A method for monitoring the running state of a downhole emulsion drilling rig

Abstract

The present application relates to the technical field of downhole intelligent monitoring, and particularly relates to a running state supervision method of a downhole emulsion drilling machine, which comprises the following steps: obtaining historical steady-state data sets of the drilling machine to pre-train a trained model; obtaining multi-dimensional parameters at the current time, including mechanical running state data, internal and external environmental state data, and electrical state data; calculating deep feature vector differences and statistical distribution offset conditions to obtain a basic working condition offset coefficient; calculating a first adjustment factor based on the mechanical running state data; calculating a second adjustment factor based on the internal and external environmental state data and the electrical state data; fusing the basic working condition offset coefficient and the two factors to dynamically correct a preset fixed threshold to obtain an adaptive dynamic threshold; and comparing the obtained fault probability value at the current time with the adaptive dynamic threshold to determine the current running state of the drilling machine. The present application effectively solves the false and missed reporting problems caused by the fixed threshold.

Description

Technical Field

[0001] This invention relates to the field of downhole intelligent monitoring technology, and in particular to a method for monitoring the operating status of downhole emulsion drilling rigs. Background Technology

[0002] In underground mining environments such as coal mines, emulsion drilling rigs are critical drilling equipment, and their operating status directly affects operational efficiency and overall safety. Due to the extremely complex and harsh underground environment, which is constantly filled with high concentrations of methane, large amounts of dust, and extremely high humidity, the internal mechanical components of the drilling rig are prone to wear and aging under long-term, high-load continuous operation, leading to a significant increase in the probability of equipment failure. If the operating status is not monitored in a timely manner, a sudden failure during operation will not only force the drilling operation to be interrupted, seriously affecting the progress of the entire mining face and causing huge economic losses, but also may lead to serious secondary safety accidents due to equipment malfunction, such as drill rod breakage causing casualties, high-pressure emulsion leakage causing fires or even gas explosions, seriously threatening the lives of front-line underground workers. Therefore, real-time monitoring of the operating status of underground emulsion drilling rigs has become an essential requirement for ensuring the safety of underground operations.

[0003] Currently, when monitoring the operational status of downhole emulsion drilling rigs, traditional methods typically involve real-time collection of key parameters of the rig's operation using sensors, such as motor current, vacuum level, and emulsion temperature. These multi-dimensional parameters are then used as input to a Softmax classifier. This new method utilizes massive historical data to label different operational states as training labels. The Softmax classifier transforms the model's linear output into a probability distribution for each classification state, selecting the state with the highest probability as the final prediction result. During the training phase, the model's internal parameters are continuously optimized using a cross-entropy loss function. Ultimately, upon deployment, this enables real-time classification and early warning of the drilling rig's operational status, assisting maintenance personnel in quickly locating and responding to problems.

[0004] However, when faced with complex and variable downhole operating conditions, traditional methods typically use a fixed threshold for the output probability of Softmax. This fixed threshold is prone to false alarms or missed alarms when dealing with dynamic fluctuations in operating conditions. For example, when the drilling rig is drilling in hard rock, the overall load on the equipment will increase significantly. At this time, the range of normal mechanical vibration and fluid pressure fluctuations will also naturally expand. If the alarm threshold remains unchanged, the model is very likely to directly misjudge this normal high-load condition as a fault alarm. Conversely, as the drilling rig equipment gradually ages, the fault characteristic signals generated in the early stages will often attenuate and weaken. If the original fixed threshold is still used, it may cause the system to miss these critical early hidden faults, making the supervision lose the meaning of early prevention. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method for monitoring the operating status of downhole emulsion drilling rigs, in order to solve the problem that the existing technology, which uses traditional fixed probability thresholds, is prone to serious false alarms or omissions in monitoring the operating status of drilling rigs.

[0006] This invention provides a method for monitoring the operational status of downhole emulsion drilling rigs, comprising the following steps: acquiring historical steady-state datasets of the drilling rig for pre-training to obtain a trained model; acquiring multi-dimensional parameters of the drilling rig at the current moment, the multi-dimensional parameters including mechanical operating status data, internal and external environmental status data, and electrical status data; extracting and calculating the differences between the multi-dimensional parameters at the current moment and the deep feature vectors of the historical steady-state dataset in the trained model, and calculating the basic operating condition offset coefficient at the current moment by combining the statistical distribution offset of each dimension parameter at the current moment relative to the corresponding dimension parameter in the historical steady-state dataset; and extracting the pump station output pressure increase and motor torque based on the mechanical operating status data. The first adjustment factor for the current moment is calculated by comprehensively considering the torque surge rate, emulsion pressure pulsation, and recent energy consumption anomalies. Based on internal and external environmental data and electrical status data, the combined influence coefficient of temperature and humidity, the high-frequency / low-frequency energy ratio of vibration, and the dispersion of power supply voltage fluctuations are extracted to comprehensively calculate the second adjustment factor for the current moment. The basic operating condition offset coefficient, the first adjustment factor, and the second adjustment factor are fused together to dynamically correct a preset fixed threshold, resulting in an adaptive dynamic threshold for the current moment. The multi-dimensional parameters for the current moment are input into the trained model to obtain the fault probability value for the current moment. The fault probability value is compared with the adaptive dynamic threshold to determine the current operating status of the drilling rig.

