acceleration data, triaxial magnetic field strength data, triaxial angular rate data, and temperature data. Based on the spectral characteristics, magnetic field standard deviation, and temperature change rate, the environmental state is classified to obtain the current environmental state identifier. An embedding module is used to extract the corresponding error weight vector from a preset environment-error weight mapping matrix according to the current environment state identifier. The error weight vector includes accelerometer noise weight, gyroscope drift weight, magnetometer interference weight, and temperature drift weight. Each weight in the error weight vector is embedded into the corresponding diagonal element positions of the reference process noise covariance matrix and the reference observation noise covariance matrix to obtain the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix. The fusion module is used to substitute the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix into the Kalman filter for data fusion to obtain compensated data; The update module is used to calculate the root mean square of the residuals of the compensation data. When the root mean square of the residuals continuously exceeds the threshold, hardware calibration is triggered to update the zero bias parameter. The measurement parameters are calculated and output based on the compensation data.

[0006] The technical solution provided in this application synchronously acquires triaxial acceleration data, triaxial magnetic field strength data, triaxial angular rate data, and temperature data. It then constructs a sliding time window from the multi-source sensor data to perform spectral analysis, extracting the spectral concentration coefficient, calculating the magnetic field standard deviation, and the temperature change rate. Based on the comparison between these three characteristic parameters and preset thresholds, real-time classification and identification of environmental conditions are achieved. This solves the problem that existing technologies cannot accurately identify complex downhole working conditions, providing a reliable environmental basis for subsequent adaptive parameter adjustments and hierarchical compensation strategies. By establishing an environment-error weight mapping matrix and extracting the corresponding error weight vector from the mapping matrix based on the current environmental state identifier, and embedding the error weight vector into the process noise covariance matrix and the observation noise covariance matrix, the Kalman filter parameters are dynamically and adaptively adjusted according to the environment. This overcomes the parameter mismatch problem caused by using fixed filter parameters in existing technologies. The filter can enhance the suppression of long-term gyroscope drift in static drilling environments and enhance the filtering of high-frequency noise from accelerometers in dynamic impact environments, significantly improving the data fusion accuracy and environmental adaptability under different working conditions. After substituting the adaptive covariance matrix into the Kalman filter for data fusion, the system switches between a long-term drift compensation layer and a transient disturbance suppression layer based on the current environmental state to provide targeted compensation for the fused data. This solves the problem that existing technologies using a single compensation algorithm cannot take into account errors of different natures. In static and temperature-changing environments, the system effectively corrects long-term drift through a temperature-zero biased linear model. In dynamic impact environments, the system quickly predicts and eliminates transient disturbances through a GRU neural network. This avoids the compensation mismatch caused by using a long-term drift algorithm to handle transient impacts or a transient suppression algorithm to handle slow temperature drift, thus improving the targeting and real-time performance of error compensation. Attached Figure Description

[0007] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of an embodiment of the real-time measurement system of the high-precision drilling dynamic measurement module in this application. Figure 2 This is a schematic diagram illustrating the error weight allocation under different environmental conditions in the embodiments of this application; Figure 3 This is a schematic diagram illustrating the verification of the layering error compensation effect in the embodiments of this application. Detailed Implementation

