Equipment fault early warning method based on well drilling real-time dynamic multi-dimensional feature modeling

Abstract

The invention relates to the technical field of petroleum engineering equipment monitoring and fault early warning, in particular to an equipment fault early warning method based on drilling real-time dynamic multi-dimensional feature modeling, which comprises the following steps of: constructing a drill string torsion and axial motion model by acquiring rotation, axial and other parameters in a drilling process in real time, and predicting torque and hook load; the actual value is compared with the predicted value, torque and hook load residual signals are generated, a residual signal vector is constructed, and statistical characteristics of the residual signal vector are analyzed; determining a standard and real-time residual signal distribution by using Gaussian distribution, calculating a probability density value, and comparing the probability density value with a fault early warning threshold to obtain target real-time residual signal distribution; and further calculating the distance of the drilling machine, and comparing with a preset alarm threshold to judge a fault alarm. The fault early warning accuracy is effectively improved, the equipment reliability is enhanced, the maintenance decision is optimized, and the drilling efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of oil engineering equipment monitoring and fault warning, and particularly to a device fault warning method based on real-time dynamic multi-dimensional feature modeling of drilling. Background Art

[0002] The failure of drilling equipment is usually caused by the abnormality of the drill string system. When faults such as stuck drill and drill string bounce occur in the drill string, it is easy to cause damage to the drilling equipment and threaten the safety of the drilling rig and the drilling platform. Usually, when a fault occurs, it will cause changes in measurement signals such as hook load, drilling speed, torque, and rotational speed. However, the interference of measurement signals and the uncertainty in the hole make it difficult to detect these changes, and even cause false alarms under normal conditions. In addition, the physical structure of the drilling process determines that there is a strong dependence relationship among the above four signals. Therefore, considering the real-time measurement signals, the mechanism model of the drill string system, and the changes in the hole environment, a fault warning method for drilling equipment is designed.

[0003] It can be seen that there are the following problems: In the prior art, due to the lack of multi-dimensional feature modeling of real-time data, the problems of low drilling efficiency and slow drilling speed occur. Summary of the Invention

[0004] Therefore, the present invention provides a device fault warning method based on real-time dynamic multi-dimensional feature modeling of drilling to overcome the problems of low drilling efficiency and slow drilling speed in the prior art due to the lack of multi-dimensional feature modeling of real-time data.

[0005] To achieve the above object, the present invention provides a device fault warning method based on real-time dynamic multi-dimensional feature modeling of drilling, including:

[0006] Step S1, collect rotation parameters, axial parameters, and other parameters of several collection points in the drilling process according to the sampling interval. The rotation parameters include rotary table speed and torque, the axial parameters include hook load and drilling speed, and the other parameters include wire rope displacement and drill string mass;

[0007] Step S2, obtain the motor moment of inertia, the rotary table moment of inertia, the effective damping, and the motor gear ratio. Determine the effective inertia according to the motor gear ratio, the motor moment of inertia, and the rotary table moment of inertia. Construct a drill string torsional motion model according to the effective damping, the effective inertia, and the rotary table speed to calculate the torque of the drill string to obtain a torque prediction value;

[0008] Step S3, construct a drill string axial motion model according to the drill string mass, the wire rope displacement, and the drilling speed to calculate the hook load of the drill string to obtain a hook load prediction value;

[0009] Step S4, compare the torque with the predicted torque value to obtain a torque residual signal, and compare the hook load with the predicted hook load value to obtain a hook load residual signal;

[0010] Step S5, construct a residual signal vector based on the hook load residual signal and the torque residual signal, and analyze the residual signal vector to obtain a mean vector and a covariance matrix;

[0011] Step S6, use the Gaussian distribution method according to the mean vector and the covariance matrix to determine the standard residual signal distribution and the real-time residual signal distribution, calculate the compliance degree of the residual signal vector in the standard residual signal distribution to obtain a probability density value, and compare the probability density value with a preset fault warning threshold to obtain the target real-time residual signal distribution;

[0012] Step S7, determine the distance of the drilling rig according to the standard residual signal distribution and the target real-time residual signal distribution;

[0013] Step S8, compare the distance of the drilling rig with a preset alarm threshold to determine whether a fault alarm is generated.

[0014] Further, the process of constructing a drill string torsional motion model based on the effective damping, the angular displacement, the effective inertia, and the rotary table speed to calculate the torque of the drill string to obtain a torque prediction value includes:

[0015]

[0016] where Tr represents the torque prediction value; c_v represents the effective damping of the drill string, θ represents the angular displacement, t_s represents the time interval, k represents the current sampling point, J represents the effective inertia, J_rt and J_m respectively represent the moments of inertia of the rotary table and the motor shaft; n0 represents the gear ratio of the rotary table motor.