[0007] Preferably, the acquisition of multi-dimensional parameters of the drilling rig at the current moment includes: mechanical operating status data, which refers to pump station output pressure, motor torque, high-frequency pressure pulsation amplitude, and power consumption; internal and external environmental status data, which refers to tank temperature, ambient humidity, and vibration; and electrical status data, which refers to power supply voltage.

[0008] Preferably, the calculation of the baseline operating condition offset coefficient at the current moment includes: constructing a target domain feature vector set from the deep feature vectors of each moment within the current time window, and constructing a source domain feature vector set from all deep feature vectors of the historical steady-state dataset; obtaining the normalized value of the inner product distance between the target domain feature vector set and the source domain feature vector set in the regeneration kernel Hilbert space, as the spatial difference; calculating the absolute difference between the mean of each dimension parameter within the time window and the mean of the corresponding dimension parameter in the historical steady-state dataset, and averaging the ratio of the absolute difference to the corresponding standard deviation, as the sensor data drift; calculating the ratio of the variance of each dimension parameter within the current time window to the variance of the corresponding dimension parameter in the historical steady-state dataset, taking the logarithm of the ratio and averaging it, as the sensor data fluctuation variation; summing and normalizing the spatial difference, sensor data drift, and sensor data fluctuation variation to obtain the baseline operating condition offset coefficient.

[0009] Preferably, the extraction of the pump station output pressure increase magnitude, motor torque surge rate, emulsion pressure pulsation, and recent energy consumption anomaly degree includes: calculating the normalized result of the absolute difference between the pump station output pressure at the current moment and the average pump station output pressure in the historical steady-state dataset, as the pump station output pressure increase magnitude; obtaining the nearest local minimum point as the motor torque valley value, and calculating the normalized result of the rate of change of the difference between the motor torque at the current moment and the motor torque valley value, as the motor torque surge rate; normalizing the amplitude of all high-frequency pressure pulsations within the time window at the current moment and taking the average value, as the emulsion pressure pulsation; and performing linear fitting on all power consumption data within the time window at the current moment, obtaining the absolute value of the slope of the fitted straight line as the recent energy consumption anomaly degree.

[0010] Preferably, the calculation of the first adjustment factor at the current moment includes: summing the motor torque surge rate with the emulsion pressure pulsation to obtain a first intermediate value; summing the recent energy consumption anomaly with a preset constant to obtain a second intermediate value; multiplying the first intermediate value and the second intermediate value to obtain a third intermediate value; dividing the increase in pump station output pressure by the third intermediate value, and performing standard normalization on the obtained quotient to obtain the first adjustment factor.

[0011] Preferably, the extraction of the joint influence coefficient of temperature and humidity, the high-frequency and low-frequency energy ratio of vibration, and the dispersion of the power supply voltage fluctuation amplitude includes: obtaining the ratio of the normalized value of ambient humidity to the normalized value of oil tank temperature at the current moment as the joint influence coefficient of temperature and humidity; performing a fast Fourier transform on the vibration sequence within the time window at the current moment to extract the total energy of the high-frequency band and the total energy of the low-frequency band, and calculating the high-frequency and low-frequency energy ratio of vibration; using an extreme value search algorithm to extract the absolute difference of all adjacent peak-valley pairs of power supply voltage within the time window at the current moment and calculating the first average value; simultaneously calculating the second average value of the absolute difference of adjacent peak-valley pairs of power supply voltage based on the historical steady-state dataset; and calculating the normalized result of the absolute value of the difference between the second average value and the first average value as the dispersion of the power supply voltage fluctuation amplitude.

[0012] Preferably, the calculation of the second adjustment factor at the current moment includes: summing the high-frequency and low-frequency energy ratio of the vibration with the dispersion of the power supply voltage fluctuation amplitude to obtain a comprehensive environmental fluctuation value; multiplying the temperature and humidity joint influence coefficient with the comprehensive environmental fluctuation value, and normalizing the product result to obtain the second adjustment factor.

[0013] Preferably, the adaptive dynamic threshold at the current moment satisfies the expression: In the formula, The adaptive dynamic threshold for the current moment; This is a preset fixed threshold; This is the basic operating condition offset coefficient at the current moment; The first adjustment factor at the current moment; The second adjustment factor at the current moment; This is the standard normalization function.

[0014] Preferably, determining the current operating status of the drilling rig includes: if the fault probability value is greater than the adaptive dynamic threshold, then determining that the drilling rig has malfunctioned and triggering an alarm; if the fault probability value is less than or equal to the adaptive dynamic threshold, then determining that the drilling rig is in normal operating condition.

