A method and apparatus for dynamically measuring gravity tool face angle of a drilling tool

Abstract

The application discloses a kind of drilling tool gravity tool face angle dynamic measurement method and device. With gravity tool face angle and gyroscope drift as state variable, with accelerometer y-axis and z-axis measurement value as observation, fusion drilling tool measurement matrix and system nonlinearity establishes nonlinear state space model, linearizes nonlinear state space model at current working point;With central symmetric polytope to the bounded component of state variable initial value, process disturbance and observation noise, with random vector to the random component of system state variable initial value, process disturbance and observation noise, establish hybrid Kalman filter observer;Through forgetting factor attenuation hybrid Kalman filter prior covariance matrix historical information, according to hybrid Kalman filter observer hybrid prior covariance matrix, observation matrix and hybrid new information covariance matrix determine hybrid Kalman filter optimal gain, utilizes the hybrid Kalman filter observer established to obtain gravity tool face angle.

Description

Technical Field

[0001] This invention belongs to the field of oil drilling measurement and control technology, specifically relating to a method and device for dynamic measurement of the gravity tool face angle of drilling tools. Background Technology

[0002] Oil and gas resources are a vital guarantee for the national economy, and their efficient extraction relies on high-performance drilling tools. The gravity toolface angle (GTA) is a crucial parameter for the precise control of the drilling trajectory guided by modern drilling tools, playing a vital role. To ensure precise control of the downhole directional drilling trajectory, dynamic measurement technology of the gravity toolface angle has become the core technology for achieving precise control. This technology achieves this by deploying measurement units such as accelerometers and gyroscopes in the drill string to collect, process, and transmit key attitude parameters in real time. Major international service providers include National Oilwell Varco, Schlumberger, and Halliburton, which offer different dynamic measurement systems in terms of accuracy and adaptability. These systems require much higher accuracy in dynamic toolface angle measurement than static measurement, especially in complex conditions such as ultra-deep wells, horizontal wells, and high-temperature, high-pressure wells, where the dynamic measurement accuracy needs to reach ±1.0°. However, the main obstacle to achieving the aforementioned high-precision dynamic measurements stems from the extremely harsh physical environment downhole. During drilling, the drill string endures high-intensity, wide-bandwidth composite vibrations and impacts. This complex mechanical dynamic environment couples with the inherent measurement noise of inertial sensors such as accelerometers and gyroscopes, forming a dual interference mechanism consisting of bounded interference and random noise. This environment severely degrades the accelerometer's perception of the true gravity vector, while the time-varying drift error of the gyroscope induced by the high temperature and pressure environment further worsens the angular velocity measurement information. This poses a severe challenge to the anti-interference and real-time signal processing capabilities of the dynamic measurement system. Against this backdrop, the core challenge of tool face angle dynamic measurement lies in extracting effective information that truly reflects the drill string's attitude from sensor signals severely contaminated by dual uncertainties. Furthermore, existing technologies generally introduce the influence of historical data, causing the algorithm to be continuously hampered by outdated information from previous operating conditions. Specifically, the estimation of key states such as tool face angle and well inclination is severely lagging. Furthermore, when encountering new formations or when operating parameters change abruptly, the filter becomes overconfident and almost ignores real-time measurements, causing state tracking failure or even divergence. This fails to meet the stringent requirements of real-time directional drilling for measurement accuracy and reliability.

[0003] Chinese Patent Publication No. CN119466718B, published on October 28, 2024, discloses an invention entitled "Dynamic Measurement Method and System for Attitude Parameters of Rotary Guided Tools." This application discloses a dynamic measurement method and system for the attitude parameters of a rotary guide tool. However, its shortcomings lie in the following: the method uses an iterative Kalman filter to process measurement noise; however, this method has certain limitations in practical applications: its noise statistical model heavily relies on prior information, while the dual uncertainties in the actual drilling environment are highly complex and difficult to characterize using a unified and accurate mathematical model.

[0004] Chinese Patent Publication No. CN115618167B, published on August 1, 2025, is entitled "A Multi-Sensor Redundancy Combined Fault-Tolerant Dynamic Measurement Model and its Establishment Method." This application discloses the establishment of a multi-sensor redundant combined fault-tolerant dynamic measurement model. This type of method mainly has the following limitations: firstly, it lacks differentiated and detailed modeling of noise interference from various sensors; secondly, when constructing a multi-sensor fault-tolerant mechanism based on deep learning, it fails to fully balance the real-time requirements of attitude calculation with the computational complexity of the model.

[0005] In summary, existing technologies have failed to systematically solve the dual uncertainty problem of bounded vibration interference and random measurement noise coexisting. The influence of historical data leads to long delays in estimation results, and it is difficult to achieve a balance between algorithm accuracy, computational complexity, and engineering feasibility. Therefore, there is an urgent need for a new method and device for dynamic measurement of tool face angles that can effectively suppress strong bounded interference and random noise while ensuring high accuracy and high reliability under limited computing power constraints. Summary of the Invention

[0006] To address the problems existing in the prior art, the present invention provides a method and apparatus for dynamic measurement of the gravity tool face angle of drilling tools.

[0007] A method for dynamic measurement of the gravity tool face angle of a drilling tool, wherein the drilling tool is equipped with a gyroscope and an accelerometer, the method comprising:

[0008] Using the gravity tool face angle and gyroscope drift as system state variables and the accelerometer y-axis and z-axis measurements as observations, a nonlinear state-space model is established by integrating the drilling tool measurement matrix and system nonlinearity, and the nonlinear state-space model is linearized at the current operating point.

[0009] A hybrid Kalman filter observer is established by using a centrosymmetric multiple cell to characterize the bounded components of the initial values ​​of the system state variables, process disturbances, and observation noise, and by using random vectors to characterize the random components of the initial values ​​of the system state variables, process disturbances, and observation noise.

[0010] By attenuating the historical information of the hybrid prior covariance matrix of the hybrid Kalman filter through the forgetting factor, the optimal gain of the hybrid Kalman filter is determined based on the hybrid prior covariance matrix, observation matrix, and hybrid innovation covariance matrix of the hybrid Kalman filter observer. The gravity tool face angle is obtained using the established hybrid Kalman filter observer.

[0011] Furthermore, the establishment of a nonlinear state-space model, which uses the gravity tool face angle and gyroscope drift as system state variables, and accelerometer y-axis and z-axis measurements as observations, and integrates the drilling tool measurement matrix and system nonlinearity, includes:

[0012] The tool facet angle and gyroscope drift together constitute the system state variables. The state equation for establishing the combined measurement model that integrates gyroscope and accelerometer data is as follows:

[0013] ;

[0014] in, For discrete-time indexing, Represents the state transition matrix. For the input matrix, For the input vector, , for Time-lapse gyroscope rotation speed measurement value For system process interference, for The gyroscope measurement noise at any given time. for White noise that constantly causes gyroscope drift;

[0015] by Accelerometer y-axis and z-axis measurements at any time , As an observation, construct the observation equation:

[0016] ;

[0017] in, for y-axis accelerometer observation at time [time] for The z-axis accelerometer observation at time [time]. for At any given moment, the y-axis accelerometer observes the true y-axis gravitational component. for At any given moment, the z-axis accelerometer observes the true z-axis gravitational component. For observing noise, including vibration noise With the inherent measurement noise of the sensor , Indicates time; For the y-axis measurement matrix of the accelerometer, For the accelerometer z-axis measurement matrix, For the nonlinearity of the accelerometer y-axis system, The nonlinearity of the accelerometer z-axis system.

[0018] Furthermore, the linearization of the nonlinear state-space model at the current operating point includes: obtaining the Jacobian observation matrix by taking the partial derivative of the observation equation.

[0019] ;

[0020] in, for The observation matrix at each time point; Indicates the inclination angle of the well. Indicates based on The posterior estimation results predict the Constant gravity tool face angle;

[0021] The linearized state-space model is obtained as follows:

[0022] ;

[0023] in, For the observation matrix, To observe the noise, These are the observation values ​​from the observation equations of the system's state-space model.

