Formation identification method and device based on vibration of rotary drilling rig drill rod and drilling data

Abstract

The application discloses a stratum identification method and device based on drill rod vibration and drilling data of a rotary drilling rig. The method comprises the following steps: obtaining drill rod vibration original data in the drilling process of the rotary drilling rig and drilling parameters and drill bit parameters of the rotary drilling rig; performing time synchronization on the drill rod vibration original data and the drilling parameters, and processing the drill rod vibration original data to obtain vibration data to be analyzed; performing Fourier transform on the vibration data to be analyzed to obtain power spectral density data; extracting time domain features and frequency domain features according to the vibration data to be analyzed and the power spectral density data; fusing the time domain features, the frequency domain features, the drilling parameters and the drill bit parameters to form feature vectors for training, taking stratum categories confirmed through geological exploration as labels, training a random forest classification model; and outputting stratum types based on the trained random forest classification algorithm model. The application improves the real-time performance and reliability of stratum identification under complex stratum conditions.

Description

Technical Field

[0001] This invention relates to the field of rotary drilling rig construction technology, specifically to a method and device for identifying strata based on rotary drilling rig drill rod vibration and drilling data. Background Technology

[0002] Rotary drilling rigs, as a highly efficient pile foundation construction equipment, are widely used in building, bridge, and municipal engineering projects. During construction, they often need to traverse complex geological formations such as clay, sand, and rock. The complexity of the geological conditions directly affects the drilling efficiency, drill bit wear, and pile quality of the rotary drilling rig. Therefore, real-time and accurate identification of geological formations during drilling is crucial for optimizing drilling parameters, mitigating construction risks, and improving construction efficiency.

[0003] Traditional formation identification techniques rely primarily on operator experience or post-drilling core analysis to determine formation types, which has several inherent drawbacks: First, poor real-time performance, failing to identify formation changes during drilling, leading to delayed adjustments in drilling parameters and increasing the risk of drilling deviation, borehole collapse, and excessive drill bit wear. Second, low accuracy, as manual judgment is heavily influenced by subjective factors, making it difficult to accurately identify weak interlayers or isolated boulders, easily resulting in misjudgments. Third, reliance on offline analysis, with time-consuming and labor-intensive core sampling processes, and analysis results failing to provide timely guidance for current drilling operations, significantly reducing construction efficiency. Fourth, limited monitoring dimensions, relying solely on single-source data such as drill pressure, rotation speed, torque, or drilling rate, making it susceptible to fluctuations in equipment status, operating methods, and working conditions, resulting in insufficient identification stability.

[0004] With the development of technology, existing technologies have also seen solutions for automatically identifying strata types using construction status data from rotary drilling rigs. For example, invention patent application number 202310855989.3 discloses a method, device, system, and rotary drilling rig for identifying strata types during rotary drilling. It identifies the working conditions of the rotary drilling rig, acquires status data associated with strata types under drilling conditions, and extracts characteristic data representing strata types based on vibration type data and non-vibration type data, thereby identifying the current strata type.

[0005] However, the vibration data in the above schemes mainly consists of power head vibration signals. Vibrations generated by the interaction between the formation and the drill bit need to be transmitted through structures such as the drill pipe to the power head position before being collected. During this transmission process, signal attenuation, coupling, and interference may occur, resulting in the power head vibration signals not directly representing changes in the formation near the drill bit. Furthermore, this scheme focuses on fusing and identifying a large number of state statistical features within a preset unit time period. There is still room for improvement in the representation of drill pipe vibration response near the drill bit, the frequency domain abrupt changes in the drill pipe's triaxial vibration signals, and the synergistic representation relationship between drilling parameters, drill bit parameters, and near-drill bit vibration features. Therefore, it is still necessary to propose a formation identification method and device that can collect vibration data from the lower part of the drill pipe near the drill bit during rotary drilling and combine it with drilling parameters and drill bit parameters for multi-source feature fusion, in order to improve the real-time performance, stability, and reliability of formation identification under complex construction conditions.

[0006] Furthermore, existing technologies also include research and application approaches that use drill rod or drill string vibration signals to reflect formation characteristics. Vibration signals can characterize the impact, friction, and rock-breaking state during the interaction between the drill bit and the formation to a certain extent. However, in actual rotary drilling, vibration signals are easily affected by factors such as equipment vibration, drill rod transmission, construction operations, and the surrounding environment. If analysis relies solely on a single vibration signal, or if the vibration signal is not effectively correlated with drilling parameters such as drill pressure, torque, and drilling rate, the formation identification characteristics may be incomplete, affecting the stability and real-time performance of the identification results under complex formation conditions. Drilling parameters such as drill pressure, torque, rotational speed, drilling rate, and drilling depth can reflect the stress state and construction response during the drill bit's rock-breaking or soil-cutting process, and are correlated with the formation's hardness, particle composition, and local abrupt changes. Therefore, it is necessary to propose a formation identification method and device that can collect vibration data of the lower part of the drill rod near the drill bit, and combine it with drilling parameters and drill bit parameters for preprocessing, frequency domain analysis, feature extraction and model recognition, so as to improve the reliability and applicability of formation identification during rotary drilling. Summary of the Invention

[0007] This invention provides a method and apparatus for stratum identification based on the vibration of rotary drilling rig drill rod and drilling data, in order to solve the problems of insufficient effective information from single-source data, difficulty in balancing real-time performance and stability, and unclear model training and inference processes in existing stratum identification methods.

[0008] According to a first aspect of the present invention, a method for formation identification based on drill rod vibration and drilling data of a rotary drilling rig is provided, comprising:

[0009] The raw data of drill rod vibration during the drilling process of rotary drilling rig are obtained. The raw data of drill rod vibration are vibration signals of three mutually perpendicular axes collected by a triaxial accelerometer set at the lower part of the drill rod near the drill bit.

[0010] Obtain the drilling parameters and drill bit parameters of the rotary drilling rig. The drilling parameters include drilling pressure, rotation speed, torque, drilling rate and drilling depth. The drill bit parameters include drill bit type and borehole diameter.

