Full-condition adaptive intelligent drilling robot system and closed-loop control method

Abstract

This invention relates to the field of intelligent drilling technology in geotechnical engineering, specifically disclosing a full-condition adaptive intelligent drilling robot system and a closed-loop control method. The method includes: real-time acquisition of drilling parameters of the intelligent drilling robot system at a preset sampling frequency to construct a drilling state feature vector; online identification of formation response parameters using an active perturbation method to form an inner-loop dynamic data stream; based on the inner-loop dynamic data stream, a model predictive control optimizer is used to solve for the optimal combination of drilling parameters satisfying multiple constraints, and the target setpoint of the outer loop is dynamically updated; real-time deviation analysis is performed between the real-time drilling state and the optimal drilling parameters to calculate anomaly factors, and control commands are triggered based on anomaly factor thresholds; drilling trajectory parameters are dynamically corrected based on control commands, and the correction information is fed back to the model predictive control optimizer to complete the iterative closed loop; this invention achieves multi-objective dynamic optimization of drilling parameters and provides automated diagnosis and timely feedback for abnormal working conditions.

Description

Technical Field

[0001] This invention relates to the field of intelligent drilling technology in geotechnical engineering, specifically to an all-condition adaptive intelligent drilling robot system and a closed-loop control method. Background Technology

[0002] In the process of underground space development and resource exploration, traditional drilling operations mainly rely on manual operation, with on-site drillers judging the sound, vibration and slag return during the drilling process, and adjusting the main drilling parameters such as drilling pressure and rotation speed based on their professional experience. However, with the continuous deep expansion and increasing complex geological conditions in geotechnical engineering construction, the limitations of traditional drilling methods have become increasingly apparent.

[0003] To address the limitations of complex geological conditions and the automation level of drilling equipment, intelligent control methods have been introduced into the drilling field. Intelligent drilling equipment using fuzzy control and neural network control, such as intelligent drilling robots, can achieve real-time simulation of drilling conditions and verification of experimental data to a certain extent. However, their application in actual drilling projects has significant limitations. The main reasons are the current lack of reliable multi-source sensing systems, the complexity of multi-source data, and the inability to obtain accurate and synchronous drilling status information. In addition, there is a lack of lightweight real-time geological identification technology, making it difficult to perform rapid inference on computing devices. Furthermore, the dynamic optimization of some drilling parameters depends on the actual engineering implementation framework, resulting in significant deployment defects in advanced intelligent algorithms for model predictive control, making it impossible to deploy computationally complex algorithms on embedded platforms. Summary of the Invention

[0004] The purpose of this invention is to provide a full-condition adaptive intelligent drilling robot system and a closed-loop control method to solve the following technical problems: (1) How to solve the problem of multi-objective dynamic optimization of drilling parameters under real-time drilling conditions of existing drilling equipment; (2) How to perform automated diagnosis and timely feedback for abnormal operating conditions, and realize the implementation, deployment and engineering application of closed-loop control.

[0005] The objective of this invention can be achieved through the following technical solutions: A closed-loop control method for a full-condition adaptive intelligent drilling robot system, comprising: Step 1: Collect drilling parameters of the intelligent drilling robot system in real time at a preset sampling frequency, construct a drilling status feature vector, identify formation response parameters online using the active perturbation method, and form an inner loop dynamic data stream; drilling parameters include drilling pressure, torque, and rotational speed; the inner loop dynamic data stream includes formation type, formation response parameters, and real-time drilling status; Step 2: Based on the inner loop dynamic data flow, the model predictive control optimizer is used to solve for the optimal combination of drilling parameters that satisfies multiple constraints, and the target setpoint of the outer loop is dynamically updated. Step 3: Perform real-time deviation analysis between the real-time drilling status and the optimal drilling parameters, calculate the anomaly factor, and trigger control commands based on the anomaly factor threshold. Step 4: Dynamically correct the drilling trajectory parameters based on control commands, and feed the correction information back to the model predictive control optimizer to complete the iterative closed loop.

[0006] Preferably, formation response parameters are identified online using the active perturbation method to form an inner-loop dynamic data stream. The specific steps are as follows: S101. Perform an active identification cycle at a preset frequency, superimposing a small sinusoidal rotation speed disturbance on the normal drilling command for a duration of 0.5 seconds; S102. Synchronously observe the torque response and calculate the amplitude attenuation and phase lag. S103. The equivalent stiffness coefficient of the formation is calculated in real time using the recursive least squares method. With damping coefficient , as formation response parameters for subsequent prediction models; S104. Perform PID servo control on drilling pressure and rotation speed, track the target setpoint of the outer loop, and determine the control frequency.

[0007] Preferably, stratigraphic category Confidence level of stratigraphic categories ,in, The control cycle is represented; the real-time drilling state is characterized based on a discrete-time model, whose equation is expressed as:

[0008] in, This represents the output drilling speed of the drilling machinery. To input drilling pressure, Input speed, For torque, This represents the current formation impedance coefficient, and , It is white noise; , , , , These are the corresponding input model parameters, and are updated online every 5 seconds using recursive least squares with a forgetting factor.