[0024] Furthermore, the description of using centrosymmetric multiple cells to characterize the bounded components of the system state variables' initial values, process disturbances, and observation noise, and using random vectors to characterize the random components of the system state variables' initial values, process disturbances, and observation noise, includes: representing the initial state... Decomposed into mixing centers Initial state centrosymmetric multicellular and initial state random vector Interference with the process Decomposed into process-disrupting centrosymmetric multicellular structures and process disturbance random vector ; Observation noise Decomposed into observation noise centrosymmetric multicells and observation noise random vector .

[0025] Furthermore, the historical information of the prior covariance matrix of the hybrid Kalman filter attenuated by the forgetting factor includes:

[0026] The centrosymmetric multicell generation matrix of the system state variables is:

[0027] ;

[0028] in, Here is the state transition matrix at each time step. , , This represents the initial state centrosymmetric multicell generation matrix. for Time-process disturbance centrosymmetric multicell generation matrix Let be the forgetting factor, and the mixture prior covariance matrix be:

[0029] ;

[0030] in For weight parameters, For based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. For based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t.

[0031] Furthermore, the optimal gain utilizes a cost function minimization. Obtain, among which For weight parameters, The trace of the matrix, for The time-state variable is a centrosymmetric multicell generating matrix. for The prediction covariance matrix of the Gaussian part of the system state variables at time t is obtained. Optimal gain at time for:

[0032] ;

[0033] in For the mixed prior covariance matrix, ; For based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. For based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t. For the observation matrix, The mixed posterior covariance matrix, ; For the mixed new information covariance matrix, .

[0034] Furthermore, prediction methods for hybrid Kalman filter observers include:

[0035] The prediction center for the system state variables is:

[0036] ;

[0037] in, for The predictive center of the system state variables at any given time. for The input vector at time step; The propagation mode of the bounded components of the system state variables at any given time in a centrosymmetric multicell is as follows:

[0038] ;

[0039] in, for The system state variables at any given time are bounded components of a centrosymmetric multicell. for The time-matter disturbance is a centrosymmetric multiple cell; the generating matrix of the bounded components of the system state variables is:

[0040] ;

[0041] The propagation method of the random components of the system state variable is as follows:

[0042] ;

[0043] in, for A random vector of states at any given time. for The time-matter disturbance random vector; the covariance matrix of the Gaussian part of the system state vector is:

[0044] ;

[0045] in, for The Gaussian part of the predicted covariance matrix of the system state vector at time step [time]. for The covariance matrix of the Gaussian noise component in the time-process disturbance.

[0046] Furthermore, hybrid Kalman filter update methods include:

[0047] The system state variable central update method is as follows:

[0048] ;

[0049] in For based on The central posterior estimate of the state variable at time point predicted The center of the system state variables at time t, It is the identity matrix, and the uncertainty update of the bounded components of the system state variables is as follows:

[0050] ;

[0051] for The system state vector at any given time is a centrally symmetric multicell. To observe the centrosymmetric many-cell structure of the noise; the generation matrix of the centrosymmetric many-cell structure of the system state variables is updated as follows:

[0052] ;

[0053] in For based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. for The noise center-symmetric multicell generation matrix is ​​observed at all times; the uncertainty of the random components of the system state variables is updated as follows:

[0054] ;

[0055] in for Time-based The posterior estimation results at time point predicted A random vector of state variables at any given time. for Observation of a random vector of noise at any given time; The covariance matrix of the system state variables at time step 1 is updated as follows:

[0056] ;

[0057] in, For based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t. To observe the covariance matrix of the Gaussian noise component in the noise.

[0058] A dynamic measurement device for the gravity tool face angle of a drilling tool includes an accelerometer and a gyroscope. The device also includes a first processor and a second processor. The first processor linearizes a nonlinear state-space model at the current operating point. The nonlinear state-space model uses the gravity tool face angle and gyroscope drift as system state variables, and accelerometer y-axis and z-axis measurements as observations, fusing the drilling tool measurement matrix and system nonlinearity to establish the nonlinear state-space model. The second processor attenuates the historical information of the prior covariance matrix of the hybrid Kalman filter through a forgetting factor, obtains the optimal gain of the hybrid Kalman filter observer by minimizing the hybrid cost function, and uses the established hybrid Kalman filter observer to obtain the gravity tool face angle. The hybrid Kalman filter observer uses a centrosymmetric polycell to characterize the bounded components of the initial values ​​of the system state variables, process disturbances, and observation noise, and uses random vectors to characterize the random components of the initial values ​​of the system state variables, process disturbances, and observation noise.

[0059] Furthermore, the measuring device includes a first processing unit and a second processing unit. The first processing unit includes a first processor and an internal communication interface, and the second processing unit includes a second processor and an external communication interface. The first processor sends data to the second processor through the internal communication interface.

[0060] This invention provides a method, device, and storage medium for dynamic measurement of the tool face angle of drilling tools, which has the following beneficial effects: The proposed method based on a forgetting factor, a centrosymmetric multicell, and a Gaussian mixture Kalman filter decomposes the influence of external factors on the measured value into bounded interference and random noise. Bounded interference is characterized by a centrosymmetric multicell, and bounded vibration interference and random measurement noise are processed collaboratively within a unified framework. By fully utilizing the domain information of bounded noise, its influence can be effectively suppressed, thereby improving estimation accuracy. In addition, by dynamically balancing historical information through the forgetting factor, the influence of historical data on the current data is attenuated while considering historical data, improving the real-time performance of the filter, thereby optimizing the filter's tracking performance and adaptive capability, and effectively suppressing error drift. Based on this, the optimal gain of the mixture Kalman filter is dynamically calculated according to the real-time updated mixture prior covariance matrix, observation matrix, and mixture innovation covariance matrix, achieving high-precision and robust dynamic tracking of the tool face angle under strong interference background, meeting the dynamic measurement requirements of the tool face angle in directional drilling.

[0061] When measuring drilling attitude data using this method, the state estimator acquires the core function of adapting to time-varying environments. It can continuously and quickly track the real changes in parameters such as tool face angle without manual reset, thus providing a stable and reliable data foundation for real-time downhole decision-making and control, and significantly improving the operational efficiency and safety of drilling in complex formations. Attached Figure Description

[0062] Figure 1 This is a flowchart of the dynamic measurement method for the tool face angle of drilling tools according to the present invention;

[0063] Figure 2 This is a schematic diagram of the attitude parameters of the drilling tool;

[0064] Figure 3 Sensor spatial layout diagram for stable platform;

[0065] Figure 4 A study figure on the selection of FF-ZGKF forgetting factor in dynamic measurement of tool face angle;

[0066] Figure 5 A study on the selection of FF-ZGKF weighting parameters in dynamic measurement of tool face angles;

[0067] Figure 6 The figure shows the simulation analysis of the tool face angle estimation using the FF-ZGKF filtering algorithm under extreme working conditions.

[0068] Figure 7 for Figure 6 Enlarged view of point A in the middle;

[0069] Figure 8 This is an experimental verification diagram of FF-ZGKF for tool face angle estimation under actual working conditions;

[0070] Figure 9 for Figure 8 Enlarged view at point B in the middle;

[0071] Figure 10 A schematic diagram of a dynamic measuring device for the tool face angle of drilling tools;

[0072] Figure 11 A schematic diagram of the module connection for a dynamic measurement device for the tool face angle of drilling tools;

[0073] Figure 12 A flowchart of the working process of a dynamic measurement device for the tool face angle of drilling tools;

[0074] Wherein: 1-First end cap, 2-Equipment frame, 3- z 4- axial accelerometer y 5- axial accelerometer x 6-Main gyroscope, 7-First processing unit, 701-Communication protection module, 702-First processor, 703-Internal communication interface, 8-Second processing unit, 801-Second processor, 802-Storage module, 803-Power interface and management module, 804-External communication interface, 9-Second end cover, 10-External communication interface. Detailed Implementation

[0075] To address the challenges of accurately tracking and estimating state information in downhole drilling operations due to their high non-stationarity and strong time-varying nature, traditional filtering methods are insufficient. This invention provides a method and apparatus for measuring the gravity tool face angle of drilling tools. In existing technologies, dynamic measurement of tool face angles typically simplifies various measurement noises into single-type interferences, and generally lacks systematic consideration of the system's nonlinear dynamic characteristics and the time-varying drift error of the gyroscope. The linear or simplified models used deviate from the actual complex motion conditions in downhole drilling, severely limiting the accuracy of dynamic measurements in environments with strong vibration and high noise.