[0017] Further, the process of step S3 includes:

[0018]

[0019] where F_h is the predicted load value, m_ds is the mass of the drill string, g represents the acceleration due to gravity, x_ds is the axial displacement of the drill string, k represents the current sampling point, v_ds represents the lowering speed of the drill string, i.e., the drilling speed.

[0020] Further, the process of step S4 includes:

[0021] Obtain a torque residual signal based on the difference between the torque and the predicted torque value;

[0022] Obtain a hook load residual signal based on the difference between the hook load and the predicted hook load value.

[0023] Further, the process of step s5 includes:

[0024] Combine the hook load residual signal and the torque residual signal at the same sampling point into a two-dimensional vector to obtain a residual signal vector;

[0025] Define the size of a time window to obtain the number of sampling points;

[0026] Collect the residual signal vectors at each time interval within the time window;

[0027] Average the residual signal vectors within the time window according to the number of sampling points to obtain the mean vector;

[0028] Perform centering processing on the mean vector to obtain a processing result;

[0029] Calculate the mean vector of the product of the processing result and the transpose of the processing result to obtain the covariance matrix.

[0030] Further, the process of step S6 includes:

[0031] Construct a multivariate Gaussian distribution model based on the mean vector and the covariance matrix;

[0032] Determine the real-time residual signal distribution according to the multivariate Gaussian distribution model;

[0033] Obtain historical data according to the real-time residual signal distribution to obtain the standard residual signal distribution;

[0034] Calculate the degree of conformity of the residual signal vector in the standard residual signal distribution according to the residual signal vector using the multivariate Gaussian distribution formula to obtain a probability density value;

[0035] Compare the probability density value with the fault warning threshold to obtain the target real-time residual signal distribution.

[0036] Further, the process of comparing the probability density value with the fault warning threshold to obtain the target real-time residual signal distribution includes:

[0037] If the probability density value is less than the fault warning threshold, adjust the mean vector and the covariance matrix to obtain the target real-time residual signal;

[0038] If the probability density value is greater than or equal to the fault warning threshold, the real-time residual signal is equal to the target real-time residual signal.

[0039] Further, the process of step S7 includes:

[0040] Calculating the Euclidean distance between the mean vector of the standard residual signal distribution and the mean vector of the target residual signal distribution;

[0041] Calculating the trace of the covariance matrix according to the covariance matrix of the standard residual signal distribution and the covariance matrix of the target residual signal distribution;

[0042] Determining the drill rig distance according to the sum of the trace of the covariance matrix and the Euclidean distance.

[0043] Further, the process of step S8 includes:

[0044] If the drill rig distance is less than or equal to the alarm threshold, it indicates normal working conditions and no fault alarm is generated;

[0045] If the drill rig distance is greater than the alarm threshold, a fault alarm is generated.

[0046] Further, the setting process of the alarm threshold includes:

[0047]

[0048] Wherein, P(w <= W1) represents the probability that the random variable w is less than or equal to W1, F represents the false alarm rate, and W1 represents the alarm threshold.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows. The present invention ensures the timeliness and accuracy of data by collecting rotational parameters, axial parameters and other parameters. Through multi-dimensional feature modeling, the operating state of the drilling equipment is more comprehensively reflected, and the accuracy and reliability of fault early warning are improved. A residual signal vector is constructed, and the mean vector and covariance matrix are analyzed to provide effective data support for fault early warning. The Gaussian distribution method is applied to determine the standard residual signal distribution and the real-time residual signal distribution. By calculating the degree of conformity of the residual signal vector in the standard residual signal distribution, the probability density value is obtained, providing a quantitative basis for fault early warning. The drill rig distance is introduced as a new judgment index to enhance the accuracy of fault early warning. The standard residual signal distribution is associated with the real-time residual signal distribution to improve the reliability of early warning. By comparing the drill rig distance with the preset alarm threshold, the false alarm and missed alarm situations are effectively reduced, and the reliability and practicability of fault alarm are improved. The application of this method can timely detect and early warn equipment faults, avoiding drilling accidents caused by equipment faults. The safety and efficiency of drilling operations are improved, and the maintenance cost and downtime are reduced.

[0050] In particular, by combining multi-dimensional features such as effective damping, angular displacement, effective inertia, and rotary table speed, a more accurate torsional motion model of the drill string is constructed. Using real-time monitoring data and physical parameters, a more accurate torque prediction value is calculated, reducing the errors caused by simplified models or ignoring certain factors in traditional methods. By adopting real-time data acquisition and calculation, the change of drill string torque can be quickly reflected, providing strong support for real-time monitoring and early warning.

