Drill bit wear detection and remaining life prediction method and apparatus

Abstract

This invention discloses a method and apparatus for drill bit wear detection and remaining life prediction. The method includes: acquiring multimodal data of the drill bit; performing time-series alignment of the multimodal data using a dynamic time warping algorithm; inputting the time-series aligned multimodal data of the drill bit into an intelligent analysis center and outputting the remaining life prediction result of the target drill bit; using a visual prediction model to segment the wear region of the target drill bit; using a dynamic rating algorithm to classify the wear region; and using a digital twin simulation model, which is a multi-scale simulation system driven by digital twin simulation technology based on wear region segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data, which can improve the detection accuracy, real-time performance, model generalization, and life prediction capability of drill bit wear.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology for oil drilling equipment, and in particular to a method and device for detecting drill bit wear and predicting remaining life. Background Technology

[0002] This section is intended to provide background or context for embodiments of the present invention. The description herein is not intended to imply that it is prior art simply because it is included in this section.

[0003] In the field of intelligent detection technology for oil drilling equipment, in order to achieve real-time detection of drill bit wear, health status classification and remaining life prediction, existing technologies typically employ manual visual inspection, traditional image processing technology and convolutional neural networks.

[0004] However, the aforementioned existing technologies suffer from one or more of the following drawbacks in practical applications: First, insufficient detection accuracy. Manual visual inspection relies on standard calipers, resulting in a measurement error of up to 23.6% for wear on the nose of PDC (Polycrystalline Diamond Compact) drill bits. Traditional image processing techniques, limited by two-dimensional imaging, cannot quantify the volume loss rate of diamond composites. Second, poor real-time performance. Detection systems require disassembling the drill bit for laboratory testing, with a single test taking over 8 hours; alternatively, fixed industrial cameras can only trigger a scan every 2 meters of drilling. Furthermore, the existing models have weak generalization capabilities. Convolutional neural network (CNN) schemes require training with over 5000 labeled samples, necessitating re-labeling for new drill bit types. Moreover, existing AI models have a false alarm rate of 37.2% for interference from mud adhesion and metal reflection. Finally, a lack of predictive ability. Traditional mean squared error (MSE) (mechanical specific energy) models can only reflect macroscopic wear trends and cannot predict sudden tooth breakage; digital twin system updates are delayed by more than 24 hours, leading to a mismatch between the simulation and the physical drill bit condition.

[0005] Therefore, improving the detection accuracy, real-time performance, model generalization, and life prediction capabilities of drill bit wear is a technical challenge currently faced by those skilled in the art. Summary of the Invention

[0006] This invention provides a method for drill bit wear detection and remaining life prediction, which improves the detection accuracy, real-time performance, model generalization, and life prediction capability of drill bit wear. The method includes:

[0007] Acquire drill bit multimodal data; drill bit multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of drill bits of the same type; historical life data of drill bits of the same type refers to the historical life data of drill bits of the same type as the target drill bit;

[0008] A dynamic time warping algorithm is used to perform time-series alignment on drill bit multimodal data.

[0009] The time-aligned multimodal data of the drill bit is input into the intelligent analysis center, which outputs the remaining life prediction results of the target drill bit. The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The visual prediction model is a large-scale visual model combining the CLIP (Contrastive Language-Image Pre-training) model and the SAM (Segment Anything Model) model. The visual prediction model is used to segment the wear area of ​​the target drill bit and generate wear area segmentation results. The dynamic rating algorithm is used to classify the wear area segmentation results and generate classification results. The digital twin simulation model is a multi-scale simulation system driven by digital twin simulation technology using wear area segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data.

[0010] This invention also provides a drill bit wear detection and remaining life prediction device to improve the detection accuracy, real-time performance, model generalization, and life prediction capability of drill bit wear. The device includes:

[0011] The acquisition module is used to acquire drill bit multimodal data; the drill bit multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of drill bits of the same type; the historical life data of drill bits of the same type is the historical life data of drill bits of the same type as the target drill bit;

[0012] The timing alignment module is used to perform timing alignment of drill bit multimodal data using a dynamic time warping algorithm.