[0015] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: This invention calculates the basic working condition offset coefficient by comparing the overall distribution differences between real-time and historical steady-state working conditions, effectively avoiding large-scale misjudgments caused by macroscopic geological changes such as drastic changes in coal seam hardness. Simultaneously, by combining the physical load characteristics of the drilling rig, it extracts multi-dimensional mechanical stress indicators such as pressure, torque, and pulsation to construct a first adjustment factor, enabling the alarm threshold to accurately rise and fall with the fluctuations of actual working intensity, significantly improving the identification accuracy under critical working conditions such as heavy loads. Furthermore, it fully integrates the characteristics of underground temperature and humidity, as well as internal and external interferences such as electrical and vibration systems, to construct a second adjustment factor, and deeply links it with the first adjustment factor and the basic working condition offset coefficient, ultimately generating an adaptive dynamic threshold. This method achieves dynamic adaptive matching between specific working conditions and thresholds, comprehensively enhancing the reliability of the status monitoring system's early warning in complex, harsh, and variable underground environments. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art 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.

[0017] Figure 1 This is a flowchart of a method for monitoring the operating status of a downhole emulsion drilling rig, provided in Embodiment 1 of the present invention. Detailed Implementation

[0018] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0019] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0020] This invention provides a method for monitoring the operational status of downhole emulsion drilling rigs, such as... Figure 1 As shown, the method includes the following steps: Step S101: Obtain the historical steady-state dataset of the drilling rig for pre-training to obtain the trained model; obtain the multi-dimensional parameters of the drilling rig at the current moment, including mechanical operation status data, internal and external environmental status data and electrical status data.

[0021] It should be noted that before monitoring and analyzing the operating status of downhole emulsion drilling rigs, it is necessary to comprehensively and in real-time grasp various physical and operational parameters of the equipment in the complex downhole environment. The operation of the drilling rig downhole is not only limited by its own mechanical hardware structure, but also affected by external environmental temperature and humidity and power supply system stability. Therefore, it is necessary to systematically collect comprehensive sensor data, including drilling rig operating mechanical parameters, emulsion fluid parameters, and electrical equipment status monitoring. Since the working principles of various sensors are different, their data acquisition frequencies have objective differences. Directly inputting them into the model will lead to dimensional errors. Therefore, it is necessary to first perform synchronization and alignment processing on the time series data. At the same time, in order to analyze the feature shift during subsequent operation, it is necessary to pre-train a deep learning classification model using steady-state data of the equipment in a healthy state before deploying the model, and extract the depth feature vector under the steady-state benchmark to provide data basis for the subsequent monitoring process.

[0022] Specifically, the system acquires pump station output pressure data at a frequency of 10Hz using an explosion-proof pressure transmitter installed at the emulsion pump station outlet; acquires motor torque data at a frequency of 10Hz using a differential pressure sensor at the hydraulic motor inlet and outlet; acquires power consumption data at a frequency of 10Hz using a motor electrical parameter measurement module; acquires high-frequency pressure signals using a specially installed high-frequency dynamic pressure sensor on the pipeline, performs bandpass filtering, and extracts pressure pulsation amplitude at a frequency of 1Hz; acquires tank temperature data at a frequency of 0.1Hz using a platinum resistance temperature sensor inside the emulsion tank; acquires vibration data at a frequency of 10Hz using a vibration sensor at the power head; acquires ambient humidity data at a frequency of once per minute using a downhole wall-mounted temperature and humidity transmitter; and acquires power supply voltage data at a frequency of 10Hz using a voltage sensor at the front end of the power module. Since the acquisition frequencies of each parameter are different, the system uses a known linear interpolation method to time-synchronize and resample all the original sensor data with different acquisition frequencies, aligning them to a uniform frequency of 10Hz to form a multi-dimensional feature sequence.

[0023] It should be added that the sampling frequency can be adjusted by the personnel based on the actual complex downhole environment; the multi-dimensional parameters include mechanical operating status data, internal and external environmental status data, and electrical status data. Among them, mechanical operating status data refers to pump station output pressure, motor torque, high-frequency pressure pulsation amplitude, and power consumption; internal and external environmental status data refers to tank temperature, ambient humidity, and vibration; electrical status data refers to power supply voltage.

[0024] Furthermore, a historical steady-state dataset of a period of stable operation of the downhole emulsion drilling rig under test without abnormal interference was extracted. The input parameters of this dataset consist of eight dimensions of parameters after unified resampling and alignment: pump station output pressure, motor torque, power consumption, pressure pulsation amplitude, tank temperature, vibration data, ambient humidity, and power supply voltage. This historical steady-state dataset was then input into a Softmax classification model for pre-training. After training, the historical steady-state dataset was input again into the trained model, allowing the model to perform forward propagation on each time sample in the historical steady-state dataset to the penultimate layer, extracting the deep feature vectors output from each time sample to form a source domain feature vector set.