[0076] Therefore, this invention aims to solve the core challenge of constructing a highly robust measurement model that can accurately characterize system nonlinearity, effectively distinguish and suppress both bounded and random disturbances, and suppress long-term error drift under the dual uncertainty interference environment of high-intensity bounded vibration and random sensor noise. The core of this invention lies in constructing a state-space model that integrates the measurement matrix and system nonlinearity, systematically processing bounded vibration interference and random measurement noise within a unified anti-interference filtering architecture, and innovatively introducing a forgetting factor. It proposes a Forgetting Factor-based Zonotopic and Gaussian Hybrid Kalman filter (FF-ZGKF) method, thereby ultimately achieving high-precision and high-reliability dynamic measurement of tool face angles.

[0077] The drilling tool is equipped with a gyroscope and three single-axis accelerometers to measure the gravity tool face angle. The three accelerometers correspond to the x-axis, y-axis, and z-axis, respectively. The specific method is as follows... Figure 1 As shown. Includes:

[0078] S101: Using the gravity tool face angle and gyroscope drift as system state variables, and the y-axis and z-axis measurements of the accelerometer as observations, a nonlinear state-space model is established by integrating the drilling tool measurement matrix and system nonlinearity, and the nonlinear state-space model is linearized at the current operating point.

[0079] Using the gravity tool face angle and gyroscope drift as system state variables, a nonlinear state-space model integrating the measurement matrix and system nonlinearity is established to achieve a high-precision mathematical description of the complex motion behavior of the drilling tool.

[0080] Directional drilling is the process of precisely guiding and controlling the three-dimensional wellbore trajectory. Its key lies in ensuring that the wellbore trajectory, starting from the vertical section, deflects in the designed direction and accurately reaches the target area underground. In this process, precise dynamic measurement of the downhole drill string's attitude parameters is fundamental to achieving accurate trajectory control. Inclination angle, gravity tool facet angle, inclination azimuth line, and elevation direction line define the drill string's spatial orientation downhole. The spatial geometric relationships of the downhole drill string's attitude parameters are as follows: Figure 2 As shown.

[0081] The well inclination angle refers to the angle between the tangent to the wellbore axis and the direction of gravity. This invention uses the symbol... Indicates the well inclination angle, in degrees (°). This parameter quantitatively describes the degree to which the wellbore deviates from the vertical direction, and is a key basis for determining the well type (such as vertical, deviated, or horizontal well) and for trajectory control. The tool face angle is the angle between the tool face and the bottom plane, rotated clockwise relative to the high side of the wellbore cross-section. This invention uses the symbol... This indicates the angle of the gravity tool face, in degrees (°). This parameter characterizes the rotational orientation of the drill string around the wellbore axis and is the core attitude information for achieving directional control of the steering tool.

[0082] like Figure 3 As shown, the specific embodiment of the present invention relates to an attitude measurement system for downhole drilling tools. The system employs a centrally symmetrical layout to achieve dynamic measurement of the attitude parameters of rotary steered drilling tools. Its core sensing unit consists of three independent single-axis quartz accelerometers and one single-axis gyroscope.

[0083] The use of quartz accelerometers is an optimized choice for the high-temperature, high-vibration, and high-precision directional drilling requirements of downhole wells. Quartz accelerometers offer engineering advantages such as high-temperature resistance, shock resistance, and long lifespan, enabling them to reliably adapt to extreme deep-well conditions. Compared to traditional integrated MEMS accelerometers, quartz accelerometers provide better structural strength, temperature resistance, and long-term reliability.

[0084] By employing a centrally symmetrical arrangement, the accelerometer configuration effectively suppresses the rotational and centripetal acceleration interference introduced by drill string rotation, thus establishing mechanical vibration as the primary bounded interference source affecting the accuracy of gravity vector sensing. This provides a clear physical basis and design foundation for subsequent hybrid filtering algorithms to specifically model and suppress the dual uncertainties of bounded interference and random measurement noise. The complete toolface angle can be calculated using three independent single-axis quartz accelerometers and gyroscopes. The accelerometers are arranged along the wellbore axis (x-axis), the toolface radius (y-axis), and the radius perpendicular to the toolface (z-axis). Based on this coordinate system, the well inclination angle... and tool face angle The following formula can be used to solve it:

[0085] ;

[0086] .

[0087] The attitude matrix is ​​represented using the Euler angle method. A northeast geodetic coordinate system is established. and the carrier coordinate system The x-axis of the carrier coordinate system coincides with the wellbore axis, representing the axial direction of the drill string. The plane formed by the y and z axes of the carrier coordinate system represents the cross-section of the drill string. Accelerometers are distributed along the coordinate axes of the carrier coordinate system. Both the geodetic coordinate system and the carrier coordinate system conform to the right-hand rule. When rotating along the coordinate axes with the origin as the starting point, counterclockwise rotation is positive, and clockwise rotation is negative. The carrier coordinate system can be obtained by transforming the geodetic coordinate system using the current attitude angle of the drill string.

[0088] During drilling, strong vibrations introduce significant vibration noise into the gravitational components of the y and z axes. This noise, combined with the inherent measurement noise of the sensors, can easily lead to significant errors in the calculated tool face angle. Simultaneously, gyroscope drift caused by the high temperature and pressure environment downhole further reduces measurement accuracy. Therefore, it is necessary to develop a highly robust dynamic measurement method for tool face angles to ensure the engineering reliability of drilling tools.

[0089] This invention combines the tool face angle and gyroscope drift to form the system state variables. , express The angle of gravity tool face at any moment. for The gyroscope drift at any given moment. The state equation for establishing the combined measurement model that integrates gyroscope and accelerometer data is:

[0090] ;

[0091] in, This is a discrete-time index used to identify the sequence of sampling times. State transition matrix, For the input matrix, For the input vector, This is a disturbance to the system process. and Depending on whether the system can be time-varying or constant, in this embodiment... and For a constant value, where, , , , for Time-lapse gyroscope rotation speed measurement value , for The gyroscope measurement noise at any given time. for White noise that constantly causes gyroscope drift; The sampling period.

[0092] Specifically, this invention utilizes the high short-time accuracy of gyroscopes to perform one-step prediction of tool face angles:

[0093] ;

[0094] in, for Constant gravity tool face angle, for Gyroscope drift at any given moment.

[0095] Gyroscope drift Influenced by factors such as temperature and rotation speed, it is modeled as white noise. Driven first-order stochastic process:

[0096] ;

[0097] The state vector is constructed by combining the tool face angle and the gyroscope drift. The state equation for establishing the combined measurement model that integrates gyroscope and accelerometer data is as follows:

[0098] .

[0099] Accelerometer measurements mainly consist of three components: gravitational acceleration component. Vibration noise introduced by drill string vibration and the inherent measurement noise of the sensor Accelerometer measurement matrix and system nonlinearity This will also affect the measurement output and needs to be compensated for through filtering. Considering the good long-term stability of the accelerometer's measurement results, and The arctangent operation alters the statistical characteristics of the noise, making it unsuitable for approximation as zero-mean Gaussian white noise. Therefore, the accelerometer's y-axis and z-axis measurements are selected... , As an observation, construct the observation equation:

[0100] ;

[0101] in, for y-axis accelerometer observation at time [time] for The z-axis accelerometer observation at time [time]. for At any given moment, the y-axis accelerometer observes the true y-axis gravitational component. for At any given moment, the z-axis accelerometer observes the true z-axis gravitational component. For observing noise, including vibration noise With the inherent measurement noise of the sensor , Indicates time; For the y-axis measurement matrix of the accelerometer, For the accelerometer z-axis measurement matrix, For the nonlinearity of the accelerometer y-axis system, The nonlinearity of the accelerometer z-axis system.

