Method for monitoring and analyzing drill string state in real time under limit drilling parameters

Abstract

The invention discloses a drill string state real-time monitoring and analyzing method under limit drilling parameters, which comprises the following steps of: acquiring multi-source sensor data of a drill string in real time, and transmitting the acquired original multi-source sensor data to a data processing center in real time; de-noising and drift correction are carried out on the multi-source sensor data by adopting a self-adaptive filtering algorithm; the preprocessed multi-source sensor data are input into the multi-parameter fusion analysis model subjected to data equalization processing optimization, and the current drill string state is recognized; outputting a drill string state identifier by using a multi-layer neural network structure and a nonlinear activation function, and transmitting the drill string state identifier to a monitoring terminal on a drilling site in real time; and the data processing center classifies and stores the information in a distributed database. According to the invention, the working state information of the drill string under the complex working condition can be comprehensively and timely obtained, and the state of the drill string can be accurately identified. Major accidents are effectively avoided, and the personal safety of drilling operators and the safety of drilling equipment are guaranteed.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of oil and gas drilling and geological exploration drilling, and particularly relates to a method for real-time monitoring and analyzing the state of a drill string under extreme drilling parameters. BACKGROUND

[0002] In the current drilling industry, the drill string, as a key component connecting the surface equipment and the bottom hole assembly, plays an irreplaceable role in the drilling process. On the one hand, it bears the responsibility of transmitting torque and axial force, enabling the drill bit to obtain sufficient power to break rock; on the other hand, it also provides a flow channel for mud, which carries cuttings back to the surface through circulation, ensuring the cleanliness of the bottom hole and maintaining the smooth progress of drilling operations.

[0003] In practical applications, the drill string is usually composed of multiple single pieces, each with a specific length and specification, which need to be selected and connected according to actual drilling needs. During drilling, the working state of the drill string directly affects drilling efficiency, cost, and safety, so effective monitoring of its state is crucial.

[0004] At present, the following conventional methods are mainly used for drill string state monitoring: I. Manual experience judgment: experienced drilling workers observe phenomena such as drill string vibration, mud return state, changes in drilling pressure and torque, etc. during drilling, and use their own experience to roughly judge whether the drill string is in a normal working state. This method relies on the individual experience and professional skill level of workers, and the results of different workers may differ, and it is difficult to detect some early and subtle fault problems. For example, when the drill string has slight fatigue damage, manual experience judgment may not be able to detect it in time.

[0005] II. Simple instrument monitoring: some simple instruments such as pressure gauges, torque gauges, and tachometers are used to monitor some parameters of the drill string. The above instruments can display some working parameters of the drill string in real time, providing some reference for operators. However, the instrument can only monitor a single or a few parameters, and cannot fully reflect the overall working state of the drill string. For example, by monitoring torque and speed alone, it is impossible to know the stress distribution of the drill string in the complex environment of the well and whether there are local deformations, etc.

[0006] Third, wired sensor monitoring: wired sensors such as stress sensors, acceleration sensors, etc. are installed at specific positions of the drill string, and the collected data is transmitted to the ground monitoring system through wired transmission. This method can obtain some key mechanical parameters of the drill string, which is more accurate and comprehensive than manual experience judgment and simple instrument monitoring. However, the installation and wiring of wired sensors are relatively complex, and during drilling, the drill string is constantly rotating and vibrating, which can easily cause cable damage, affecting the stability and reliability of data transmission. In addition, wired transmission is also limited by transmission distance, and it is difficult to meet the monitoring needs for some ultra-deep wells or complex well conditions.

[0007] In extreme drilling parameters, such as ultra-deep well drilling (well depth exceeding 8000 meters), deep sea drilling (water depth exceeding 1000 meters), etc. special scenarios, the drill string faces extreme conditions such as high temperature, high pressure, high torque, large axial force and complex formation environment, etc. The above existing monitoring methods expose many shortcomings as follows: First, real-time monitoring is not possible: manual experience judgment and simple instrument monitoring cannot achieve real-time and continuous monitoring of the drill string state, and there are monitoring blind spots. When the drill string has a sudden failure in a short period of time, it cannot be detected in time, resulting in serious drilling accidents. For example, in ultra-deep well drilling, the drill string suddenly breaks, and due to the lack of timely monitoring, the drill bit may fall to the bottom of the well, causing complex accidents such as sticking, which not only delays drilling progress, but also increases significant economic losses.