[0009] Preferably, step two, which involves using a model predictive control optimizer to solve for the optimal combination of drilling parameters that satisfies multiple constraints based on the inner loop dynamic data flow, and dynamically updating the target setpoint of the outer loop, specifically includes the following steps: S201. An autoregressive moving average model with exogenous variables is used as a predictive model to describe the real-time drilling status, and it is updated online in real time. S202. Construct a multi-objective optimization function for the prediction model and set the constraints of the multi-objective optimization function; S203. Perform rolling optimization on the prediction model to generate the target given value.

[0010] Preferably, the multi-objective optimization function is:

[0011] Constraints:

[0012] in, Indicates based on stratigraphic category The target drilling rate is adaptively set according to the working mode; This refers to the actual drilling speed; To control the cycle, To increase the number of cycles, , , , These are the weight coefficients corresponding to the target parameters. This represents the minimum drill pressure constraint related to the formation. This represents the maximum drill pressure constraint related to the formation. This is the minimum rotational speed. This represents the maximum rotational speed. This represents the maximum torque.

[0013] Preferably, the formula for calculating the anomaly factor in step three is as follows:

[0014] in, As an abnormal factor, This refers to the actual drilling pressure. This refers to the actual rotational speed. This is the actual torque. The actual drilling mechanical drilling speed; read the optimal combination of drilling parameters output by the model predictive control optimizer of the prediction model, including the optimal drilling pressure. Optimal speed Optimal torque Optimal drilling speed ; The value must be a very small positive number to prevent division by zero errors; The correlation coefficient between drilling pressure and rotation speed. It is the absolute value of the difference between the actual formation impedance coefficient and the predicted optimal impedance coefficient; , , , , , These are the adaptive weighting coefficients for each deviation value, based on the stratigraphic category. and its confidence level Dynamic adjustment.

[0015] Preferably, the judgment triggered by the control command for the abnormal factor threshold is as follows: Three outlier thresholds are set to form two intervals. , And satisfy ;abnormal factors Compare with these two intervals: like If the system is in a steady state, no control commands will be triggered, and the current drilling parameters will be maintained. like If so, a fine-tuning control command is triggered, and the target setpoint is slightly corrected according to a preset proportional coefficient; like If this occurs, a correction control command is triggered, pausing the rolling optimization of the model predictive control optimizer, forcibly initiating active disturbance identification, reacquiring the corresponding formation parameters, and resetting the predictive model. like If this happens, an emergency control command will be triggered, immediately executing a drilling parameter rollback operation to reduce drilling pressure and rotation speed to preset safe values, while simultaneously triggering an alarm.

[0016] Preferably, the method for dynamically correcting the drilling trajectory parameters in step four is as follows: S301. Define the drilling trajectory parameter vector based on the three-dimensional spatial coordinates of the drill bit; S302. Obtain the current actual trajectory parameters in real time through the drilling measurement system, and obtain the correction vector from the preset trajectory correction rule library; S303. Update the target trajectory parameters based on the correction vector, encode the correction information into a trajectory correction command frame, and send it to the robot actuator.

[0017] This invention also provides a full-condition adaptive intelligent drilling robot system, for realizing a closed-loop control method for the full-condition adaptive intelligent drilling robot system, the system comprising: The multi-source sensing module is used to collect drilling parameters of the intelligent drilling robot system in real time through a preset sampling frequency; the drilling parameters include drilling pressure, torque, and rotational speed. The edge computing control module is used to implement dual-closed-loop adaptive control. The edge computing control module includes an inner-loop dynamic data flow construction unit, an outer-loop target setpoint construction unit, and a deviation analysis unit. The inner loop dynamic data stream construction unit is used to construct the drilling state feature vector, identify formation response parameters online through the active perturbation method, and form the inner loop dynamic data stream; the inner loop dynamic data stream includes formation type, formation response parameters and real-time drilling status; The outer ring target setpoint construction unit is used to solve the optimal drilling parameter combination that satisfies multiple constraints based on the inner ring dynamic data flow using a model predictive control optimizer, and dynamically update the outer ring target setpoint. The deviation analysis unit is used to perform real-time deviation analysis between the real-time drilling status and the optimal drilling parameters, and to calculate the anomaly factor. The multi-degree-of-freedom drilling execution module is used to trigger control commands based on the threshold of abnormal factors; it receives correction information and feeds the correction information back to the model predictive control optimizer to complete the loop iteration. The mobile walking module is used to dynamically correct drilling trajectory parameters based on control commands.

