A method for controlling wellbore pushing force

Abstract

This invention relates to the field of wellbore pushing force control technology and discloses a wellbore pushing force control method, comprising: acquiring fluid pressure gradient data and simultaneously acquiring the contact stress distribution at the working face, performing synchronous correlation, obtaining spatiotemporally aligned correlation data, fusing the data to eliminate irregular fluctuations, and obtaining a smoothed pressure and stress correlation dataset; real-time detection of local stress accumulation state to generate control commands for adjusting the displacement of the contact actuator; constructing a nonlinear relationship model between the pressure and stress correlation dataset and the contact force output, embedding a dynamic compensation mechanism to correct and fuse the input real-time data to generate a collaborative feedback quantity, adjusting the displacement of the contact actuator in real time through the collaborative feedback quantity, comparing the response delay during the adjustment process with a preset time range, and adjusting the parameters of the dynamic compensation mechanism according to the comparison result to form a closed-loop control. This invention can achieve high-precision and high-robust control of wellbore pushing force.

Description

Technical Field

[0001] This invention relates to the field of push-in force control technology, and in particular to a wellbore push-in force control method. Background Technology

[0002] Currently, in the wellbore trajectory control process of fully rotating vertical drilling systems, precise control of the wellbore pushing force is crucial to ensuring vertical drilling accuracy and preventing wellbore deviation. Especially in formations with alternating soft and hard geological conditions, uneven distribution of the pushing force between the wellbore and the drill string can easily lead to localized stress overload, resulting in serious consequences such as loss of wellbore trajectory control, wellbore instability, or even tool face jamming.

[0003] In existing technologies, wellbore thrust control largely relies on monitoring and feedback from a single signal source, such as downhole pressure sensors or simple strain gauges. These methods have limitations: the mud medium within the wellbore annulus experiences severe pressure pulsations and turbulent disturbances, generating noise signals that significantly interfere with the extraction of accurate pressure gradient information, leading to perception distortion. Drill string rotation, lateral vibration, and formation debris introduce high-frequency interference, causing deviations in the interpretation of thrust state; traditional filtering methods struggle to effectively remove these irregular fluctuations. The density, viscosity, and other fluid properties of the mud dynamically change during drilling, causing hysteresis in pressure signal transmission and resulting in delayed actuator response, hindering real-time and precise thrust control. Traditional PID control algorithms are insufficiently adaptable to such nonlinear, strongly coupled, and high-hysteresis control systems, making it difficult to establish a precise mapping between the pressure gradient and the hydraulic actuator output, and lacking adaptability to complex operating conditions.

[0004] Therefore, this application provides a wellbore pushing force control method to solve the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this invention is to provide a wellbore pushing force control method to solve the technical problem that the existing technology has poor anti-interference ability and is difficult to adapt to complex working conditions, resulting in low wellbore pushing force control accuracy.

[0006] To address the aforementioned technical problems, this invention provides a wellbore pushing force control method, comprising:

[0007] Collect fluid pressure gradient data at different locations within the work area and simultaneously acquire the contact stress distribution on the work surface;

[0008] The fluid pressure gradient data is synchronously correlated with the contact stress distribution of the working surface to obtain spatiotemporally aligned correlation data. The fluid pressure gradient data in the correlation data is then fused to eliminate irregular fluctuations, resulting in a smoothed pressure and stress correlation dataset.

[0009] The local stress accumulation state between the contact actuator and the working surface is detected in real time, and control commands for adjusting the displacement of the contact actuator are generated based on the dynamic changes of the local stress accumulation state.

[0010] A nonlinear relationship model is constructed between the pressure and stress correlation dataset and the contact force output, and a dynamic compensation mechanism for compensating for signal transmission hysteresis deviation is embedded in the nonlinear relationship model.

[0011] The dynamic compensation mechanism is used to correct the input real-time data, and the corrected real-time data is fused to generate a collaborative feedback quantity. The displacement of the contact actuator is adjusted in real time through the collaborative feedback quantity, and the response delay during the adjustment process is compared with a preset time range. The parameters of the dynamic compensation mechanism are adjusted according to the comparison result to form a closed-loop control.

[0012] In some specific embodiments, fluid pressure gradient data at different locations within the working area are collected, and the contact stress distribution on the working surface is acquired simultaneously, further including:

[0013] Instantaneous fluid pressure values ​​at multiple points spatially distributed within the work area are collected using a sensor array;

[0014] Spatial difference calculation is performed on the instantaneous value of the fluid pressure to generate fluid pressure gradient data that reflects the rate of change of pressure in space;

[0015] A specific frequency sound wave signal is emitted directionally toward the working surface, and reflected sound wave signals are received from the working surface;

[0016] The reflected acoustic wave signal is processed using a tomographic imaging algorithm to obtain the contact stress distribution of the working surface, which reflects the stress conditions in each region of the working surface.

[0017] In some specific embodiments, the fluid pressure gradient data is synchronously correlated with the contact stress distribution of the working surface to obtain spatiotemporally aligned correlation data. The fluid pressure gradient data in the correlation data is then fused to eliminate irregular fluctuations, resulting in a smoothed pressure-stress correlation dataset. This further includes:

[0018] Anomaly detection is performed on the fluid pressure gradient data to identify and remove distorted data points caused by transient interference;

[0019] A multi-scale decomposition algorithm is used to process the fluid pressure gradient data after outlier removal, separating the low-frequency components that characterize the true flow state from the high-frequency components that characterize noise.

[0020] The low-frequency components are dynamically smoothed using a state estimation algorithm to obtain denoised fluid pressure gradient data.

[0021] Spatial correlation analysis is performed on the denoised fluid pressure gradient data and the synchronously corresponding contact stress distribution on the working surface to construct the smoothed pressure and stress correlation dataset.

[0022] In some specific embodiments, the local stress concentration state between the contact actuator and the working surface is detected in real time, and control commands for adjusting the displacement of the contact actuator are generated based on the dynamic changes in the local stress concentration state, further including:

[0023] Real-time measurement of the micro-deformation signal generated by the contact actuator due to force;

[0024] Based on the micro-deformation signal and combined with the time delay characteristics of sound wave propagation at the contact interface, the local stress accumulation state between the contact actuator and the working surface is calculated.

[0025] The local stress concentration state is compared with a preset safety threshold in real time;

[0026] When the local stress accumulation state exceeds a preset safety threshold, a corresponding displacement adjustment command is generated based on the dynamic change and output as the control command.

[0027] In some specific embodiments, a nonlinear relationship model is constructed between the pressure and stress correlation dataset and the contact force output, and a dynamic compensation mechanism for compensating for signal transmission hysteresis is embedded in the nonlinear relationship model, further including:

[0028] Collect the changing trends of the pressure and stress correlation dataset and the corresponding contact force output response data from historical operations;

[0029] The change trend and response data are trained based on fuzzy control rules to generate a nonlinear relationship model that characterizes the input and output properties.

[0030] Real-time monitoring of the physical properties of the fluid medium, and extraction of corresponding signal hysteresis compensation parameters based on changes;

[0031] The signal hysteresis compensation parameters are dynamically matched with the collected pressure and stress correlation dataset, and the matching results are embedded into the nonlinear relationship model to achieve real-time correction of the output contact force prediction value.

[0032] In some specific embodiments, after dynamically matching the signal hysteresis compensation parameters with the acquired pressure and stress correlation dataset and embedding the matching results into the nonlinear relationship model to achieve real-time correction of the output contact force prediction value, the method further includes:

[0033] The predicted contact force and the actual contact force output value after the correction of the nonlinear relationship model are continuously acquired;

[0034] Calculate the real-time deviation between the predicted contact force value and the actual contact force output value;

[0035] The logical rules for generating compensation parameters in the dynamic compensation mechanism are updated using an iterative algorithm based on the real-time deviation.

[0036] Closed-loop optimization stabilizes the deviation between the predicted contact force output by the nonlinear relationship model and the actual contact force output within a preset error band.

[0037] In some specific embodiments, the input real-time data is corrected based on the dynamic compensation mechanism, and the corrected real-time data is fused to generate a collaborative feedback quantity, further including:

[0038] Based on the real-time fluctuation of the fluid medium characteristics, the preset hysteresis feature library is queried to obtain the signal hysteresis compensation benchmark under the current operating condition.

[0039] The compensation benchmark amount is weighted and superimposed with the changes in the contact stress distribution of the working surface and the local stress concentration state detected in real time to generate the displacement adjustment compensation amount.

[0040] The displacement adjustment compensation amount is decomposed into a pressure compensation component for correcting the fluid pressure gradient data and a stress compensation component for correcting the contact stress distribution on the working surface.

[0041] After applying the pressure compensation component and stress compensation component for correction, respectively, they are fused with the local stress concentration state to generate the cooperative feedback quantity used to control displacement.