[0051] In particular, the motion state of the drill string can be monitored in real time by calculating the lowering speed of the drill string at the current sampling point. The update of the real-time load prediction value helps to adjust the drilling parameters in a timely manner, ensuring the stability and safety of the drilling process. By adjusting the load in real time, the friction between the drill string and the wellbore can be reduced, the energy consumption can be lowered, and the drilling efficiency can be improved.

[0052] In particular, the torque residual signal and the hook load residual signal directly reflect the deviation between the actual operation and the prediction model, which helps to more accurately detect potential equipment failures or abnormal working conditions. By analyzing the residual signal, the drilling parameters can be optimized, the drilling efficiency can be improved, and the efficiency loss caused by equipment problems can be avoided.

[0053] In particular, by combining the hook load and torque residual signals into a two-dimensional vector, the mutual influence of the two signals is fully considered, improving the comprehensiveness and accuracy of data processing. Defining a time window and collecting the residual signal vectors at corresponding time intervals enables the analysis to be based on a time series, better capturing the dynamic changes of the signal.

[0054] In particular, by constructing a multivariate Gaussian distribution model, the statistical characteristics of the residual signal can be more accurately described, thereby improving the accuracy of fault detection. The comparison between the real-time residual signal distribution and the standard residual signal distribution further ensures the reliability of fault judgment. By obtaining historical data to get the standard residual signal distribution, the historical normal working condition data is fully utilized, enhancing the representativeness and accuracy of the model.

[0055] In particular, when the probability density value is less than the fault warning threshold, by adjusting the mean vector and covariance matrix, the system can adaptively reflect the change of the equipment state in real time, enhancing the sensitivity and response speed to equipment anomalies. When the probability density value is greater than or equal to the fault warning threshold, the real-time residual signal is confirmed as the target real-time residual signal, avoiding unnecessary parameter adjustment and maintaining the stability and consistency of the system operation. By dynamically adjusting the mean vector and covariance matrix, the fault detection model is made closer to the real-time data distribution, thereby improving the accuracy and reliability of fault detection.

[0056] In particular, the drill distance comprehensively considers the changes in the mean vector and covariance matrix, and can more sensitively detect minor changes in the equipment state, thereby improving the sensitivity of fault detection. By accurately calculating the drill distance, it is possible to more accurately determine whether the equipment deviates from the normal state, thus improving the accuracy of fault warning.

[0057] In particular, by comparing the drill distance with the alarm threshold in real time, potential faults can be discovered and processed in a timely manner, effectively reducing the unexpected downtime of the equipment and improving the reliability and stability of the equipment. This helps to extend the service life of the equipment and reduce maintenance costs.

[0058] In particular, specifically, by calculating the integral based on the probability density function, the alarm threshold can be accurately determined, significantly improving the accuracy of fault warning. The false alarm rate is reduced, and the ineffective maintenance and resource waste caused by false alarms are decreased. Setting a reasonable alarm threshold helps to detect potential equipment faults in a timely manner, prevent small problems from evolving into major accidents, improve the operating efficiency and safety of the equipment, and extend the service life of the equipment. Brief Description of the Drawings

[0059] Figure 1 It is a schematic flow chart of the equipment fault warning method based on drilling real-time dynamic multi-dimensional feature modeling provided by an embodiment of the present invention;

[0060] Figure 2 It is a schematic flow chart of step S5 in the equipment fault warning method based on drilling real-time dynamic multi-dimensional feature modeling provided by an embodiment of the present invention;

[0061] Figure 3 It is a schematic flow chart of step S6 in the equipment fault warning method based on drilling real-time dynamic multi-dimensional feature modeling provided by an embodiment of the present invention;

[0062] Figure 4 It is a schematic flow chart of step S7 in the equipment fault warning method based on drilling real-time dynamic multi-dimensional feature modeling provided by an embodiment of the present invention. Detailed Embodiments

[0063] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0065] It should be noted that in the description of the present invention, the terms indicating the direction or positional relationship such as "upper", "lower", "left", "right", "inner", "outer", etc. are based on the direction or positional relationship shown in the drawings. This is only for convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to the present invention.