[0009] This application provides a high-precision real-time measurement system for a dynamic measurement while drilling module. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0010] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the real-time measurement system for the high-precision drilling dynamic measurement module in this application includes: The data acquisition module is used to simultaneously acquire triaxial acceleration data, triaxial magnetic field strength data, triaxial angular rate data, and temperature data. Based on the spectral characteristics, magnetic field standard deviation, and temperature change rate, the environmental state is classified to obtain the current environmental state identifier. The acquisition module synchronously reads the raw voltage signals from the triaxial MEMS accelerometer, magnetometer, gyroscope, and temperature sensor at a sampling frequency of 1000Hz via the MCU's SPI digital interface. After AD conversion, it obtains digital triaxial acceleration data, triaxial magnetic field strength data, triaxial angular rate data, and temperature data. All data are timestamped using a hardware synchronization clock to ensure time alignment error is less than 1 millisecond. A 100-millisecond sliding window containing 100 consecutive sampling points is constructed for the acquired triaxial acceleration data. After calculating the composite acceleration value, a Fast Fourier Transform is performed to extract the ratio of the energy of the dominant frequency component with the highest energy in the 0-500Hz frequency band to the total energy of the entire frequency band as the spectral concentration coefficient. Simultaneously, the mean and standard deviation of the magnetic field strength modulus within the window are calculated as magnetic field fluctuation characteristics, and the temperature change rate is obtained by dividing the difference between two consecutive temperature data samples by the time interval. The system is classified as follows: static drilling state is determined based on a spectral concentration coefficient greater than 0.6, a magnetic field standard deviation less than 50 nanoteslas, and an absolute temperature change rate less than 3 degrees Celsius per minute; dynamic impact state is determined based on a spectral concentration coefficient less than 0.3 and a peak acceleration greater than twice the gravitational acceleration; electromagnetic interference state is determined based on a magnetic field standard deviation greater than 200 nanoteslas; and temperature drastic change state is determined based on an absolute temperature change rate greater than 5 degrees Celsius per minute. Finally, the current environmental state identifier is output as the index basis for subsequent error weight extraction.

[0011] An embedding module is used to extract the corresponding error weight vector from a preset environment-error weight mapping matrix according to the current environment state identifier. The error weight vector includes accelerometer noise weight, gyroscope drift weight, magnetometer interference weight, and temperature drift weight. Each weight in the error weight vector is embedded into the corresponding diagonal element positions of the reference process noise covariance matrix and the reference observation noise covariance matrix to obtain the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix. The embedded module maintains a four-row, six-column environment-error weight mapping matrix pre-established through offline calibration experiments. The four rows correspond to four environmental states: static drilling, dynamic impact, electromagnetic interference, and rapid temperature change. The six columns correspond to six error sources: accelerometer noise, gyroscope drift, magnetometer interference, temperature drift, bias error, and scaling error. The value of each element in the matrix is ​​obtained by normalizing the residual variance contribution after collecting data for two hours in a simulated downhole environment in the laboratory using the least squares method. The current environmental state identifier is used as the row index to directly look up the six weight coefficients for the corresponding row. For example, the weight vector corresponding to the static drilling state is gyroscope drift weight 0.65, accelerometer noise weight 0.15, magnetometer interference weight 0.10, and temperature drift weight 0.10; the weight vector corresponding to the dynamic impact state is accelerometer noise weight 0.70 and gyroscope drift weight 0.15. The first three weight values ​​in the extracted weight vector are multiplied by the diagonal elements of the reference process noise covariance matrix, and the last three weight values ​​are multiplied by the diagonal elements of the reference observation noise covariance matrix. This element-wise multiplication operation completes the weight embedding, and the output adaptive covariance matrix can dynamically adjust the suppression strength of the Kalman filter for different error sources according to the current environment.

[0012] The fusion module is used to substitute the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix into the Kalman filter for data fusion to obtain compensated data; Specifically, the fusion module substitutes the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix into the five recursive formulas of Kalman filtering, and sequentially performs state prediction, covariance prediction, Kalman gain calculation, state update, and covariance update. The state vector contains six components: well inclination angle, azimuth angle, rotation angle, and three-axis angular rate. The measurement vector contains six components: three-axis acceleration and three-axis magnetic field strength. By iteratively fusing multi-sensor data, the optimal estimated attitude angle and angular rate are obtained. The compensation algorithm level is selected based on the transmitted current environmental state identifier. When the identifier is static drilling or a sudden temperature change state, the long-term drift compensation layer is activated. The compensation amount is calculated using a linear model of temperature and zero bias drift and subtracted from the fused data. When the identifier is dynamic impact state, the transient disturbance suppression layer is activated. Nine values, including the residuals of the current fused data and the previous three cycles, are input into a pre-trained GRU neural network. The network outputs a six-dimensional disturbance prediction vector, which is then subtracted from the fused data to complete the compensation. Finally, high-precision attitude and angular rate data after environmental adaptive compensation are output.