[0025] Thus, the source domain feature vector set of the downhole emulsion drilling rig and the real-time multi-dimensional parameter data during operation were obtained.

[0026] Step S102: Extract and calculate the differences between the multi-dimensional parameters at the current moment and the deep feature vectors of the historical steady-state dataset in the trained model. Combine the statistical distribution offset of each dimension parameter at the current moment with respect to the corresponding dimension parameter in the historical steady-state dataset to calculate the basic working condition offset coefficient.

[0027] It should be noted that the actual operating conditions of downhole emulsion drilling rigs are often complex and variable due to different geological conditions. For example, when the drilling rig suddenly transitions from a soft coal seam to a rock formation with extremely high hardness, or when the internal mechanical components undergo irreversible aging and wear due to continuous high-load use of the equipment, these macroscopic factors will cause a significant drift between the underlying data distribution generated by the actual operation of the drilling rig and the historical steady-state data distribution on which the pre-trained model relies. If the model's default fixed probability threshold is used directly for fault alarm judgment at this time, the classifier is very likely to produce large-scale false alarms or false alarms at the moment the operating condition shifts or during long-term aging. Therefore, it is necessary to analyze this overall distribution difference from a macroscopic statistical perspective, and calculate a comprehensive basic operating condition offset coefficient by extracting spatial distance differences, the degree of offset of various data center locations, and changes in the overall data fluctuation amplitude, so as to provide a reliable benchmark amplitude for subsequent dynamic adjustment of the threshold.

[0028] Specifically, the spatial difference at the current moment is calculated by: extracting multi-dimensional parameter data within the time window of the current moment and inputting it into the trained model; propagating forward to the penultimate layer; extracting the deep feature vectors of each moment within the time window of the current moment to form a target domain feature vector set; then mapping the target domain feature vector set and the source domain feature vector set to the regenerating kernel Hilbert space; calculating the inner product distance between the two in the regenerating kernel Hilbert space; and performing max-min normalization to obtain the spatial difference at the current moment.

[0029] The sensor data drift at the current moment is calculated by: calculating the mean of any dimension parameter within the time window at the current moment; calculating the absolute difference between the mean and the mean of the corresponding dimension parameter in the historical steady-state dataset; calculating the ratio of the absolute difference to the standard deviation of the corresponding dimension parameter in the historical steady-state dataset; and averaging the ratios calculated for all dimensions to obtain the sensor data drift at the current moment.

[0030] The calculation method for the degree of fluctuation of sensor data at the current moment includes: calculating the ratio of the variance of any dimension parameter within the time window at the current moment to the variance of the corresponding dimension parameter in the historical steady-state dataset, taking the logarithm of the ratio, and averaging the logarithmic results obtained from all dimensions to obtain the degree of fluctuation of sensor data at the current moment.

[0031] It should be added that the time window is a time series of data extracted backward from the current moment (including the data at the current moment). The size of the time window reflects the statistical length of the system for recent operating conditions. In this embodiment, it is set to 10 minutes to ensure that it can effectively filter out extremely short-term accidental fluctuation noise and keep up with the macroscopic distribution trend of geological environment and equipment status in a timely manner. The implementers can flexibly adjust it according to the sensitivity requirements of actual mine geological changes.

[0032] The spatial difference, sensor data drift, and sensor data fluctuation at the current moment are integrated to calculate the basic operating condition offset coefficient at the current moment; the specific calculation formula is as follows: In the formula, This is the basic operating condition offset coefficient at the current moment; The spatial difference at the current moment; The current sensor data drift. The degree of fluctuation in the sensor data at the current moment; This is the standard normalization function.

[0033] in, It measures the macroscopic shift of the overall distribution of depth characteristics such as pressure and vibration when drilling into coal seams of different hardness at the current moment. The larger the value, the more drastic the change in the underground geological conditions or equipment status at the current moment. It reflects the current degree of sensor data drift. The larger the value, the more the current average state of the device deviates from the historical steady-state safety benchmark. This represents the degree of data fluctuation at the current moment, reflecting the overall volatility of the current operating conditions compared to the historical steady-state benchmark. Combining these three dimensions and normalizing them, the larger the overall value, the more significant the overall difference is likely to exist between the current downhole operating conditions and the historical steady-state conditions during model pre-training.

[0034] At this point, the basic operating condition offset coefficients for each moment are obtained.

[0035] Step S103: Based on the mechanical operation status data, extract the pump station output pressure increase rate, motor torque surge rate, emulsion pressure pulsation and recent energy consumption anomaly degree, and calculate the first adjustment factor.