[0066] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and defined, the terms "installation", "connection", "coupling" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0067] Please refer to Figure 1 As shown, a device fault warning method based on real-time dynamic multi-dimensional feature modeling of drilling includes:

[0068] Step S1, collect the rotation parameters, axial parameters and other parameters of several sampling points in the drilling process in real time according to the sampling interval. The rotation parameters include the rotary table speed and torque, the axial parameters include the hook load and drilling speed, and the other parameters include the wire rope displacement and drill string mass;

[0069] Step S2, obtain the motor moment of inertia, the rotary table moment of inertia, the effective damping and the motor gear ratio. Determine the effective inertia according to the motor gear ratio, the motor moment of inertia and the rotary table moment of inertia. Construct a drill string torsional motion model according to the effective damping, the effective inertia and the rotary table speed to calculate the torque of the drill string to obtain a torque prediction value;

[0070] Step S3, construct a drill string axial motion model according to the drill string mass, the wire rope displacement and the drilling speed to calculate the hook load of the drill string to obtain a hook load prediction value;

[0071] Step S4, compare the torque with the predicted torque value to obtain a torque residual signal, and compare the hook load with the hook load prediction value to obtain a hook load residual signal;

[0072] Step S5, construct a residual signal vector according to the hook load residual signal and the torque residual signal, and analyze the residual signal vector to obtain a mean vector and a covariance matrix;

[0073] Step S6: Using the Gaussian distribution method based on the mean vector and the covariance matrix, determine the standard residual signal distribution and the real-time residual signal distribution, calculate the degree of conformity of the residual signal vector in the standard residual signal distribution to obtain a probability density value, and compare the probability density value with a preset fault warning threshold to obtain the target real-time residual signal distribution;

[0074] Step S7: Determine the distance of the drilling rig based on the standard residual signal distribution and the target real-time residual signal distribution;

[0075] Step S8: Compare the distance of the drilling rig with a preset alarm threshold to determine whether a fault alarm is generated.

[0076] Specifically, according to the system requirements and the dynamic characteristics of the drilling process, set an appropriate sampling interval (such as 1 second, 5 seconds, etc.). Measure and record the rotary speed (unit: rpm) and torque (unit: Nm) of the rotary table in real time. Measure and record the hook load (unit: kN) and drilling speed (unit: m / h) in real time. Measure and record the wireline displacement (unit: m) and drill string mass (unit: kg) in real time. Obtain the motor moment of inertia, rotary table moment of inertia, effective damping, and motor gear ratio from the system database or real-time input. Ensure the accuracy and real-time update of the parameters.

[0077] Specifically, the timeliness and accuracy of the data are ensured by collecting rotation parameters, axial parameters, and other parameters. Through multi-dimensional feature modeling, the operating state of the drilling equipment is more comprehensively reflected, improving the accuracy and reliability of fault warning. Construct a residual signal vector, and analyze to obtain the mean vector and covariance matrix, providing effective data support for fault warning. Apply the Gaussian distribution method to determine the standard residual signal distribution and the real-time residual signal distribution. By calculating the degree of conformity of the residual signal vector in the standard residual signal distribution, a probability density value is obtained, providing a quantitative basis for fault warning. Introduce the distance of the drilling rig as a new judgment index to enhance the accuracy of fault warning. Associate the standard residual signal distribution with the real-time residual signal distribution to improve the reliability of the warning. By comparing the distance of the drilling rig with a preset alarm threshold, false alarms and missed alarms are effectively reduced, improving the reliability and practicality of fault alarms. The application of this method can timely detect and warn of equipment failures, avoiding drilling accidents caused by equipment failures. It improves the safety and efficiency of drilling operations, reduces maintenance costs and downtime.

[0078] Specifically, the process of constructing a drill string torsional motion model based on the effective damping, the angular displacement, the effective inertia, and the rotary speed of the rotary table to calculate the torque of the drill string to obtain a torque prediction value includes:

[0079]

[0080] Among them, Tr represents the predicted torque value; c_v represents the effective damping of the drill string, θ represents the angular displacement, t_s represents the time interval, k represents the current sampling point, J represents the effective inertia, J_rt and J_m represent the moments of inertia of the rotary table and the motor shaft respectively; n0 represents the gear ratio of the rotary table motor.

[0081] Specifically, extract the effective damping value of the drill string from the real-time monitoring system or the pre-set database. Ensure that the damping value matches the current drilling conditions. In case of special situations (such as changes in mud properties), re-calibration is required. Use a high-precision rotational speed sensor to monitor the rotational speed of the rotary table in real time. The sensor should be installed near the rotary table to ensure accurate measurement data. The rotational speed data is in rpm and is transmitted in real time through the data acquisition system. Determine the time interval according to the sampling frequency set by the system. For example, if the sampling frequency is 1 Hz, the time interval t_s is 1 second. The time interval should be kept constant to ensure the continuity and accuracy of the calculation. Extract the moments of inertia of the rotary table and the motor shaft from the equipment's technical manual or parameter database. These values are usually fixed, but need to be reconfirmed after equipment replacement or repair. Also extract the gear ratio of the rotary table motor from the equipment parameter database. The gear ratio is a design parameter and usually does not change, but needs to be checked regularly to ensure accuracy.

[0082] Specifically, considering that θ(k) cannot be directly measured, the angular velocity can be calculated as:

[0083]

[0084] Among them, r(k) represents the rotational speed of the rotary table.