[0011] The raw data of drill pipe vibration is synchronized with the drilling parameters in time, and the raw data of drill pipe vibration is subjected to outlier removal, filtering and noise reduction and windowing to obtain the vibration data to be analyzed.

[0012] Fourier transform is performed on each axial vibration signal of the vibration data to be analyzed to obtain the corresponding power spectral density data; time-domain features are extracted from the vibration data to be analyzed, and frequency-domain features are extracted from the power spectral density data. The time-domain features include at least one of root mean square value, peak value, peak-to-peak value, kurtosis, and skewness; the frequency-domain features include at least one of frequency band energy proportion, main peak frequency, main peak amplitude, spectral centroid, and spectral kurtosis. These features are used to represent the differences in power spectral density maps when drilling into different formations.

[0013] The time-domain features, frequency-domain features, drilling parameters, and drill bit parameters are fused to form a training feature vector, and the stratigraphic categories confirmed by geological surveys are used as labels to train a random forest classification model.

[0014] During the construction identification phase, the time-domain features, frequency-domain features, drilling parameters, and drill bit parameters acquired in real time are fused to form an inference feature vector, and the inference feature vector is input into a trained random forest classification model to output the formation category corresponding to the current drilling depth.

[0015] During construction, drill rod vibration data, drilling parameters and drill bit parameters are acquired in real time. Time domain features and frequency domain features are extracted from the vibration data to be analyzed. The time domain features, frequency domain features, drilling parameters and drill bit parameters are fused to form an inference feature vector.

[0016] The inference feature parameters are input into the trained random forest classification algorithm model, and the output is the stratum category corresponding to the current drilling depth.

[0017] Optionally, the triaxial accelerometer is installed inside the vibration data acquisition module, which is fixedly installed on the bottom side wall of the drill pipe and maintains a preset distance from the drill bit; the triaxial accelerometer has a range of ±50g, a frequency response range of 0.5kHz to 10kHz, and a sampling frequency not lower than the acquisition frequency of the drilling parameters.

[0018] Optionally, the time synchronization includes aligning the original drill pipe vibration data with the drilling parameters based on a timestamp, or aligning the original drill pipe vibration data with the drilling parameters based on a drilling depth range; the windowing process includes extracting data segments according to a preset time window or a preset depth window.

[0019] Optionally, the filtering and noise reduction includes at least one of bandpass filtering, wavelet thresholding, adaptive Kalman filtering, and moving average filtering; the outlier removal includes removing outlier data points based on amplitude thresholds, rate of change thresholds, or statistical outlier rules.

[0020] Optionally, the Fourier transform is a 1024-point or 2048-point fast Fourier transform, and the power spectral density data includes the energy distribution of each axial vibration signal within multiple preset frequency bands.

[0021] Optionally, the drill bit type is incorporated into the training feature vector or the inference feature vector using category coding or one-hot coding, and the drilling parameters are smoothed and normalized before being incorporated into the training feature vector or the inference feature vector.

[0022] Optionally, the training process of the random forest classification model includes: collecting training samples from multiple pile foundation construction sites; labeling and filtering the vibration data, drilling parameters, and drill bit parameters in the training samples; dividing the filtered samples into training set, validation set, and test set; adjusting the number of trees, tree depth, and node splitting parameters of the random forest classification model according to the validation set results; and determining whether the model has reached the preset recognition accuracy threshold according to the test set results.

[0023] Optionally, the stratigraphic category includes at least one of clay, sand, gravel, rock, and boulders; when the output stratigraphic category is inconsistent with the stratigraphic category of the previous identification window, and multiple consecutive identification windows point to the same new stratigraphic category, a stratigraphic abrupt change prompt is generated.

[0024] According to a second aspect of the present invention, a formation identification device based on rotary drilling rig drill rod vibration and drilling data is provided for implementing the formation identification method described in any of the above claims, comprising:

[0025] The vibration data acquisition module is located at the lower part of the drill pipe near the drill bit and is used to collect raw vibration data of the drill pipe in three mutually perpendicular axes.

[0026] The drilling parameter acquisition module is used to communicate with the rotary drilling rig control system and acquire drilling pressure, rotation speed, torque, drilling rate and drilling depth.

[0027] The drill bit parameter acquisition module is used to obtain the drill bit type and borehole diameter;

[0028] The data preprocessing module is used to synchronize the original vibration data with the drilling parameters in time, and to perform outlier removal, filtering and noise reduction, and windowing processing on the original vibration data to form vibration data to be analyzed.

[0029] The Fourier transform module is used to perform Fourier transform on the vibration data to be analyzed and generate power spectral density data based on the Fourier transform result.

[0030] The feature extraction module is used to extract time-domain features and frequency-domain features based on the vibration data to be analyzed and the power spectral density data;

[0031] The identification module is used to fuse the time-domain features, frequency-domain features, drilling parameters and drill bit parameters into a feature vector, and input the feature vector into a trained formation classification model to output the formation category corresponding to the current drilling depth.

[0032] The display module is used to display the formation type, drilling parameters, and power spectral density data.

[0033] Optionally, the vibration data acquisition module includes a triaxial accelerometer, a signal conditioning unit, a microcontroller, a wireless communication unit, and a protective housing; the wireless communication unit is used to send the acquired vibration data to the data preprocessing module.

[0034] The drilling parameter acquisition module is connected to the rotary drilling rig control system via industrial Ethernet, CAN bus or serial communication interface; the display module is located in the rotary drilling rig control room and is used to trigger alarm prompts when there is a sudden change in the formation type or data abnormality.

[0035] It also includes a data storage and export module and a power supply module; the data storage and export module is used to store and export vibration data, drilling parameters, feature vectors, identification results and power spectral density data, and the power supply module is used to provide vehicle power supply adaptation and power protection for each module.