[0013] The output module is used to input time-aligned multimodal data of the drill bit into the intelligent analysis center and output the remaining life prediction results of the target drill bit. The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The visual prediction model is a large-scale visual model combining the CLIP model and the SAM model. The visual prediction model is used to segment the wear area of ​​the target drill bit and generate wear area segmentation results. The dynamic rating algorithm is used to classify the wear area segmentation results and generate classification results. The digital twin simulation model is a multi-scale simulation system driven by digital twin simulation technology using wear area segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data.

[0014] Compared with existing technologies for drill bit wear detection and remaining life prediction, this invention acquires multimodal data of the drill bit. This multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of similar drill bits. The historical life data of similar drill bits refers to the historical life data of drill bits of the same type as the target drill bit. A dynamic time warping algorithm is used to perform time-series alignment on the multimodal data. The time-series aligned multimodal data is then input into an intelligent analysis center, which outputs the remaining life prediction result of the target drill bit. The intelligent analysis center includes a visual prediction model and a pre-established dynamic evaluation system. The system includes a multi-scale algorithm and a digital twin simulation model; a visual prediction model that combines the CLIP and SAM models; a visual prediction model used to segment the wear area of ​​the target drill bit and generate wear area segmentation results; a dynamic rating algorithm used to classify the wear area segmentation results and generate classification results; and a digital twin simulation model that is a multi-scale simulation system driven by digital twin simulation technology using wear area segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data, which can improve the detection accuracy, real-time performance, model generalization, and life prediction capabilities of drill bit wear. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0016] Figure 1 This is a flowchart of a drill bit wear detection and remaining life prediction method provided in an embodiment of the present invention;

[0017] Figure 2 This is a schematic diagram of a drill bit wear detection and remaining life prediction system provided in an embodiment of the present invention;

[0018] Figure 3 This is a schematic diagram of a drill bit wear detection and remaining life prediction device provided in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of a specific example of a drill bit wear detection and remaining life prediction device provided in an embodiment of the present invention;

[0020] Figure 5 This is a schematic diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0022] The acquisition, storage, use, and processing of data in this application all comply with relevant regulations.

[0023] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0024] In the description of this specification, the terms "comprising," "including," "having," and "containing" are open-ended terms, meaning that they include but are not limited to. The terms "an embodiment," "a specific embodiment," "some embodiments," and "for example," etc., refer to specific features, structures, or characteristics described in connection with that embodiment or example that are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. The order of steps involved in the various embodiments is used to illustrate the implementation of this application, and the order of steps is not limited and can be adjusted appropriately as needed.

[0025] Due to the shortcomings of existing technologies, such as insufficient detection accuracy, poor real-time performance, weak model generalization, and lack of predictive ability, a well in a certain year failed to detect a microcrack (0.15mm deep) on the nose of a φ215.9mm PDC drill bit in a timely manner, resulting in the collapse of three cutting teeth during drilling and a direct economic loss of 4.5 million yuan.

[0026] To address the aforementioned issues, this invention aims to provide a drill bit lifecycle management method that integrates high-precision 3D visual scanning, multi-physics dynamic sensing, zero-shot transfer learning, and digital twin simulation. It is applicable to real-time monitoring of drill bit wear, health status grading, and remaining life prediction under complex operating conditions such as onshore drilling, offshore platforms, and ultra-deep wells.

[0027] Figure 1 This is a flowchart of a drill bit wear detection and remaining life prediction method provided in an embodiment of the present invention, such as... Figure 1 As shown, the method may include:

[0028] Step 101: Obtain drill bit multimodal data; drill bit multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of drill bits of the same type; historical life data of drill bits of the same type refers to the historical life data of drill bits of the same type as the target drill bit.