[0036] It should be noted that although the basic working condition offset coefficient reflects the offset of the macroscopic geological distribution, the specific mechanical stress state has a very concrete impact on the early warning judgment. For example, in heavy-load drilling operations, the pressure output by the pump station will naturally increase due to the high resistance of the formation. If the threshold is not actively raised to tolerate this physical heavy load, it is very easy to cause false alarms. Conversely, if the driving torque of the hydraulic motor increases sharply in a very short period of time, this is often a high-risk precursor to drill bit jamming in the borehole. Or, if the high-frequency pressure pulsation and energy consumption of the emulsion system increase significantly at the same time, it indicates that the pipeline has ruptured and leaked. Faced with these fatal latent faults, it is necessary to immediately lower the alarm threshold to maximize the alarm sensitivity of the classifier. Therefore, by analyzing the natural increase in pressure, the sudden increase rate of torque, and the abnormal pulsation state of the emulsion pipeline, a first adjustment factor is constructed to characterize the mechanical micro-stress state and balance the judgment threshold.

[0037] Specifically, the increase in pump station output pressure at the current moment is calculated as follows: the absolute difference between the current pump station output pressure and the mean of the pump station output pressure at all moments in the historical steady-state dataset is calculated, and the normalized result of the absolute difference is taken as the increase in pump station output pressure at the current moment.

[0038] The current motor torque surge rate is obtained by continuously monitoring the motor torque time series using a sliding window algorithm, searching for the nearest local minimum point as the motor torque valley value, calculating the difference between the current motor torque value and the motor torque valley value, and the ratio of this difference to the time interval between the current time and the time corresponding to the motor torque valley value. The normalized result of this ratio is taken as the current motor torque surge rate.

[0039] It should be added that high-risk physical changes such as stuck drill pipe in downhole often go through a rapid stress deterioration process of tens of seconds from the initial appearance of warning signs to complete jamming. In this embodiment, the sliding window is selected as 1 minute. The 1-minute window can not only fully cover the transient evolution of the sudden increase after the sudden drop in torque to accurately locate the real torque trough value, but also avoid the introduction of early irrelevant historical fluctuations due to an excessively long window.

[0040] The emulsion pressure pulsation at the current moment is calculated by performing maximum and minimum normalization on all high-frequency pressure pulsation amplitudes within the current time window. The emulsion pressure pulsation is equal to the average of the normalized results of all high-frequency pressure pulsation amplitudes within the current time window.

[0041] The recent energy consumption anomaly level at the current moment is obtained by extracting all power consumption data within the current time window, performing a linear fit on it, and obtaining the absolute value of the slope of the fitted line as the recent energy consumption anomaly level at the current moment.

[0042] Based on the current pump station output pressure increase, motor torque surge rate, emulsion pressure pulsation, and recent energy consumption anomalies, the first adjustment factor for the current moment is calculated; the specific calculation formula is as follows: In the formula, The first adjustment factor at the current moment; This represents the increase in pump station output pressure at the current moment. This represents the rate of increase in motor torque at the current moment. This represents the current emulsion pressure pulsation. The degree of recent energy consumption anomaly at the current moment; This is a preset constant to prevent the denominator from approaching zero when the equipment power is extremely stable; its value range is [value range missing]. In this embodiment, we take ; This is the standard normalization function.

[0043] in, This indicates the physical increase in the pump station output pressure at the current moment compared to the historical normal baseline value. The larger this value, the more the equipment is currently under normal heavy load requiring maximum thrust. At this time, the normal fluctuation boundary of the system pressure expands, which should prompt the adjustment factor to increase as a whole and guide the subsequent system to actively raise the alarm threshold. This means that whether the equipment faces a sudden surge in torque that could cause the drill to jam or abnormal high-frequency pulsations due to pipeline rupture and leakage, any local high-risk physical precursor will increase the basic risk. Moreover, by using abnormal energy consumption at the system level as a deteriorating amplifier, once the stress characteristics of local drill jamming or leakage are superimposed on the abnormal fluctuations in overall energy consumption, the denominator will expand rapidly. The entire structure constructs a balancing mechanism: when the pressure increase caused by normal thrust dominates, the adjustment factor increases; but once high-risk risks such as a sudden increase in torque or abnormal fluid pulsation are accompanied by abnormal energy consumption and dominate, the expansion of the denominator will cause the factor to decrease rapidly as a whole. This characterizes the bidirectional comprehensive traction effect of multidimensional complex load characteristics on the safety warning threshold.

[0044] Thus, the first adjustment factor at each moment is obtained.

[0045] Step S104: Based on the internal and external environmental status data and electrical status data, extract the joint influence coefficient of temperature and humidity, the high-frequency and low-frequency energy ratio of vibration, and the dispersion of power supply voltage fluctuation amplitude, and calculate the second adjustment factor.