[0036] The beneficial effects of this invention are:

[0037] 1. Existing automatic formation identification methods only collect vibration data from the drill bit during vibration data acquisition. Furthermore, the vibration generated by the interaction between the formation and the drill bit needs to be transmitted to the drill bit through the drill pipe. During this transmission, the vibration signal attenuates, resulting in insufficient reflection of formation characteristics by the drill bit vibration data. This invention collects raw triaxial vibration data of the drill pipe near the drill bit at the lower part of the drill pipe, bringing the vibration signal acquisition location closer to the direct interaction area between the drill bit and the formation. Compared to only collecting equipment vibration signals from locations far from the drill bit, this method can more directly obtain the vibration response generated during drill bit rock breaking, cutting, or soil disturbance, which is beneficial for improving the ability of vibration characteristics to characterize formation changes.

[0038] 2. The drill rod vibration data collected in this invention consists of vibration data along three mutually perpendicular axes. Preprocessing, Fourier transform, and power spectral density analysis are performed on the vibration data for each axis. The triaxial vibration data reflects the impact, friction, and fracturing responses generated by the interaction between the drill bit and the formation from different directions. The power spectral density data further characterizes the distribution of vibration energy across different frequency ranges, thereby improving the ability to distinguish between different formation states such as clay, sand, gravel, rock, and boulders.

[0039] 3. Before feature extraction, this invention synchronizes the raw triaxial vibration data of the drill rod with the drilling parameters in time. It also performs outlier removal, filtering and noise reduction, and windowing on the raw vibration data. This ensures that each set of vibration data, drilling parameters, and drill bit parameters corresponds to a defined time window or drilling depth range, which helps reduce the impact of equipment vibration, drill rod transmission interference, construction operation disturbances, and environmental noise on the identification results. Simultaneously, this invention extracts time-domain and frequency-domain features from the vibration data and power spectral density data to be analyzed, respectively. This comprehensively reflects the amplitude changes, impact intensity, and frequency energy distribution characteristics of the vibration signal, thereby improving the ability to distinguish between different strata such as clay, sand, gravel, rock, and boulders.

[0040] 4. This invention not only extracts time-domain and frequency-domain features from drill pipe vibration signals, but also integrates drilling parameters such as drill pressure, rotational speed, torque, drilling rate, and drilling depth, as well as drill bit parameters such as drill bit type and borehole diameter. Drill pipe vibration characteristics reflect the dynamic response generated by the interaction between the drill bit and the formation; drilling parameters reflect the stress state and construction response during drilling; and drill bit parameters reflect the influence of different drilling tool conditions on vibration and drilling conditions. Through the fusion of these multi-source features, more comprehensive formation identification information can be obtained than by relying solely on vibration signals or drilling parameters.

[0041] 6. The device of this invention has a simple structure and is easy to install. It can communicate with the control system of a rotary drilling rig and realize data acquisition, processing, identification, and display through a vibration data acquisition module, a drilling parameter acquisition module, a drill bit parameter acquisition module, a data preprocessing module, a Fourier transform module, a feature extraction module, an identification module, and a display module. The device has a clear structure, which facilitates integration with existing rotary drilling rigs. The vibration data acquisition module can transmit data wirelessly, which helps to reduce the interference of wired lines on drilling operations and improves the convenience of on-site installation and use. Attached Figure Description

[0042] Figure 1 This is a flowchart of an embodiment of the formation identification method based on rotary drilling rig drill rod vibration and drilling data of the present invention;

[0043] Figure 2 This is a schematic diagram of the installation of the vibration sensor of the present invention;

[0044] Figure 3 This is a schematic diagram of a data preprocessing module according to an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of an embodiment of the drill pipe triaxial vibration data acquisition method of the present invention;

[0046] Figure 5 This is a flowchart of drill pipe vibration data preprocessing according to an embodiment of the present invention;

[0047] Figure 6 This is a flowchart of a multi-source feature fusion and stratigraphic classification model construction according to an embodiment of the present invention;

[0048] Figure 7 This is a schematic diagram of an embodiment of the stratigraphic identification device of the present invention. Detailed Implementation

[0049] To enable those skilled in the art to better understand the technical solutions of the present invention, the embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. All equivalent substitutions, modifications, or improvements made based on the technical concept of the present invention should be included within the scope of protection of the present invention.

[0050] Example 1

[0051] See Figure 1 This embodiment provides a formation identification method based on rotary drilling rig drill rod vibration and drilling data. The method collects triaxial vibration data from the lower part of the drill rod near the drill bit, combines this data with drilling parameters and drill bit parameters for feature fusion, and then uses a trained formation classification model to output the formation category corresponding to the current drilling depth. The method specifically includes the following steps.

[0052] Step S100: Obtain raw data of drill rod vibration during the drilling process of the rotary drilling rig. The raw data of drill rod vibration consists of vibration signals from three mutually perpendicular axes collected by a triaxial accelerometer located at the lower part of the drill rod near the drill bit.

[0053] See Figure 2 A vibration data acquisition module is installed at the lower part of the rotary drilling rig's drill rod, near the drill bit. This module can be fixed to the bottom sidewall or the outer side of the lower part of the drill rod using clamps, bolts, or a protective housing, maintaining a preset distance from the drill bit. The vibration data acquisition module includes a triaxial accelerometer, which is used to collect raw vibration data of the drill rod in three mutually perpendicular axes.

[0054] In one embodiment, the three axes can correspond to the drill pipe axis, the first radial direction, and the second radial direction, respectively. The drill pipe axis reflects the impact and indentation response experienced by the drill bit along the drilling direction, while the two radial directions reflect the lateral vibration response generated when the drill bit interacts with the borehole wall, formation particles, or localized hard interlayers. By simultaneously acquiring data along these three axes, the vibration state during the interaction between the drill bit and the formation can be comprehensively reflected.

[0055] In one embodiment, the triaxial accelerometer has a range of ±50g and a frequency response range of 0.5–10kHz, with a sampling frequency not lower than the acquisition frequency of the drilling parameters. The vibration data acquisition module may further include a signal conditioning unit, a data acquisition unit, a wireless communication unit, and a protective housing. The signal conditioning unit is used to perform preliminary amplification, filtering, or analog-to-digital conversion on the signal output by the triaxial accelerometer; the data acquisition unit is used to acquire raw vibration data according to a preset sampling frequency; the wireless communication unit is used to transmit the raw vibration data to the data preprocessing module; and the protective housing provides waterproof, dustproof, and impact-resistant protection for the triaxial accelerometer and data acquisition unit to adapt to the mud, impact, and vibration environment during rotary drilling.