[0018] The beneficial effects of this invention are: (1) The drilling parameters collected in real time, the formation categories and their confidence levels identified by the 1D-CNN-LSTM neural network, and the formation response parameters identified by the active perturbation method are fused to form a unified inner loop dynamic data stream, realizing the spatiotemporal alignment and structured output of multi-source heterogeneous information, solving the problem of fragmented perception information during drilling, and providing a standardized input interface that can be used at any time for the outer loop model predictive control optimizer. At the same time, the inner loop runs at a high frequency of 100-500Hz, ensuring the real-time and continuous nature of the data, enabling the outer loop optimizer to obtain the latest formation state in each control cycle. The model predictive control optimizer superimposes the optimal control increment obtained by the solution with the current value to determine the target setpoint of the outer loop, and sends it to the inner loop PID servo controller for execution. By constructing a dual closed-loop control architecture of inner and outer loop coordination—the inner loop achieves accurate tracking of the target value at a high frequency, and the outer loop achieves adaptive optimization of parameters at a low frequency; the coordinated control process of the inner and outer loops is realized.

[0019] (2) Based on the inner loop dynamic data flow, the autoregressive moving average model with exogenous variables (ARMAX) is used as the prediction model. The model parameters are updated in real time by online recursive least squares method. A multi-objective optimization function including drilling speed tracking error, torque penalty, drilling pressure change rate penalty, and rotation speed change rate penalty is constructed. The optimal control increment is solved in a rolling manner under drilling pressure constraints, rotation speed constraints, and torque change rate constraints. Dynamic optimization of drilling parameters under multiple constraints is realized. Unlike the traditional PID control system, it explicitly handles multivariate coupling and constraint boundaries, avoids drilling pressure and torque exceeding limits, has multi-step forward prediction capability, and responds to formation changes in advance. Moreover, the rolling optimization mechanism enables the system to recalculate the optimal parameters according to the latest state in each control cycle, ensuring real-time adaptive adjustment of the drilling state under the control cycle.

[0020] (3) The real-time drilling status (actual drilling pressure, rotation speed, torque, and drilling speed) and the optimal drilling parameters output by the model predictive control optimizer are analyzed in multiple dimensions. The weighted fusion formula is used to calculate the comprehensive anomaly factor, which realizes the upgrade from single parameter monitoring to multi-dimensional comprehensive anomaly assessment. The correlation coefficient can detect the abnormal coupling between drilling pressure and rotation speed, and the impedance deviation can quickly identify formation changes. Compared with the alarm method of only monitoring single parameter over-limit in the traditional system, the comprehensive anomaly factor of the present invention has higher sensitivity and lower false alarm rate, and can be identified in the anomaly in the incipient stage.

[0021] Of course, any product implementing this invention does not necessarily need to achieve all the advantages described above at the same time. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the closed-loop control method of a full-condition adaptive intelligent drilling robot system according to the present invention. Figure 2 This is a module framework diagram of an all-condition adaptive intelligent drilling robot system according to the present invention; Figure 3 This is a step diagram of the method for forming the inner loop dynamic data stream in step one of the present invention; Figure 4 This is a step diagram illustrating the method for obtaining the target given value of the outer ring in step two of this invention; Figure 5 This is a step diagram of the method for dynamically correcting drilling trajectory parameters in step four of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 As shown, this invention provides a closed-loop control method for a full-condition adaptive intelligent drilling robot system, the method comprising: Step 1: Collect drilling parameters of the intelligent drilling robot system in real time at a preset sampling frequency, construct a drilling status feature vector, identify formation response parameters online using the active perturbation method, and form an inner loop dynamic data stream; drilling parameters include drilling pressure, torque, and rotational speed; the inner loop dynamic data stream includes formation type, formation response parameters, and real-time drilling status; Step 2: Based on the inner loop dynamic data flow, the model predictive control optimizer is used to solve for the optimal combination of drilling parameters that satisfies multiple constraints, and the target setpoint of the outer loop is dynamically updated. Step 3: Perform real-time deviation analysis between the real-time drilling status and the optimal drilling parameters, calculate the anomaly factor, and trigger control commands based on the anomaly factor threshold. Step 4: Dynamically correct the drilling trajectory parameters based on control commands, and feed the correction information back to the model predictive control optimizer to complete the iterative closed loop.

[0026] In the above technical solution, the operator inputs the borehole coordinates on the cloud platform. The robot automatically walks to the borehole, automatically levels its four legs, and loads default drilling parameters (drilling pressure 15kN, rotation speed 80rpm, flushing fluid flow rate 80L / min). Drilling pressure, torque, rotation speed, and vibration signals are collected at a frequency of 200Hz. The formation identification module outputs the formation type as "soft plastic clay" with a confidence level of 92%, and the outer ring is optimized at a rolling frequency of 5Hz. After the system starts, the multi-source sensing module is initialized through the edge computing control module in step one, setting the sampling frequency to 100-500Hz (200Hz is selected in this embodiment). Drilling parameters, including drilling pressure W, torque M, rotation speed N, and drilling speed v, are collected synchronously, and corresponding parameters are configured for these parameters. Measurement range and accuracy; all sensor data are transmitted to the edge computing control module in real time via EtherCAT bus, with a time synchronization error of less than 1ms; a drilling state feature vector is constructed, and the corresponding formation parameters are identified online using the active perturbation method. The original sensor data is preprocessed using a sliding window (window length 0.5 seconds, step size 0.1 seconds); the vibration signal is bandpass filtered (0.5-200Hz) and subjected to short-time Fourier transform (STFT) to extract the low-frequency energy (0-20Hz), high-frequency energy (20-100Hz), and torsional vibration energy of axial vibration; the corresponding mechanical specific energy, torque fluctuation coefficient, and flushing fluid pressure difference are calculated, and an 8-dimensional drilling state feature vector is constructed based on these values. Then, in step two, the drilling dynamics are described by using an autoregressive moving average model with exogenous variables (ARMAX). Based on the drilling dynamics, the model parameters are updated every 5 seconds using recursive least squares (RLS) with a forgetting factor. A multi-objective optimization function is constructed for each control cycle (sampling frequency 1-10Hz, in this embodiment, the frequency is set to 5Hz). The following quadratic programming (QP) problem is constructed, and the optimal combination of drilling parameters is selected according to the constraints. The optimal drilling parameters are obtained by rolling optimization. The above QP problem is solved by calling the OSQP solver, and the optimal control sequence is extracted to calculate the target given value.