[0102] Vibration noise With the inherent measurement noise of the sensor Both together constitute observation noise. :

[0103] .

[0104] Because the observation equations exhibit nonlinear time-varying characteristics, nonlinear filtering methods are required for state estimation. Therefore, the observation equations are linearized, transforming them into a linear time-varying system. Linearization of the nonlinear state-space model at the current operating point includes: obtaining the Jacobian observation matrix by taking the partial derivative of the observation equations.

[0105] ;

[0106] in, for The observation matrix at each time point; Indicates the inclination angle of the well. Indicates based on The posterior estimation results predict the Constant gravity tool face angle;

[0107] The linearized state-space model is obtained as follows:

[0108] ;

[0109] in, for The observed values ​​of the state-space model observation equations of the system at time t. For the observation matrix, To observe noise.

[0110] S102: A hybrid Kalman filter observer is established by using a centrosymmetric multiple cell to characterize the bounded components of the initial values ​​of the system state variables, process disturbances, and observation noise, and by using a random vector to characterize the random components of the initial values ​​of the system state variables, process disturbances, and observation noise.

[0111] The Zonotopic Kalman filter (ZKF) aims to provide a deterministic bounding boundary for the state of a system subjected to unknown but bounded noise by recursively optimizing a banded set represented by a Zonotopic polytope.

[0112] Given, Let n be a real vector space. for 3D real matrix space, for A 3D real matrix space.

[0113] A centrosymmetric multicellular body is essentially a convex multicellular body with centrosymmetry, defined as: an m-order centrosymmetric multicellular body It is a hypercube Affine transformation:

[0114] ;

[0115] in, Minkowski and, for The center for The generating matrix determines The shape and size. For centrosymmetric multicellular organisms, the following properties hold:

[0116] ;

[0117] ;

[0118] ;

[0119] in , , . ,in , .Notice It is a box, that is, an interval vector, which represents A box-shaped external environment. Higher-order symmetric multicells can be compactly enveloped from lower-order symmetric multicells by dimensionality reduction operations, while preserving their geometric characteristics. Given a multicell... and a selected integer s (satisfying) ), for matrix The column vectors are sorted in descending order according to their Euclidean norm, and the resulting matrix is ​​denoted as . .but ,in , Depend on The former Composed of columns, and It is a diagonal matrix with the following elements:

[0120] ;

[0121] For a matrix of appropriate dimension The following equations hold true:

[0122] ;

[0123] ;

[0124] .

[0125] The hybrid Kalman filter employs a unified anti-interference filtering architecture, the core of which lies in constructing a hybrid estimator capable of handling bounded set uncertainties and random statistical uncertainties in parallel. Based on the aforementioned nonlinear state-space model, this architecture innovatively integrates centrosymmetric polytope set operations with Gaussian Kalman filtering. Specifically, it utilizes centrosymmetric polytopes to provide deterministic boundary descriptions and propagation of initial system states and bounded process disturbances; simultaneously, it models the statistical characteristics of the sensor's inherent measurement noise using a random noise covariance matrix.

[0126] Specifically, the initial state Decomposed into mixing centers Initial state centrosymmetric multicellular and initial state random vector Interference with the process Decomposed into process-disrupting centrosymmetric multicellular structures and process disturbance random vector ; Observation noise Decomposed into observation noise centrosymmetric multicells and observation noise random vector .

[0127] For discrete-time linear time-varying systems, a hybrid modeling framework is used to handle system uncertainties, including the initial state. Process interference With observation noise Decomposed into a combination of bounded and random components:

[0128] ;

[0129] ;

[0130] ;

[0131] in, Initial state The mixing center. , and These are the bounded parts of the state vector, centrosymmetric multiple cells. Process interference bounded partial centrosymmetric multicellular bodies Bounded part of observation noise centrosymmetric multicell Their generation matrices are respectively , and .in, for 3D real matrix space, for 3D real matrix space, for A 3D real matrix space.

[0132] , and Let these be random vectors, which can be characterized by known, uncorrelated, zero-mean random variables, and their covariances are respectively... , and ,in For positive determination Initial estimate of the error covariance matrix, positive semidefinite dimensional process noise covariance matrix, For positive determination The measurement noise covariance matrix is ​​3D. Each random variable is uncorrelated across different time points.

[0133] Within a unified filtering framework, bounded disturbances caused by mechanical vibration and inherent random measurement noise from sensors are systematically modeled and characterized. A hybrid noise model that better reflects actual physical constraints is established, simultaneously considering the statistical characteristics of random noise and the range of bounded noise sets. For the bounded noise component, deterministic upper and lower bounds or feasible sets of estimation errors are provided, thus offering performance guarantees under worst-case conditions and enhancing the robustness of theoretical results. Based on this, a robust integrated design of the filter is completed to achieve synergistic suppression of dual uncertainty disturbances.

[0134] This invention utilizes a single filter structure to simultaneously handle two types of uncertainties, significantly reducing the complexity of the system architecture and the difficulty of integration; while simultaneously improving practical reliability and accuracy. During the drilling process of downhole drilling tools, the noise encountered by the drilling tool is often a mixed form, including bounded disturbances and random noise. This framework directly matches this mixed characteristic. By fully utilizing the domain information of bounded noise, its influence can be effectively suppressed, achieving a higher and more stable estimation accuracy than traditional methods that only consider random noise and set-membership methods that only consider bounded disturbances.

[0135] In summary, this unified framework theoretically achieves a more rigorous characterization of mixed noise and a more comprehensive performance guarantee. In drilling engineering, it brings direct benefits such as structural simplification, improved accuracy, and enhanced reliability, and has important methodological value and broad application prospects.

[0136] S103: By attenuating the historical information of the prior covariance matrix of the hybrid Kalman filter through the forgetting factor, the optimal gain of the hybrid Kalman filter is determined based on the hybrid prior covariance matrix, observation matrix and hybrid innovation covariance matrix of the hybrid Kalman filter observer, and the gravity tool face angle is obtained using the established hybrid Kalman filter observer.

[0137] By introducing a forgetting factor, the confidence weight of historical estimation information in the current filtering cycle is dynamically adjusted through an adaptive forgetting factor. This effectively balances the dual uncertainty handling of bounded sets and random statistical distributions, suppresses long-term error drift, and ultimately achieves high-precision and robust dynamic estimation of tool face angles.

[0138] The centrosymmetric multicell generating matrix of the system state variables at time t is:

[0139] ;

[0140] in, Here is the state transition matrix at each time step. , , This represents the initial state centrosymmetric multicell generation matrix. for Time-process disturbance centrosymmetric multicell generation matrix The forgetting factor is used to attenuate the influence of the initial state centrosymmetric multicell generation matrix and the historical process disturbance centrosymmetric multicell generation matrix in the centrosymmetric multicell generation matrix of the system state variables. In other words, it attenuates the historical information of the hybrid Kalman filter's hybrid prior covariance matrix, thereby affecting the hybrid Kalman filter gain.

[0141] The hybrid Kalman filter observer obtains its optimal gain by minimizing the hybrid cost function. Consider the above discrete-time linear time-varying system, where the input... and measurement Given. The central estimate Bounded partial centrosymmetric multicell of state vectors Process interference bounded partial centrosymmetric multicellular bodies Bounded part of observation noise centrosymmetric multicell and Gaussian random vectors , and The goal is to determine the mixed-state estimate based on the minimum variance criterion, that is, for ,in, Given the set of positive integers, solve... , and To achieve this goal, the specific steps are as follows:

[0142] By minimizing the cost function ,in Given the weight parameters, The trace of the matrix, To minimize The gain matrix, Let the trace of the matrix be denoted as . The optimal gain is obtained as . for:

[0143] ;

[0144] in For the mixed prior covariance matrix, ; The centrosymmetric multicell generating matrix of system state variables The effect, thus attenuating the optimal gain, is The influence of historical information in China. For the observation matrix, The mixed posterior covariance matrix, ; For the mixed new information covariance matrix, .