[0056] Step S200: Obtain the drilling parameters and drill bit parameters of the rotary drilling rig. The drilling parameters include drilling pressure, rotation speed, torque, drilling rate and drilling depth. The drill bit parameters include drill bit type and borehole diameter.

[0057] Drilling parameters can be read from the rotary drilling rig control system via the drilling parameter acquisition module. These drilling parameters include drilling pressure, rotational speed, torque, drilling rate, and drilling depth. The drilling parameter acquisition module can establish a communication connection with the rotary drilling rig control system via industrial Ethernet, CAN bus, serial communication, or other communication methods. In one embodiment, the drilling parameter acquisition module communicates with the PLC controller or construction data recording unit in the rotary drilling rig control system via industrial Ethernet and reads the drilling parameters according to the output frequency of the control system.

[0058] Drill bit parameters include drill bit type and borehole diameter. These parameters can be obtained through manual input, preset construction parameters, QR code recognition, reading from an equipment parameter database, or importing from a construction management system. Drill bit types can include various types such as cutting bit barrels, sand dredging buckets, auger bits, and rock drill barrels. Since different drill bit types and borehole diameters affect vibration response and drilling resistance, using drill bit parameters as input parameters for formation identification helps reduce the impact of different drilling tool conditions on the identification results.

[0059] Step S300: Synchronize the raw data of drill pipe vibration with the drilling parameters in time, and perform outlier removal, filtering and noise reduction, and windowing processing on the raw data of drill pipe vibration to obtain the vibration data to be analyzed.

[0060] Since the sampling frequency of vibration data is usually higher than that of drilling parameters, it is necessary to synchronize the vibration data, drilling parameters, and drill bit parameters in time. This time synchronization includes aligning the raw drill pipe vibration data with the drilling parameters based on timestamps, or aligning the raw drill pipe vibration data with the drilling parameters based on a drilling depth range. The windowing process includes extracting data segments according to preset time windows or preset depth windows. In one embodiment, both the vibration data acquisition module and the drilling parameter acquisition module bind timestamps to the acquired data. The data preprocessing module matches the vibration data, drilling parameters, and drill bit parameters within the same time window based on the timestamps, ensuring that each set of vibration data to be analyzed corresponds to a specific time window or drilling depth range.

[0061] In another embodiment, synchronization can also be performed based on the drilling depth. This involves dividing the drilling depth into several depth intervals, for example, each interval being 0.1m, 0.2m, or 0.5m as a depth window. Vibration data collected within each depth window is then correlated with drilling parameters to form a data sample corresponding to that depth interval.

[0062] After time synchronization is completed, the data preprocessing module performs outlier removal, filtering and noise reduction, and windowing processing on the raw triaxial vibration data of the drill pipe. Outlier removal includes removing abnormal data points based on amplitude thresholds, rate of change thresholds, or statistical outlier rules. Specifically, outlier removal may include removing data points exceeding the sensor's measurement range, removing data points with obvious abrupt changes that do not conform to continuous vibration patterns, or correcting abnormal data based on statistical rules such as mean, standard deviation, and interquartile range. Filtering and noise reduction can employ one or more of the following: bandpass filtering, wavelet thresholding, adaptive Kalman filtering, and moving average filtering, to reduce the impact of equipment vibration, drill pipe transmission interference, construction operation disturbances, and environmental noise. Windowing processing can be performed according to preset time windows, such as 1s, 5s, or 10s, or according to preset drilling depth windows. A certain overlap rate can be set between adjacent windows to enhance the continuity of the identification results.

[0063] After the above processing, vibration data to be analyzed is generated. The vibration data to be analyzed includes three vibration data segments corresponding to three axes, and each vibration data segment is associated with the corresponding drilling parameters, drill bit parameters, and drilling depth.

[0064] Step S400: Perform a Fourier transform on each axial vibration signal of the vibration data to be analyzed to obtain the corresponding power spectral density data; extract time-domain features from the vibration data to be analyzed, and extract frequency-domain features from the power spectral density data. The time-domain features include at least one of root mean square value, peak value, peak-to-peak value, kurtosis, and skewness; the frequency-domain features include at least one of frequency band energy proportion, main peak frequency, main peak amplitude, spectral centroid, and spectral kurtosis. These features are used to represent the differences in power spectral density maps when drilling into different formations.

[0065] The data preprocessing module transmits the vibration data to be analyzed to the Fourier transform module. The Fourier transform module performs Fast Fourier Transform (FFT) on the vibration data along the three axes, converting the time-domain vibration signal into a frequency-domain signal. In one embodiment, a 1024-point or 2048-point FFT can be used. Before performing the FFT, a Hanning window, Hamming window, or other window functions can be applied to the vibration data to reduce spectral leakage.

[0066] The Fourier transform module calculates power spectral density data based on the frequency domain signal. This power spectral density data includes the energy distribution of vibration signals along each axis within multiple preset frequency bands. Specifically, the power spectral density data can include frequency points, the power spectral density values ​​corresponding to each frequency point, the frequency corresponding to the main peak, the amplitude of the main peak, and energy distribution information within the preset frequency bands. For vibration data along three axes, the power spectral density data for each of the three axes can be obtained separately, or the power spectral density features of the three axes can be spliced ​​or statistically fused in the subsequent feature fusion stage.

[0067] The feature extraction module extracts time-domain features from the vibration data to be analyzed. These time-domain features may include at least one of the following: root mean square (RMS), peak value, kurtosis, variance, peak-to-peak value, and skewness. The RMS value can be used to reflect the overall energy level of the vibration signal; the peak value and peak-to-peak value can be used to reflect the impact response intensity; the kurtosis can be used to reflect the sudden impact component in the signal; and the variance and skewness can be used to reflect the fluctuation degree and distribution characteristics of the vibration signal.