[0027] It should be noted that, by combining Step 1 with Step 2 using a dual closed-loop architecture (i.e., inner loop 100-500Hz servo tracking and outer loop 1-10Hz MPC rolling optimization), the time for the drilling speed to recover to the optimal value after a sudden formation change is ≤1.5 seconds, which is more than 70% shorter than traditional PID control (5-10 seconds). This process also employs an active disturbance method, which can identify formation response parameters every 10 seconds without interrupting drilling, ensuring that the MPC prediction model always reflects the current formation characteristics and avoiding control lag caused by model mismatch. It ensures that the lag caused by adjusting parameters based on traditional manual experience is not relied upon, but rather that rapid tracking is ensured by the high-frequency servo in the inner loop, and the optimal parameter change trajectory is calculated in advance by the prediction model in the outer loop MPC, thus realizing the basis for coordinated control of feedforward and feedback.

[0028] Next, step three involves multi-dimensional deviation calculation, including obtaining the actual drilling response for each control cycle, which includes the actual drilling pressure. Actual speed Actual torque Actual drilling speed Then, the optimal value output by the MPC optimizer is read, and the relative deviation of each component is calculated. The correlation coefficient between drilling pressure and rotation speed is calculated, and the impedance coefficient deviation is calculated. Based on the parameters calculated above, further comprehensive calculation of abnormal factors is performed. The comprehensive abnormal factors are calculated by using a weighted fusion formula, and a corresponding dynamic adjustment strategy for weight coefficients is implemented. The weight coefficients are determined according to the formation type and confidence level, and threshold classification judgment and control command trigger response instructions are performed, including fine-tuning instructions, correction instructions and emergency instructions. The parameters in step three can adapt to changes in the formation and dynamically adjust the target drilling speed and its corresponding weight coefficients according to the formation type information, ensuring adaptive optimization within the drill bit operation control cycle.

[0029] Finally, in step four, the drill bit's three-dimensional spatial trajectory parameter vector is defined, and the corresponding drill bit spatial coordinates are obtained. The current actual trajectory parameters are obtained in real time through the measurement while drilling (MWD) system. The target trajectory parameters are read, and the corresponding trajectory correction rule base is queried according to the anomaly type obtained in step three. The corresponding correction information is fed back to the MPC optimizer. After completing step four, the system also automatically feeds back to step one to execute the next control cycle, completing the complete closed-loop iteration process.

[0030] It should be noted that the autonomous operation process consisting of the active disturbance identification and formation classification in step one, the automatic optimization of MPC in step two, the anomaly self-healing in step three, and the trajectory correction in step four ensures that remote takeover can be achieved through the cloud in extremely complex situations, without the need for manual risk assessment.

[0031] Please see Figure 3As shown, in one embodiment of the present invention, formation response parameters are identified online using an active perturbation method to form an inner-loop dynamic data stream. The specific steps are as follows: S101. Perform an active identification cycle at a preset frequency, superimposing a small sinusoidal rotation speed disturbance on the normal drilling command for a duration of 0.5 seconds; S102. Synchronously observe the torque response and calculate the amplitude attenuation and phase lag. S103. The equivalent stiffness coefficient of the formation is calculated in real time using the recursive least squares method. With damping coefficient , as formation response parameters for subsequent prediction models; S104. Perform PID servo control on drilling pressure and rotation speed, track the target setpoint of the outer loop, and determine the control frequency.