[0008] Second, poor accuracy: manual experience judgment is greatly influenced by subjective factors, and it is difficult to accurately determine the fault type and location for some complex fault conditions. Simple instrument monitoring cannot comprehensively analyze the working state of the drill string due to the single parameter, resulting in low accuracy of the monitoring results. In deep sea drilling, due to the corrosion of seawater and the complex action of sea currents, the stress on the drill string becomes more complex, and it is difficult to accurately determine whether the drill string has fatigue cracks or other damage relying on simple monitoring means.

[0009] Third, cannot be fully analyzed: the existing monitoring methods cannot fully collect and comprehensively analyze the mechanical parameters of the drill string, drilling fluid parameters, and downhole environmental parameters, making it difficult to accurately assess the working state of the drill string under complex working conditions. For example, in high temperature and high pressure formation environments, the mechanical properties of the drill string will change, and the performance of the drilling fluid will also be affected. Monitoring only part of the parameters cannot fully understand the working state of the drill string, and cannot provide effective decision support for drilling operations.

[0010] Therefore, there is an urgent need to develop a method that can monitor and analyze the drill string state in real time, accurately and comprehensively under extreme drilling parameters. SUMMARY

[0011] The present application solves the problem of providing a drilling string state real-time monitoring and analysis method under extreme drilling parameters, which can realize real-time, accurate and comprehensive monitoring and analysis of the drilling string state under extreme drilling parameters.

[0012] To solve the above technical problems, the technical scheme adopted by the present application is as follows: a drilling string state real-time monitoring and analysis method under extreme drilling parameters, comprising the following steps, S1: acquiring multi-source sensor data of the drilling string in real time, and transmitting the collected original multi-source sensor data to a data processing center in real time; S2: using an adaptive filtering algorithm to denoise and correct drift of the multi-source sensor data; S3: inputting the preprocessed multi-source sensor data into a multi-parameter fusion analysis model optimized by data equalization processing, and identifying the current drilling string state; S4: outputting the drilling string state identification by using a multi-layer neural network structure and a nonlinear activation function, and transmitting the state information to a monitoring terminal at the drilling site in real time by the data processing center; S5: the data processing center classifies and stores the collected original data, preprocessed data, multi-parameter fusion analysis results and state warning information in a distributed database, so as to perform backtracking analysis on the drilling string state in the future.

[0013] Further, in the S1, the multi-source sensor data includes drilling string mechanical parameters, drilling fluid parameters and downhole environment parameters, and the multi-source sensor data is acquired by a sensor group, which includes but is not limited to stress sensors, torque sensors and acceleration sensors installed at different positions of the drilling string, pressure sensors and flow sensors installed in the drilling fluid circulation system, and temperature sensors and vibration sensors installed in the downhole.

[0014] Further, in the S2, the adaptive filtering algorithm is based on a first-order Kalman filter, and the update formula of the state estimation value thereof is: wherein, Xk represents the state estimation value at the current time; Xk-1 represents the estimation value at the last time; Yk represents the observation value at the current time; K represents the Kalman gain coefficient, which is determined by the prediction error covariance and the observation error covariance; The calculation formula of the Kalman gain is: wherein, denotes the estimation error covariance of the last time, and R denotes the observation error covariance, and by dynamically adjusting the gain, the estimation stability and response speed can be maintained when strong noise interference or sensor performance fluctuation occurs.

[0015] Further, in the S3, the multi-parameter fusion analysis model synthesizes the abnormal state samples by using an SMOTE algorithm, generates new samples by interpolating between the minority class samples and their neighbors in the feature space, and the generation mode is as follows: wherein, denotes the original minority class sample, denotes the k-nearest neighbor sample in the feature space, and δ is a random coefficient with a value between 0 and 1, which is used to generate new points between samples and enhance the representativeness of abnormal samples.