[0042] Based on the same concept, the present invention also provides a wellbore pushing force control system, comprising:

[0043] The data acquisition module is configured to collect fluid pressure gradient data at different locations within the work area and simultaneously acquire the contact stress distribution on the work surface.

[0044] The associated dataset generation module is configured to synchronously associate the fluid pressure gradient data with the contact stress distribution of the working surface to obtain spatiotemporally aligned associated data, and to perform fusion processing on the fluid pressure gradient data in the associated data to eliminate irregular fluctuations, thereby obtaining a smoothed pressure and stress associated dataset.

[0045] The control command generation module is configured to detect the local stress accumulation state between the contact actuator and the working surface in real time, and generate control commands for adjusting the displacement of the contact actuator based on the dynamic changes of the local stress accumulation state.

[0046] The dynamic compensation mechanism embedding module is configured to construct a nonlinear relationship model between the pressure and stress correlation dataset and the contact force output, and to embed a dynamic compensation mechanism for compensating for signal transmission hysteresis deviation in the nonlinear relationship model.

[0047] The control execution module is configured to correct the input real-time data based on the dynamic compensation mechanism, fuse the corrected real-time data to generate a collaborative feedback quantity, adjust the displacement of the contact actuator in real time through the collaborative feedback quantity, compare the response delay during the adjustment process with a preset time range, and adjust the parameters of the dynamic compensation mechanism according to the comparison result to form a closed-loop control.

[0048] Based on the same concept, the present invention also provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of a wellbore pushing force control method.

[0049] Based on the same concept, the present invention also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a wellbore push-in force control method.

[0050] Compared with existing technologies, its advantages are as follows:

[0051] This invention discloses a wellbore pushing force control method that can integrate multi-source information, suppress strong interference, dynamically compensate for signal lag, and has strong adaptive capability, enabling high-precision and high-robust control of wellbore pushing force. Attached Figure Description

[0052] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0053] Figure 1 This is a flowchart illustrating some specific embodiments of the wellbore pushing force control method of the present invention;

[0054] Figure 2 This is a schematic diagram of the structure of a wellbore pushing force control system according to some specific embodiments of the present invention;

[0055] Figure 3 This is a schematic diagram of the structure of an electronic device according to some specific embodiments of the present invention;

[0056] In the diagram, 710 is the processor; 720 is the memory; 730 is the input device; and 740 is the output device. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0059] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0060] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0061] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0062] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0063] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.

[0064] Reference Figure 1 A wellbore pushing force control method, comprising:

[0065] S101, collects fluid pressure gradient data at different locations within the work area and simultaneously acquires the contact stress distribution on the work surface;

[0066] S102, synchronously correlate the fluid pressure gradient data with the contact stress distribution of the working surface to obtain spatiotemporally aligned correlation data, and perform fusion processing on the fluid pressure gradient data in the correlation data to eliminate irregular fluctuations, and obtain a smoothed pressure and stress correlation dataset.

[0067] S103, Real-time detection of the local stress accumulation state between the contact actuator and the working surface, and generation of control commands for adjusting the displacement of the contact actuator based on the dynamic changes of the local stress accumulation state.

[0068] S104, Construct a nonlinear relationship model between the pressure and stress correlation dataset and the contact force output, and embed a dynamic compensation mechanism for compensating for signal transmission hysteresis deviation in the nonlinear relationship model;

[0069] S105, the input real-time data is corrected based on the dynamic compensation mechanism, and the corrected real-time data is fused to generate a collaborative feedback quantity. The displacement of the contact actuator is adjusted in real time through the collaborative feedback quantity, and the response delay during the adjustment process is compared with a preset time range. The parameters of the dynamic compensation mechanism are adjusted according to the comparison result to form a closed-loop control.

[0070] Specifically, in this embodiment of the invention, a sensor array arranged in a specific topology is deployed at key locations in the work area to synchronously collect instantaneous fluid pressure values ​​at multiple spatially distributed points within the area at high frequency. A spatial difference calculation method is used to generate fluid pressure gradient data reflecting the spatial rate of pressure change. A specific frequency acoustic signal is directionally emitted towards the work surface and its reflected signal is received. An inversion algorithm based on wave theory and regularization optimization is used to process the reflected signal, thereby obtaining the contact stress distribution reflecting the stress conditions in each area of ​​the work surface. A high-precision timestamp mechanism ensures that the above two types of data remain synchronized in time and space, obtaining spatiotemporally aligned correlated data. The fluid pressure gradient data in the correlated data is fused to eliminate irregular fluctuations. This process includes: outlier detection and removal from the original pressure gradient data to remove transient interference; separation of low-frequency components characterizing the true flow state and filtering out high-frequency noise using a multi-scale decomposition algorithm; dynamic smoothing of the retained low-frequency components using a state estimation algorithm to obtain denoised data; and spatial correlation analysis of the smoothed pressure gradient data and the synchronized contact stress distribution data to construct a smoothed pressure and stress correlated dataset. The system monitors the local stress accumulation state between the contact actuator and the working surface in real time. This is achieved by measuring the micro-deformation signal generated by the actuator body under stress and combining this with the time delay characteristics of sound wave propagation at the contact interface to comprehensively calculate the local stress accumulation state. This state quantity is then compared in real time with a preset safety threshold. If the threshold is exceeded, a corresponding displacement adjustment command is generated based on the dynamic change and output as a control command to the servo actuator system. A nonlinear relationship model is constructed between the smoothed pressure and stress correlation dataset and the system's historical contact force output. This model is trained based on fuzzy rules and system identification methods. Simultaneously, a dynamic compensation mechanism is embedded in the model. This mechanism monitors the changes in the physical properties of the fluid medium in real time and extracts corresponding signal transmission hysteresis deviation compensation parameters from a preset hysteresis feature library. These parameters are dynamically matched with real-time data and embedded into the model to achieve real-time correction of the model's output prediction values. Based on this dynamic compensation mechanism, the fluid pressure gradient and stress distribution data input to the system in real time are corrected. The corrected data is then weighted and fused with the local stress accumulation state data detected in real time to generate a comprehensive collaborative feedback quantity. This feedback quantity is used to adjust the displacement of the contact actuator in real time, and the response delay of the entire adjustment process from command generation to displacement execution is continuously monitored. This delay is compared with a preset time range, and the parameter generation logic and strength of the compensation mechanism are dynamically adjusted according to the comparison results, thereby forming a continuously optimized adaptive closed-loop control link.

[0071] For example, when deploying a sensor array, the drill bit diameter is 216mm, 12 sensors are symmetrically arranged around the axis, with an angle of 30° between adjacent sensors, and the radial installation depth of the sensors is 15mm from the drill bit surface. During dynamic calibration, the drill bit angular velocity is assumed to be 5rad / s, the mud density is 1200kg / m³, and the sensor installation radius is 0.015m. The corrected pressure value is then calculated using the centrifugal force compensation formula. During acoustic stress inversion, the center frequency of the acoustic wave transmitter is 1MHz, the Young's modulus of the shale formation is 15GPa, and the distance from the contact surface to the sound source in the acoustic wave propagation path is 50mm. Inversion calculations are performed using a time-delay stress model and a regularization algorithm with a regularization parameter of 0.07. The inversion error does not exceed 8kPa. When calculating the thrust fusion, the pressure gradient signal-to-noise ratio was 20dB and the acoustic inversion confidence level was 0.9. The calculated pressure gradient fusion coefficient was approximately 0.69, and the acoustic stress fusion coefficient was approximately 0.31. The thrust of the hydraulic actuator was calculated by integrating the finite element discretization of the drill bit-well wall contact surface into 256 micro-elements. For wavelet packet denoising, a 5-level decomposition was performed using the Symlet8 wavelet basis. Low-frequency components at the 4th level and below were processed using a threshold function with a threshold of 1.2 multiplied by the square root of the logarithm of the number of nodes. The measured signal-to-noise ratio improved to 28dB after denoising. For Kalman dynamic smoothing, the state variables were selected as the mean pressure, slope of change, and acceleration. The process noise covariance matrix was set as a diagonal matrix with diagonal elements of 0.01, 0.05, and 0.1, respectively. The observation noise covariance was set to 0.3. After filtering, the root mean square error of the pressure gradient data decreased from ±0.7MPa / m to ±0.2MPa / m. When constructing the fuzzy proportional-integral-differential model, the input variable pressure gradient was divided into three fuzzy sets: low, medium, and high, with inflection points of 1.5, 2.0, and 2.5 MPa / m, respectively. The contact force error was divided into three fuzzy sets: negative, zero, and positive, with support sets of -10%, 0, and +10%, respectively. Forty-nine control rules were generated through training. During dynamic compensation parameter updates, if the mud density increased from 1.2 g / cm³ to 1.35 g / cm³, the viscosity change was estimated online based on the Hagen-Poiseuille equation, and a transfer function was constructed for phase compensation, reducing the phase lag from 45° to 12°. Signal lag dynamic compensation employed a dual-loop parameter update mechanism. The inner loop used a least mean square algorithm every 10 ms with a step size of 0.02 to fine-tune the compensation weights in real time. The outer loop used a particle swarm optimization algorithm with a population size of 50 every 1 second to iterate 10 times to update the compensation structure parameters, causing the objective function to converge to 0.15. This embodiment ensures that the response time of the closed-loop control link is constrained within the preset drilling process cycle, and achieves high-precision adaptive matching of the contact force between the drill bit and the wellbore.