[0046] It should be noted that, in addition to mechanical and fluid loads, harsh external environments and the hidden health status of the system also continuously affect the baseline reliability of sensor data. When the temperature inside the emulsion tank rises significantly due to continuous operation, the fluid viscosity decreases drastically, and the pressure drop response of minor leaks becomes extremely sluggish. To more sensitively detect leaks, higher temperatures actually require the system to actively lower the threshold to improve detection sensitivity. Conversely, extremely high ambient humidity corrodes electrical insulation and significantly increases the background white noise of the sensing circuit. To prevent noise-induced false alarms, higher humidity actually requires raising the threshold to filter interference. On the other hand, an abnormally high ratio of high- and low-frequency mechanical vibration energy is a core indicator of early wear failure, while frequent fluctuations in the power supply voltage can cause false drift in sensor readings. Therefore, it is necessary to integrate this antagonistic logic of temperature and humidity and combine it with the internal health status to construct a second adjustment factor to ensure the accuracy of subsequent threshold determination.

[0047] Specifically, the temperature and humidity joint influence coefficient at the current moment is obtained by performing maximum and minimum normalization on the fuel tank temperature and ambient humidity collected synchronously by the sensors at the current moment. Since the threshold needs to be lowered to cover up the leakage due to the temperature increase, and the threshold needs to be raised to introduce noise due to the increased humidity, the ratio of the normalized ambient humidity to the normalized fuel tank temperature is used as the temperature and humidity joint influence coefficient at the current moment.

[0048] The high-frequency to low-frequency energy ratio of vibration at the current moment is obtained by performing a fast Fourier transform on the vibration sequence within the time window of the current moment, extracting the total energy of the high-frequency band and the total energy of the low-frequency band, and taking the ratio of the total high-frequency energy to the total low-frequency energy as the high-frequency to low-frequency energy ratio of vibration at the current moment.

[0049] The dispersion of the power supply voltage fluctuation amplitude at the current moment is calculated as follows: An extreme value search algorithm is used to extract all peak values ​​of the power grid power supply voltage time series data within the current time window, and the next trough value immediately adjacent to each peak value in the time series is located sequentially, thus forming multiple sets of adjacent peak-trough value pairs; the absolute difference of each set of peak-trough value pairs within the time window is calculated, and the first average value is obtained from all absolute differences; similarly, based on the power grid power supply voltage sequence of the historical steady-state dataset, the mean of the absolute differences of all adjacent peak-trough value pairs in the power grid power supply voltage sequence is calculated as the second average value; the absolute value of the difference between the second average value and the first average value is calculated, and maximum-minimum normalization is performed to obtain the dispersion of the power supply voltage fluctuation amplitude at the current moment.

[0050] The second adjustment factor is calculated based on the combined influence coefficient of temperature and humidity at the current moment, the high-frequency to low-frequency energy ratio of vibration, and the dispersion of power supply voltage fluctuations; the specific calculation formula is as follows: In the formula, The second adjustment factor at the current moment; This represents the combined influence coefficient of temperature and humidity at the current moment. The ratio of high-frequency to low-frequency energy of the vibration at the current moment; The degree of dispersion of the power supply voltage fluctuation at the current moment; This is the standard normalization function.

[0051] in, It incorporates the counter-logic of the current fluid state and environmental severity. The higher the tank temperature, the smaller the coefficient, which causes the second adjustment factor to shrink downward, guiding the system to lower the threshold and improve the detection sensitivity. The higher the ambient humidity, the larger the coefficient, which requires the system to raise the threshold to avoid false alarms. The larger the value, the more rapidly the probability of hidden faults such as abnormal wear of the drill bit or bearing breakage is increasing at the current moment. The larger the value, the more dispersed the recent power supply voltage fluctuations in the mine are, and the more serious the drift phenomenon of the bottom sensor readings is. The higher the threshold needs to be to avoid misjudgment.

[0052] This reflects the degree of degradation and interference in the system's internal infrastructure at the current moment: on the one hand, This reflects the internal high-frequency energy transfer of mechanical vibration at the current moment. The larger this value, the stronger the structural high-frequency mechanical noise caused by drill bit wear or bearing breakage at the current moment; on the other hand... This reflects the dispersion of the power supply network at the current moment. The larger this value, the more severe the drift of the underlying sensor electrical signals caused by voltage instability at the current moment. The sum of these two values ​​represents the total basic interference and drift amount superimposed within the current system. Since the degree of harm of internal interference to the final state determination is affected by external temperature and humidity: The construction logic is based on the evolution of the humidity-to-temperature ratio. When the humidity in the downhole environment increases sharply, the high humidity environment will severely erode the electrical insulation performance, multiplying the risk of false alarms caused by power supply drift and mechanical noise. Therefore, the multiplication of the two parts will cause the overall adjustment factor to be affected. Rapidly increasing in size guides subsequent models to raise alarm thresholds to rigorously prevent false alarms; conversely, when the tank temperature is extremely high, the emulsion viscosity drops significantly, and the pressure drop characteristics of minor leaks become extremely subtle and slow. At this point, even if there is strong voltage drift or vibration noise within the system, the multiplicative structure will utilize the reduced... Forced suppression of the whole item The expansion prevents the alarm threshold from rising.