[0085] Specifically, by combining multi-dimensional features such as effective damping, angular displacement, effective inertia, and rotational speed of the rotary table, a more accurate drill string torsional motion model is constructed. Using real-time monitoring data and physical parameters, a more accurate predicted torque value is calculated, reducing the errors caused by simplified models or ignoring certain factors in traditional methods. With real-time data acquisition and calculation, the change of the drill string torque can be quickly reflected, providing strong support for real-time monitoring and early warning.

[0086] Specifically, the process of step S3 includes:

[0087]

[0088] Among them, F_h is the predicted load value, m_ds is the mass of the drill string, g represents the acceleration due to gravity, x_ds is the axial displacement of the drill string, k represents the current sampling point, v_ds represents the lowering speed of the drill string, i.e., the drilling speed.

[0089] Specifically, considering that the axial displacement x_ds of the drill string is difficult to accurately measure, the displacement x_dl of the wire rope is used for conversion. According to the structure of the movable pulley, the relationship between x_dl and x_ds is:

[0090] v ds (k) = f·v dl (k)

[0091] where f represents the rope coefficient of the pulley.

[0092] Specifically, the motion state of the drill string can be monitored in real time by calculating the lowering speed of the drill string at the current sampling point. The update of the real-time load prediction value helps to adjust the drilling parameters in a timely manner, ensuring the stability and safety of the drilling process. By adjusting the load in real time, the friction between the drill string and the wellbore can be reduced, the energy consumption can be lowered, and the drilling efficiency can be improved.

[0093] Specifically, the process of step S4 includes:

[0094] Obtaining a torque residual signal based on the difference between the torque and the predicted torque value;

[0095] Obtaining a hook load residual signal based on the difference between the hook load and the predicted hook load value.

[0096] Specifically, the actual torque value and the actual hook load value at the current sampling point are obtained from the real-time monitoring system or sensors. The calculation formula for the torque residual signal is: Torque residual signal = Actual torque value - Predicted torque value. This formula represents the difference between the actual torque value and the predicted torque value. The calculation formula for the hook load residual signal is: Hook load residual signal = Actual hook load value - Predicted hook load value. The calculated torque residual signal and hook load residual signal are filtered or smoothed to reduce noise and fluctuations. Optionally, the torque residual signal is normalized or standardized for subsequent analysis and comparison.

[0097] Specifically, the torque residual signal and the hook load residual signal directly reflect the deviation between the actual operation and the prediction model, which helps to more accurately detect potential equipment failures or abnormal operating conditions. By analyzing the residual signals, the drilling parameters can be optimized, the drilling efficiency can be improved, and the efficiency loss caused by equipment problems can be avoided.

[0098] Specifically, as Figure 2 shown, the process of step S5 includes:

[0099] Step 51, combining the hook load residual signal and the torque residual signal at the same sampling point into a two-dimensional vector to obtain a residual signal vector;

[0100] Step 52, defining the size of a time window to obtain the number of sampling points;

[0101] Step 53, collect the residual signal vectors at each of the time intervals within the time window;

[0102] Step 54, average the residual signal vectors within the time window according to the number of sampling points to obtain the mean vector;

[0103] Step 55, perform a centering process on the mean vector to obtain a processing result;

[0104] Step 56, calculate the mean vector of the product of the processing result and the transpose of the processing result to obtain the covariance matrix.

[0105] Specifically, according to system requirements or preset parameters, define the size of a time window, that is, the number of sampling points (N). For example, the time window can be set to the data within the past 10 seconds. If the sampling frequency is 1 Hz, then N = 10. The calculation formula for the mean vector is:

[0106]

[0107] where μ2 represents the mean vector, k represents the current sampling point, L_1 represents the size of the time window, and E_t represents the residual signal vector at time t. After accumulating all the residual signal vectors within the time window and then dividing by the number of sampling points within the time window, the mean vector is obtained.

[0108] Specifically, the calculation formula for the covariance is:

[0109] Σ2 = E[(E - E[E] T )(E - E[E])]

[0110] where Σ2 represents the covariance matrix, E represents the residual signal vector, E[E] represents the mean of the residual signal vector (i.e., μ2), and E[] represents the expectation function. The meaning of the formula is to calculate the product of the difference between each residual signal vector and the mean vector, and then find the expectation value to obtain the covariance matrix.

[0111] Specifically, by combining the hook load and torque residual signals into a two-dimensional vector, the mutual influence of the two signals is fully considered, improving the comprehensiveness and accuracy of data processing. Defining a time window and collecting the residual signal vectors at the corresponding time intervals enables the analysis to be based on a time series, better capturing the dynamic changes of the signals.