[0145] The centrosymmetric multicell generating matrix of the system state variables decays through a forgetting factor, thereby affecting the gain of the hybrid Kalman filter observer from historical information.

[0146] Furthermore, prediction methods for hybrid Kalman filter observers include:

[0147] The prediction center for the system state variables is:

[0148] ;

[0149] The propagation mode of the bounded components of the system state variables in a centrosymmetric multicell is as follows:

[0150] ;

[0151] in, for The system state variables at any given time are bounded components of a centrosymmetric multicell. for The time-matter disturbance is a centrosymmetric multiple cell; the generating matrix of the bounded components of the system state variables is:

[0152] ;

[0153] ;

[0154] The propagation method of the random components of the system state variable is as follows:

[0155] ;

[0156] in, for A random vector of states at any given time. for The time-matter disturbance random vector; the covariance matrix of the Gaussian part of the system state vector is:

[0157] ;

[0158] in, for The Gaussian part of the predicted covariance matrix of the system state vector at time step [time]. for The covariance matrix of the Gaussian noise component in the time-process disturbance.

[0159] Hybrid Kalman filter update methods include:

[0160] The system state variable central update method is as follows:

[0161] ;

[0162] in For based on The central posterior estimate of the state variable at time point predicted The center of the system state variables at time t, It is the identity matrix, and the uncertainty update of the bounded components of the system state variables is as follows:

[0163] ;

[0164] for The system state vector at any given time is a centrally symmetric multicell. To observe the centrosymmetric many-cell structure of the noise; the generation matrix of the centrosymmetric many-cell structure of the system state variables is updated as follows:

[0165] ;

[0166] in For based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. for The noise center-symmetric multicell generation matrix is ​​observed at all times; the uncertainty of the random components of the system state variables is updated as follows:

[0167] ;

[0168] in for Time-based The posterior estimation results at time point predicted A random vector of state variables at any given time. for Observation of a random vector of noise at any given time; The covariance matrix of the system state variables at time step 1 is updated as follows:

[0169] .

[0170] in, For based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t. To observe the covariance matrix of the Gaussian noise component in the noise.

[0171] The core of the filtering update mechanism lies in the gain matrix. The optimized design of this matrix directly determines the fusion ratio of predicted state and observation information in the estimation. This is achieved by minimizing the cost function. The optimal gain can be obtained as for:

[0172] ;

[0173] in:

[0174] ;

[0175] ;

[0176] ;

[0177] in, For the mixed prior covariance matrix, For the observation matrix, The mixed posterior covariance matrix, Let be the mixed-information covariance matrix. It has a significant impact on the calculation of the mixed prior covariance matrix, and the optimal gain automatically adjusts the information fusion strategy:

[0178] ;

[0179] The core of the update mechanism lies in the gain matrix. The optimized design of this matrix This directly determines the balance between predicted and observed information. Based on the aforementioned theoretical derivation, this invention establishes a prediction and estimation method that introduces a forgetting factor. This method strictly satisfies the boundedness of centrosymmetric multicells and the definition criteria of Gaussian random vectors. Within a unified observation fusion framework, the complete theoretical system and implementation process of steps such as state prediction, measurement update, and covariance recursion are systematically elaborated. Through rigorous mathematical derivation, this invention verifies the stability of the filtering system. The introduction of the forgetting factor can effectively improve the algorithm's tracking ability and robustness to time-varying systems without affecting the accuracy of steady-state estimation, thus theoretically guaranteeing the reliability and convergence of the proposed method in complex dynamic environments.

[0180] In the Forgetting Factor-based Zonotopic and Gaussian Hybrid Kalman filter (FF-ZGKF), the forgetting factor... By adjusting the mixed prior covariance matrix Directly affects Kalman gain .when When the norm of the prior covariance matrix shrinks, the Kalman gain decreases. As the filter decreases, it enhances the confidence in the predicted value while reducing its dependence on the observed value. This dynamic adjustment mechanism enables a gradual forgetting of historical information, allowing the system to better adapt to time-varying characteristics and effectively solving the practical engineering problem of decreased state estimation accuracy under highly time-varying conditions.

[0181] Compared with existing filtering methods, this method significantly reduces the requirements for prior information at the engineering implementation level, thereby greatly improving its field applicability and deployability. Specifically, for the complex vibration noise experienced by the drilling tool, only the bounded range of its amplitude needs to be obtained, without the need for precise identification of its statistical distribution characteristics beforehand. Simultaneously, the inherent measurement noise parameters of the sensor can be easily obtained through conventional calibration procedures or by directly consulting the device's technical manual, eliminating the need for cumbersome online estimation. Furthermore, the introduced forgetting factor only affects the norm adjustment of the mixed prior covariance matrix; its calculation process is simple, introducing no additional computational burden and maintaining the lightweight and efficient overall structure of the algorithm. This invention does not require precise statistical modeling and complex field calibration of downhole vibration noise and sensor measurement noise, making it easy to implement in highly time-varying and non-stationary drilling environments with a relatively low computational burden.

[0182] This invention effectively overcomes the engineering implementation bottlenecks caused by the complexity and calibration difficulties of traditional methods, enabling the algorithm to achieve both estimation accuracy and real-time processing capabilities in real-world downhole, highly time-varying, and non-stationary operating conditions. Therefore, while ensuring state estimation accuracy, this method possesses higher engineering applicability and real-time processing efficiency, effectively solving the practical engineering problem of accurately acquiring and updating noise models in real-time in the field environment.

[0183] The following specific data illustrates the effectiveness of this invention. Forgetting factor. Used to regulate the degree of decay of historical uncertainty, usually Values This invention employs a systematic parameter scanning method to optimize the FF-ZGKF filter to determine the optimal parameter configuration for the hybrid filter. Specifically, within the parameter range... Within the range, the parameters are discretized with a step size of 0.0001, and the filtering performance is calculated for each parameter value. The optimal parameters are then selected by comprehensively evaluating the parameters based on two indicators: root mean square error (RMSE) and maximum estimation error.

[0184] like Figure 4 As shown: With As the value increases, the estimated RMSE and maximum estimation error show a trend of first decreasing and then increasing. At this point, the system achieves optimal estimation performance; when Afterwards, the error increased significantly. Experiments show that... The optimal value is suitable for the FF-ZGKF algorithm and provides parameter basis for subsequent simulations and experiments.

[0185] Based on the completed research on the selection of forgetting factor parameters, further research was conducted on the weighting parameters. Conduct systematic selection research. The contribution of bounded uncertainty and random uncertainty in equilibrium state estimation is used. Based on weighted fusion theory, when prior information is lacking, the contribution of bounded uncertainty and random uncertainty is usually taken as the percentage. As the initial value for the uniform weighting strategy, this study integrates measured vibration and noise boundary data, key sensor parameters, and previous calibration results, and employs a systematic parameter scanning method to conduct parameter optimization analysis. Specifically, within the closed interval... The system discretizes the parameters in increments of 0.001, calculates the system performance corresponding to each parameter value point by point, and performs a comprehensive evaluation based on the set evaluation indicators to determine the parameter configuration that optimizes the overall system performance.

[0186] like Figure 5 As shown: With As the value increases, the estimated RMSE and maximum estimation error show a trend of first decreasing and then increasing. At this point, the system achieves optimal estimation performance, effectively coordinating the two types of uncertainty. When Then, the error increased rapidly. The algorithm's performance deteriorates significantly over time, approaching that of the extended Kalman filter algorithm. Experiments confirm this. To achieve the optimal value, high-precision dynamic estimation of the tool face angle can be guaranteed in subsequent simulations and experiments.