[0016] Further, the multi-parameter fusion analysis model optimizes the model parameters by minimizing the loss function, and the loss function expression is as follows: wherein, N is the total number of samples, is the true label of the sample, is the probability that the model predicts that the sample belongs to the positive class, and the function is used to evaluate the prediction effect of the model under the current parameters.

[0017] Further, in the S4, the drill string state identification method is converted by a multi-layer neural network structure and a nonlinear activation function, and the formula is as follows: wherein, denotes the activation output of the lth layer, is the weight matrix of the lth layer, is the bias vector, and f is the nonlinear activation function.

[0018] Further, in the S4, the output layer of the drill string state identification method performs probability normalization on the classification results by using a softmax function, which is used for state recognition, and the formula is as follows: wherein, denotes the output value corresponding to the jth state category, K is the total number of states, and in the present model, three state categories are corresponding, which are normal drilling, slight abnormality and serious abnormality, denotes the probability that the input data belongs to the state j.

[0019] Further, the present application also provides a device for running the above-mentioned data processing method.

[0020] Further, the present application also provides a device comprising a memory, a processor and an algorithm stored in the memory and executable on the processor, wherein the processor implements the above-mentioned data processing method when executing the computer program.

[0021] Further, the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer algorithm, and the computer algorithm implements the above-mentioned data processing when executed by a processor.

[0022] The present application has the advantages and positive effects that: The present application can comprehensively and timely acquire the working state information of the drill string under complex working conditions, accurately identify whether the drill string is in a normal drilling state, a slight abnormal state or a serious abnormal state. Once an abnormal state is detected, a warning can be quickly sent and corresponding shutdown measures can be taken, so that major accidents such as drill string fracture and pipe sticking can be effectively avoided, the personal safety of the drilling operation personnel and the safety of the drilling equipment can be ensured, the efficiency of the drilling operation can be improved, the operation cost can be reduced, and strong support can be provided for the smooth performance of oil and gas exploitation and geological exploration activities. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 is a model online updating mechanism flowchart of an embodiment of the present application.

[0024] Figure 2 is a multi-parameter fusion analysis model construction flowchart of an embodiment of the present application.

[0025] Figure 3 is a whole flowchart schematic diagram of an embodiment of the present application.

[0026] Figure 4 is a result comparison diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions of the present application will be described clearly and completely below with reference to the drawings, obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0028] The embodiments of the present application will be further described below with reference to the drawings: As shown in Figures 1-4 a real-time monitoring and analysis method for the state of a drill string under extreme drilling parameters, comprising the following steps, S1: acquiring the multi-source sensor data of the drill string in real time, and transmitting the collected original multi-source sensor data to a data processing center in real time.

[0029] Specifically, the multi-source sensor data includes drilling string mechanical parameters, drilling fluid parameters, and downhole environmental parameters. The multi-source sensor data is obtained through a sensor group. The sensor group includes, but is not limited to, stress sensors, torque sensors, and acceleration sensors installed at different positions of the drilling string, pressure sensors and flow sensors installed in the drilling fluid circulation system, and temperature sensors and vibration sensors installed downhole. These sensors work continuously during the drilling process, real-time acquire the mechanical parameters of the drilling string, the parameters of the drilling fluid, and the downhole environmental parameters, and transmit the acquired raw data to the data processing center in real time.

[0030] S2: Adopting an adaptive filtering algorithm to denoise and drift correct the multi-source sensor data acquired in real time.