[0112] Specifically, as Figure 3 shown, the process of step S6 includes:

[0113] Step S61, construct a multivariate Gaussian distribution model according to the mean vector and the covariance matrix;

[0114] Step S62: Determine the real-time residual signal distribution according to the multivariate Gaussian distribution model;

[0115] Step S63: Obtain historical data according to the real-time residual signal distribution to obtain the standard residual signal distribution;

[0116] Step S64: Calculate the compliance degree of the residual signal vector in the standard residual signal distribution using the multivariate Gaussian distribution formula according to the residual signal vector to obtain the probability density value;

[0117] Step S65: Compare the probability density value with the fault warning threshold to obtain the target real-time residual signal distribution.

[0118] Specifically, extract the residual signal data under normal working conditions from the historical database. Using the historical normal working condition data, calculate the mean vector (μ1) and the covariance matrix (Σ1). The standard residual signal distribution is expressed as N1(μ1,Σ1). Determine the target real-time residual signal distribution (N2(μ2,Σ2) according to the mean vector (μ2) and the covariance matrix (Σ2) obtained in step S5. The calculation formula for the probability density value is:

[0119]

[0120] where p(e) represents the probability density value of the real-time residual signal vector e under the standard residual signal distribution N1(μ1,Σ1). n is the dimension of the residual signal vector, |Σ1| represents the determinant of the covariance matrix Σ1, (Σ1)^(-1) represents the inverse matrix of the covariance matrix Σ1, (e - μ1) represents the centered residual signal vector, [(e - μ1)^T(Σ1)^(-1)(e - μ1)] represents the exponential function of the square of the Mahalanobis distance between the residual signal vector e and the mean vector μ1, reflecting the degree of deviation of e in the multivariate Gaussian distribution. (2π)^(n / 2) represents the normalization factor, which is related to the dimension n of the residual signal vector. |Σ1|^(1 / 2) represents the square root of the determinant of the covariance matrix Σ1 and is also part of the normalization factor, representing the volume of the covariance matrix. e^(-1 / 2[(e - μ1)^T(Σ1)^(-1)(e - μ1)]) represents the exponential function of the square of the Mahalanobis distance between the residual signal vector e and the mean vector μ1, reflecting the degree of deviation of e in the multivariate Gaussian distribution.

[0121] Specifically, by constructing a multivariate Gaussian distribution model, the statistical characteristics of the residual signal can be more accurately described, thereby improving the accuracy of fault detection. The comparison between the real-time residual signal distribution and the standard residual signal distribution further ensures the reliability of fault judgment. By obtaining historical data to get the standard residual signal distribution, the historical normal condition data is fully utilized, enhancing the representativeness and accuracy of the model.

[0122] Specifically, the process of obtaining the target real-time residual signal distribution by comparing the probability density value with the fault warning threshold includes:

[0123] If the probability density value is less than the fault warning threshold, adjust the mean vector and the covariance matrix to obtain the target real-time residual signal;

[0124] If the probability density value is greater than or equal to the fault warning threshold, the real-time residual signal is equal to the target real-time residual signal.

[0125] Specifically, a preset fault warning threshold is determined based on historical data, expert experience, and system requirements. Compare the calculated probability density value p(e) with the preset fault warning threshold W. When the probability density value is less than the fault warning threshold, calculate a new mean vector μ_new. The following method can be used: perform a weighted average on the real-time residual signals in the recent period, giving higher weights to the newer data. Update the mean vector using a recurrence formula, such as μ_new = α * μ_old + (1 - α) * e, where α is the forgetting factor and e is the current residual signal vector.

[0126] Specifically, to calculate a new covariance matrix Σ_new, the following method can be used: update the covariance matrix according to the deviation between the real-time residual signals and the mean vector in the recent period. Update the covariance matrix using a recurrence formula, such as Σ_new = α * Σ_old + (1 - α) * (e - μ_new) * (e - μ_new)W, where α is the forgetting factor. Use the adjusted mean vector μ_new and covariance matrix Σ_new to construct a new multivariate Gaussian distribution model as the target real-time residual signal distribution. When the probability density value is greater than or equal to the fault warning threshold, directly use the current real-time residual signal e as the target real-time residual signal. Keep the mean vector and the covariance matrix unchanged: consider the current device state normal and there is no need to adjust the mean vector and the covariance matrix.

[0127] Specifically, when the probability density value is less than the fault warning threshold, by adjusting the mean vector and covariance matrix, the system can adaptively reflect the changes in the device state in real time, enhancing the sensitivity and response speed to device anomalies. When the probability density value is greater than or equal to the fault warning threshold, the real-time residual signal is confirmed as the target real-time residual signal, avoiding unnecessary parameter adjustments and maintaining the stability and consistency of the system operation. By dynamically adjusting the mean vector and covariance matrix, the fault detection model is made closer to the real-time data distribution, thereby improving the accuracy and reliability of fault detection.