[0029] Step 102: Perform time-series alignment of drill bit multimodal data using a dynamic time warping algorithm;

[0030] Step 103: Input the time-aligned multimodal data of the drill bit into the intelligent analysis center and output the remaining life prediction result of the target drill bit. The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The visual prediction model is a large-scale visual model combining the CLIP model and the SAM model. The visual prediction model is used to segment the wear area of ​​the target drill bit and generate wear area segmentation results. The dynamic rating algorithm is used to classify the wear area segmentation results and generate classification results. The digital twin simulation model is a multi-scale simulation system driven by digital twin simulation technology using wear area segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data.

[0031] The embodiments of the present invention can improve the efficiency and quality of work scenario configuration through the above steps, avoid resource waste, meet users' personalized needs, and improve user experience.

[0032] In one embodiment, in step 101, drill bit multimodal data is acquired. The drill bit multimodal data may include downhole data, point cloud data of the target drill bit, sensor data, and historical lifespan data of similar drill bits. Point cloud data of the drill bit surface is acquired using a high-precision 3D scanner employing blue light structured light scanning technology, with a resolution of 2048×2048, a scanning frame rate ≥120fps, a point cloud density ≥200 points / mm², and an accuracy ≤5μm. The 3D scanner includes high-temperature resistant optical components: a sapphire protective window (transmittance >92%@500nm) + an active cooling system (temperature control accuracy ±0.5℃).

[0033] In one embodiment, the sensor data includes torque data acquired by a mechanical sensor and acoustic emission signals acquired by an acoustic emission array. The mechanical sensor can be a six-axis force sensor with a measurement range of 0-50 kN·m for torque and 0-200 kN for axial force. The acoustic emission array has a frequency range of 10 kHz to 1 MHz and a sensitivity of -65 dB.

[0034] Step 101 provides a wealth of multi-dimensional drill bit data, offering a reliable foundation for subsequent analysis and prediction.

[0035] In one embodiment, after acquiring the drill bit multimodal data, the method may further include: denoising the point cloud data of the target drill bit using a moving least squares (MLS) filtering algorithm improved by an adaptive parameter adjustment mechanism and a weighting function; and registering the denoised point cloud data of the target drill bit using an NDT point cloud registration algorithm based on path planning.

[0036] In this embodiment, the improved moving least squares filtering algorithm mainly enhances the accuracy and efficiency of point cloud denoising by introducing an adaptive parameter adjustment mechanism and optimizing the weighting function. This improvement effectively overcomes the shortcomings of the standard MLS algorithm in terms of noise sensitivity and detail loss in drill bit point cloud data, thereby improving the reliability of subsequent registration and analysis.

[0037] In one embodiment, after acquiring the drill bit multimodal data, the method may further include: synchronously acquiring the time-frequency characteristics of torque, axial force, and acoustic emission signals.

[0038] In one embodiment, step 102, which involves temporally aligning the drill bit multimodal data using a dynamic time warping algorithm, may include: aligning sensor data, historical lifespan data of similar drill bits, and point cloud data of the registered target drill bit using the dynamic time warping algorithm. The dynamic time warping (DTW) algorithm can be used to align the drill bit multimodal data. This step eliminates temporal deviations caused by different sampling frequencies or transmission delays, improving data availability.

[0039] First, an improved moving least squares filtering algorithm is used to denoise the point cloud data of the target drill bit to reduce the impact of noise on subsequent analysis. Then, the denoised point cloud data is registered using a path planning-based NDT point cloud registration algorithm to ensure data consistency and accuracy. These denoising and registration processes significantly improve the quality of the point cloud data, thereby enhancing the accuracy and reliability of subsequent analysis. Finally, the Dynamic Time Warping (DTW) algorithm is used to perform time-series alignment of the registered point cloud data, sensor data (such as torque data and acoustic emission signals), and historical lifespan data of similar drill bits, eliminating time-series biases from multiple data sources.

[0040] In one embodiment, in step 103, the time-aligned drill bit multimodal data is input into the intelligent analysis center, and the remaining life prediction result of the target drill bit is output.