[0031] Specifically, in the data processing center, the preprocessing module first performs filtering processing on the raw data, and adopts an adaptive filtering algorithm to denoise and drift correct the multi-source sensor data. According to the data characteristics collected by different sensors during the drilling process and the dynamic characteristics of noise changes in the environment, the filtering parameters are adjusted in real time, thereby improving the adaptability and accuracy of the filtering processing. Specifically, the core of the adaptive filtering is to dynamically update the filtering gain according to the error between the observation value and the predicted value. As shown in FIG. 1, based on the first-order Kalman filter, the update formula of the state estimation value is: Figure 1 wherein, represents the state estimation value at the current time, is the estimation value at the last time, is the observation value at the current time, is the Kalman gain coefficient, which is determined by the prediction error covariance and the observation error covariance. The calculation formula of the Kalman gain is: wherein, represents the estimation error covariance at the last time, and R represents the observation error covariance. By dynamically adjusting the gain, the stability and response speed of the estimation can be maintained when there is strong noise interference or sensor performance fluctuation. After completing the filtering, the abnormal value detection stage is entered, and the system uses a statistical analysis method for multi-dimensional feature screening. The mean μ and the standard deviation σ of the data sequence are calculated for each type of physical quantity, and μ ± 3σ is used as the boundary to identify outliers, that is, all data points satisfying are regarded as abnormal values. When repairing these abnormal values, the system performs fitting calculation or interpolation processing according to the adjacent normal data before and after the abnormal value, and uses the linear interpolation method to replace the abnormal point with the weighted average of the adjacent normal values, thereby preserving the continuity and structural integrity of the data sequence and providing stable and reliable input for subsequent analysis. ​

[0032] S3: During real-time analysis, the pre-processed drilling string mechanical parameters, drilling fluid parameters and downhole environment parameters are input into the multi-parameter fusion analysis model optimized by data balancing processing to identify the current drilling string state.

[0033] Specifically, as shown in Figure 2 the construction process of the multi-parameter fusion analysis model first systematically cleans and pre-processes historical drilling data, eliminates redundant data and invalid information caused by collection errors, and ensures the accuracy and integrity of the data. After pre-processing, the data set is divided into a training set and a test set, and the data imbalance problem existing in the training set is processed, in which the amount of normal state data is much larger than the amount of abnormal state data. To avoid class bias in the training process of the model, the SMOTE algorithm is used to synthesize abnormal state samples, and new samples are generated by interpolating between the minority class samples and their neighbors in the feature space. The specific generation method is wherein, represents the original minority class sample, represents its k-nearest neighbor sample in the feature space, and δ is a random coefficient with a value between 0 and 1, which is used to generate new points between samples to enhance the representativeness of abnormal samples.

[0034] To further balance the data distribution, undersample the majority class data, randomly delete a part of the normal state samples, so that the number of two types of samples is close, thereby reducing the model deviation in the training process. After completing the data balancing processing, select a machine learning algorithm suitable for the characteristics of the multi-source complex data to construct an initial model, such as a multilayer perception or a convolutional neural network, and use the processed training set to train the parameters. The training process optimizes the model parameters by minimizing the loss function, and the typical cross-entropy loss function is expressed as wherein, N is the total number of samples, is the true label of the sample, is the probability that the model predicts that the sample belongs to the positive class, and the function is used to evaluate the prediction effect of the model under the current parameters.

[0035] After the model training is completed, the independent test set is used for verification, and the generalization ability and robustness of the model are evaluated through accuracy, recall rate, F1 score and other indicators to ensure the performance stability and sensitive identification ability of the model on unseen samples.

[0036] S4: Use a multi-layer neural network structure and a nonlinear activation function to output the drilling string state identifier, and the data processing center transmits the state information to the monitoring terminal at the drilling site in real time.

[0037] Specifically, in real-time analysis, the pre-processed drilling string mechanical parameters, drilling fluid parameters and downhole environment parameters are input into the multi-parameter fusion analysis model optimized by data balancing processing. The multi-parameter fusion analysis model adopts a multi-layer neural network structure to perform feature extraction and fusion calculation on the multi-dimensional data related to the drilling string state in the real-time analysis stage. The input data includes the pre-processed drilling string mechanical parameters, drilling fluid parameters and downhole environment parameters. These inputs are first transmitted to the neural network through the input layer. In each layer, the neurons perform weighted summation on the information transmitted from the previous layer and add a bias term, and then perform conversion through a nonlinear activation function, which has the expression form wherein, represents the activation output of the lth layer, is the weight matrix of the lth layer, is the bias vector, and f is a nonlinear activation function, and ReLU function is often used to enhance the expression ability of the model to complex patterns.