[0013] The update module is used to calculate the root mean square of the residuals of the compensation data. When the root mean square of the residuals continuously exceeds the threshold, hardware calibration is triggered to update the zero bias parameter. The measurement parameters are calculated and output based on the compensation data.

[0014] Specifically, the update module uses the compensated triaxial acceleration data to independently calculate the reference well inclination angle using an inverse cosine function. The inclination angle in the compensated data is subtracted from the reference well inclination angle to obtain the residual. A sliding window containing the residual values ​​of the most recent 50 sampling periods is constructed. The square root of the residual is obtained by calculating the sum of squares of the residuals within the window, dividing by 50, and then taking the square root. The root root of the residual is compared with the dynamic thresholds corresponding to the current environmental state. The threshold for static drilling is set to 0.05 degrees, the threshold for dynamic impact is relaxed to 0.3 degrees, the threshold for electromagnetic interference is set to 0.2 degrees, and the threshold for sudden temperature change is set to 0.15 degrees. When the root root of the residual exceeds the corresponding threshold for 50 consecutive sampling periods (50 milliseconds), the software compensation is deemed saturated, triggering the hardware calibration process. A pause command is sent to the drilling control system to stop the drill string. In the static state, 500 consecutive samples are collected, and the arithmetic mean of the three-axis zero bias of the magnetometer and gyroscope is calculated as the update parameter and written to the MCU's non-volatile memory. Drilling resumes after calibration. Based on the acceleration and magnetic field components in the compensation data, the well inclination angle and tool face angle are calculated using the inverse cosine function and the four-quadrant arctangent function, respectively. The drill string rotation speed is calculated by weighted fusion of the gyroscope angular rate integral value and the acceleration vibration frequency analysis results. The quaternion attitude and three-dimensional position coordinates are calculated by fusing the three sensor data through the Mahony complementary filter. All parameters are encapsulated into a 49-byte data packet according to a predetermined protocol and sent to the external control system every 10 milliseconds via the RS422 interface at a baud rate of 115200.

[0015] In one specific embodiment, the acquisition module is used for: The raw sensor data is synchronously acquired through the SPI interface and timestamp aligned to obtain multi-source sensor data. A sliding time window is constructed for the multi-source sensor data, a fast Fourier transform is performed on the acceleration data to extract the spectral concentration coefficient, and the standard deviation of the magnetic field data and the rate of change of the temperature data are calculated. The current environmental state identifier is obtained by classifying and determining the relationship between the spectral concentration coefficient, the standard deviation, the rate of change, and a preset threshold.

[0016] Specifically, the acquisition module synchronously reads the raw digital signals output by the triaxial MEMS accelerometer, magnetometer, gyroscope, and temperature sensor at a fixed sampling frequency through the MCU's SPI digital interface. All sensor data are sampled at the same time through a hardware synchronization clock mechanism and marked with a unified timestamp to ensure that the data from different sensors are strictly aligned in the time dimension, avoiding data association errors caused by inconsistent sampling times. The multi-source sensor data after timestamp alignment includes ten physical quantities, including triaxial acceleration, triaxial magnetic field strength, triaxial angular velocity, and temperature. These data serve as the raw input for subsequent environmental state identification.

[0017] The acquisition module constructs a fixed-length sliding time window for multi-source sensor data. The window stores continuous data from the most recent sampling periods in a first-in-first-out manner. After calculating the composite magnitude of the triaxial acceleration data within the window, a fast Fourier transform is performed. The ratio of the dominant frequency energy to the total spectral energy is extracted through spectral analysis as the spectral concentration coefficient, which reflects the periodic strength of the vibration signal. At the same time, the mean and standard deviation of the magnetic field strength magnitude sequence within the window are calculated. The standard deviation reflects the intensity of magnetic field fluctuations. For temperature data, the difference between the current sample value and the previous sample value is divided by the time interval to obtain the temperature change rate. The acquisition module compares the extracted spectral concentration coefficient, magnetic field standard deviation, and temperature change rate with multiple preset thresholds. Based on different combinations of comparison results, the current downhole working condition is classified into one of the following: static drilling state, dynamic impact state, electromagnetic interference state, or temperature drastic change state, and the corresponding environmental state identification code is output.