[0072] In some applications, fluid pressure gradient data at different locations within the working area are collected, and the contact stress distribution of the working surface is acquired simultaneously. This includes collecting instantaneous fluid pressure values ​​at multiple spatially distributed points within the working area using a sensor array; performing spatial difference calculations on the instantaneous fluid pressure values ​​to generate fluid pressure gradient data reflecting the spatial rate of change of pressure; directionally transmitting a specific frequency acoustic wave signal to the working surface and receiving reflected acoustic wave signals from the working surface; and processing the reflected acoustic wave signals based on a tomographic imaging algorithm to invert and obtain the contact stress distribution of the working surface reflecting the stress conditions in each region of the working surface.

[0073] Understandably, a sensor array arranged in a specific spatial topology is deployed within the work area. Each sensor in this array collects instantaneous fluid pressure values ​​at its location at a millisecond-level synchronous sampling frequency, ensuring the acquisition of multi-point pressure data distributed spatially. These synchronously acquired instantaneous fluid pressure values ​​are processed using a spatial differential calculation method. By calculating the ratio of the pressure difference between adjacent measuring points to the spatial distance, fluid pressure gradient data reflecting the rate of pressure change in space is generated. This data characterizes the non-uniformity of the flow field pressure distribution. Simultaneously, to obtain the contact stress distribution on the work surface, an ultrasonic transmitting device is installed directly above the work surface. This device directionally emits sound wave signals of a specific frequency towards the work surface. After the sound waves are reflected from the work surface, these reflected sound wave signals are captured by a receiving array. Based on the captured reflected sound wave signals, a tomographic imaging algorithm is used for processing. This algorithm analyzes the changes in the time delay and attenuation characteristics of sound waves propagating in the medium, and reconstructs the contact stress distribution pattern in each region of the work surface, thereby obtaining contact stress distribution data reflecting the stress condition of the work surface. By aligning the acquisition time of fluid pressure gradient data with the transmission and reception time of acoustic signals through a high-precision timestamp synchronization mechanism, the two types of data are kept synchronized in time and space, laying the foundation for subsequent data fusion and correlation analysis.

[0074] For example, within a 216mm diameter annular area at the drill bit tip, 12 pressure sensors are symmetrically arranged around the axis, with an angle of 30° between adjacent sensors. The sensors are installed at a depth of 15mm from the drill bit surface. The sensors synchronously acquire instantaneous mud pressure values ​​at their locations using a 1000Hz sampling frequency, obtaining pressure values ​​of P1=20.5MPa, P2=20.7MPa, P3=20.3MPa…P12=20.6MPa. The pressure gradient is calculated using the central difference method. Taking sensors 1 and 2 as an example, with a distance ΔL=15mm between them, the pressure gradient is (20.7-20.5) / 0.015=13.3MPa / m. The built-in acoustic wave transmitter in the drill pipe directionally emits acoustic pulses towards the wellbore at a center frequency of 1MHz. After reflection at the wellbore-formation interface, the acoustic waves are captured by the receiving array, and the echo delay sequence is measured. Based on the tomographic inversion algorithm, a mapping relationship between time delay and stress is established, assuming the propagation speed of sound waves in shale formations (E=15GPa). Then based on the time delay Calculate the propagation path length The contact stress value σ1 = 8.5 MPa in this region was then obtained through inversion. A nanosecond-level precision timestamp synchronization mechanism was used to ensure that the time alignment error between the pressure gradient data and the stress distribution data was less than 1 ms, thereby generating a spatiotemporally synchronized associated dataset.

[0075] In some applications, the fluid pressure gradient data is synchronously correlated with the contact stress distribution of the working surface to obtain spatiotemporally aligned correlated data. The fluid pressure gradient data in the correlated data is then fused to eliminate irregular fluctuations, resulting in a smoothed pressure-stress correlated dataset. This process includes outlier detection of the fluid pressure gradient data, identifying and removing distorted data points caused by transient disturbances; processing the outlier-removed fluid pressure gradient data using a multi-scale decomposition algorithm to separate low-frequency components representing the true flow state from high-frequency components representing noise; dynamically smoothing the low-frequency components using a state estimation algorithm to obtain denoised fluid pressure gradient data; and performing spatial correlation analysis between the denoised fluid pressure gradient data and the synchronously corresponding contact stress distribution of the working surface to construct the smoothed pressure-stress correlated dataset.

[0076] It is understood that the first part of the embodiment

[0077] In this embodiment of the invention, outlier detection is performed on the collected fluid pressure gradient data. A threshold determination method based on statistical distribution is used to identify and remove distorted data points caused by instantaneous interference such as bubble bursting and impurity collision, ensuring the validity of the data. A multi-scale decomposition algorithm is used to process the cleaned fluid pressure gradient data. By decomposing the signal in different scale spaces, stable low-frequency components representing the real flow state and high-frequency components representing measurement noise and turbulence disturbances are separated, achieving effective separation of signal and noise. The separated low-frequency components are dynamically smoothed using a state estimation algorithm. This algorithm is based on a state-space model and performs optimal estimation of the signal through prediction and update steps, thereby obtaining denoised and smoothed fluid pressure gradient data. The processed fluid pressure gradient data is spatially correlated with the contact stress distribution data of the working surface that is strictly synchronized through timestamps. The coupling relationship and statistical characteristics of the two in spatial distribution are analyzed. Based on this, a pressure and stress correlation dataset with spatiotemporal alignment, smooth data and clear physical correlation is constructed to provide data input for subsequent control decisions.

[0078] For example, outlier detection was performed on the initially acquired pressure gradient dataset [∇P1=13.3MPa / m,∇P2=12.8MPa / m,∇P3=15.1MPa / m,∇P4=8.5MPa / m (outliers),...,∇P12=13.0MPa / m]. The threshold was set to the mean ± 3 times the standard deviation (mean μ=13.1MPa / m, standard deviation σ=1.5MPa / m), and the outlier data point ∇P4=8.5MPa / m was identified and removed. The remaining data was decomposed into 5 levels using the Symlet8 wavelet basis, decomposing the signal into different frequency subbands. The low-frequency approximation coefficients cA4=[12.9,13.2,12.7,...,13.1]MPa / m at the 4th level (62.5-125Hz) and below were extracted, and high-frequency detail coefficients (>125Hz) were filtered out. Kalman filtering was applied to the low-frequency coefficients for dynamic smoothing. The state variable was set as X=[P,∇P,∇²P]ᵀ, the process noise covariance was Q=diag[0.01,0.05,0.1], and the observation matrix was H=[0,1,0]. After filtering, smooth pressure gradient data were obtained [∇P1_sm=13.1MPa / m,∇P2_sm=12.9MPa / m,∇P3_sm=13.0MPa / m,...,∇P12_sm=12.95MPa / m]. Spatial correlation analysis was performed on the above data and the wellbore contact stress distribution data [σ1=8.5MPa,σ2=8.3MPa,σ3=8.7MPa,...,σ12=8.4MPa] obtained by synchronous inversion. The correlation coefficient r between pressure gradient and stress was calculated to be 0.89, indicating a strong positive correlation. Finally, a smooth pressure-stress correlation dataset {(13.1,8.5),(12.9,8.3),(13.0,8.7),...,(12.95,8.4)} was constructed.

[0079] In some applications, the local stress accumulation state between the contact actuator and the working surface is detected in real time, and control commands for adjusting the displacement of the contact actuator are generated based on the dynamic changes of the local stress accumulation state. This includes real-time measurement of the micro-deformation signal generated by the contact actuator due to force; calculation of the local stress accumulation state between the contact actuator and the working surface based on the micro-deformation signal and the time delay characteristics of sound wave propagation at the contact interface; real-time comparison of the local stress accumulation state with a preset safety threshold; and when the local stress accumulation state exceeds the preset safety threshold, a corresponding displacement adjustment command is generated based on the dynamic changes and output as the control command.