[0068] The feature extraction module also extracts frequency domain features from the power spectral density data. These frequency domain features may include at least one of the following: peak frequency, peak amplitude, band energy percentage, spectral centroid, and spectral energy entropy. Peak frequency and peak amplitude can be used to reflect the main vibrational frequencies and their intensity; band energy percentage can be used to reflect the energy distribution within different frequency ranges; spectral centroid can be used to reflect the centroid location of the spectral energy distribution; and spectral energy entropy can be used to reflect the complexity of the spectral distribution.

[0069] In one embodiment, the frequency domain can be divided into multiple frequency bands, such as low-frequency, mid-frequency, and high-frequency bands, and the proportion of energy in each frequency band to the total energy can be calculated. Under different geological conditions, the interaction between the drill bit and the formation may vary, which may lead to changes in the distribution of vibration energy in different frequency bands. Therefore, the proportion of energy in each frequency band can be used as one of the important features for formation identification.

[0070] Step S500: The time-domain features, frequency-domain features, drilling parameters, and drill bit parameters are fused to form a training feature vector, and the stratigraphic categories confirmed by geological survey are used as labels to train the random forest classification model.

[0071] During the model training phase, data samples from multiple actual construction scenarios are collected. Each data sample includes raw data of drill pipe triaxial vibration, drilling parameters, drill bit parameters, and corresponding formation category labels. The formation category labels can be confirmed through geological survey data, field sampling, rock and soil core sample analysis, or construction records. Formation categories may include at least one of clay, sand, gravel, rock, and boulders, and may be expanded to other formation categories according to specific engineering needs.

[0072] The drill bit type is incorporated into the training or inference feature vector using either categorical encoding or one-hot encoding. The drilling parameters are smoothed and normalized before being incorporated into the training or inference feature vector. Specifically, before forming the feature vector, the time-domain features, frequency-domain features, and drilling parameters can be normalized or standardized to reduce the impact of different dimensions and numerical ranges on model training. For categorical parameters such as drill bit type, number encoding, one-hot encoding, or other categorical variable encoding methods can be used. Subsequently, the time-domain features, frequency-domain features, drilling parameters, and drill bit parameters within the same time window or the same drilling depth interval are concatenated to form the feature vector.

[0073] In one embodiment, the feature vector can be represented as:

[0074] X = [Time domain characteristics, Frequency domain characteristics, Drilling parameters, Drill bit parameters]

[0075] The time-domain features include the root mean square value, peak value, kurtosis, and other features corresponding to the three axial vibration signals; the frequency-domain features include the main peak frequency and frequency band energy ratio corresponding to the three axial power spectral density data; the drilling parameters include drilling pressure, rotational speed, torque, drilling rate, and drilling depth; and the drill bit parameters include drill bit type and borehole diameter.

[0076] Feature vectors labeled with geological strata categories are used as training data and divided into training, validation, and test sets. The training set is used to train the geological strata classification model, the validation set is used to adjust model parameters, and the test set is used to evaluate the model's recognition performance. In one embodiment, the geological strata classification model is a random forest classification model. The random forest classification model includes multiple decision trees, each trained based on randomly selected data samples and feature subsets, and finally outputs the geological strata category through majority voting or probability averaging. The random forest classification model is suitable for multi-source feature inputs and has a certain degree of resistance to overfitting.

[0077] The training process of the random forest classification model may include: collecting training samples from multiple pile foundation construction sites; labeling and filtering the vibration data, drilling parameters, and drill bit parameters in the training samples; dividing the filtered samples into training set, validation set, and test set; adjusting the number of trees, tree depth, and node splitting parameters of the random forest classification model according to the validation set results; and determining whether the model has reached the preset recognition accuracy threshold according to the test set results.

[0078] In other embodiments, the stratigraphic classification model may also employ a support vector machine model, a gradient boosting tree model, an extreme random tree model, or a neural network classification model. The specific model type can be selected based on the scale of the training data, computational resources, and real-time requirements in the field.

[0079] Step S600: During the construction identification stage, the time-domain features, frequency-domain features, drilling parameters, and drill bit parameters acquired in real time are fused to form an inference feature vector. The inference feature vector is then input into a trained random forest classification model to output the formation category corresponding to the current drilling depth.

[0080] During the real-time identification phase at the construction site, the vibration data acquisition module continuously collects raw triaxial vibration data of the drill pipe, the drilling parameter acquisition module continuously reads drilling pressure, rotational speed, torque, drilling rate, and drilling depth, and the drill bit parameter acquisition module provides the current drill bit type and borehole diameter. The data preprocessing module performs time synchronization, outlier removal, filtering and noise reduction, and windowing processing in the same or corresponding manner as the training phase. The Fourier transform module generates the corresponding power spectral density data, the feature extraction module extracts time-domain and frequency-domain features, and the identification module inputs the real-time generated feature vectors into the trained formation classification model.

[0081] The formation classification model outputs the formation category corresponding to the current drilling depth. The output can be a single formation category or include probability values ​​for each formation category. When the output results fluctuate across multiple consecutive time windows or depth windows, the identification module can use sliding window voting, probability smoothing, or adjacent depth result correction methods to improve the continuity of the identification results. Finally, the identification result is bound to the current drilling depth to form a depth-formation identification result.

[0082] Step S700: During the construction process, real-time acquisition of drill rod vibration data, drilling parameters and drill bit parameters is performed. Time domain features and frequency domain features are extracted from the vibration data to be analyzed. The time domain features, frequency domain features, drilling parameters and drill bit parameters are fused to form an inference feature vector.

[0083] Step S800 involves inputting the inferred feature parameters into the trained random forest classification algorithm model and outputting the stratigraphic category corresponding to the current drilling depth. When the output stratigraphic category differs from the stratigraphic category of the previous identification window, and multiple consecutive identification windows point to the same new stratigraphic category, a stratigraphic abrupt change warning is generated.

[0084] Step S900: Display and store the stratigraphic identification results.

[0085] The identification module transmits the formation identification results to the display module. The display module can display the current drilling depth, drilling pressure, rotation speed, torque, drilling rate, power spectral density curve or spectrum, and the identified formation type in real time. The display module can also store vibration data, drilling parameters, feature extraction results, and formation identification results for subsequent construction analysis and quality traceability.