[0041] The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The input-output relationship among the three is as follows: the output of the visual prediction model (wear area segmentation) is directly used as the input of the dynamic rating algorithm; the output of the dynamic rating algorithm (grading results) is used as one of the inputs of the digital twin simulation model, combined with other data (such as sensor data) to drive simulation and prediction.

[0042] First, we introduce the visual prediction model, which is a large visual model combining the CLIP model and the SAM model, used to segment the wear area of ​​the target drill bit.

[0043] In one embodiment, the CLIP model extracts visual features from point cloud data using a first image encoder and semantically annotates these visual features using preset text prompts; the SAM model performs zero-sample segmentation using the visual features extracted by the CLIP model using a second image encoder and outputs the wear area of ​​the target drill bit through the SAM model's decoder; the first and second image encoders share weights; the visual prediction model fine-tunes its parameters using LoRA fine-tuning technology, drill bit multimodal data, and user-input prompts; the visual prediction model is trained using drill bit multimodal data and the preset text prompts during fine-tuning.

[0044] In one embodiment, the visual prediction model is based on a combination of the CLIP (Contrastive Language-Image Pre-trained) model and the SAM (Segment All Model). The CLIP model extracts visual features from point cloud data through an image encoder and guides semantic attention with textual prompts (such as 'wear area'). The SAM model uses the features extracted by CLIP to perform zero-shot segmentation and generate accurate wear area masks. The combination method is as follows: the CLIP image encoder and the SAM image encoder share weights, global and local features are integrated through a feature fusion module (such as an attention mechanism), and finally the segmentation result is output through the SAM decoder.

[0045] In one embodiment, the visual prediction model is a zero-shot transfer learning model, segmenting the wear region, with the number of parameters fine-tuned being less than 0.5%. The visual prediction model utilizes LoRA (Low-Rank Adaptation) fine-tuning technology, drill bit multimodal data, and user-input prompts to fine-tune the model parameters. Only a portion of the intermediate layer parameters (less than 0.5%) are updated. Fine-tuning is performed using drill bit multimodal data and user-input prompts to adapt to the specific characteristics of drill bit wear. This step allows for rapid adaptation to the characteristics of different drill bit types, improving the model's generalization ability and prediction accuracy.

[0046] In one embodiment, the dynamic rating algorithm may include: determining the volume loss of the cutting teeth at each location in the wear region based on the point cloud data of the target drill bit; determining the energy integral of the acoustic emission signal at each location in the wear region in a preset frequency band based on the acoustic emission signal; determining the rate of change of the torque fluctuation coefficient at each location in the wear region based on torque data; determining the grading result of each location in the wear region based on the volume loss of the cutting teeth, the energy integral, the rate of change of the torque fluctuation coefficient, and the dynamic weighting coefficient; and adaptively adjusting the dynamic weighting coefficient according to the cumulative working time of the target drill bit.

[0047] In this embodiment, the dynamic rating algorithm specifically includes:

[0048] ;

[0049] Among them, V loss E represents the volume loss of the cutting teeth. AE ΔTWF is the energy integral of the acoustic emission signal in the preset frequency band; V0 is the initial volume of the cutting teeth of the target drill bit type in a brand new state; E0 is the initial reference value of the energy integral of the acoustic emission signal measured in the preset frequency band of the target drill bit type in a brand new state; α(t) is the dynamic weighting coefficient of the volume loss of the cutting teeth; β(t) is the dynamic weighting coefficient of the energy integral; γ(t) is the dynamic weighting coefficient of the rate of change of the torque fluctuation coefficient; and t is the cumulative working time of the target drill bit.

[0050] In this embodiment, the dynamic weighting coefficient is adaptively adjusted according to the following formula:

[0051] ;

[0052] ;

[0053] ;

[0054] Where α(t) is the dynamic weighting coefficient of the volume loss of the cutting teeth; β(t) is the dynamic weighting coefficient of the energy integral; γ(t) is the dynamic weighting coefficient of the rate of change of the torque fluctuation coefficient; and t is the cumulative working time of the target drill bit.