[0038] The stacking of the multi-layer structure enables the model to extract abstract features layer by layer, and extracts key information reflecting the drilling string state from the original sensor data. After processing by several hidden layers, the last output layer normalizes the classification results by the softmax function for state recognition, and the formula is: wherein, represents the output value corresponding to the jth state category, and K is the total number of states, which is three in this model, corresponding to normal drilling, slight abnormality and severe abnormality, represents the probability that the input data belongs to state j.

[0039] The model finally outputs the state category with the highest probability as the state identifier of the current drilling string, and this state information is then transmitted by the data processing center to the field monitoring terminal in real time to support state perception and decision intervention during drilling operations.

[0040] S5: The data processing center classifies and stores the collected raw data, pre-processed data, multi-parameter fusion analysis results and state early warning information in a distributed database for subsequent backtracking analysis of the drilling string state.

[0041] The present application will be specifically described below in conjunction with specific embodiments, including the following steps: S1: Real-time Data Acquisition: When the drilling equipment starts and enters the drilling process, the sensor array installed on the drill string, drilling fluid circulation system, and downhole tools begins to operate. Specifically, in this embodiment, five stress sensors and three torque sensors are installed at the connection section between the drill string and the drill bit. At locations in the middle of the drill string prone to bending deformation, one stress sensor and one torque sensor are installed every 100 meters. In the drilling fluid circulation system, pressure sensors and flow sensors are installed at the drilling pump outlet, riser, and drill bit nozzle, respectively. Downhole, near the drill bit and the bottom of the drill string, high-temperature and high-pressure resistant temperature sensors, vibration sensors, and formation pressure sensors are installed. These sensors acquire the mechanical parameters of the drill string, drilling fluid parameters, and downhole environmental parameters in real time, and transmit the raw data to the data processing center in real time via a combination of wireless and wired transmission.

[0042] S2: Data Preprocessing: Upon receiving the raw data, the data processing center immediately performs preprocessing. First, filtering is performed. For low-frequency mechanical data collected by stress and torque sensors, a Butterworth low-pass filter is used with a cutoff frequency of 50Hz to remove high-frequency noise caused by equipment vibration. For high-frequency vibration data collected by accelerometers and vibration sensors, a wavelet denoising algorithm is used with the db4 wavelet basis function and a decomposition level of 5 to remove noise and retain the true vibration characteristics. Next, outlier detection is performed. A statistical analysis-based method is used to calculate the mean and standard deviation of each parameter, and data exceeding the mean ± 3 times the standard deviation are considered outliers. Simultaneously, an outlier detection rule base is established by combining drilling technology knowledge and historical data patterns. For example, when drilling fluid pressure suddenly drops significantly while flow rate increases significantly at the same time, it is judged as a possible well leakage anomaly, and related data is subject to focused detection and repair. For detected outliers, interpolation or adjacent data fitting methods are used for repair, thus obtaining the preprocessed effective data.