[0086] When there is a sudden change in the geological stratum type, abnormal identification results, or drilling parameters exceed the preset threshold, the display module can issue an audible and visual alarm or an interface prompt to remind construction personnel to adjust drilling parameters such as drilling pressure, rotation speed, and torque in a timely manner, thereby improving the safety and hole quality of rotary drilling construction.

[0087] As can be seen from the above embodiments, existing automatic formation type identification methods only collect vibration data from the power head during vibration data acquisition. Furthermore, the vibration generated by the interaction between the formation and the drill bit needs to be transmitted to the power head through the drill pipe, during which the vibration signal attenuates, resulting in the power head vibration data not significantly reflecting formation characteristics. This invention collects raw triaxial vibration data of the drill pipe near the drill bit at the lower part of the drill pipe, making the vibration signal acquisition location closer to the direct interaction area between the drill bit and the formation. Compared to only collecting equipment vibration signals from locations far from the drill bit, this method can more directly obtain the vibration response generated during the drill bit's rock breaking, cutting, or soil disturbance, which is beneficial for improving the ability of vibration characteristics to characterize formation changes.

[0088] The drill rod vibration data collected in this invention consists of vibration data along three mutually perpendicular axes. Each axis's vibration data is preprocessed, subjected to Fourier transform, and analyzed by power spectral density. The triaxial vibration data reflects the impact, friction, and fracturing responses generated by the interaction between the drill bit and the formation from different directions. The power spectral density data further characterizes the distribution of vibration energy across different frequency ranges, thereby improving the ability to distinguish between different formation states such as clay, sand, gravel, rock, and boulders.

[0089] Before feature extraction, this invention synchronizes the raw triaxial vibration data of the drill rod with the drilling parameters in time. It also performs outlier removal, filtering, noise reduction, and windowing on the raw vibration data, ensuring that each set of vibration data, drilling parameters, and drill bit parameters corresponds to a defined time window or drilling depth range. This helps reduce the impact of equipment vibration, drill rod transmission interference, construction operation disturbances, and environmental noise on the identification results. Simultaneously, this invention extracts time-domain and frequency-domain features from the vibration data and power spectral density data to be analyzed, respectively. This comprehensively reflects the amplitude changes, impact intensity, and frequency energy distribution characteristics of the vibration signal, thereby improving the ability to distinguish different strata such as clay, sand, gravel, rock, and boulders. Furthermore, this invention not only utilizes drill rod vibration signals to extract time-domain and frequency-domain features but also integrates drilling parameters such as drill pressure, rotational speed, torque, drilling rate, and drilling depth, as well as drill bit parameters such as drill bit type and borehole diameter. Drill pipe vibration characteristics can reflect the dynamic response generated by the interaction between the drill bit and the formation; drilling parameters can reflect the stress state and construction response during drilling; and drill bit parameters can reflect the influence of different drilling tool conditions on vibration and drilling status. By fusing these multi-source features, more comprehensive formation identification information can be obtained than by relying solely on vibration signals or drilling parameters.

[0090] The device of this invention has a simple structure and is easy to install. It can communicate with the control system of a rotary drilling rig and achieves data acquisition, processing, identification, and display through a vibration data acquisition module, a drilling parameter acquisition module, a drill bit parameter acquisition module, a data preprocessing module, a Fourier transform module, a feature extraction module, an identification module, and a display module. The device has a clear structure, which facilitates integration with existing rotary drilling rigs. The vibration data acquisition module can transmit data wirelessly, which helps to reduce the interference of wired lines on drilling operations and improves the convenience of on-site installation and use.

[0091] Example 2

[0092] See Figures 3 to 7 This embodiment provides a formation identification device based on the vibration and drilling data of a rotary drilling rig drill rod. This device is used to implement the formation identification method described in Embodiment 1. The device includes a vibration data acquisition module, a drilling parameter acquisition module, a drill bit parameter acquisition module, a data preprocessing module, a Fourier transform module, a feature extraction module, an identification module, and a display module.

[0093] The vibration data acquisition module is located at the lower part of the drill pipe near the drill bit, and is used to collect raw vibration data of the drill pipe in three mutually perpendicular axes. The vibration data acquisition module may include a triaxial accelerometer, a signal conditioning unit, a data acquisition unit, a wireless communication unit, and a protective housing. The triaxial accelerometer is used to sense axial and two radial vibration signals of the drill pipe; the signal conditioning unit is used to condition the sensor signals; the data acquisition unit is used to collect and buffer raw vibration data; the wireless communication unit is used to send the vibration data to the data preprocessing module; and the protective housing is used to protect the internal sensors and circuitry. The wireless communication unit can use Bluetooth, WiFi, LoRa, ZigBee, or other industrial wireless communication methods. In one embodiment, an industrial-grade Bluetooth communication module can be used to reduce interference from wired lines during drilling operations.

[0094] The drilling parameter acquisition module establishes a communication connection with the rotary drilling rig control system and acquires drilling pressure, rotational speed, torque, drilling rate, and drilling depth. The module can read drilling parameters from the rotary drilling rig control system via industrial Ethernet, CAN bus, or serial communication. The read drilling parameters are then transmitted to the data preprocessing module and the identification module.

[0095] The drill bit parameter acquisition module is used to obtain the drill bit type and borehole diameter. This module can be a manual input interface, a parameter reading interface, or a construction database call unit. When construction personnel change the drill bit or adjust the borehole diameter, they can update the current drill bit parameters through the drill bit parameter acquisition module, enabling the identification module to identify the formation based on the current drilling conditions.

[0096] The data preprocessing module is used to synchronize the raw vibration data with drilling parameters in time, and to perform outlier removal, filtering and noise reduction, and windowing processing on the raw vibration data to form the vibration data to be analyzed. The data preprocessing module may include a time synchronization unit, an outlier handling unit, a filtering and noise reduction unit, and a windowing processing unit. The time synchronization unit matches the vibration data and drilling parameters based on timestamps or drilling depths; the outlier handling unit removes or corrects abnormal vibration data; the filtering and noise reduction unit reduces interference signals; and the windowing processing unit divides the continuous vibration data into data windows suitable for Fourier transform and feature extraction.