[0187] To improve the realism of the simulation, a constant drift term and a noise term are introduced into the gyroscope measurement model. Based on engineering measurement data, the boundary values ​​for gyroscope measurement noise and drift noise are set as follows: The accelerometer's y-axis and z-axis measurement matrices are respectively... and The corresponding system nonlinearities are respectively and This set of parameters was used in all subsequent simulations.

[0188] To evaluate the adaptability of the filtering model under different drilling conditions, three typical operating stages were simulated: (1) the build-up stage, where the tool face angle is constant at 90°; (2) the stick-slip stage, where the tool face angle oscillates sinusoidally at a frequency of 0.05Hz and an amplitude of 20°; and (3) the stabilization stage, where the drill collar rotates at a speed of 18° / s, and the tool face angle is within a certain range. Continuous changes within the range. Each phase lasts 200 seconds.

[0189] Drill string vibration is a major factor causing distortion in tool face angle measurements. High-intensity vibration can mask the gravity component in the accelerometer signal, severely affecting the accuracy of dynamic measurements. According to Baker Hughes' drill string vibration classification standard, lateral vibrations acting on the YZ plane and exceeding level 4 are classified as strong vibrations, and their noise can be approximated as variance. Gaussian white noise, this invention is based on statistical... The criterion is to set the vibration noise boundary as a diagonal matrix. The boundary values ​​are expressed in units of gravitational acceleration g. These are then superimposed with the sensor's inherent noise (variance 0.2) on the accelerometer's y and z axis outputs to create a strong vibration interference environment. Using this setup, the system evaluates the estimation accuracy of the FF-ZGKF filtering method under extreme conditions and analyzes its adaptability and robustness at different drilling stages.

[0190] To address the mixed uncertainty characteristics of the tool face angle dynamic measurement system, this study proposes four key improvements: using a centrosymmetric multiple cell to represent the initial state of the system; establishing a disturbance model for a purely bounded uncertainty process; implementing set propagation of prediction steps based on the centrosymmetric multiple cell; and optimizing the calculation of mixed covariance through dimensionality reduction generating matrices. These improvements enable state estimation even in complex noisy environments.

[0191] Initial settings parameters: , , , , , , , , .

[0192] Given constant ,cycle Execute, then obtain input data. and observation data The prediction steps are as follows:

[0193] ;

[0194] ;

[0195] ;

[0196] The observation and prediction steps are as follows:

[0197] ;

[0198] ;

[0199] Optimal gain calculation:

[0200] ;

[0201] ;

[0202] ;

[0203] ;

[0204] The update steps are as follows:

[0205] ;

[0206] ;

[0207] ;

[0208] Perform dimensionality reduction operation:

[0209] ;

[0210] return , , The algorithm loop ends, and the dynamic estimate of the tool face angle is obtained.

[0211] Figure 6 and Figure 7 Simulation results of the FF-ZGKF filtering algorithm for tool face angle estimation under extreme conditions are presented. Analysis shows that the FF-ZGKF algorithm can achieve effective state estimation and exhibits good adaptability to different drilling conditions, with an estimation error RMSE of 1.5481° and a dynamic measurement accuracy of ±1.5123°. All uncertainties were effectively controlled. This result verifies the theoretical advantages of the FF-ZGKF algorithm in coordinating the decay of historical uncertainty and the fusion of bounded sets and random distribution uncertainties, and demonstrates the algorithm's engineering applicability in environments with strong disturbances.

[0212] The experiment superimposed vibration noise on the gravitational components of the y-axis and z-axis of the accelerometer, with the boundary being... The vibration noise and sensor inherent noise with a variance of 0.2 were used to construct a strong vibration interference environment in the well. Figure 8 and Figure 9 The results demonstrate that the FF-ZGKF algorithm can effectively estimate the tool face angle and adapt to various drilling conditions in practical tests, with an estimation error RMSE of 1.1982° and a dynamic measurement accuracy of ±1.1973°. This algorithm demonstrates significant effectiveness in improving filtering accuracy, while also exhibiting good engineering applicability and robustness under various complex operating conditions.

[0213] Experimental verification shows that the FF-ZGKF filtering algorithm can achieve a dynamic measurement accuracy of ±1.1973° for tool face angles in a laboratory environment. This accuracy has initially met the technical requirements of mainstream international drilling tools.

[0214] The above content provides detailed examples of specific drilling tool stabilization platforms, and it should not be assumed that the specific implementation of this invention is limited to these examples.

[0215] This invention also provides a dynamic measurement device for the gravity tool face angle of drilling tools. This device is suitable for extreme working conditions such as high temperature, strong vibration, and low computing power in downhole environments. The device includes an accelerometer and a gyroscope, as well as a first processor 702 and a second processor 801. The first processor 702 is used to linearize the nonlinear state-space model at the current working point. The nonlinear state-space model is established by using the gravity tool face angle and gyroscope drift as system state variables, and accelerometer y-axis and z-axis measurements as observations, and fusing the drilling tool measurement matrix and system nonlinearity. The second processor 801 is used to attenuate the historical information of the prior covariance matrix of the hybrid Kalman filter through a forgetting factor, obtain the optimal gain of the hybrid Kalman filter observer by minimizing the hybrid cost function, and obtain the gravity tool face angle using the established hybrid Kalman filter observer. The hybrid Kalman filter observer uses a centrosymmetric multiple cell to characterize the initial values ​​of the system state variables, process disturbances, and bounded components of bounded noise, and uses random vectors to characterize the random components of the initial values ​​of the system state variables, process disturbances, and bounded noise.

[0216] Specifically, such as Figure 10 and Figure 11 As shown, the device consists of a device frame 2 and a first end cover 1, a second end cover 9 and an external communication interface 10 connected to its two ends. The device frame 2 is equipped with a first processing unit 7, a second processing unit 8, an accelerometer and a gyroscope. The accelerometer includes an x-axis accelerometer 5, a y-axis accelerometer 4 and a z-axis accelerometer 3. The accelerometer is a single-axis quartz accelerometer. The gyroscope includes a main gyroscope 6 and an auxiliary gyroscope.

[0217] After assembly, the entire sensing and circuit module is encapsulated using a modular vacuum encapsulation process based on a centrally symmetrical layout. This process eliminates air bubbles within the encapsulation material and employs an independent cavity design for differentiated sealing of each functional module, significantly improving the overall dielectric strength, heat dissipation performance, withstand voltage, and waterproof rating of the module. Furthermore, the centrally symmetrical sensor layout itself possesses redundancy characteristics, and combined with the modular, detachable design, it allows for partial repair or replacement in the event of a single sensor failure, enhancing the device's field maintainability and overall lifecycle economics.

[0218] The first processing unit 7 includes a communication protection module 701, a first processor 702, and an internal communication interface 703. The first processor 702 acquires accelerometer and gyroscope signals, performs anti-aliasing filtering and preliminary noise reduction, and linearizes the nonlinear state-space model at the current operating point to generate accurate linear time-varying system model parameters for subsequent filtering algorithms. The nonlinear state-space model uses the gravity tool facet angle and gyroscope drift as system state variables, and the measurements from the y-axis and z-axis accelerometers as observations, fusing the drilling tool measurement matrix and system nonlinearity to establish the nonlinear state-space model.

[0219] The auxiliary gyroscope's sensitive axis is parallel to the main gyroscope 6 and connected to the first processor 702. When the main gyroscope malfunctions or data is abnormal, it can provide backup angular velocity data, significantly enhancing the system's fault tolerance and reliability during long-term operation.

[0220] The communication protection module 701 is connected to the communication interface of the first processor 702. It adopts electrical isolation technology to effectively block surges and common-mode interference from complex underground power lines, ensuring the safe operation of the core controller in harsh electrical environments.

[0221] The internal communication interface 703 is connected to the second processing unit 8, and the first processing unit 7 communicates with the second processing unit 8 through the internal communication interface 703.

[0222] The second processing unit 8 includes a second processor 801, a storage module 802, a power interface and management module 803, and an external communication interface 804.