[0018] In one specific embodiment, the embedded module includes: A mapping unit is used to establish an environment-error weight mapping matrix, wherein the row index of the mapping matrix corresponds to the environment state and the column index corresponds to the error source type; The fitting unit is used to collect data through offline calibration experiments, fit the residual distribution using the least squares method and normalize it to obtain the weight coefficients in the mapping matrix; The extraction unit is used to extract the corresponding row from the mapping matrix using the current environment state identifier as the row index, and obtain the error weight vector.

[0019] Specifically, the mapping unit is responsible for establishing a four-row, six-column environment-error weight mapping matrix data structure within the embedding module. Row indices 1 to 4 correspond to four downhole environment categories: static drilling, dynamic impact, electromagnetic interference, and temperature drastic change. Column indices 1 to 6 correspond to six types of sensor error sources: accelerometer noise, gyroscope drift, magnetometer interference, temperature drift, zero bias error, and scaling error. Each element in the matrix stores the contribution weight coefficient of a specific error source to the total measurement error under a specific environmental condition. The fitting unit simulates four typical downhole working conditions in a laboratory environment, continuously collects sensor data under each condition, and records the corresponding true reference. It uses the least squares method to fit and decompose the residual between the actual measurement value and the true value, calculates the variance contribution ratio of each error source in the residual, and normalizes the variance contribution ratio of all error sources so that their sum is 1. The normalized value is the weight coefficient of the corresponding position in the mapping matrix.

[0020] The extraction unit receives the current environmental state identifier output by the acquisition module as a lookup index. It uses this identifier value as a row index to directly access the storage address of the corresponding row in the mapping matrix. From this row, it sequentially reads the weight coefficient values ​​stored in the six column positions. These six weight coefficients form an error weight vector in sequence. The value of each element in this vector reflects the severity of the corresponding error source under the current environmental state. The larger the weight value, the more significant the impact of the error source on the measurement accuracy in the current environment, requiring stronger compensation. The error weight vector output by the extraction unit changes dynamically with the switching of environmental states, enabling the subsequent covariance matrix to adaptively adjust the filter parameters according to the real-time operating conditions. The embedding module multiplies each weight in the extracted error weight vector with the diagonal elements of the reference process noise covariance matrix and the reference observation noise covariance matrix element by element. After completing the weight embedding operation, the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix are obtained.

[0021] In one specific embodiment, the error weight vector includes accelerometer noise weight, gyroscope drift weight, magnetometer interference weight, and temperature drift weight, including: setting the gyroscope drift weight to 0.65 and the accelerometer noise weight to 0.15 in static drilling state; setting the accelerometer noise weight to 0.70 and the gyroscope drift weight to 0.15 in dynamic impact state; and forming the error weight vector according to the weight coefficients corresponding to different environmental states.

[0022] Specifically, the error weight vector adopts a differentiated weight configuration strategy according to different environmental states. When the system is in a static drilling state, the long-term zero-bias drift of the gyroscope becomes the main source of error because the drill bit moves smoothly and the vibration interference is small. Therefore, the gyroscope drift weight is set to 0.65 to dominate, while the accelerometer noise weight is only set to 0.15. Conversely, when the system switches to a dynamic impact state, the drill bit is subjected to severe impact and generates high-frequency vibration. The instantaneous noise of the accelerometer increases sharply and becomes the main source of error. At this time, the accelerometer noise weight is increased to 0.70, while the gyroscope drift weight is reduced to 0.15. The extraction unit reads all the weight coefficients of the row corresponding to the current environmental state from the mapping matrix and arranges the six values ​​of gyroscope drift weight, accelerometer noise weight, magnetometer interference weight, temperature drift weight, etc., in a fixed order to form the error weight vector. The values ​​of each element of this vector are dynamically adjusted according to the changes in the environmental state indicator to ensure that the compensation algorithm always applies the strongest suppression force to the most significant error source under the current working condition.