[0128] Specifically, as Figure 4 shown, the process of step S7 includes:

[0129] Step S71, calculate the Euclidean distance between the mean vector of the standard residual signal distribution and the mean vector of the target residual signal distribution;

[0130] Step S72, calculate the trace of the covariance matrix according to the covariance matrix of the standard residual signal distribution and the covariance matrix of the target residual signal distribution;

[0131] Step S73, determine the drill rig distance according to the sum of the trace of the covariance matrix and the Euclidean distance.

[0132] Specifically, the drill rig distance is the 2nd-order Wasserstein distance or Wasserstein-2 distance, which is an index to measure the difference between two probability distributions. The drill rig distance (p = 2) between the standard residual signal distribution N1(μ1,1) and the target real-time residual signal distribution N2(μ2,2) is:

[0133]

[0134] where is the Euclidean distance between μ1 and μ2, and tr() represents the trace of the covariance matrix.

[0135] Specifically, use a numerical calculation library (such as NumPy in Python) for matrix operations, including the calculation of the Euclidean distance, the calculation of the trace of the covariance matrix, and the calculation of the square root and product of matrices. Ensure that numerical stability is considered during the calculation to avoid calculation errors caused by numerical problems.

[0136] Specifically, the drill rig distance comprehensively considers the changes in the mean vector and covariance matrix, can more sensitively detect small changes in the device state, thereby improving the sensitivity of fault detection. By accurately calculating the drill rig distance, it is possible to more accurately determine whether the device deviates from the normal state, thereby improving the accuracy of fault warning.

[0137] Specifically, the process of step S8 includes:

[0138] If the distance of the drilling machine is less than or equal to the alarm threshold, it indicates normal working conditions and no fault alarm is generated;

[0139] If the distance of the drilling machine is greater than the alarm threshold, a fault alarm is generated.

[0140] Specifically, if the difference distance is less than or equal to the alarm threshold, it indicates that the equipment is operating normally. The system records the current status as "normal" and does not trigger a fault alarm. Optionally, update the equipment operation status monitoring interface to display the "normal" status. If the difference distance is greater than the alarm threshold, it indicates that there may be a fault with the equipment. The system triggers a fault alarm to notify relevant personnel. Record the fault alarm information, including the time, difference distance value, alarm threshold, etc. Optionally, perform automatic shutdown, speed reduction, or other safety measures. Or update the equipment operation status monitoring interface to display the "fault" status and highlight relevant parameters.

[0141] Specifically, by comparing the distance of the drilling machine with the alarm threshold in real time and promptly discovering and handling potential faults, the accidental shutdown time of the equipment can be effectively reduced, and the reliability and stability of the equipment can be improved. It helps to extend the service life of the equipment and reduce the maintenance cost.

[0142] Specifically, the process of setting the alarm threshold includes:

[0143]

[0144] Where P(w <= W1) represents the probability that the random variable w is less than or equal to W1, F represents the false alarm rate, and W1 represents the alarm threshold. represents the estimated value P(W) of the probability density function integrated from 0 to W1, that is, calculates the probability of w within this interval.

[0145] Specifically, by integrating to calculate the area under the probability density function from 0 to W1, this area is the probability that the random variable w is less than or equal to W1, and at the same time this probability should be equal to 1 minus the false alarm rate. Determine an acceptable false alarm rate F according to the equipment characteristics, operation requirements, and safety standards. The false alarm rate should not be set too high or too low. Too high will lead to frequent false alarms, and too low may miss real faults.

[0146] Specifically, through the integral calculation based on the probability density function, the alarm threshold can be accurately determined, significantly improving the accuracy of fault early warning. Reducing the false alarm rate and minimizing the ineffective maintenance and resource waste caused by false alarms. Setting a reasonable alarm threshold helps to promptly discover potential equipment faults, prevent small problems from evolving into major accidents. Improving the operation efficiency and safety of the equipment and extending the service life of the equipment.