[0080] Understandably, a multi-channel stress sensing unit embedded in the key stress points of the contact actuator measures the micro-deformation signal generated by the mechanism under stress in real time. The signal is output in the form of electrical parameters, reflecting the micro-deformation of the contact interface. Based on the micro-deformation signal and combined with the time delay characteristics of the sound waves emitted by the sound wave emitting device linked to the actuator propagating at the contact interface, a physical model of the sound wave propagation speed and interface stress state is established to calculate the local stress concentration state between the contact actuator and the working surface. This state quantity characterizes the actual stress concentration degree at the contact point. The calculated local stress concentration state is compared in real time with a preset safety threshold, which is a stress upper limit preset according to the material properties of the working surface, the tool design strength, and process requirements. When the local stress concentration state is detected to exceed the preset safety threshold, a corresponding displacement adjustment command is generated according to the amplitude and trend of its dynamic change through preset control rules. This command includes the displacement direction and the magnitude of the adjustment, and is output as a control command to the servo execution system, thereby realizing real-time closed-loop adjustment of the displacement of the contact actuator.

[0081] For example, a set of silicon piezoresistive micro-strain sensors (range 0-50MPa, sampling frequency 2kHz) is integrated on the piston surfaces of each of the three hydraulic cylinders of the drill bit pushing actuator to measure micro-deformation signals in real time. The measured output voltage values ​​at a certain moment are U1=3.21V, U2=3.45V, and U3=4.02V (full scale 5V corresponds to 50MPa). An acoustic wave emitting device (center frequency 1MHz) transmits pulse signals to the contact interface, and the measured echo delays are... Based on the calibration curve σ=0.25×(Δτ-100)×U (where... The local stress values ​​at the three contact points were calculated as follows: σ1 = 0.25 × (5.2) × 3.21 = 4.17 MPa, σ2 = 0.25 × (4.8) × 3.45 = 4.14 MPa, and σ3 = 0.25 × (3.5) × 4.02 = 3.52 MPa. A safety threshold of σ_safe = 10 MPa was set for the shale well wall, and real-time comparison revealed that the stress at all measuring points did not exceed the limit. Further monitoring showed that σ3 rapidly increased from 3.52 MPa to 10.82 MPa within 10 ms, exceeding the safety threshold with a rate of change Δσ / Δt = 0.73 MPa / ms. According to the preset control rules, when Δσ / Δt > 0.5 MPa / ms, a displacement callback command was generated: ΔL = -1.5 mm (the negative sign indicates contraction). The command is output to the servo controller via the CAN bus in 0x23ID data frame format, driving the hydraulic cylinder to retract the pushing mechanism by 1.5mm, reducing σ3 to 8.95MPa and restoring it to the normal safe range.

[0082] In some applications, a nonlinear relationship model is constructed between the pressure and stress correlation dataset and the contact force output. A dynamic compensation mechanism for compensating for signal transmission hysteresis is embedded in the nonlinear relationship model. This includes collecting the changing trend of the pressure and stress correlation dataset and the corresponding contact force output response data in historical operations; training the changing trend and response data based on fuzzy control rules to generate a nonlinear relationship model characterizing the input and output characteristics; monitoring the physical property parameters of the fluid medium in real time and extracting the corresponding signal hysteresis compensation parameters according to the changes; dynamically matching the signal hysteresis compensation parameters with the collected pressure and stress correlation dataset and embedding the matching result into the nonlinear relationship model to achieve real-time correction of the predicted output contact force value.

[0083] Understandably, the process involves collecting historical data on the changing trends of pressure and stress correlation datasets over time, along with their corresponding contact force output response data, to form a historical dataset for model training. This historical dataset is then used to train the model. Fuzzy control rules are applied to this dataset, and fuzzy clustering is used to categorize operating conditions and establish local linear sub-models for each condition. A neural network is then used to fuse these sub-models, generating an initial nonlinear relationship model that characterizes the complex nonlinear relationship between system input and output. Real-time monitoring of the physical properties of the fluid medium, including density and viscosity, is conducted. Based on the magnitude and trend of these parameter changes, corresponding signal transmission hysteresis compensation parameters are extracted using system identification methods. These extracted hysteresis compensation parameters are dynamically matched with the real-time collected pressure and stress correlation dataset to establish association rules between the compensation parameters and data features. The matching results are then embedded as weighted adjustments into the parameter adjustment module of the nonlinear relationship model, enabling real-time correction of the model's output contact force predictions and ensuring that the predictions adapt to the effects of changes in fluid medium properties.

[0084] For example, historical data from the past 50 drilling operations were collected, including pressure gradient data [∇P1=12.8MPa / m, ∇P2=13.2MPa / m,...], stress distribution data [σ1=8.3MPa, σ2=8.6MPa,...], and their corresponding push force output values ​​[F1=42.3kN, F2=43.1kN,...]. Fuzzy C-means clustering was used to divide the data into 5 operating condition categories. After establishing a local linear model for each category, an initial nonlinear relationship model F_out=f(∇P,σ) was generated by fusing the data through a 3-layer BP neural network (4-6-1 structure). Real-time monitoring of mud characteristic parameters revealed that when density increased from 1.20 g / cm³ to 1.35 g / cm³ and viscosity increased from 30 cP to 45 cP, the pressure signal hysteresis time constant calculated based on the Hagen-Poiseuille equation increased from 0.6 s to 1.2 s. Compensation parameters were extracted, including the hysteresis constant τ = 1.2 s and the gain coefficient K = 0.85. These compensation parameters were matched with the real-time acquired ∇P = 13.5 MPa / m and σ = 9.2 MPa. When the pressure change rate d(∇P) / dt > 1.5 MPa / m / s, a proactive compensation strategy was adopted, converting the compensation parameters into weight adjustment amounts ΔW = [0.12, -0.08, 0.15] to update the model parameter matrix. After correction, the predicted pushing force output by the model was adjusted from 41.8kN to 43.5kN, and the deviation from the actual measured value of 43.2kN was reduced from 3.4% to 0.7%, thus improving the prediction accuracy.

[0085] In some applications, after dynamically matching the signal hysteresis compensation parameters with the acquired pressure and stress correlation dataset and embedding the matching results into the nonlinear relationship model to achieve real-time correction of the output contact force prediction value, the process further includes continuously acquiring the contact force prediction value and the actual contact force output value after correction by the nonlinear relationship model; calculating the real-time deviation between the contact force prediction value and the actual contact force output value; updating the logic rules used to generate compensation parameters in the dynamic compensation mechanism using an iterative algorithm based on the real-time deviation; and stabilizing the deviation between the contact force prediction value and the actual contact force output value output by the nonlinear relationship model within a preset error band through closed-loop optimization.

[0086] Understandably, the predicted contact force output after the nonlinear relationship model is corrected is continuously acquired, along with the actual contact force output value measured by the sensor, forming a prediction-actual data pair. The real-time deviation between the predicted contact force value and the actual contact force output value is calculated, reflecting the model's prediction accuracy under the current operating conditions. Based on the real-time deviation, an iterative algorithm is used to update the logic rules for generating compensation parameters in the dynamic compensation mechanism. The weighting coefficients and calculation logic in the compensation parameter generation rules are adjusted by analyzing the trend and characteristics of the deviation. Through the above closed-loop optimization process, the deviation between the predicted contact force output by the nonlinear relationship model and the actual contact force output value is stabilized within a preset error band, forming a complete adaptive optimization loop from data acquisition, model prediction, deviation calculation to rule updating, ensuring that the control system can maintain stable control accuracy under various operating conditions.

[0087] For example, the model-corrected contact force prediction values ​​[F_p1=43.5kN, F_p2=43.8kN, F_p3=44.1kN,...] and actual contact force output values ​​[F_a1=43.2kN, F_a2=43.5kN, F_a3=44.5kN,...] are continuously acquired at a frequency of 100Hz, and the real-time deviation [e1=0.3kN, e2=0.3kN, e3=-0.4kN,...] is calculated. When the absolute value of the average deviation over 10 consecutive sampling periods exceeds the preset threshold of 0.5kN, the gradient descent algorithm is initiated to update the compensation parameter generation rule: learning rate. Weight adjustment amount ,in The error was calculated using the partial derivative of the deviation with respect to the weight parameters. After five iterations, the time delay compensation weight in the compensation rule was adjusted from 0.75 to 0.68, and the gain coefficient was adjusted from 1.05 to 0.97. The updated rule was written into the knowledge base of the dynamic compensation mechanism, and the deviation between the predicted and actual contact force output by the model decreased from ±1.2kN to ±0.4kN, stabilizing within the preset ±0.5kN error band. Continuous monitoring of deviation changes automatically triggers a new round of rule update when the operating conditions change again, causing the deviation to exceed the threshold, forming a complete closed-loop optimization cycle.