[0023] Figure 2 This is a schematic diagram illustrating the error weight allocation under different environmental conditions in the embodiments of this application; Figure 2 The figure shows the weight coefficient distribution of six error sources (accelerometer noise, gyroscope drift, magnetometer interference, temperature drift, zero bias error, and scaling error) under four environmental conditions (static drilling, dynamic impact, electromagnetic interference, and rapid temperature change). The horizontal axis represents the error source type, and the vertical axis represents the weight coefficient values ​​ranging from 0 to 0.8. The bars with different filling patterns correspond to the four environmental conditions. As can be seen from the figure, under the static drilling condition, the gyroscope drift weight reaches 0.65, which is dominant, while the accelerometer noise weight is only 0.15. Under the dynamic impact condition, the accelerometer noise weight rises sharply to 0.70, becoming the most significant error source, while the gyroscope drift weight drops to 0.15. Under the electromagnetic interference condition, the magnetometer interference weight reaches 0.45, which is significantly higher than other error sources. Under the rapid temperature change condition, the temperature drift weight is 0.35, which is at a relatively high level. This weight allocation strategy enables the error compensation algorithm to dynamically adjust the suppression intensity of different error sources according to the real-time environmental conditions, achieving high-precision measurement with environmental adaptability.

[0024] In one specific embodiment, the embedding module further includes: The multiplication unit is used to extract the gyroscope drift weight and accelerometer noise weight from the error weight vector, construct a diagonal matrix, and perform element-wise multiplication with the reference process noise covariance matrix to obtain the adaptive process noise covariance matrix. The extraction unit is used to extract the magnetometer interference weight and temperature drift weight from the error weight vector, construct a diagonal matrix and perform element-wise multiplication with the reference observation noise covariance matrix to obtain the adaptive observation noise covariance matrix.

[0025] Specifically, the multiplication unit in the embedded module selects two values ​​from the error weight vector: gyroscope drift weight and accelerometer noise weight. These two weights are copied into six diagonal elements according to the sensor axes to construct a six-dimensional diagonal matrix. The first three diagonal elements of this matrix fill the gyroscope drift weights corresponding to the three gyroscope axes, and the last three diagonal elements fill the accelerometer noise weights corresponding to the three accelerometer axes. The multiplication unit reads a pre-calibrated reference process noise covariance matrix from the system memory. This reference matrix is ​​a six-dimensional diagonal matrix storing the reference values ​​of process noise variance for each sensor under standard conditions. The multiplication unit performs element-wise multiplication between the constructed weight diagonal matrix and the reference process noise covariance matrix, that is, multiplying the diagonal elements at corresponding positions of the two matrices to obtain new diagonal elements. The resulting adaptive process noise covariance matrix reflects the actual intensity of process noise under the current environmental conditions. The covariance elements corresponding to error sources with larger weight coefficients are amplified, causing the Kalman filter to reduce its confidence in the state prediction model.

[0026] The extraction unit in the embedded module selects two values, magnetometer interference weight and temperature drift weight, from the same error weight vector. These two weights are then copied and constructed into a six-dimensional diagonal matrix. The first three elements of the diagonal matrix are filled with the magnetometer interference weight corresponding to the three axes of the magnetometer, and the last three elements are filled with the temperature drift weight corresponding to the temperature-affected measurement channels. The extraction unit reads the reference observation noise covariance matrix from the memory. This reference matrix is ​​also a six-dimensional diagonal matrix storing the variance reference values ​​of the measurement noise of each sensor. The extraction unit performs element-wise multiplication between the weight diagonal matrix and the reference observation noise covariance matrix, multiplying the diagonal elements one by one. The resulting adaptive observation noise covariance matrix reflects the true level of measurement noise in the current environment. Error sources with larger weights have larger covariance elements, causing the Kalman filter to reduce its confidence in the current measurement value and rely more on historical state predictions. The two adaptive covariance matrices generated by the multiplication unit and the extraction unit, respectively, are passed to the fusion module for Kalman filter recursive calculation.

[0027] In one specific embodiment, the fusion module is used for: Substituting the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix into the Kalman filter recursive formula, state prediction, covariance prediction, Kalman gain calculation, state update and covariance update are performed to obtain the fused attitude angle and fused angular rate. Based on the current environmental state identifier, a long-term drift compensation layer or a transient disturbance suppression layer is selected to perform error compensation on the fusion attitude angle and the fusion angular rate, thereby obtaining compensation data.