[0097] The Fourier transform module performs Fourier transforms on the vibration data to be analyzed and generates power spectral density data based on the Fourier transform results. The Fourier transform module can process the vibration data along three axes separately and output the corresponding power spectral density data. This power spectral density data can be transmitted to the feature extraction module for extracting frequency domain features, or to the display module for displaying power spectral density curves or graphs.

[0098] The feature extraction module is used to extract time-domain and frequency-domain features from the vibration data and power spectral density data to be analyzed. The feature extraction module can extract time-domain features such as root mean square value, peak value, kurtosis, variance, peak-to-peak value, and skewness from the vibration data to be analyzed, and frequency-domain features such as main peak frequency, main peak amplitude, frequency band energy proportion, spectral centroid, and spectral energy entropy from the power spectral density data. The extracted time-domain and frequency-domain features are then transmitted to the recognition module.

[0099] The identification module fuses time-domain features, frequency-domain features, drilling parameters, and drill bit parameters into a feature vector, inputs this feature vector into a trained formation classification model, and outputs the formation category corresponding to the current drilling depth. The identification module may include a feature fusion unit, a model storage unit, and a model inference unit. The feature fusion unit normalizes, encodes, and concatenates multi-source features; the model storage unit stores the trained formation classification model; and the model inference unit invokes the formation classification model and outputs the identification result. In one embodiment, the formation classification model is a random forest classification model.

[0100] The display module is used to display formation type, drilling parameters, and power spectral density data. It can be installed in the rotary drilling rig's control room and includes a display screen, a data storage unit, a data export interface, and an alarm unit. The display screen shows the current drilling depth, drill pressure, rotational speed, torque, drilling rate, power spectral density spectrum, and formation type; the data storage unit stores identification data generated during the drilling process; the data export interface exports data to external storage devices; and the alarm unit issues alerts when there are sudden changes in formation, abnormal drilling parameters, or abnormal data quality.

[0101] In one embodiment, the device further includes a power supply module. The power supply module supplies power to the vibration data acquisition module, drilling parameter acquisition module, drill bit parameter acquisition module, data preprocessing module, Fourier transform module, feature extraction module, identification module, and display module. The power supply module can be powered by the rotary drilling rig's onboard power supply or by an independent battery. The power supply module may also include surge protection, short-circuit protection, and voltage stabilization protection units to adapt to the power fluctuation environment at the construction site.

[0102] With the aforementioned device, the rotary drilling rig can collect triaxial vibration data and drilling parameters in real time near the drill bit position of the drill rod during construction. Through preprocessing, frequency domain analysis, feature extraction, and model recognition, it can determine the formation type corresponding to the current drilling depth. This device can be used in conjunction with existing rotary drilling rig control systems, facilitating formation identification without significantly altering the original equipment structure.

[0103] As can be seen from the above embodiments, existing automatic formation type identification methods only collect vibration data from the power head during vibration data acquisition. Furthermore, the vibration generated by the interaction between the formation and the drill bit needs to be transmitted to the power head through the drill pipe. During this transmission, the vibration signal attenuates, resulting in the power head vibration data not significantly reflecting the formation characteristics. This invention, by collecting and processing drill pipe vibration data, obtains analysis data that is only affected by the drill bit model and formation type. Research has confirmed that when different drill bits drill into different formations, the power spectral density map corresponding to the analysis data will exhibit unique abrupt changes. Based on this, the time-domain and frequency-domain features representing these abrupt changes in the power spectral density map can be used to reflect the differences between the drill bit model and the formation type. Given a fixed drill bit parameter, the formation type can be accurately identified using the time-domain and frequency-domain features corresponding to these abrupt changes. Compared to existing methods that directly characterize stratigraphic categories through data, this invention employs an indirect data characterization approach. Specifically, it identifies stratigraphic layers by combining abrupt changes in power spectral density maps with time-domain and frequency-domain features. This significantly reduces the types of data required for stratigraphic identification, simultaneously improving accuracy, efficiency, and success rate. Existing automatic stratigraphic identification methods often suffer from single-source data dependence. Whether relying solely on drilling parameters or a single vibration signal, these methods suffer from limited information and susceptibility to interference, leading to incomplete feature extraction and a difficulty in balancing accuracy and efficiency. Some solutions, while increasing data volume, create redundancy, further reducing identification efficiency and success rate. Power; This invention integrates time-domain and frequency-domain features, drilling parameters, and drill bit parameters extracted from drill pipe vibration data, eliminates irrelevant interference, and retains only the effective data of the interaction between the drill bit and the formation. It utilizes the complementarity of multi-source data to comprehensively reflect the formation characteristics, ensuring recognition accuracy without redundant data, while also considering recognition efficiency and success rate; This invention clearly divides the model training and model inference into two core processes, standardizes the requirements for training data collection, the proportion of dataset division, and the accuracy standard, and adopts the random forest classification algorithm, which has the advantages of anti-overfitting and strong ability to process high-dimensional data, adapts to complex construction conditions, and is convenient for real-time application in actual construction.

[0104] The device of this invention has a simple structure and is easy to install. The vibration data acquisition module adopts wireless communication, avoiding the interference of wired lines during construction. Each module is compatible with the existing rotary drilling rig control system, requiring no large-scale modification and resulting in low cost. The display module can provide real-time feedback on the identification results and trigger abnormal alarms, facilitating timely adjustment of drilling parameters by operators, thereby improving construction efficiency, drilling quality, and construction safety.

[0105] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only.

[0106] It should be understood that the above description is only a preferred embodiment of the present invention, and the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. For those skilled in the art, adjustments or substitutions can be made to the module structure, communication method, sampling parameters, filtering method, feature type, and classification model in the above embodiments without departing from the technical concept of the present invention, and such adjustments or substitutions should all fall within the protection scope of the present invention.