[0088] In some applications, the input real-time data is corrected based on the dynamic compensation mechanism, and the corrected real-time data is fused to generate a collaborative feedback quantity. This includes querying a preset hysteresis feature library based on the real-time fluctuation of the fluid medium characteristics to obtain the signal hysteresis compensation benchmark quantity under the current working condition; weighting and superimposing the compensation benchmark quantity with the changes in the contact stress distribution of the working surface and the local stress accumulation state detected in real time to generate a displacement adjustment compensation quantity; decomposing the displacement adjustment compensation quantity into a pressure compensation component for correcting the fluid pressure gradient data and a stress compensation component for correcting the contact stress distribution of the working surface; applying the pressure compensation component and the stress compensation component for correction respectively, and then performing feature fusion with the local stress accumulation state to generate the collaborative feedback quantity used to control the displacement.

[0089] Understandably, based on the fluctuations in fluid medium characteristic parameters monitored in real time by online sensors, a pre-set hysteresis feature library is queried. This feature library stores the correspondence between different medium characteristics and signal hysteresis deviations, obtaining a signal hysteresis compensation benchmark quantity that matches the current working condition. The obtained compensation benchmark quantity is then weighted and superimposed with the real-time detected working surface contact stress distribution data and the changes in local stress accumulation state. The weighting coefficients are dynamically adjusted according to the reliability and real-time performance of each data point to generate a comprehensive displacement adjustment compensation quantity. A modal decomposition algorithm is used to decompose the displacement adjustment compensation quantity into two independent components: a pressure compensation component for correcting the fluid pressure gradient data and a stress compensation component for correcting the working surface contact stress distribution. The pressure compensation component is applied to perform phase lead and amplitude correction on the real-time collected fluid pressure gradient data, while the stress compensation component is applied to correct the model parameters of the working surface contact stress distribution data. The corrected fluid pressure gradient data, the corrected working surface contact stress distribution data, and the uncorrected local stress accumulation state data are then fused at the feature level. Principal component analysis is used to extract key feature vectors, ultimately generating a collaborative feedback quantity for controlling the displacement of the contact actuator.

[0090] For example, real-time monitoring showed that the mud density increased from 1.20 g / cm³ to 1.28 g / cm³, and the viscosity increased from 32 cP to 38 cP. Querying the hysteresis feature library yielded the corresponding signal hysteresis compensation benchmarks: time delay compensation coefficient K_t = 1.15, amplitude compensation coefficient K_a = 0.92. Real-time monitoring showed the contact stress distribution at the working face as [σ_1 = 8.7 MPa, σ_2 = 8.9 MPa, σ_3 = 9.2 MPa], with a local stress concentration change Δσ = +0.8 MPa / ms. Using weighted coefficients w1 = 0.6 (compensation benchmark), w2 = 0.3 (stress distribution), and w3 = 0.1 (stress change), a weighted summation was performed to generate a displacement adjustment compensation amount ΔD = 1.15 × 0.6 + 8.9 × 0.3 + 0.8 × 0.1 = 3.52 mm. The compensation amount was decomposed into a pressure compensation component ΔD_p = 2.35 mm (67% of the total) and a stress compensation component ΔD_σ = 1.17 mm (33% of the total) through modal decomposition. The pressure compensation component was used to correct the collected pressure gradient data, changing ∇P = 13.8 MPa / m to ∇P' = 13.8 × 0.92 + 2.35 = 15.04 MPa / m; the stress compensation component was used to correct the stress distribution data, changing the average stress σ_avg = 8.93 MPa to σ_avg' = 8.93 × 1.15 - 1.17 = 9.10 MPa. The corrected pressure gradient data of 15.04 MPa / m, the corrected stress distribution data of [8.97, 9.15, 9.38] MPa, and the local stress concentration data of 9.2 MPa are fused. Principal component analysis yields the first principal component coefficients as [0.72, 0.58, 0.38]. A collaborative feedback quantity C_f = 0.72 × 15.04 + 0.58 × 9.23 + 0.38 × 9.2 = 22.15 is generated and converted into a 4-20mA control signal, which is output to the servo execution system to drive the pushing mechanism for precise displacement adjustment.

[0091] Another embodiment of the present invention is described below:

[0092] This embodiment uses a distributed fluid pressure sensor array to collect fluid pressure gradient data at different locations within the work area in real time, and simultaneously utilizes acoustic wave reflection signals to invert the contact stress distribution on the work surface. The distributed fluid pressure sensor array is deployed at key locations within the work area, with each sensor collecting fluid pressure values ​​at a millisecond-level synchronous sampling frequency. Fluid pressure gradient data is generated through spatial difference calculation. An ultrasonic transmitter is installed directly above the work surface, directionally emitting acoustic waves of a specific frequency towards the work surface. After the receiving array captures the reflected signals, a tomographic imaging algorithm is used to invert the contact stress distribution in different areas of the work surface. A timestamp synchronization mechanism aligns the two types of data streams, establishing a spatial mapping relationship between pressure gradient and stress distribution, providing a foundational dataset for subsequent fusion processing.

[0093] After synchronously correlating the fluid pressure gradient data with the contact stress distribution at the working surface, wavelet denoising and state estimation algorithms are used to process irregular fluctuations in the fluid pressure gradient data. Outlier detection is performed on the synchronously acquired fluid pressure gradient data to remove transient spikes caused by bubbles or impurities. Wavelet packet transform is used to decompose the gradient data into multiple scales, retaining low-frequency components reflecting the true flow state and filtering out high-frequency noise. The true state of the system is estimated using a Kalman filter, and spatial correlation analysis is performed between the denoised pressure gradient and the contact stress distribution at the working surface to identify the transmission delay characteristics between them. A joint pressure-stress feature vector is constructed and stored in a circular buffer for real-time control retrieval.

[0094] A multi-channel stress sensing unit is integrated within the contact actuator. Combined with an acoustic wave emitting device linked to the actuator, it continuously monitors the local stress concentration between the actuator and the working surface. Based on the dynamic changes in this local stress concentration, the servo actuator system adjusts the displacement of the contact actuator. Multi-channel stress sensing units are embedded at key stress points of the actuator, and strain gauge arrays measure the micro-deformation of the actuator body in real time. The acoustic wave emitting device, linked to the mechanism, operates in pulse mode, measuring the propagation delay of sound waves at the contact interface to infer the local stress concentration. When the detected stress concentration exceeds a safety threshold, the servo actuator system is driven to adjust the actuator's forward displacement according to a preset gradient, and the stress distribution uniformity is reassessed after each adjustment. The correlation between displacement adjustment and stress relief effect is recorded, forming an experience base for intelligent decision-making reference.

[0095] A nonlinear relationship model between the fluid pressure gradient data and the contact force output is constructed based on the fuzzy proportional-integral-differential algorithm, and a dynamic compensation mechanism is embedded in this model. A historical dataset containing pressure gradient input and contact force output is established, and an initial model reflecting the nonlinear characteristics of the system is trained using the fuzzy proportional-integral-differential algorithm. The dynamic compensation mechanism is designed into the model, including a pressure signal hysteresis compensation module and a contact force prediction correction module. Changes in the density and viscosity parameters of the fluid medium are monitored in real time; when a change in medium characteristics is detected, an online update process for the compensation parameters is automatically triggered. The updated compensation coefficients are written to the model's memory mapping area to ensure that the control command generation stage can call the latest parameters in real time.

[0096] The dynamic compensation mechanism eliminates signal transmission lag caused by changes in fluid medium properties, enabling the coordinated feedback of fluid pressure gradient data, contact stress distribution on the working surface, and local stress accumulation state to act in real time on the displacement adjustment process of the servo actuator, forming a closed-loop control link. This achieves adaptive matching between the contact force output by the contact actuator and the actual stress state of the working surface, with the response time of the closed-loop control link constrained within a preset time range. The dynamic compensation mechanism performs phase lead correction on the collected fluid pressure gradient data to eliminate time lag caused by the signal transmission pipeline. The compensated pressure data, stress distribution data, and local stress state are weighted and fused to generate a comprehensive feedback quantity input to the servo actuator. The system dynamically adjusts the displacement command of the actuator based on the feedback quantity, while strictly monitoring the response delay of the entire closed-loop control link. When the response delay exceeds a preset time threshold, a degradation protection strategy is activated. The contact force matching accuracy after each adjustment is recorded for continuous optimization of the compensation mechanism.

[0097] A nonlinear relationship model between the fluid pressure gradient data and the contact force output is constructed based on the fuzzy proportional-integral-differential algorithm. A dynamic compensation mechanism is embedded in this nonlinear relationship model, including collecting real-time change rates of the fluid pressure gradient data and historical response data of the contact force output. An initial nonlinear relationship model between fluid pressure and contact force is generated based on fuzzy rules. Samples of the change rates of the fluid pressure gradient data and their corresponding contact force output response curves are extracted from the historical database, and data cleaning and feature annotation are performed. The input and output data are divided into several operating condition categories using a fuzzy C-means clustering algorithm, and a local linear sub-model is established for each category. The sub-models are fused using a neural network to generate an initial nonlinear relationship model that covers all operating conditions. This model includes the core mapping rules for pressure-force conversion. An online learning interface is configured for the model to support the real-time embedding of subsequent compensation parameters.