[0028] Specifically, the fusion module receives the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix output by the embedding module, and substitutes these two matrices into the five recursive formulas of the Kalman filter to perform iterative calculations.

[0029] First, the state prediction step is performed, calculating the prior state estimate for the current time step based on the state estimate from the previous time step and the state transition matrix. ,in This indicates that the prior state estimation vector at time k contains six state components: well inclination angle, azimuth angle, rotation angle, and triaxial angular rate. The state transition matrix is ​​obtained by discretizing the quaternion differential equation. This is the posterior state estimate from the previous time step. To control the input matrix, To control the input vector.

[0030] Then, the covariance prediction step is performed to calculate the prior covariance matrix. ,in The prior state covariance matrix reflects the uncertainty of the state estimate. Let the posterior covariance be the value from the previous time step. This is the adaptive process noise covariance matrix passed in by the embedded module.

[0031] Next, calculate the Kalman gain. ,in The Kalman gain matrix determines the fusion ratio of the measured and predicted values. The observation matrix describes the mapping relationship from the state vector to the measurement vector. This is the adaptive observation noise covariance matrix passed in by the embedded module.

[0032] Then perform a state update. ,in This is the final output posterior state estimation vector. The sensor measurement vector at the current moment includes triaxial acceleration and triaxial magnetic field strength. The values ​​in parentheses represent the measurement residuals, i.e., the difference between the actual and predicted measurements. Finally, the covariance matrix is ​​updated. ,in It is the identity matrix. The updated posterior covariance is calculated using the above five recursive steps and then fused into a fusion module. The first three elements are extracted as the fused attitude angle, including the well inclination angle, azimuth angle, and rotation angle. The last three elements are extracted as the fused angular rate, including the three-axis angular velocity components.

[0033] The fusion module executes compensation level selection logic based on the current environmental state identifier transmitted by the acquisition module. When the environmental state identifier is static drilling or a sudden temperature change, the fusion module activates the long-term drift compensation layer. This compensation layer maintains a linear relationship model between temperature and sensor zero-bias drift. It calculates the gyroscope zero-bias drift and magnetometer sensitivity drift coefficient based on the current temperature value, subtracts the calculated drift compensation from the fused angular rate, and multiplies the reciprocal of the magnetometer sensitivity coefficient by the azimuth component in the fused attitude angle to complete temperature compensation. When the environmental state identifier is dynamic impact, the fusion module switches to the transient disturbance suppression layer. The compensation layer calls a pre-trained gated recurrent unit neural network. The network input layer receives a nine-dimensional vector including three components of the fused attitude angle and three components of the fused angular rate at the current moment, as well as the residual values ​​of the previous three sampling periods. The data undergoes a nonlinear transformation through 32 GRU units in the hidden layer. The output layer generates a six-dimensional perturbation prediction vector. The fusion module subtracts the first three elements of the prediction vector from the fused attitude angle and the last three elements from the fused angular rate to complete the real-time suppression of transient perturbations. After compensation, the compensation data output by the fusion module, which includes high-precision attitude angles and angular rates after environmental adaptive compensation, is passed to the update module.

[0034] Figure 3 This is a schematic diagram illustrating the verification of the layering error compensation effect in an embodiment of this application; Figure 3 The data shows the trends of attitude angle error changes within a 60-second time window using three curves: before compensation, after long-term drift compensation, and after layered compensation. The horizontal axis represents time in seconds, and the vertical axis represents attitude angle error in degrees. The error curve before compensation shows a large fluctuation range between 0.6 and 1.8 degrees. After long-term drift compensation, the error is reduced to 0.2 to 0.5 degrees. After adopting the layered compensation strategy of this invention, the error is further reduced to 0.05 to 0.3 degrees with a significant reduction in fluctuation amplitude. From the time-domain characteristics, it can be seen that the layered compensation strategy not only reduces the average level of error but also effectively suppresses transient fluctuations in error. Especially in the dynamic impact environment of the 20 to 30-second and 40 to 50-second intervals, layered compensation successfully reduced the peak error from 1.1 degrees to 0.15 degrees through the transient disturbance suppression layer. This verifies the effectiveness of switching between the long-term drift compensation layer and the transient disturbance suppression layer according to the environmental state indicator, and the accuracy is improved by about 75% compared to the single compensation algorithm.