[0043] S3: Model Building and Analysis: Before drilling operations, historical drilling data from the past five years for the region was collected, covering various sensor data corresponding to normal and abnormal drill string conditions (such as drill string fatigue, fracture, stick-slip vibration, etc.) under different extreme drilling parameters. First, the historical drilling data was cleaned and preprocessed to remove duplicate and invalid data, and the data was divided into training and testing sets. Given that the amount of normal drill string data in the training set was much larger than the amount of abnormal data, data augmentation techniques were used to oversample the minority class of data (abnormal data). For example... Figure 4As shown, the new abnormal state data samples are specifically synthesized by the SMOTE algorithm, the 5 nearest neighbor samples of the minority class samples are analyzed, and new synthetic samples are randomly generated between the minority class samples and their 5 nearest neighbor samples. The number of generated synthetic samples is determined according to the imbalance ratio of normal state and abnormal state data, so that the ratio of normal state and abnormal state data in the oversampled training set reaches 1:1. At the same time, undersampling is performed on the majority class data (normal state data) to further reduce the number of normal state data and make the data distribution more balanced. Then, a convolutional neural network algorithm is selected to build an initial model. The initial model is trained using the processed training set, and the parameters and structure of the model are adjusted, such as adjusting the number of layers of the neural network to 5 layers, the number of nodes to 128, and the learning rate to 0.001, etc. The prediction accuracy of the model on the training set reaches 95%. Finally, the trained model is verified by the test set to evaluate the generalization ability and robustness of the model, and to ensure that the model can accurately identify the state of the drill string under extreme drilling parameters. In real-time analysis, the preprocessed drill string mechanics parameters, drilling fluid parameters and downhole environment parameters are input into the multi-parameter fusion analysis model optimized by data balancing, and the model performs feature extraction and fusion calculation on multi-dimensional data through multiple layers of neural networks to identify the current state of the drill string, determine whether it is in a normal drilling state, a slight abnormal state or a serious abnormal state, and output the corresponding state identifier.

[0044] S4: State judgment and early warning: After the multi-parameter fusion analysis model outputs the drill string state identifier, the data processing center transmits the state information to the monitoring terminal at the drilling site in real time. The monitoring terminal displays the state of the drill string in a visual manner on the display screen, with green representing normal state, yellow representing slight abnormal state, and red representing serious abnormal state, and different graphical identifiers are used to distinguish different states. At the same time, the real-time curves and values of key parameters such as the stress, torque, drilling fluid pressure, flow rate, etc. of the drill string are displayed, as well as the current specific values. When the drill string is judged to be in a slight abnormal state, the monitoring terminal issues an audible warning to remind the operator to pay attention, and at the same time displays the detailed information of the abnormal parameter and the possible cause analysis on the display screen. When it is in a serious abnormal state, in addition to the audible warning, an emergency stop signal is automatically triggered to stop drilling operations to prevent the accident from further expanding. At this time, the display screen displays a detailed abnormal report including the type of abnormality, the time of occurrence, the possible range of influence and the recommended treatment measures, etc.

[0045] S5: Data storage and backtracking: The data processing center classifies and stores the collected raw data, preprocessed data, multi-parameter fusion analysis results, and state warning information in a distributed database. During storage, data compression techniques are used to compress the massive drilling data. For raw data, lossless compression algorithms are used to ensure data accuracy during backtracking analysis. For preprocessed data and analysis results, lossy compression algorithms are used within the allowable error range to reduce data storage space. At the same time, a data index and metadata management system is established to facilitate fast retrieval and access to stored data. After drilling operations are completed or when historical data needs to be analyzed, the relevant data can be retrieved from the database through the data query module to study the state change law of the drill string under different extreme drilling parameters. Using data mining and machine learning algorithms, historical data can be analyzed in depth to uncover the relationship between drill string state changes and drilling parameters, downhole environments, providing data support for optimizing drilling technology and drill string design.

[0046] The present application has the advantages and positive effects that: The present application can comprehensively and timely obtain the working state information of the drill string under complex working conditions, accurately identify whether the drill string is in a normal drilling state, a slight abnormal state, or a serious abnormal state. Once an abnormal state is detected, a warning can be quickly issued and appropriate shutdown measures can be taken to effectively prevent major accidents such as drill string rupture and sticking, ensuring the safety of drilling personnel and drilling equipment, improving drilling efficiency and reducing operating costs, and providing strong support for the smooth conduct of oil and gas exploration and geological exploration activities.

[0047] The above describes one embodiment of the present application in detail, but the content described is only a preferred embodiment of the present application and cannot be considered as limiting the scope of the present application. Any equivalent changes and improvements made in accordance with the scope of the present application should still be within the scope of the present application.