[0098] Compensation parameters for pressure signal hysteresis are extracted based on real-time detection results of fluid medium characteristic changes. These parameters are then dynamically matched with the real-time change rate of fluid pressure gradient data. Fluid medium characteristic changes are monitored in real-time using online viscometers and densitometers. When a deviation of the medium parameter from the baseline value exceeds 5%, the compensation process is triggered. A system identification method is used to calculate the hysteresis deviation of the pressure signal under the current medium conditions, extracting compensation parameters that include the time delay constant and gain coefficient. A correlation rule is established between the compensation parameters and the real-time pressure change rate; when the pressure fluctuates drastically, the lead compensation is increased, while a conservative compensation strategy is adopted for gradual changes. The parameter matching table is loaded into the shared memory area for real-time access by the model adjustment module.

[0099] The matched compensation parameters are embedded into the parameter adjustment module of the initial nonlinear relationship model. The output response curve of the model is corrected by adjusting the weight allocation coefficients within the module. The structure of the initial nonlinear relationship model is analyzed to locate the weight coefficient matrix of its parameter adjustment module. The matched compensation parameters are converted into weight adjustment values, and the connection weights within the model are updated through matrix operations. Sensitivity analysis is used to verify the effect of the weight adjustment, ensuring that the slope change of the model's output response curve at key operating points meets expectations. A version snapshot is created for the adjusted model to retain rollback capability in response to unexpected operating conditions.

[0100] The compensation parameter generation logic in the dynamic compensation mechanism is updated based on the difference between the corrected output response curve and the real-time contact force output data, forming a closed-loop parameter update chain. The difference between the corrected output response curve and the measured contact force output data is compared in real time, and the root mean square error is calculated as a feedback indicator. When the error continuously exceeds a threshold, the compensation parameter iterative algorithm is initiated: the weighting coefficients in the compensation parameter generation rules are adjusted using the gradient descent method. The updated rules are written into the knowledge base, and optimal parameter combinations are tagged and stored for different working conditions. A parameter version management mechanism is established to ensure a smooth transition during the rule update process.

[0101] The dynamic compensation mechanism of the nonlinear relationship model is continuously optimized through a closed-loop parameter update link, ensuring that the deviation between the model's predicted contact force and the actual contact force feedback signal remains within a preset range. The closed-loop parameter update link continuously monitors the deviation trend between the model's predicted values ​​and the actual feedback, triggering a model retraining process when a systematic deviation is detected. Incremental learning algorithms are used to optimize the deep parameters of the nonlinear relationship model, focusing on adjusting the trigger threshold and compensation intensity of the dynamic compensation mechanism. The control effect after the model update is verified using a digital twin system, and deployment to the physical system is only made after safety is confirmed. This forms a complete evolutionary closed loop from data acquisition and model optimization to control execution, ensuring that the contact force control accuracy is consistently maintained within a preset range of ±2%.

[0102] The dynamic compensation mechanism eliminates signal transmission lag caused by changes in fluid medium properties, enabling the coordinated feedback of fluid pressure gradient data, working surface contact stress distribution, and local stress accumulation to act in real-time on the displacement adjustment process of the servo execution system, forming a closed-loop control link. This includes real-time detection of the correlation between the fluctuation amplitude of fluid medium properties and signal transmission lag, and extraction of the lag compensation benchmark based on the current rate of change of fluid pressure gradient data. Instantaneous changes in fluid medium properties are monitored in real-time using online viscosity sensors and densitometers, recording fluctuation curves of parameters such as temperature and viscosity. Next, cross-correlation analysis is used to calculate the time delay relationship between medium parameter changes and pressure signal transmission lag, establishing a lag feature library corresponding to different fluctuation amplitudes. Based on the slope of the current fluid pressure gradient data, the closest operating mode is matched from the feature library, and the benchmark compensation parameter under this mode is extracted as the compensation benchmark. The compensation benchmark is then subjected to moving average filtering to eliminate parameter jitter caused by random interference, ensuring the stability of the compensation amount.

[0103] The compensation reference value is superimposed with real-time data of the contact stress distribution on the working surface and the dynamic changes in the local stress concentration state to generate a displacement adjustment compensation value for the servo actuator. A real-time cloud map of the contact stress distribution on the working surface is obtained from a distributed sensor network to identify the location and intensity of stress concentration areas. Stress change rate data is collected through a local stress concentration state monitoring unit to calculate the dynamic stress change at the current moment. Then, a weighted superposition algorithm is used to fuse the compensation reference value, stress distribution data, and stress change: higher weights are assigned to high-pressure areas, and lower weights are assigned to edge areas. The superposition result is converted into displacement adjustment compensation values ​​for each degree of freedom of the servo actuator through coordinate transformation, generating a control package containing X / Y / Z three-axis adjustment commands.

[0104] A dynamic compensation mechanism decomposes the displacement adjustment compensation amount into pressure compensation and stress compensation components. The pressure compensation component is used to correct the fluid pressure gradient data, while the stress compensation component is used to correct the contact stress distribution at the working surface. The decomposition module of the dynamic compensation mechanism receives the complete displacement adjustment compensation amount and separates it into low-frequency pressure compensation and high-frequency stress compensation components using a modal decomposition algorithm. The pressure component is input to the fluid control subsystem, where a PID algorithm is used to adjust the reference value of the pressure sensor, eliminating measurement deviations caused by signal transmission lag. The stress component is transmitted to the contact force control system, where a fuzzy logic regulator corrects the calculation model parameters of the contact stress distribution at the working surface. The evaluation indicators of the adjustment effect of each component are recorded to provide data support for subsequent optimization of the compensation strategy.

[0105] The corrected fluid pressure gradient data, the contact stress distribution at the working surface, and the data on local stress accumulation are synchronously fused to generate a collaborative feedback quantity, which is then input into the displacement adjustment module of the servo actuator. Spatial interpolation is performed on the corrected fluid pressure gradient data to generate a continuous pressure field distribution map. The updated contact stress distribution data at the working surface is time-aligned with the monitoring results of local stress accumulation to ensure consistent data timeliness. Feature-level fusion technology is used to project the three types of data onto a unified state space. Principal component analysis is used to extract key feature vectors, synthesizing a collaborative feedback quantity that characterizes the overall system state. This feedback quantity is encoded into a standard control protocol format and transmitted in real time to the displacement adjustment command queue of the servo actuator.

[0106] The compensation benchmark generation logic in the dynamic compensation mechanism is updated based on the real-time changes in the collaborative feedback quantity, ensuring that the output response of the displacement adjustment module remains synchronized with the preset time range of the closed-loop control link, forming a closed-loop feedback iteration. The temporal variation characteristics of the collaborative feedback quantity are monitored; when the frequency of feedback fluctuation exceeds a threshold, the compensation logic update process is triggered. An incremental learning algorithm is used to analyze the matching degree between the feedback quantity changes and the current compensation benchmark quantity, dynamically adjusting the benchmark quantity generation rules: increasing the response weight for rapidly changing conditions and reducing sensitivity to steady-state conditions. The updated rules are written into the knowledge base of the dynamic compensation mechanism, and the calculation efficiency of the compensation parameters is optimized to ensure that the response speed of the closed-loop control link always meets the preset time constraints.

[0107] The following describes this embodiment in conjunction with an application scenario:

[0108] During the drill bit advancement process of the fully rotating vertical drilling system, a distributed fluid pressure sensor array is embedded in a ring layout at the drill bit's front end to collect real-time mud pressure gradient data in different areas of the wellbore. Simultaneously, a high-frequency pulse signal is directionally emitted towards the wellbore via a built-in acoustic transmitter in the drill pipe, and the reflected echo signal is received to invert the stress distribution pattern in the contact area between the wellbore and the drill bit. After the collected mud pressure gradient data and the working face contact stress distribution inverted by the acoustic wave are synchronized with a timestamp, a wavelet denoising algorithm is used to filter out pressure abrupt noise caused by mud turbulence disturbances. Furthermore, a Kalman state estimator is used to dynamically smooth the pressure gradient data, eliminating irregular fluctuations caused by changes in drilling fluid viscosity. Inside the drill bit pushing actuator, three sets of multi-channel stress sensing units are integrated. These units use piezoelectric ceramic sensors to capture the local stress accumulation at the contact point between the drill bit and the wellbore in real time. When uneven hardness of the wellbore rock formations leads to an abnormal increase in local stress, a linked acoustic emission device immediately triggers a high-frequency detection mode. This dynamically corrects the stress accumulation data and inputs the change into the displacement controller of the servo actuator, driving the hydraulic cylinder to adjust the extension of the pushing mechanism to balance the local pressure. Based on the fuzzy proportional-integral-differential algorithm, the system models the nonlinear response relationship between real-time mud pressure gradient data and the pushing force as a multivariable coupled model. A dynamic compensation module is embedded in this model: based on the signal hysteresis characteristics caused by changes in downhole mud temperature and density, the compensation weights of the pressure gradient data are adaptively adjusted. The corrected model predicts the pushing force output curve, and the deviation is iterated with the actual contact stress feedback value, updating the compensation parameter generation logic in real time. During this process, the dynamic compensation mechanism integrates the mud pressure correction component, the wellbore stress correction component, and the local stress accumulation increment into a collaborative feedback quantity, which directly affects the opening adjustment of the hydraulic servo valve. This ensures that the adjustment response time of the drill bit push force is always constrained within the preset drilling process cycle, ultimately achieving adaptive matching of the contact force between the drill bit and the wellbore, and avoiding the risk of wellbore collapse or drilling pressure overload.