[0147] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

[0148] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An equipment fault warning method based on real-time dynamic multi-dimensional feature modeling of drilling, characterized in that Including: Step S1: Real-time collect the rotation parameters, axial parameters, and other parameters of several collection points during the drilling process according to the sampling interval. The rotation parameters include the rotary table speed and torque, the axial parameters include the hook load and drilling speed, and the other parameters include the wire rope displacement and drill string mass; Step S2: Obtain the motor moment of inertia, rotary table moment of inertia, effective damping, and motor gear ratio. Determine the effective inertia according to the motor gear ratio, the motor moment of inertia, and the rotary table moment of inertia. Construct a drill string torsional motion model based on the effective damping, the effective inertia, and the rotary table speed to calculate the torque of the drill string to obtain a torque prediction value; Step S3: Construct a drill string axial motion model based on the drill string mass, the wire rope displacement, and the drilling speed to calculate the hook load of the drill string to obtain a hook load prediction value; Step S4: Compare the torque with the predicted torque value to obtain a torque residual signal, and compare the hook load with the hook load prediction value to obtain a hook load residual signal; Step S5: Construct a residual signal vector based on the hook load residual signal and the torque residual signal, and analyze the residual signal vector to obtain a mean vector and a covariance matrix; Step S6: Use the Gaussian distribution method to determine the standard residual signal distribution and the real-time residual signal distribution according to the mean vector and the covariance matrix. Calculate the degree of conformity of the residual signal vector in the standard residual signal distribution to obtain a probability density value, and compare the probability density value with a preset fault warning threshold to obtain the target real-time residual signal distribution; Step S7: Determine the drilling rig distance according to the standard residual signal distribution and the target real-time residual signal distribution; Step S8: Compare the drilling rig distance with a preset alarm threshold to determine whether a fault alarm is generated.

2. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 1, characterized in that The process of constructing a drill string torsional motion model based on the effective damping, the angular displacement, the effective inertia, and the rotary table speed to calculate the torque of the drill string to obtain a torque prediction value includes: Where Tr represents the predicted torque value; cv represents the effective damping of the drill string, θ represents the angular displacement, ts represents the time interval, and k represents the current sampling point. J represents the effective inertia, J_rt and J_m represent the moments of inertia of the rotary table and the motor shaft respectively; n0 represents the gear ratio of the rotary table motor.

3. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 2, wherein The process of step S3 includes: Among them, \(F_h\) is the load prediction value, \(m_{ds}\) is the mass of the drill string, \(g\) represents the acceleration due to gravity, \(x_{ds}\) is the axial displacement of the drill string, \(k\) represents the current sampling point, \(v_{ds}\) represents the lowering speed of the drill string, that is, the drilling speed.

4. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 3, characterized in that, The process of step S4 includes: Obtain a torque residual signal according to the difference between the torque and the predicted torque value; Obtain a hook load residual signal according to the difference between the hook load and the hook load prediction value.

5. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 4, characterized in that, The process of step s5 includes: Combine the hook load residual signal and the torque residual signal at the same sampling point into a two-dimensional vector to obtain a residual signal vector; Define the size of a time window to obtain the number of sampling points; Collect the residual signal vectors at each time interval within the time window; Average the residual signal vectors within the time window according to the number of sampling points to obtain the mean vector; Perform a centering process on the mean vector to obtain a processing result; Calculate the mean vector of the product of the processing result and the transpose of the processing result to obtain the covariance matrix.

6. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 5, characterized in that, The process of step S6 includes: Construct a multivariate Gaussian distribution model according to the mean vector and the covariance matrix; Determine the real-time residual signal distribution according to the multivariate Gaussian distribution model; Obtain historical data based on the real-time residual signal distribution to obtain a standard residual signal distribution; Use the multivariate Gaussian distribution formula according to the residual signal vector to calculate the degree of conformity of the residual signal vector in the standard residual signal distribution to obtain a probability density value; Compare the probability density value with the fault warning threshold to obtain the target real-time residual signal distribution.

7. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 6, characterized in that, The process of comparing the probability density value with the fault warning threshold to obtain the target real-time residual signal distribution includes: If the probability density value is less than the fault warning threshold, adjust the mean vector and the covariance matrix to obtain a target real-time residual signal; If the probability density value is greater than or equal to the fault warning threshold, the real-time residual signal is equal to the target real-time residual signal.

8. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 7, characterized in that, The process of step S7 includes: Calculate the Euclidean distance between the mean vector of the standard residual signal distribution and the mean vector of the target residual signal distribution; Calculate the trace of the covariance matrix according to the covariance matrix of the standard residual signal distribution and the covariance matrix of the target residual signal distribution; Determine the drilling rig distance according to the sum of the trace of the covariance matrix and the Euclidean distance.

9. The equipment fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 8, characterized in that The process of step S8 includes: If the drilling rig distance is less than or equal to the alarm threshold, it indicates normal working conditions and no fault alarm is generated; If the drilling rig distance is greater than the alarm threshold, a fault alarm is generated.

10. The device fault warning method based on real-time dynamic multi-dimensional feature modeling according to claim 9, characterized in that The setting process of the alarm threshold includes: Where, P(w<=W1) represents the probability that the random variable w is less than or equal to W1, F represents the false alarm rate, and W1 represents the alarm threshold.

Images