[0032] In the aforementioned technical solution, conditions for triggering active disturbance are set, and the disturbance is automatically executed every 10 seconds. A small-amplitude sinusoidal disturbance is superimposed on the normal speed command, with a disturbance duration of 0.5 seconds. The torque response is then recorded synchronously. Calculate the amplitude attenuation relative to the speed disturbance. Phase lag was used, and the equivalent stiffness coefficient of the formation was calculated online using the recursive least squares (RLS) method. With damping coefficient As a formation response parameter for subsequent prediction models, the torque response characteristics are determined by calculating amplitude attenuation and phase lag. The acquired torque signal is bandpass filtered to remove high-frequency noise, and Hilbert transform is used to extract instantaneous amplitude and instantaneous phase to improve noise immunity. Based on the previous step, the formation equivalent stiffness coefficient is further calculated by configuring the RLS algorithm parameters. With damping coefficient The drill bit-formation torsion system is simplified into a spring-damped model, and the calculation method is as follows: The calculation determines that, among which This represents the drill bit angular displacement. The drill bit angular velocity is determined by integrating the rotational speed. The RLS algorithm is implemented by calculating the RLS recursive formula, setting parameters, and then determining model convergence: convergence is determined when the parameter change rate at five consecutive sampling points is ≤1%. After the disturbance ends, the estimated value at the last moment is taken as the identification result to determine the corresponding formation equivalent stiffness coefficient. With damping coefficient and obtain the formation impedance coefficient The identification result serves as a key input parameter for the ditch MPC prediction model. Finally, PID servo tracking control is performed. Specifically, the inner loop servo control is responsible for quickly tracking the target setpoint issued by the outer loop MPC optimizer. It is a high-bandwidth execution link. The control frequency selects the control cycle and is synchronized with the sensing acquisition cycle. The corresponding timestamps are aligned, and the inner loop output drive signal directly controls the multi-degree-of-freedom drilling execution module to execute subsequent frequency control.

[0033] As one embodiment of the present invention, stratigraphic categories Confidence level of stratigraphic categories ,in, The control cycle is represented; the real-time drilling state is characterized based on a discrete-time model, whose equation is expressed as:

[0034] in, This represents the output drilling speed of the drilling machinery. To input drilling pressure, Input speed, For torque, This represents the current formation impedance coefficient, and , It is white noise; , , , , These are the corresponding input model parameters, and are updated online every 5 seconds using recursive least squares with a forgetting factor.

[0035] In the aforementioned technical solution, the stratigraphic category is determined based on the current control cycle. and confidence level And based on a discrete-time model, it is characterized to complete the prediction and determination of the real-time drilling state, and the drilling mechanical drilling rate. As an output variable, drilling pressure Rotation speed Torque The formation impedance coefficient obtained in step one and white noise All input values ​​are used for online parameter updates. Specifically, the model parameters are updated every 5 seconds using recursive least squares (RLS) with a forgetting factor. , , , , The RLS algorithm is used to obtain historical data vectors containing these online parameters, model parameter vectors, and gain matrices to update model parameters, covariance, and forgetting factor, ensuring the construction of multi-objective optimization functions based on the inner loop dynamic data flow using a model predictive control optimizer.

[0036] Please see Figure 4 As shown, in one embodiment of the present invention, step two, which involves using a model predictive control optimizer to solve for the optimal combination of drilling parameters that satisfies multiple constraints based on the inner loop dynamic data flow, and dynamically updating the target setpoint of the outer loop, specifically includes the following steps: S201. An autoregressive moving average model with exogenous variables is used as a predictive model to describe the real-time drilling status, and it is updated online in real time. S202. Construct a multi-objective optimization function for the prediction model and set the constraints of the multi-objective optimization function; S203. Perform rolling optimization on the prediction model to generate the target given value.

[0037] In the above technical solution, an Auto-Regressive Moving Average (ARMAX) model with exogenous inputs is used as the discrete-time model to describe the dynamic characteristics of the drilling system. Using a first-order ARMAX model balances model complexity and prediction accuracy. The online parameter update cycle is chosen to be once every 5 seconds, and then a data vector is constructed. parameter vector The RLS recursive formula with forgetting factor is used to calculate the prediction error. Calculate the gain vector Update parameter estimation And update the covariance matrix. In each control cycle (Sampling frequency 1-10Hz), solve for the optimization variables, including drilling pressure increment. and speed increment The multi-objective optimization function and constraints are determined. Finally, the optimization problem in S202 is transformed into a standard quadratic programming (QP) form. The open-source OSQP (Operator Splitting Quadratic Program) solver, designed specifically for embedded real-time optimization, is used to implement a rolling optimization process. The predictive model is solved to determine the drill pressure increment and rotation speed increment. These are then combined with the actual drill pressure and rotation speed values ​​to generate the corresponding target setpoints. The system automatically adjusts the drill pressure and rotation speed to adapt to the fractured formation.

[0038] It should be noted that the calculation of the target given value is implemented by solving the QP problem (i.e., calculating the multi-objective optimization function of a quadratic programming problem) using the OSQP solver, and extracting its optimal control sequence. , Finally, calculate the target given value. ; The target setpoint is sent to the inner loop servo controller in step one. If the solution fails, the system automatically switches to safe mode. , And report it to the cloud. This represents the minimum drill pressure constraint related to the formation. This is the minimum rotational speed.

[0039] As one embodiment of the present invention, the multi-objective optimization function is further defined as follows:

[0040] Constraints:

[0041] in, Indicates based on stratigraphic category The target drilling rate is adaptively set according to the working mode; This refers to the actual drilling speed; To control the cycle, To increase the number of cycles, , , , These are the weight coefficients corresponding to the target parameters. This represents the minimum drill pressure constraint related to the formation. This represents the maximum drill pressure constraint related to the formation. This is the minimum rotational speed. This represents the maximum rotational speed. This represents the maximum torque.