Claims

1. A method for formation identification based on rotary drilling rig drill rod vibration and drilling data, characterized in that, include: The raw data of drill rod vibration during the drilling process of rotary drilling rig are obtained. The raw data of drill rod vibration are vibration signals of three mutually perpendicular axes collected by a triaxial accelerometer set at the lower part of the drill rod near the drill bit. Obtain the drilling parameters and drill bit parameters of the rotary drilling rig. The drilling parameters include drilling pressure, rotation speed, torque, drilling rate and drilling depth. The drill bit parameters include drill bit type and borehole diameter. The raw data of drill pipe vibration is synchronized with the drilling parameters in time, and the raw data of drill pipe vibration is subjected to outlier removal, filtering and noise reduction and windowing to obtain the vibration data to be analyzed. Fourier transform is performed on each axial vibration signal of the vibration data to be analyzed to obtain the corresponding power spectral density data; time-domain features are extracted from the vibration data to be analyzed, and frequency-domain features are extracted from the power spectral density data. The time-domain features include at least one of root mean square value, peak value, peak-to-peak value, kurtosis, and skewness; the frequency-domain features include at least one of frequency band energy proportion, main peak frequency, main peak amplitude, spectral centroid, and spectral kurtosis. These features are used to represent the differences in power spectral density maps when drilling into different formations. The time-domain features, frequency-domain features, drilling parameters, and drill bit parameters are fused to form a training feature vector, and the stratigraphic categories confirmed by geological surveys are used as labels to train a random forest classification model. During the construction identification phase, the time-domain features, frequency-domain features, drilling parameters, and drill bit parameters acquired in real time are fused to form an inference feature vector, and the inference feature vector is input into a trained random forest classification model to output the formation category corresponding to the current drilling depth. During construction, drill rod vibration data, drilling parameters and drill bit parameters are acquired in real time. Time domain features and frequency domain features are extracted from the vibration data to be analyzed. The time domain features, frequency domain features, drilling parameters and drill bit parameters are fused to form an inference feature vector. The inference feature parameters are input into the trained random forest classification algorithm model, and the output is the stratum category corresponding to the current drilling depth.

2. The stratigraphic identification method according to claim 1, characterized in that, The triaxial accelerometer is installed inside the vibration data acquisition module, which is fixedly installed on the bottom side wall of the drill pipe and maintains a preset distance from the drill bit. The triaxial accelerometer has a range of ±50g, a frequency response range of 0.5kHz to 10kHz, and a sampling frequency not lower than the acquisition frequency of the drilling parameters.

3. The stratigraphic identification method according to claim 1, characterized in that, The time synchronization includes aligning the original drill pipe vibration data with the drilling parameters based on timestamps, or aligning the original drill pipe vibration data with the drilling parameters based on a drilling depth range; the windowing process includes extracting data segments according to a preset time window or a preset depth window.

4. The stratigraphic identification method according to claim 1, characterized in that, The filtering and noise reduction includes at least one of bandpass filtering, wavelet thresholding, adaptive Kalman filtering, and moving average filtering; the outlier removal includes removing outlier data points based on amplitude thresholds, rate of change thresholds, or statistical outlier rules.

5. The stratigraphic identification method according to claim 1, characterized in that, The Fourier transform is a 1024-point or 2048-point fast Fourier transform, and the power spectral density data includes the energy distribution of each axial vibration signal within multiple preset frequency bands.

6. The stratigraphic identification method according to claim 1, characterized in that, The drill bit type is incorporated into the training feature vector or the inference feature vector using either category coding or one-hot coding. The drilling parameters are smoothed and normalized before being incorporated into the training feature vector or the inference feature vector.

7. The stratigraphic identification method according to claim 1, characterized in that, The training process of the random forest classification model includes: collecting training samples from multiple pile foundation construction sites; labeling and filtering the vibration data, drilling parameters, and drill bit parameters in the training samples; dividing the filtered samples into training set, validation set, and test set; adjusting the number of trees, tree depth, and node splitting parameters of the random forest classification model according to the validation set results; and determining whether the model has reached the preset recognition accuracy threshold according to the test set results.

8. The stratigraphic identification method according to claim 1, characterized in that, The stratigraphic category includes at least one of clay, sand, gravel, rock, and boulders; when the output stratigraphic category is inconsistent with the stratigraphic category of the previous identification window, and multiple consecutive identification windows point to the same new stratigraphic category, a stratigraphic abrupt change prompt is generated.

9. A formation identification device based on rotary drilling rig drill rod vibration and drilling data according to any one of claims 1 to 8, characterized in that, include: The vibration data acquisition module is located at the lower part of the drill pipe near the drill bit and is used to collect raw vibration data of the drill pipe in three mutually perpendicular axes. The drilling parameter acquisition module is used to communicate with the rotary drilling rig control system and acquire drilling pressure, rotation speed, torque, drilling rate and drilling depth. The drill bit parameter acquisition module is used to obtain the drill bit type and borehole diameter; The data preprocessing module is used to synchronize the original vibration data with the drilling parameters in time, and to perform outlier removal, filtering and noise reduction, and windowing processing on the original vibration data to form vibration data to be analyzed. The Fourier transform module is used to perform Fourier transform on the vibration data to be analyzed and generate power spectral density data based on the Fourier transform result. The feature extraction module is used to extract time-domain features and frequency-domain features based on the vibration data to be analyzed and the power spectral density data; The identification module is used to fuse the time-domain features, frequency-domain features, drilling parameters and drill bit parameters into a feature vector, and input the feature vector into a trained formation classification model to output the formation category corresponding to the current drilling depth. The display module is used to display the formation type, drilling parameters, and power spectral density data.

10. The stratigraphic identification device according to claim 9, characterized in that, The vibration data acquisition module includes a triaxial accelerometer, a signal conditioning unit, a microcontroller, a wireless communication unit, and a protective housing; the wireless communication unit is used to send the acquired vibration data to the data preprocessing module. The drilling parameter acquisition module is connected to the rotary drilling rig control system via industrial Ethernet, CAN bus or serial communication interface; the display module is located in the rotary drilling rig control room and is used to trigger alarm prompts when there is a sudden change in the formation type or data abnormality. It also includes a data storage and export module and a power supply module; the data storage and export module is used to store and export vibration data, drilling parameters, feature vectors, identification results and power spectral density data, and the power supply module is used to provide vehicle power supply adaptation and power protection for each module.

Images