[0035] In one specific embodiment, the update module is used to: The reference well inclination angle is calculated using the compensated acceleration data. The residual between the well inclination angle in the compensated data and the reference well inclination angle is calculated. A residual sliding window is constructed and the root mean square of the residual is calculated. When the root mean square of the residual continuously exceeds the dynamic threshold corresponding to the current environmental state identifier, drilling is paused and stationary state sensor data is collected. The zero bias parameter is calculated, updated, and written to the memory. Based on the compensation data and the updated zero bias parameters, the well inclination angle, tool face angle, drill string rotation speed, and real-time pose are calculated, encapsulated into a data packet, and output through the communication interface.

[0036] Specifically, the update module uses the compensated triaxial acceleration data to independently calculate the reference well inclination angle using the inverse cosine function as a verification benchmark. The well inclination angle in the compensated data is subtracted from the reference well inclination angle to obtain the residual. A sliding window containing the residual values ​​of the most recent 50 sampling periods is constructed, and the square root of the sum of squares of the residuals is calculated to obtain the root mean square of the residual. When the root mean square of the residual continuously exceeds the dynamic threshold corresponding to the current environmental state for 50 consecutive sampling periods, the hardware calibration process is triggered. The update module sends a pause command to the drilling control system to make the drill string stop and then collects 500 consecutive samples. The arithmetic mean of the three-axis data of the magnetometer and gyroscope is calculated and written into the memory as the updated zero bias parameter. After calibration, the well inclination angle and tool face angle are calculated using the inverse cosine function and the arctangent function respectively based on the compensated data. The drill string rotation speed is calculated by fusing the integral value of the gyroscope angular rate and the acceleration vibration frequency according to the weight. Quaternion attitude and three-dimensional position coordinates are calculated by fusing multi-sensor data through a complementary filter. All measurement parameters are encapsulated into data packets and periodically sent to the external control system through the communication interface at a fixed baud rate.

[0037] In one specific embodiment, the spectral characteristics are obtained by performing a Fast Fourier Transform on the triaxial acceleration data and extracting the ratio of the dominant frequency energy to the total spectral energy. The magnetic field standard deviation is obtained by calculating the standard deviation of the magnetic field intensity magnitude within the sliding time window. The measurement parameters include well inclination angle, tool face angle, drill string rotation speed, and real-time pose.

[0038] The acquisition module performs a fast Fourier transform on the triaxial acceleration data to convert the time-domain signal into a frequency-domain signal. Through spectrum analysis, it identifies the dominant frequency component with the largest energy and calculates the ratio of its energy to the total energy of the entire frequency band to obtain the spectrum characteristic parameters. The acquisition module calculates the sum of the squares of the deviations of all sampling points from the mean value of the magnetic field intensity magnitude sequence within the sliding time window, divides it by the number of sampling points, and then takes the square root to obtain the standard deviation of the magnetic field. The measurement parameters output by the update module include four key parameters: well inclination angle, which reflects the angle of borehole deviation from the vertical direction; tool face angle, which reflects the orientation of the drill string's high side; drill string rotation speed, which reflects the speed of drill string rotation; and real-time pose, which includes quaternion attitude and three-dimensional position coordinates.

[0039] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-precision real-time measurement system for a dynamic measurement module while drilling, characterized in that the system... include: The data acquisition module is used to simultaneously acquire triaxial acceleration data, triaxial magnetic field strength data, triaxial angular rate data, and temperature data. Based on the spectral characteristics, magnetic field standard deviation, and temperature change rate, the environmental state is classified to obtain the current environmental state identifier. An embedding module is used to extract the corresponding error weight vector from a preset environment-error weight mapping matrix according to the current environment state identifier. The error weight vector includes accelerometer noise weight, gyroscope drift weight, magnetometer interference weight, and temperature drift weight. Each weight in the error weight vector is embedded into the corresponding diagonal element positions of the reference process noise covariance matrix and the reference observation noise covariance matrix to obtain the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix. The fusion module is used to substitute the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix into the Kalman filter for data fusion to obtain compensated data; The update module is used to calculate the root mean square of the residuals of the compensation data. When the root mean square of the residuals continuously exceeds the threshold, hardware calibration is triggered to update the zero bias parameter. The measurement parameters are calculated and output based on the compensation data.

2. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, The data acquisition module is used for: The raw sensor data is synchronously acquired through the SPI interface and timestamp aligned to obtain multi-source sensor data. A sliding time window is constructed for the multi-source sensor data, a fast Fourier transform is performed on the acceleration data to extract the spectral concentration coefficient, and the standard deviation of the magnetic field data and the rate of change of the temperature data are calculated. The current environmental state identifier is obtained by classifying and determining the relationship between the spectral concentration coefficient, the standard deviation, the rate of change, and a preset threshold.

3. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, Embedded modules, including: A mapping unit is used to establish an environment-error weight mapping matrix, wherein the row index of the mapping matrix corresponds to the environment state and the column index corresponds to the error source type; The fitting unit is used to collect data through offline calibration experiments, fit the residual distribution using the least squares method and normalize it to obtain the weight coefficients in the mapping matrix; The extraction unit is used to extract the corresponding row from the mapping matrix using the current environment state identifier as the row index, and obtain the error weight vector.

4. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 3, characterized in that, The error weight vector includes accelerometer noise weight, gyroscope drift weight, magnetometer interference weight, and temperature drift weight. Specifically, it includes setting the gyroscope drift weight to 0.65 and the accelerometer noise weight to 0.15 in static drilling conditions; and setting the accelerometer noise weight to 0.70 and the gyroscope drift weight to 0.15 in dynamic impact conditions. The error weight vector is composed of weight coefficients corresponding to different environmental conditions.

5. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 4, characterized in that, The embedded module also includes: The multiplication unit is used to extract the gyroscope drift weight and accelerometer noise weight from the error weight vector, construct a diagonal matrix, and perform element-wise multiplication with the reference process noise covariance matrix to obtain the adaptive process noise covariance matrix. The extraction unit is used to extract the magnetometer interference weight and temperature drift weight from the error weight vector, construct a diagonal matrix and perform element-wise multiplication with the reference observation noise covariance matrix to obtain the adaptive observation noise covariance matrix.

6. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, The fusion module is used for: Substituting the adaptive process noise covariance matrix and the adaptive observation noise covariance matrix into the Kalman filter recursive formula, state prediction, covariance prediction, Kalman gain calculation, state update and covariance update are performed to obtain the fused attitude angle and fused angular rate. Based on the current environmental state identifier, a long-term drift compensation layer or a transient disturbance suppression layer is selected to perform error compensation on the fusion attitude angle and the fusion angular rate, thereby obtaining compensation data.

7. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, The update module is used for: The reference well inclination angle is calculated using the compensated acceleration data. The residual between the well inclination angle in the compensated data and the reference well inclination angle is calculated. A residual sliding window is constructed and the root mean square of the residual is calculated. When the root mean square of the residual continuously exceeds the dynamic threshold corresponding to the current environmental state identifier, drilling is paused and stationary state sensor data is collected. The zero bias parameter is calculated, updated, and written to the memory. Based on the compensation data and the updated zero bias parameters, the well inclination angle, tool face angle, drill string rotation speed, and real-time pose are calculated, encapsulated into a data packet, and output through the communication interface.

8. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, The spectral characteristics are obtained by performing a Fast Fourier Transform on the triaxial acceleration data and extracting the ratio of the main frequency energy to the total spectral energy.

9. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, The standard deviation of the magnetic field is obtained by calculating the standard deviation of the magnetic field strength modulus within the sliding time window.

10. The high-precision real-time measurement system for drilling dynamic measurement module according to claim 1, characterized in that, The measurement parameters include well inclination angle, tool face angle, drill string rotation speed, and real-time position and attitude.

Images