[0109] The above content will be explained in detail below:

[0110] Real-time monitoring of the push force distribution using a distributed sensor array and acoustic wave inversion, and sensor array topology optimization: In a fully rotating vertical drilling system, the distributed fluid pressure sensor array with a ring layout at the drill bit tip needs to meet the requirements of high-density measurement and vibration resistance.

[0111] Assuming a drill bit diameter of Φ216mm, 12 sensors (n=12) are symmetrically arranged around the axis, with an angle θ=30° between adjacent sensors. Through Fisher information matrix optimization, the radial installation depth of the sensors is determined to be r=15mm (from the drill bit surface), improving the pressure gradient measurement sensitivity by 37%. Dynamic calibration: When the drill bit rotates, a centrifugal force compensation algorithm is used to correct the sensor readings.

[0112] P_{corr}=P_{raw}+0.5ρ(ωr)^2;

[0113] Where ω is the drill bit angular velocity (rad / s), ρ is the mud density, and r is the sensor installation radius.

[0114] The acoustic stress inversion algorithm uses a high-frequency pulse acoustic wave transmitter (center frequency 1MHz) built into the drill pipe to directionally transmit signals towards the wellbore. A time-delay-stress model is then used for shale formations (E=15GPa, β=2.8×10⁻⁶). -6 Pa -1 The sound wave propagation path Γ is drill bit-wellbore-reflecting surface. A quantitative relationship between time delay τ and contact stress σ is established:

[0115] ;

[0116] In the formula, L=50mm is the distance from the contact surface to the sound source, and Δτ_{geo} is the wellbore geometric distortion correction term.

[0117] The contact stress distribution σ was solved using Tikhonov regularization. The regularization parameter α was determined to be α=0.07 by the L-curve method, and the inversion error was ≤8kPa.

[0118] The calculation of pushing force fusion involves the fusion of multi-source data of mud pressure gradient and sonic inversion stress: the fusion coefficients w1 and w2 are dynamically adjusted according to the confidence level.

[0119] w1=SNR_p / (SNR_p+SNR_s),w2=1-w1;

[0120] SNR_p is the pressure gradient signal-to-noise ratio, and SNR_s is the acoustic wave inversion confidence (calculated from the echo intensity).

[0121] The thrust F_k of the hydraulic actuator is synthesized by spatial vector:

[0122] ;

[0123] The integration region A_c is the contact surface between the drill bit and the well wall, which is discretized into 256 micro-elements using the finite element method.

[0124] Wavelet packet-Kalman joint noise reduction algorithm for mud turbulence noise suppression, targeting high-frequency pressure disturbances (frequency band 100-500Hz) caused by drilling fluid turbulence: A 5-level decomposition using the Symlet8 wavelet basis is performed to extract the fourth level (62.5-125Hz) and lower low-frequency components. Thresholding formula:

[0125] ;

[0126] Where λ=1.2, N_i is the number of nodes in the i-th layer, and the measured signal-to-noise ratio is improved to 28dB after noise reduction.

[0127] Kalman dynamic smoothing addresses the gradual pressure variation interference caused by abrupt changes in drilling fluid viscosity (e.g., from 30 cP to 45 cP): x = [P_mean, P_slope, P_acc]^T, representing the mean pressure, slope of change, and acceleration, respectively. The process noise covariance Q = diag[0.01, 0.05, 0.1], and the observation noise covariance R = 0.3. After Kalman filtering, the root mean square error (RMSE) of the pressure gradient data decreases from ±0.7 MPa / m to ±0.2 MPa / m.

[0128] Fuzzy PID nonlinear relationship modeling and fuzzy rule base construction are employed to address the nonlinear control requirements of alternating hard and soft strata (sandstone-shale alternation): fuzzification of input variables is implemented.

[0129] Pressure gradient ΔP: divided into {Low, Medium, High}, with a trapezoidal membership function, and the inflection point ΔP = [1.5, 2.0, 2.5] MPa / m.

[0130] Contact force error e: divided into {Negative, Zero, Positive}, the membership function is triangularly distributed, and the support set e=[-10%,0,+10%].

[0131] Forty-nine rules were generated by training a neural network using downhole measured data. A typical rule example is shown below.

[0132] IFΔP=High AND e=Positive THEN ΔKp=-0.15,Ki=0.6,Kd=0.3;

[0133] Dynamic compensation parameter updates, time-delay compensation when mud density increases from 1.2 g / cm³ to 1.35 g / cm³: online viscosity estimation based on the Hagen-Poiseuille equation.

[0134] ;

[0135] D_h=80mm is the hydraulic diameter, L=5m is the length of the measurement section, and Q=0.8m³ / s is the flow rate. The transfer function G_c(s)=(1+3.5×0.8s) / (1+0.8s) is used, and the phase lag is reduced from 45° to 12° after compensation.

[0136] Signal lag dynamic compensation mechanism, dual-loop parameter update, inner loop (10ms level): uses LMS algorithm to fine-tune compensation weights in real time.

[0137] w(k+1)=w(k)+0.02·e(k)·[ΔP(k),σ(k)]^T;

[0138] Where e(k) = F_actual(k) - F_model(k) is the contact force deviation.

[0139] Outer loop (1s level): The compensation structure parameters are updated based on particle swarm optimization (PSO), with a population size of 50. After 10 iterations, the objective function converges to 0.15.

[0140] For the purpose of simplicity, the method steps disclosed in the above embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0141] like Figure 2 As shown, the present invention also provides a wellbore pushing force control system, comprising:

[0142] The data acquisition module 201 is configured to collect fluid pressure gradient data at different locations within the work area and simultaneously acquire the contact stress distribution on the work surface.

[0143] The associated dataset generation module 202 is configured to synchronously associate the fluid pressure gradient data with the contact stress distribution of the working surface to obtain spatiotemporally aligned associated data, and to perform fusion processing on the fluid pressure gradient data in the associated data to eliminate irregular fluctuations, thereby obtaining a smoothed pressure and stress associated dataset.

[0144] The control command generation module 203 is configured to detect the local stress accumulation state between the contact actuator and the working surface in real time, and generate control commands for adjusting the displacement of the contact actuator based on the dynamic changes of the local stress accumulation state.

[0145] The dynamic compensation mechanism embedding module 204 is configured to construct a nonlinear relationship model between the pressure and stress correlation dataset and the contact force output, and to embed a dynamic compensation mechanism for compensating for signal transmission hysteresis deviation in the nonlinear relationship model.

[0146] The control execution module 205 is configured to correct the input real-time data based on the dynamic compensation mechanism, and fuse the corrected real-time data to generate a collaborative feedback quantity. The displacement of the contact actuator is adjusted in real time through the collaborative feedback quantity, and the response delay during the adjustment process is compared with a preset time range. The parameters of the dynamic compensation mechanism are adjusted according to the comparison result to form a closed-loop control.

[0147] It is worth noting that although only some basic functional modules are disclosed in the embodiments of this invention, it does not mean that the composition of this system is limited to the above-mentioned basic functional modules. On the contrary, what this embodiment intends to express is that, based on the above-mentioned basic functional modules, those skilled in the art can arbitrarily add one or more functional modules in combination with existing technology to form an infinite number of embodiments or technical solutions. That is to say, this system is open rather than closed. The fact that this embodiment only discloses a few basic functional modules should not be considered as the scope of protection of the claims of this invention being limited to the disclosed basic functional modules. At the same time, for the convenience of description, the above device is described separately according to its functions as various units and modules. Of course, in implementing this invention, the functions of each unit and module can be implemented in one or more software and / or hardware.

[0148] like Figure 3 As shown, the present invention also provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of a wellbore pushing force control method.