[0042] In the aforementioned technical solution, a multi-objective optimization function is designed, in which the drilling speed tracking error is considered. Pursuing drilling efficiency, torque penalty Suppressing torque fluctuations and protecting drill bits; penalty for rate of change in drilling pressure. Prevent sudden changes in drilling pressure; penalty for rate of change in drilling speed Rotational speed change rate penalty, based on formation type Adaptive setting of target drilling rate with working mode: Further, through setting constraints: drill pressure constraint (formation-related). Speed ​​constraint and torque change rate constraint The optimal combination of drilling parameters that satisfies multiple constraints is selected. In addition to these constraints, constraints on torque fluctuations can also be set and determined by adding them to the current multi-objective optimization function as needed. This invention actively suppresses torque impact while pursuing drilling efficiency through the torque penalty term in the multi-objective optimization function. Through the constraints in the MPC optimization function, the transient rate and fluctuation coefficient of torque are dynamically limited to avoid the excitation of drill string torsional vibration (stick-slip phenomenon).

[0043] In one embodiment of the present invention, the formula for calculating the anomaly factor in step three is as follows:

[0044] in, As an abnormal factor, This refers to the actual drilling pressure. This refers to the actual rotational speed. This is the actual torque. The actual drilling mechanical drilling speed; read the optimal combination of drilling parameters output by the model predictive control optimizer of the prediction model, including the optimal drilling pressure. Optimal speed Optimal torque Optimal drilling speed ; The value must be a very small positive number to prevent division by zero errors; The correlation coefficient between drilling pressure and rotation speed. It is the absolute value of the difference between the actual formation impedance coefficient and the predicted optimal impedance coefficient; , , , , , These are the adaptive weighting coefficients for each deviation value, which are actually based on the stratigraphic category. and its confidence level Dynamic adjustment.

[0045] In the above technical solution, the actual parameters are all collected from real-time sensors, the optimal parameter is obtained based on the output of the MPC optimizer, and the comprehensive anomaly factor is calculated based on the fusion of the difference between the actual parameters and the optimal parameter. The actual parameters involved include: actual drilling pressure. Actual speed Actual torque Actual drilling speed The following involve optimal values: optimal drilling pressure. Optimal speed Optimal torque Optimal drilling speed ; Obtain the ratio of the absolute value of the difference between the corresponding values ​​to the optimal value, and configure a minimum positive number for adaptive adjustment. It can take the value 10. -6 In addition, it also includes the correlation coefficient between drilling pressure and rotational speed. The calculation is performed within a sliding window (length L = 50 sampling points, corresponding to 0.5 seconds), and the calculation formula is as follows: ,in, , The mean within the window, its The magnitude of these parameters relates to whether there is a mutual influence between changes in drilling pressure and rotation speed. If it is approximately equal to 0, it means that the drilling pressure and rotation speed change independently. If the value is greater than 0.5, then the drilling pressure and rotation speed are positively correlated. If the value is less than -0.5, then drilling pressure and rotation speed are negatively correlated; in addition, the impedance coefficient deviation is also included. In this process, based on step S103 of step one, the currently identified formation impedance coefficient is output. , This represents the optimal impedance coefficient used in the MPC prediction model, which was determined based on the most recent identification results; the impedance coefficient deviation... The size can determine the degree of matching between the formation characteristics and the prediction model. If the value is small, it indicates that the formation characteristics match the prediction model. A large value indicates a sudden change in the formation, leading to a mismatch in the prediction model. Adaptive weighting coefficients. , , , , , It is based on confidence level The weights are adjusted according to the magnitude. When the confidence level is low, the weight of the corresponding weight coefficient needs to be increased, and the normalization process ensures that... .

[0046] As one embodiment of the present invention, the judgment for triggering the control instruction of the abnormal factor threshold is as follows: Three outlier thresholds are set to form two intervals. , And satisfy ;abnormal factors Compare with these two intervals: like If the system is in a steady state, no control commands will be triggered, and the current drilling parameters will be maintained. like If so, a fine-tuning control command is triggered, and the target setpoint is slightly corrected according to a preset proportional coefficient; like If this occurs, a correction control command is triggered, pausing the rolling optimization of the model predictive control optimizer, forcibly initiating active disturbance identification, reacquiring the corresponding formation parameters, and resetting the predictive model. like If this happens, an emergency control command will be triggered, immediately executing a drilling parameter rollback operation to reduce drilling pressure and rotation speed to preset safe values, while simultaneously triggering an alarm.