[0223] The second processor 801 receives preprocessed data and linearized model parameters from the first processing unit 7, attenuates historical information of the prior covariance matrix of the hybrid Kalman filter by a forgetting factor, obtains the optimal gain of the hybrid Kalman filter observer by minimizing the hybrid cost function, and obtains the gravity tool face angle using the established hybrid Kalman filter observer. The hybrid Kalman filter observer uses a centrosymmetric polycell to characterize the initial values ​​of the system state variables, process disturbances, and bounded components of bounded noise, and uses random vectors to characterize the random components of the initial values ​​of the system state variables, process disturbances, and bounded noise.

[0224] The storage module 802 is connected to the second processor 801 and uses non-volatile memory to cyclically store complete dynamic measurement process data, including raw sensor sample values, intermediate estimated states, final calculated tool face angle results, and their corresponding timestamps and quality identifiers. This data can be retrieved by the ground system after drilling is completed for in-depth performance analysis, fault diagnosis, and algorithm optimization.

[0225] The power interface and management module 803 connects to the second processor 801 and the first processor 702, and is responsible for connecting to the DC power supplied by the downhole instrument string. This module integrates dynamic voltage and frequency adjustment management functions, enabling real-time monitoring of the second processor's computational load. Based on the load level, it dynamically adjusts the processor core's operating voltage and clock frequency, thereby achieving intelligent optimization of system power consumption while ensuring algorithm real-time performance and computational accuracy. The power management module monitors the system load throughout the process. During data acquisition and simple preprocessing stages, the system operates in low-power mode; when entering the core solution stage of the FF-ZGKF algorithm, the system automatically increases processor performance to meet the demands of complex mathematical operations, achieving a balance between performance and power consumption.

[0226] The external communication interface 804 is connected to both the second processor 801 and the external communication interface 10, and exchanges data with other downhole modules or surface systems in the drill string network through the external communication interface 10. All external communication links are isolated and protected by circuitry to ensure the stability and security of data transmission.

[0227] The device described in this embodiment achieves efficient allocation of computing power through a distributed dual-circuit board collaborative processing architecture. It overcomes the challenge of high-precision calculation under dual uncertainty interference by hardware implementation of a centrally symmetric multicell based on the forgetting factor and a Gaussian mixture Kalman filter method. Through highly reliable isolated communication, intelligent power management, and enhanced packaging technology, it ensures stable and reliable high-precision and robust dynamic measurement of tool face angles under extremely complex working conditions of high temperature, strong vibration, and low computing power in downhole drilling, providing solid technical equipment support for precise trajectory control of directional drilling.

[0228] In another embodiment, the measurement process of a drilling tool gravity tool face angle dynamic measuring device is as follows: Figure 12 As shown.

[0229] S1001: The first processing unit acquires the raw output signals of three single-axis quartz accelerometers and one single-axis gyroscope in real time.

[0230] S1002: After filtering and calibrating the acquired signals, the first processor uses this real-time data to linearize the nonlinear state-space model near the operating point to obtain an accurate linear time-varying system model; it then runs the prediction step of the FF-ZGKF hybrid filtering algorithm to generate prior estimates of the system state variables.

[0231] S1003: The preprocessed sensor data, linearized model parameters, and prior estimates of system state variables are sent to the second processing unit 8 via internal high-speed transmission.

[0232] The internal communication interface 703 is a CAN bus communication interface, through which the first processor 702 and the second processor 801 communicate. The CAN bus communication protocol allocates high priority and defined bandwidth to internal communication, ensuring low latency and high reliability of data transmission. Both internal and external communication are based on the CAN bus protocol. Internal inter-board communication uses high-speed CAN to ensure the interaction requirements of large data volumes and low latency. The external communication interface also adopts a CAN bus design with electrical isolation, effectively suppressing downhole common-mode interference and ensuring the stability of long-distance transmission. The communication protocol assigns the highest priority to toolface angle dynamic measurement data frames, ensuring the timely transmission of critical information during bus contention.

[0233] S1004: After receiving the data, the second processor combines it with the latest real-time sensor measurements and executes the measurement update step of the FF-ZGKF algorithm. In this step, the adaptive forgetting factor comes into play, dynamically balancing the contributions of bounded set uncertainty and random statistical uncertainty, and finally calculating the optimal posterior estimate of the tool facet angle.

[0234] S1005: The second processor performs a rationality check and data quality assessment on the calculated tool face angle. To address the constraint of limited downhole communication bandwidth, the device employs a tiered reporting mechanism: verified high-precision tool face angle data and key status indicators are uploaded in real-time with priority; while complete raw and process data are stored in the storage module.

[0235] Enhanced Reliability: After assembly, the entire sensing and circuit module is encapsulated using a modular vacuum encapsulation process based on a centrally symmetrical layout. This process eliminates air bubbles within the encapsulation material and employs an independent cavity design for differentiated sealing of each functional module, significantly improving the overall dielectric strength, heat dissipation performance, withstand voltage, and waterproof rating of the module. Furthermore, the centrally symmetrical sensor layout itself possesses redundancy characteristics, and combined with the modular, detachable design, it allows for partial repair or replacement in the event of a single sensor failure, enhancing the device's field maintainability and overall lifecycle economics.

[0236] In summary, the device described in this embodiment achieves efficient allocation of computing power through a distributed dual-circuit board collaborative processing architecture. It overcomes the challenge of high-precision calculation under dual uncertainty interference by hardware implementation of a centrally symmetric multicell based on the forgetting factor and a Gaussian mixture Kalman filter method. Furthermore, through highly reliable isolated communication, intelligent power management, and enhanced packaging technology, it ensures stable and reliable high-precision and robust dynamic measurement of tool face angles under extremely complex working conditions of high temperature, strong vibration, and low computing power in downhole drilling. This provides solid technical equipment support for precise trajectory control in directional drilling.

[0237] This invention achieves high-reliability sensing under extreme conditions: by employing a centrally symmetrical sensing unit composed of three independent single-axis quartz accelerometers and a single-axis gyroscope, the system's environmental adaptability is fundamentally improved. The selected quartz accelerometers possess significant advantages in high temperature resistance and strong vibration resistance, enabling them to operate stably directly in the extreme physical environment of strong vibration and high temperature in downhole applications. This provides a high-quality, highly reliable raw signal source for dynamic measurements, overcoming the bottleneck of traditional MEMS sensors being prone to failure under harsh conditions.

[0238] To ensure the real-time performance of complex algorithms under limited resources, the device employs a distributed dual-circuit board collaborative processing architecture. Through hardware function decoupling, the tasks of sensor signal acquisition, preprocessing, and core filtering calculation are distributed to two microcontrollers for parallel processing. This design effectively reduces the computational load of a single processor, ensuring that the complex algorithm based on the forgetting factor-based centrosymmetric multicell and Gaussian mixture Kalman filtering method can be executed smoothly and in real-time under the constraints of low computing power in downhole environments. This resolves the engineering contradiction of balancing algorithm accuracy and real-time performance.

[0239] This invention improves the long-term engineering applicability and maintainability of the device. Based on the modular design of the vacuum potting process and the detachable packaging structure, it ensures the device's pressure resistance, sealing and heat dissipation performance while improving its fault tolerance and on-site maintenance efficiency, providing structural protection for the long-term stability of dynamic measurement accuracy.

[0240] In summary, this invention systematically innovates from four aspects: anti-interference sensing, high-precision algorithms, high-efficiency computing, and high-reliability packaging, forming a complete technical solution that ultimately achieves high-precision, high-reliability, and real-time dynamic measurement of tool face angles under complex and harsh downhole conditions.