[0149] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. For example... Figure 3 The structure shown in this embodiment of the invention includes an electronic device comprising one or more processors 710 and a memory 720; the processors 710 in this electronic device may be one or more. Figure 3 Taking a processor 710 as an example; a memory 720 is used to store one or more programs; the one or more programs are executed by the one or more processors 710, so that the one or more processors 710 implement a wellbore pushing force control method as described in any one of the embodiments of the present invention.

[0150] The electronic device may also include an input device 730 and an output device 740.

[0151] The processor 710, memory 720, input device 730, and output device 740 in this electronic device can be connected via a bus or other means. Figure 3Taking the example of a connection between China and Israel via a bus.

[0152] The memory 720 in this electronic device serves as a computer-readable storage medium, capable of storing one or more programs. These programs can be software programs, computer-executable programs, or modules, such as the program instructions / modules corresponding to the wellbore pushing force control method provided in this embodiment of the invention. The processor 710 executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory 720, thereby implementing the wellbore pushing force control method described in the above embodiment.

[0153] The memory 720 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 720 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 720 may further include memory remotely located relative to the processor 710, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0154] Input device 730 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the electronic device. Output device 740 may include display devices such as a display screen.

[0155] Furthermore, when one or more programs included in the aforementioned electronic device are executed by one or more processors 710, the programs perform the following operations:

[0156] The present invention also provides a computer-readable storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of a wellbore push-in force control method.

[0157] Specifically, the computer storage medium in this embodiment of the invention can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be—but is not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0158] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; 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 or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling wellbore pushing force, characterized in that, include: Collect fluid pressure gradient data at different locations within the work area and simultaneously acquire the contact stress distribution on the work surface; The fluid pressure gradient data is synchronously correlated with the contact stress distribution of the working surface to obtain spatiotemporally aligned correlation data. The fluid pressure gradient data in the correlation data is then fused to eliminate irregular fluctuations, resulting in a smoothed pressure and stress correlation dataset. The local stress accumulation state between the contact actuator and the working surface is detected in real time, and control commands for adjusting the displacement of the contact actuator are generated based on the dynamic changes of the local stress accumulation state. A nonlinear relationship model is constructed between the pressure and stress correlation dataset and the contact force output, and a dynamic compensation mechanism for compensating for signal transmission hysteresis deviation is embedded in the nonlinear relationship model. The dynamic compensation mechanism is used to correct the input real-time data, and the corrected real-time data is fused to generate a collaborative feedback quantity. The displacement of the contact actuator is adjusted in real time through the collaborative feedback quantity, and the response delay during the adjustment process is compared with a preset time range. The parameters of the dynamic compensation mechanism are adjusted according to the comparison result to form a closed-loop control.

2. The wellbore pushing force control method according to claim 1, characterized in that, Collect fluid pressure gradient data at different locations within the work area, and simultaneously acquire the contact stress distribution on the work surface, further including: Instantaneous fluid pressure values ​​at multiple points spatially distributed within the work area are collected using a sensor array; Spatial difference calculation is performed on the instantaneous value of the fluid pressure to generate fluid pressure gradient data that reflects the rate of change of pressure in space; A specific frequency sound wave signal is emitted directionally toward the working surface, and reflected sound wave signals are received from the working surface; The reflected acoustic wave signal is processed using a tomographic imaging algorithm to obtain the contact stress distribution of the working surface, which reflects the stress conditions in each region of the working surface.

3. The wellbore pushing force control method according to claim 1, characterized in that, The fluid pressure gradient data is synchronously correlated with the contact stress distribution of the working surface to obtain spatiotemporally aligned correlation data. The fluid pressure gradient data in the correlation data is then fused to eliminate irregular fluctuations, resulting in a smoothed pressure-stress correlation dataset, which further includes: Anomaly detection is performed on the fluid pressure gradient data to identify and remove distorted data points caused by transient interference; A multi-scale decomposition algorithm was used to process the fluid pressure gradient data after outlier removal, separating the low-frequency components that characterize the real flow state from the high-frequency components that characterize noise. The low-frequency components are dynamically smoothed using a state estimation algorithm to obtain denoised fluid pressure gradient data. Spatial correlation analysis is performed on the denoised fluid pressure gradient data and the synchronously corresponding contact stress distribution on the working surface to construct the smoothed pressure and stress correlation dataset.

4. The wellbore pushing force control method according to claim 1, characterized in that, Real-time detection of the local stress concentration state between the contact actuator and the working surface, and generation of control commands for adjusting the displacement of the contact actuator based on the dynamic changes of the local stress concentration state, further including: Real-time measurement of the micro-deformation signal generated by the contact actuator due to force; Based on the micro-deformation signal and combined with the time delay characteristics of sound wave propagation at the contact interface, the local stress accumulation state between the contact actuator and the working surface is calculated. The local stress concentration state is compared with a preset safety threshold in real time; When the local stress accumulation state exceeds a preset safety threshold, a corresponding displacement adjustment command is generated based on the dynamic change and output as the control command.

5. The wellbore pushing force control method according to claim 1, characterized in that, A nonlinear relationship model is constructed between the pressure and stress correlation dataset and the contact force output, and a dynamic compensation mechanism for compensating for signal transmission hysteresis is embedded in the nonlinear relationship model, further including: Collect the changing trends of the pressure and stress correlation dataset and the corresponding contact force output response data from historical operations; The change trend and response data are trained based on fuzzy control rules to generate a nonlinear relationship model that characterizes the input and output properties. Real-time monitoring of the physical properties of the fluid medium, and extraction of corresponding signal hysteresis compensation parameters based on changes; The signal hysteresis compensation parameters are dynamically matched with the collected pressure and stress correlation dataset, and the matching results are embedded into the nonlinear relationship model to achieve real-time correction of the output contact force prediction value.

6. The wellbore pushing force control method according to claim 5, characterized in that, After dynamically matching the signal hysteresis compensation parameters with the acquired pressure and stress correlation dataset and embedding the matching results into the nonlinear relationship model to achieve real-time correction of the output contact force prediction value, the method further includes: The predicted contact force and the actual contact force output value after the correction of the nonlinear relationship model are continuously acquired; Calculate the real-time deviation between the predicted contact force value and the actual contact force output value; The logical rules for generating compensation parameters in the dynamic compensation mechanism are updated using an iterative algorithm based on the real-time deviation. Closed-loop optimization stabilizes the deviation between the predicted contact force output by the nonlinear relationship model and the actual contact force output within a preset error band.

7. The wellbore pushing force control method according to claim 6, characterized in that, Based on the aforementioned dynamic compensation mechanism, the input real-time data is corrected, and the corrected real-time data is fused to generate a collaborative feedback quantity, further including: Based on the real-time fluctuation of the fluid medium characteristics, the preset hysteresis feature library is queried to obtain the signal hysteresis compensation benchmark under the current operating condition. The compensation benchmark amount is weighted and superimposed with the changes in the contact stress distribution of the working surface and the local stress concentration state detected in real time to generate the displacement adjustment compensation amount. The displacement adjustment compensation amount is decomposed into a pressure compensation component for correcting the fluid pressure gradient data and a stress compensation component for correcting the contact stress distribution on the working surface. After applying the pressure compensation component and stress compensation component for correction, respectively, they are fused with the local stress concentration state to generate the cooperative feedback quantity used to control displacement.

8. A wellbore pushing force control system, characterized in that, include: The data acquisition module is configured to collect fluid pressure gradient data at different locations within the work area and simultaneously acquire the contact stress distribution on the work surface. The associated dataset generation module is configured to synchronously associate the fluid pressure gradient data with the contact stress distribution of the working surface to obtain spatiotemporally aligned associated data, and to perform fusion processing on the fluid pressure gradient data in the associated data to eliminate irregular fluctuations, thereby obtaining a smoothed pressure and stress associated dataset. The control command generation module is configured to detect the local stress accumulation state between the contact actuator and the working surface in real time, and generate control commands for adjusting the displacement of the contact actuator based on the dynamic changes of the local stress accumulation state. The dynamic compensation mechanism embedding module is configured to construct a nonlinear relationship model between the pressure and stress correlation dataset and the contact force output, and to embed a dynamic compensation mechanism for compensating for signal transmission hysteresis deviation in the nonlinear relationship model. The control execution module is configured to correct the input real-time data based on the dynamic compensation mechanism, fuse the corrected real-time data to generate a collaborative feedback quantity, adjust the displacement of the contact actuator in real time through the collaborative feedback quantity, compare the response delay during the adjustment process with a preset time range, and adjust the parameters of the dynamic compensation mechanism according to the comparison result to form a closed-loop control.

9. An electronic device, characterized in that, include: The system includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus; the memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the steps of the method according to any one of claims 1 to 7.

Images