[0053] At this point, the second adjustment factor for the current moment has been obtained.

[0054] Step S105: Input the multi-dimensional parameters of the current moment into the trained model to obtain the fault probability value of the current moment; fuse the basic working condition offset coefficient, the first adjustment factor and the second adjustment factor to dynamically correct the preset fixed threshold and obtain the adaptive dynamic threshold; compare the fault probability value with the adaptive dynamic threshold to determine the current operating status of the drilling rig.

[0055] It should be noted that traditional classifiers use fixed thresholds for state monitoring, ignoring the dynamic effects of sudden geological changes, drastic load fluctuations, and severe temperature and humidity interference in the well. Under complex operating conditions, this can easily lead to widespread false alarms or fatal missed alarms. The basic operating condition offset coefficient obtained in the aforementioned steps assesses the magnitude of macroscopic distribution changes, while the first and second adjustment factors address the microscopic mechanical load stress characteristics and the health status of the internal and external environment, respectively, compensating for the judgment blind spots of the fixed threshold. Therefore, this step integrates the basic operating condition offset coefficient and the two core adjustment factors to deeply fuse and dynamically correct the fixed basic threshold, enabling the alarm threshold to follow the real complex downhole operating conditions in real time, thereby improving the anti-interference capability and recognition accuracy of the early warning system.

[0056] Specifically, a preset fixed threshold is set in advance. To facilitate the balanced adjustment of the two sides of the algorithm, a median value of 0.5 is reasonably chosen in this embodiment. The implementer can adjust it based on the actual monitoring accuracy. The basic working condition offset coefficient, the first adjustment factor, and the second adjustment factor at the current moment are weighted and fused to obtain the adaptive dynamic threshold at the current moment. The specific calculation formula is as follows: In the formula, The adaptive dynamic threshold for the current moment; This is a preset fixed threshold; This is the basic operating condition offset coefficient at the current moment; The first adjustment factor at the current moment; The second adjustment factor at the current moment; This is the standard normalization function.

[0057] Among them, if the equipment is facing an extremely high risk of sudden stuck drill or fluid leakage at the current moment, the first adjustment factor It will increase. The value will decrease sharply, thereby driving the fixed threshold to shrink proportionally, forcing the system to capture early, weak, real fault signals more sensitively, and preventing the failure to detect fatal faults to the extreme; if the background noise caused by the high humidity environment downhole or the reading drift caused by power grid fluctuations dominates at the current moment, the second adjustment factor It will increase. The value will expand accordingly, driving the fixed threshold to be passively raised, thereby giving the system a stronger tolerance to interference and resolutely blocking false alarms caused by environmental noise; at the same time, the current basic operating condition offset coefficient is multiplied before the two adjustment factors as a leverage weight, which means that the more serious the overall underlying distribution offset at the current moment, the more intense the threshold adjustment response of the system to micro-force anomalies or environmental disturbances should be.

[0058] Furthermore, after obtaining the adaptive dynamic threshold, the multidimensional parameter data collected at the current moment is input into the trained model to obtain the fault probability value at the current moment output by the model; the fault probability value is compared with the adaptive dynamic threshold at the current moment: if the fault probability value is greater than the adaptive dynamic threshold at the current moment, it is determined that the drilling rig has malfunctioned, an alarm is triggered, and relevant personnel are notified to detect it in time; if the fault probability value is less than or equal to the adaptive dynamic threshold at the current moment, it is determined that the drilling rig is in normal operating condition, and real-time monitoring continues.

[0059] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for monitoring the operational status of a downhole emulsion drilling rig, characterized in that, include: The historical steady-state dataset of the drilling rig is obtained for pre-training to obtain the trained model; Obtain multi-dimensional parameters of the drilling rig at the current moment, including mechanical operation status data, internal and external environmental status data, and electrical status data; Extract and calculate the differences between the multi-dimensional parameters at the current moment and the deep feature vectors of the historical steady-state dataset in the trained model. Combine the statistical distribution offset of each dimension parameter at the current moment with the corresponding dimension parameter in the historical steady-state dataset to calculate the basic working condition offset coefficient at the current moment. Based on mechanical operation data, the pump station output pressure rise rate, motor torque surge rate, emulsion pressure pulsation and recent energy consumption anomaly are extracted, and the first adjustment factor at the current moment is calculated in combination. Based on internal and external environmental status data and electrical status data, the combined influence coefficient of temperature and humidity, the high-frequency and low-frequency energy ratio of vibration, and the dispersion of power supply voltage fluctuation amplitude are extracted, and the second adjustment factor at the current moment is calculated in a comprehensive manner. The basic operating condition offset coefficient, the first adjustment factor, and the second adjustment factor are fused together to dynamically correct the preset fixed threshold and obtain the adaptive dynamic threshold at the current moment. Input the multi-dimensional parameters of the current moment into the trained model to obtain the fault probability value of the current moment; The fault probability value is compared with the adaptive dynamic threshold to determine the current operating status of the drilling rig.

2. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 1, characterized in that, The acquisition of the drilling rig's multi-dimensional parameters at the current moment includes: The mechanical operating status data refers to the pump station output pressure, motor torque, high-frequency pressure pulsation amplitude, and power consumption; the internal and external environmental status data refers to the oil tank temperature, ambient humidity, and vibration; and the electrical status data refers to the power supply voltage.

3. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 1, characterized in that, The calculation of the basic operating condition offset coefficient at the current moment includes: The deep feature vectors at each moment within the current time window are used to construct the target domain feature vector set, and all deep feature vectors from the historical steady-state dataset are used to construct the source domain feature vector set. The normalized value of the inner product distance between the target domain feature vector set and the source domain feature vector set in the regeneration kernel Hilbert space is obtained as the spatial difference. The absolute difference between the mean of each dimension parameter within the time window and the mean of the corresponding dimension parameter in the historical steady-state dataset is calculated, and the ratio of the absolute difference to the corresponding standard deviation is averaged as the sensor data drift. The ratio of the variance of each dimension parameter within the current time window to the variance of the corresponding dimension parameter in the historical steady-state dataset is calculated, and the logarithm of the ratio is taken and averaged as the sensor data fluctuation variation. The spatial difference, sensor data drift, and sensor data fluctuation variation are summed and normalized to obtain the basic operating condition offset coefficient.

4. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 1, characterized in that, The increase in output pressure of the extraction pump station, the rate of increase in motor torque, the pulsation of emulsion pressure, and the degree of recent energy consumption anomalies include: The normalized absolute difference between the pump station output pressure at the current moment and the mean pump station output pressure in the historical steady-state dataset is calculated as the increase in pump station output pressure. The nearest local minimum point to the current moment is obtained as the motor torque valley value, and the normalized rate of change of the difference between the motor torque at the current moment and the motor torque valley value is calculated as the motor torque surge rate. The average value of all high-frequency pressure pulsation amplitudes within the current time window is normalized and taken as the emulsion pressure pulsation. Linear fitting is performed on all power consumption data within the current time window, and the absolute value of the slope of the fitted line is obtained as the degree of recent energy consumption anomaly.

5. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 4, characterized in that, The calculation of the first adjustment factor at the current moment includes: The first intermediate value is obtained by summing the motor torque surge rate with the emulsion pressure pulsation; the second intermediate value is obtained by summing the recent energy consumption anomaly with a preset constant; the third intermediate value is obtained by multiplying the first intermediate value with the second intermediate value; the third intermediate value is obtained by dividing the pump station output pressure increase by the third intermediate value and standardizing the obtained quotient to obtain the first adjustment factor.

6. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 1, characterized in that, The degree of dispersion of the extracted temperature and humidity combined influence coefficient, the high-frequency and low-frequency energy ratio of vibration, and the power supply voltage fluctuation amplitude includes: The ratio of the normalized ambient humidity value to the normalized fuel tank temperature value at the current moment is obtained as the joint influence coefficient of temperature and humidity. A fast Fourier transform is performed on the vibration sequence within the current time window to extract the total energy of the high-frequency band and the total energy of the low-frequency band, and the high-frequency-to-low-frequency energy ratio of the vibration is calculated. The absolute difference of all adjacent peak-valley pairs of the power supply voltage within the current time window is extracted using an extreme value search algorithm, and the first average value is obtained. At the same time, the second average value of the absolute difference of adjacent peak-valley pairs of the power supply voltage is calculated based on the historical steady-state dataset. The normalized result of the absolute value of the difference between the second average value and the first average value is calculated as the degree of dispersion of the power supply voltage fluctuation amplitude.

7. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 6, characterized in that, The calculation of the second adjustment factor at the current moment includes: The environmental fluctuation comprehensive value is obtained by summing the high-frequency and low-frequency energy ratio of the vibration with the dispersion of the power supply voltage fluctuation amplitude; the environmental fluctuation comprehensive value is obtained by multiplying the temperature and humidity joint influence coefficient with the environmental fluctuation comprehensive value and normalizing the product result.

8. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 1, characterized in that, The adaptive dynamic threshold at the current moment satisfies the expression: ； In the formula, The adaptive dynamic threshold for the current moment; This is a preset fixed threshold; This is the basic operating condition offset coefficient at the current moment; The first adjustment factor at the current moment; The second adjustment factor at the current moment; This is the standard normalization function.

9. The method for monitoring the operating status of a downhole emulsion drilling rig according to claim 1, characterized in that, The determination of the current operating status of the drilling rig includes: If the fault probability value is greater than the adaptive dynamic threshold, the drilling rig is determined to have malfunctioned and an alarm is triggered; if the fault probability value is less than or equal to the adaptive dynamic threshold, the drilling rig is determined to be in normal operating condition.

Images