[0047] In the aforementioned technical solution, the correction ratio coefficient for the fine-tuning control command is: Calculate the corrected and , specific , For correction instructions, the confidence level of step one is used as a basis. When the rate is less than 90%, the trigger condition is activated, and the active disturbance period is shortened from 10 seconds to 2 seconds (pre-set) to facilitate the rapid acquisition of formation response parameters, the reset of prediction model parameters, and the update and adjustment of the covariance matrix. For emergency commands, the drilling pressure and rotation speed are reduced proportionally (to a safe range), and an audible and visual alarm is triggered and uploaded to the cloud. If the alarm continues for two seconds without being restored, an emergency drilling rig is automatically pulled to reduce current risk losses.

[0048] Please see Figure 5 As shown, in one embodiment of the present invention, the method for dynamically correcting the drilling trajectory parameters in step four is as follows: S301. Define the drilling trajectory parameter vector based on the three-dimensional spatial coordinates of the drill bit; S302. Obtain the current actual trajectory parameters in real time through the drilling measurement system, and obtain the correction vector from the preset trajectory correction rule library; S303. Update the target trajectory parameters based on the correction vector, encode the correction information into a trajectory correction command frame, and send it to the robot actuator.

[0049] In the aforementioned technical solution, a three-dimensional spatial parameter vector for the drill bit is first defined. This defined trajectory parameter vector includes the drill bit's spatial coordinates (x, y, z), as well as the inclination and azimuth angles. The actual trajectory parameters and the target trajectory parameters of the virus area are acquired in real time through the measurement while drilling (MWD) system. These target trajectory parameters are imported from the exploration and design documents. Finally, the trajectory correction rule base is queried based on the anomaly type in step three. This rule base defines a rich set of anomaly types, correction strategies, and correction vector information. The correction information is then fed back to the MPC optimizer to determine the node information for feedback correction. After the feedback correction operation is completed, the system automatically returns to step one and continues to execute the operation of the next control cycle, thus forming a complete inner and outer loop iterative process.

[0050] Please see Figure 2 As shown, another embodiment of the present invention provides a full-condition adaptive intelligent drilling robot system for implementing a closed-loop control method for a full-condition adaptive intelligent drilling robot system. The system includes: The multi-source sensing module is used to collect drilling parameters of the intelligent drilling robot system in real time through a preset sampling frequency; the drilling parameters include drilling pressure, torque, and rotational speed. The edge computing control module is used to implement dual-closed-loop adaptive control. The edge computing control module includes an inner-loop dynamic data flow construction unit, an outer-loop target setpoint construction unit, and a deviation analysis unit. The inner loop dynamic data stream construction unit is used to construct the drilling state feature vector, identify formation response parameters online through the active perturbation method, and form the inner loop dynamic data stream; the inner loop dynamic data stream includes formation type, formation response parameters and real-time drilling status; The outer ring target setpoint construction unit is used to solve the optimal drilling parameter combination that satisfies multiple constraints based on the inner ring dynamic data flow using a model predictive control optimizer, and dynamically update the outer ring target setpoint. The deviation analysis unit is used to perform real-time deviation analysis between the real-time drilling status and the optimal drilling parameters, and to calculate the anomaly factor. The multi-degree-of-freedom drilling execution module is used to trigger control commands based on the threshold of abnormal factors; it receives correction information and feeds the correction information back to the model predictive control optimizer to complete the loop iteration. The mobile walking module is used to dynamically correct drilling trajectory parameters based on control commands.

[0051] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, and therefore described more simply; relevant parts can be referred to the descriptions of the method embodiments.

[0052] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this application, they should all fall within the protection scope of the present invention.

Claims

1. A closed-loop control method for a full-condition adaptive intelligent drilling robot system, characterized in that, The method includes: Step 1: Collect drilling parameters of the intelligent drilling robot system in real time at a preset sampling frequency, construct a drilling status feature vector, identify formation response parameters online using the active perturbation method, and form an inner loop dynamic data stream; the drilling parameters include drilling pressure, torque, and rotational speed; the inner loop dynamic data stream includes formation type, formation response parameters, and real-time drilling status; Step 2: Based on the inner loop dynamic data flow, the model predictive control optimizer is used to solve for the optimal combination of drilling parameters that satisfies multiple constraints, and the target setpoint of the outer loop is dynamically updated. Step 3: Perform real-time deviation analysis between the real-time drilling status and the optimal drilling parameters, calculate the anomaly factor, and trigger control commands based on the anomaly factor threshold. Step 4: Dynamically correct the drilling trajectory parameters based on control commands, and feed the correction information back to the model predictive control optimizer to complete the iterative closed loop.

2. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 1, characterized in that, The specific steps for identifying formation response parameters online using the active perturbation method to form an inner-loop dynamic data stream are as follows: S101. Perform an active identification cycle at a preset frequency, superimposing a small sinusoidal rotation speed disturbance on the normal drilling command for a duration of 0.5 seconds; S102. Synchronously observe the torque response and calculate the amplitude attenuation and phase lag. S103. The equivalent stiffness coefficient of the formation is calculated in real time using the recursive least squares method. With damping coefficient , as formation response parameters for subsequent prediction models; S104. Perform PID servo control on drilling pressure and rotation speed, track the target setpoint of the outer loop, and determine the control frequency.

3. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 2, characterized in that, The stratigraphic category is represented as follows: The confidence level of stratigraphic categories is expressed as: ,in, The control cycle is indicated; the real-time drilling state is characterized based on a discrete-time model, and its equation is expressed as: in, This represents the output drilling speed of the drilling machinery. To input drilling pressure, Input speed, For torque, This represents the current formation impedance coefficient, and , It is white noise; , , , , These are the corresponding input model parameters, and are updated online every 5 seconds using recursive least squares with a forgetting factor.

4. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 3, characterized in that, Step two, which involves using a model predictive control optimizer to solve for the optimal combination of drilling parameters that satisfies multiple constraints based on the inner loop dynamic data flow, and dynamically updating the target setpoint of the outer loop, specifically includes the following steps: S201. An autoregressive moving average model with exogenous variables is used as a predictive model to describe the real-time drilling status, and it is updated online in real time. S202. Construct a multi-objective optimization function for the prediction model and set the constraints of the multi-objective optimization function; S203. Perform rolling optimization on the prediction model to generate the target given value.

5. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 4, characterized in that, The multi-objective optimization function is: Constraints: in, Indicates based on stratigraphic category The target drilling rate is adaptively set according to the working mode; This refers to the actual drilling speed; To control the cycle, To increase the number of cycles, , , , These are the weight coefficients corresponding to the target parameters. This represents the minimum drill pressure constraint related to the formation. This represents the maximum drill pressure constraint related to the formation. This is the minimum rotational speed. This represents the maximum rotational speed. This represents the maximum torque.

6. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 3, characterized in that, The formula for calculating the anomaly factor in step three is as follows: in, As an abnormal factor, This refers to the actual drilling pressure. This refers to the actual rotational speed. This is the actual torque. The actual drilling mechanical drilling speed; read the optimal combination of drilling parameters output by the model predictive control optimizer of the prediction model, including the optimal drilling pressure. Optimal speed Optimal torque Optimal drilling speed ; The value must be a very small positive number to prevent division by zero errors; The correlation coefficient between drilling pressure and rotation speed. It is the absolute value of the difference between the actual formation impedance coefficient and the predicted optimal impedance coefficient; , , , , , These are the adaptive weighting coefficients for each deviation value, based on the stratigraphic category. and its confidence level Dynamic adjustment.

7. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 6, characterized in that, The judgment triggered by the control command for the abnormal factor threshold is as follows: Three outlier thresholds are set to form two intervals. , And satisfy ;abnormal factors Compare with these two intervals: like If the system is in a steady state, no control commands will be triggered, and the current drilling parameters will be maintained. like If so, a fine-tuning control command is triggered, and the target setpoint is slightly corrected according to a preset proportional coefficient; like If this occurs, a correction control command is triggered, pausing the rolling optimization of the model predictive control optimizer, forcibly initiating active disturbance identification, reacquiring the corresponding formation parameters, and resetting the predictive model. like If this happens, an emergency control command will be triggered, immediately executing a drilling parameter rollback operation to reduce drilling pressure and rotation speed to preset safe values, while simultaneously triggering an alarm.

8. The closed-loop control method for the all-condition adaptive intelligent drilling robot system according to claim 1, characterized in that, The method for dynamically correcting the drilling trajectory parameters in step four is as follows: S301. Define the drilling trajectory parameter vector based on the three-dimensional spatial coordinates of the drill bit; S302. Obtain the current actual trajectory parameters in real time through the drilling measurement system, and obtain the correction vector from the preset trajectory correction rule library; S303. Update the target trajectory parameters based on the correction vector, encode the correction information into a trajectory correction command frame, and send it to the robot actuator.

9. A full-condition adaptive intelligent drilling robot system, characterized in that, A closed-loop control method for implementing the all-condition adaptive intelligent drilling robot system as described in any one of claims 1-8, the system comprising: A multi-source sensing module is used to collect drilling parameters of the intelligent drilling robot system in real time at a preset sampling frequency; the drilling parameters include drilling pressure, torque, and rotational speed. The edge computing control module is used to implement dual closed-loop adaptive control. The edge computing control module includes an inner-loop dynamic data flow construction unit, an outer-loop target setpoint construction unit, and a deviation analysis unit. The inner loop dynamic data stream construction unit is used to construct a drilling state feature vector, identify formation response parameters online through an active perturbation method, and form an inner loop dynamic data stream; the inner loop dynamic data stream includes formation type, formation response parameters, and real-time drilling status. The outer ring target setpoint construction unit is used to solve the optimal drilling parameter combination that satisfies multiple constraints based on the inner ring dynamic data flow using a model predictive control optimizer, and to dynamically update the outer ring target setpoint. The deviation analysis unit is used to perform real-time deviation analysis between the real-time drilling status and the optimal drilling parameters, and to calculate the anomaly factor. The multi-degree-of-freedom drilling execution module is used to trigger control commands based on the threshold of abnormal factors; it receives correction information and feeds the correction information back to the model predictive control optimizer to complete the loop iteration. The mobile walking module is used to dynamically correct drilling trajectory parameters based on control commands.

Images