[0241] The above embodiments are used to explain the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A method for dynamically measuring the gravity tool face angle of drilling tools, characterized in that, The drilling tool is equipped with a gyroscope and an accelerometer, and the method includes: Using the face angle of the gravity tool and the drift of the gyroscope as system state variables, and the y-axis and z-axis measurements of the accelerometer as observations, a nonlinear state-space model is established by integrating the drilling tool measurement matrix and the nonlinearity of the system, and the nonlinear state-space model is linearized at the current operating point; A hybrid Kalman filter observer is established by using a centrosymmetric multiple cell to characterize the bounded components of the initial values ​​of the system state variables, process disturbances, and observation noise, and by using random vectors to characterize the random components of the initial values ​​of the system state variables, process disturbances, and observation noise. By attenuating the historical information of the hybrid prior covariance matrix of the hybrid Kalman filter through the forgetting factor, the optimal gain of the hybrid Kalman filter is determined based on the hybrid prior covariance matrix, observation matrix and hybrid innovation covariance matrix of the real-time hybrid Kalman filter observer, and the gravity tool face angle is obtained using the established hybrid Kalman filter observer. The historical information of the prior covariance matrix of the hybrid Kalman filter attenuated by the forgetting factor includes: The centrosymmetric multicell generation matrix of the system state variables is: ； in, Here is the state transition matrix at each time step. , , This represents the initial state centrosymmetric multicell generation matrix. for Time-process disturbance centrosymmetric multicell generation matrix Let be the forgetting factor, and the mixture prior covariance matrix be: ； in For weight parameters, For based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. For based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t.

2. The method for dynamic measurement of the gravity tool face angle of drilling tools according to claim 1, characterized in that... The nonlinear state-space model established by using the gravity tool face angle and gyroscope drift as system state variables, and accelerometer y-axis and z-axis measurements as observations, and integrating the drilling tool measurement matrix and system nonlinearity, includes: The tool facet angle and gyroscope drift together constitute the system state variables. The state equation for establishing the combined measurement model that integrates gyroscope and accelerometer data is as follows: ； in, For discrete-time indexing, Represents the state transition matrix. express The angle of gravity tool face at any moment. for Gyroscope drift over time The sampling period is For the input matrix, For the input vector, , for Time-lapse gyroscope rotation speed measurement value For system process interference, for The gyroscope measurement noise at any given time. for White noise that constantly causes gyroscope drift; by Accelerometer y-axis and z-axis measurements at any time , As an observation, construct the observation equation: ； in, for y-axis accelerometer observation at time [time] for z-axis accelerometer observation at time 10:00 for At any given moment, the y-axis accelerometer observes the true y-axis gravitational component. for At any given moment, the z-axis accelerometer observes the true z-axis gravitational component. For observing noise, including vibration noise With the inherent measurement noise of the sensor , Indicates time; For the y-axis measurement matrix of the accelerometer, For the accelerometer z-axis measurement matrix, For the nonlinearity of the accelerometer y-axis system, The nonlinearity of the accelerometer z-axis system.

3. The method for dynamic measurement of the gravity tool face angle of drilling tools according to claim 2, characterized in that... The linearization of the nonlinear state-space model at the current operating point includes: obtaining the Jacobian observation matrix by taking the partial derivative of the observation equation. ； in, for The observation matrix at each time point; Indicates the inclination angle of the well. Indicates based on The posterior estimation results predict the Constant gravity tool face angle; The linearized state-space model is obtained as follows: ； in, For the observation matrix, To observe the noise, These are the observation values ​​from the observation equations of the system's state-space model.

4. The method for dynamic measurement of the gravity tool face angle of drilling tools according to claim 1, characterized in that... The description of using centrosymmetric multiple cells to characterize the bounded components of the system state variables' initial values, process disturbances, and observation noise, and using random vectors to characterize the random components of the system state variables' initial values, process disturbances, and observation noise, includes: representing the initial state... Decomposed into mixing centers Initial state centrosymmetric multicellular and initial state random vector Interference with the process Decomposed into process-disrupting centrosymmetric multicellular structures and process disturbance random vector ; Observation noise Decomposed into observation noise centrosymmetric multicells and observation noise random vector .

5. The method for dynamic measurement of the gravity tool face angle of drilling tools according to claim 1, characterized in that... The optimal gain is obtained by minimizing the cost function, where For weight parameters, The trace of the matrix, for The time-state variable is a centrosymmetric multicell generating matrix. for The prediction covariance matrix of the Gaussian part of the system state variables at time t is obtained. Optimal gain at time for: ； in For the mixed prior covariance matrix, ; Based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. Based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t. For the observation matrix, The mixed posterior covariance matrix, ; for The generation matrix of the centrosymmetric multicell with constant-time noise observation. To observe the covariance matrix of the Gaussian noise component in the noise, For the mixed new information covariance matrix, .

6. The method for dynamic measurement of the gravity tool face angle of drilling tools according to claim 1, characterized in that... The prediction methods of hybrid Kalman filter observers include: The prediction center for the system state variables is: ； in, for The predictive center of the system state variables at any given time. for The input vector at time step; The propagation mode of the bounded components of the system state variables at any given time in a centrosymmetric multicell is as follows: ； in, for The system state variables at any given time are bounded components of a centrosymmetric multicell. for The time-matter disturbance is a centrosymmetric multiple cell; the generating matrix of the bounded components of the system state variables is: ； in, for The system state variables at time t are generated by a centrally symmetric multicell matrix. The propagation mode of the random components of the system state variables is as follows: ； in, for A random vector of states at any given time. for The time-matter disturbance random vector; the covariance matrix of the Gaussian part of the system state vector is: ； in, for The Gaussian part of the predicted covariance matrix of the system state vector at time step [time]. for The covariance matrix of the Gaussian noise component in the time-process disturbance.

7. The method for dynamic measurement of the gravity tool face angle of drilling tools according to claim 1, characterized in that... Hybrid Kalman filter update methods include: The system state variable central update method is as follows: ； in Based on The central posterior estimate of the state variable at time point predicted The center of the system state variables at time t, It is the identity matrix, and the uncertainty update of the bounded components of the system state variables is as follows: ； for The system state vector at any given time is a centrally symmetric multicell. To observe the centrosymmetric many-cell structure of the noise; the generation matrix of the centrosymmetric many-cell structure of the system state variables is updated as follows: ； in Based on The posterior estimation results at time point predicted The time-state variable is a centrosymmetric multicell generating matrix. for The noise center-symmetric multicell generation matrix is ​​observed at all times; the uncertainty of the random components of the system state variables is updated as follows: ； in for Time-based The posterior estimation results at time point predicted A random vector of state variables at any given time. for Observation of a random vector of noise at any given time; The covariance matrix of the system state variables at time step 1 is updated as follows: ； in, For based on The posterior estimation results at time point predicted The prediction covariance matrix of the Gaussian part of the system state variables at time t. To observe the covariance matrix of the Gaussian noise component in the noise.

8. A dynamic measurement device for the gravity tool face angle of drilling tools, comprising an accelerometer and a gyroscope, characterized in that... The apparatus is used to implement the measurement method according to any one of claims 1-7. The apparatus includes a first processor and a second processor. The first processor is used to linearize the nonlinear state-space model at the current operating point. The nonlinear state-space model is established by fusing the drilling tool measurement matrix and the system nonlinearity, with the gravity tool face angle and gyroscope drift as system state variables and the accelerometer y-axis and z-axis measurements as observations. The second processor is used to attenuate the historical information of the prior covariance matrix of the hybrid Kalman filter by a forgetting factor, obtain the optimal gain of the hybrid Kalman filter observer by minimizing the hybrid cost function, and obtain the gravity tool face angle using the established hybrid Kalman filter observer. The hybrid Kalman filter observer uses a centrosymmetric polycell to characterize the bounded components of the initial values ​​of the system state variables, process disturbances, and observation noise, and uses a random vector to characterize the random components of the initial values ​​of the system state variables, process disturbances, and observation noise.

9. A dynamic measuring device for the gravity tool face angle of drilling tools according to claim 8, characterized in that... The measuring device includes a first processing unit and a second processing unit. The first processing unit includes a first processor and an internal communication interface, and the second processing unit includes a second processor and an external communication interface. The first processor sends data to the second processor through the internal communication interface.

Images