Claims

1. A method for real-time monitoring and analysis of drill string status under extreme drilling parameters, characterized in that: Includes the following steps, S1: Acquire multi-source sensor data from the drill string in real time, and transmit the collected raw multi-source sensor data to the data processing center in real time; S2: An adaptive filtering algorithm is used to denoise and correct drift in the multi-source sensor data; S3: Input the preprocessed multi-source sensor data into the multi-parameter fusion analysis model that has been optimized by data equalization processing to identify the current drill string status; S4: The drill string status identifier is output using a multi-layer neural network structure and a non-linear activation function, and the data processing center transmits the status information to the monitoring terminal at the drilling site in real time. S5: The data processing center classifies and stores the collected raw data, preprocessed data, multi-parameter fusion analysis results, and status warning information in a distributed database for subsequent retrospective analysis of the drill string status.

2. The method for real-time monitoring and analysis of drill string status under extreme drilling parameters according to claim 1, characterized in that: In S1, the multi-source sensor data includes drill string mechanical parameters, drilling fluid parameters, and downhole environmental parameters. The multi-source sensor data is acquired through a sensor group, which includes, but is not limited to, stress sensors, torque sensors, and acceleration sensors installed at different locations on the drill string, pressure sensors and flow sensors installed in the drilling fluid circulation system, and temperature sensors and vibration sensors installed downhole.

3. A method for real-time monitoring and analysis of drill string status under extreme drilling parameters according to claim 1 or 2, characterized in that: In S2, the adaptive filtering algorithm is based on a first-order Kalman filter, and the update formula for its state estimate is: in, This represents the estimated state value at the current moment; It is the estimated value from the previous moment; It is the observation value at the current moment; It is the Kalman gain coefficient, whose value is determined by both the prediction error covariance and the observation error covariance; The formula for calculating Kalman gain is: in, R represents the estimation error covariance at the previous moment; R represents the observation error covariance. By dynamically adjusting this gain, the stability and response speed of the estimation can be maintained under strong noise interference or sensor performance fluctuations.

4. A method for real-time monitoring and analysis of drill string status under extreme drilling parameters according to claim 1 or 2, characterized in that: In step S3, the multi-parameter fusion analysis model uses the SMOTE algorithm to synthesize abnormal state samples. New samples are generated by interpolating between minority class samples and their nearest neighbors in the feature space. The generation method is as follows: in, Represents the original minority class samples, It represents the k nearest neighbor sample in the feature space, and δ is a random coefficient between 0 and 1, used to generate new points between samples and enhance the representativeness of abnormal samples.

5. The method for real-time monitoring and analysis of drill string status under extreme drilling parameters according to claim 4, characterized in that: The multi-parameter fusion analysis model optimizes the model parameters by minimizing a loss function, the expression of which is as follows: Where N is the total number of samples, For the true label of the sample, This function predicts the probability that the sample belongs to the positive class and evaluates the model's prediction performance under the current parameters.

6. A method for real-time monitoring and analysis of drill string status under extreme drilling parameters according to claim 1 or 2, characterized in that: In step S4, the method for identifying the drill string state is achieved through a multi-layer neural network structure and a non-linear activation function, as shown in the following formula: in, This represents the activation output of the l-th layer. Let be the weight matrix of the l-th layer. Let f be the bias vector and f be the nonlinear activation function.

7. The method for real-time monitoring and analysis of drill string status under extreme drilling parameters according to claim 6, characterized in that: In step S4, the output layer of the drill string state identification method uses a softmax function to probabilistically normalize the classification results for state recognition. The formula is as follows: in, This represents the output value corresponding to the j-th state category, where K is the total number of states. In this model, there are three state categories: normal drilling, minor anomaly, and severe anomaly. This represents the probability that the input data belongs to state j.

8. An apparatus, characterized in that: The data processing method described in any one of claims 1 to 7 is executed.

9. An apparatus comprising a memory, a processor, and an algorithm stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the data processing method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer algorithm, characterized in that, When the computer algorithm is executed by the processor, it performs the data processing as described in any one of claims 1 to 7.

Images