[0055] By dynamically adjusting the weighting coefficients, we can better adapt to drill bit wear under different working conditions, thereby improving the accuracy and robustness of predictions.

[0056] The dynamic rating algorithm dynamically fuses multimodal data to generate a health score, with weighting coefficients adaptively adjusted according to drill bit lifespan. The digital twin simulation model is a multi-scale simulation system driven by real-time point cloud data and sensor data of the target drill bit. Based on the digital twin simulation, it predicts remaining lifespan with a lead time of ≥4 hours. Through the dynamic rating algorithm, a comprehensive and detailed assessment of drill bit wear can be performed, providing a scientific basis for subsequent decision-making.

[0057] By integrating and driving wear zone segmentation results from a visual prediction model, grading results (quantitative scoring) generated by a dynamic rating algorithm, and real-time point cloud and sensor data of the target drill bit, a virtual simulation system synchronously mapped and dynamically interacting with the physical drill bit—a digital twin simulation model—was constructed. This model is essentially a multi-scale simulation system capable of high-fidelity simulation of the drill bit's wear process and mechanical behavior at different scales, from molecular dynamics and discrete element method to finite element method. Through this model, accurate and forward-looking predictions of the drill bit's remaining lifespan are achieved. Its ultimate goal is to transform traditional offline, delayed diagnostics into online, real-time, and early warning-enabled intelligent insights, thereby significantly improving the safety, economy, and decision-making efficiency of drilling operations and solving the core problem of mismatch between simulation and physical systems.

[0058] The visual prediction model utilizes a large-scale visual model combining the CLIP and SAM models to accurately segment the wear areas of the drill bit. A dynamic rating algorithm classifies the wear areas based on various features, while a digital twin simulation model uses real-time data for multi-scale simulation, ultimately outputting a prediction of the remaining life of the target drill bit. This process enables accurate prediction of drill bit life, providing crucial decision support for drilling operations.

[0059] The embodiments of the present invention also provide a visualization interface that displays a three-dimensional wear heat map and uses red, yellow and green to indicate the degree of drill bit wear.

[0060] In one embodiment, the drill bit wear detection and remaining life prediction method may further include: outputting maintenance recommendations based on the remaining life prediction results of the target drill bit and a decision tree generated from historical drill bit maintenance cases. By combining the decision tree with over 3000 historical maintenance cases, more scientific and reasonable maintenance recommendations can be generated, helping operators take timely measures to avoid economic losses caused by drill bit failures. This step can improve the safety and economy of drilling operations.

[0061] Figure 2 This is a schematic diagram of a drill bit wear detection and remaining life prediction system provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the system includes: a data acquisition layer, an edge computing layer, an intelligent analysis layer, and a decision output layer.

[0062] In this embodiment, the data acquisition layer includes downhole data acquisition, a 3D scanner, a force sensor, and an acoustic emission probe. The 3D scanner uses a blue light structured light scanner to acquire point cloud data with an accuracy of 5μm and a resolution of 2048×2048. The blue light structured light scanner uses high-temperature resistant optical components: a sapphire protective window (transmittance >92%@500nm) + an active cooling system (temperature control accuracy ±0.5℃). The force sensor is a six-axis force sensor that measures torque or vibration data, with a range of: torque 0-50kN·m, axial force 0-200kN. The acoustic emission probe outputs a high-frequency signal with a frequency range of 10kHz-1MHz and a sensitivity of -65dB.

[0063] In this embodiment, the edge computing layer is implemented using an edge computing hardware platform, and the key algorithms are: point cloud denoising: an improved moving least squares (MLS) filtering algorithm; and time alignment: multi-source data synchronization based on dynamic time warping (DTW).

[0064] In this embodiment, the intelligent analysis center in the intelligent analysis layer includes a large visual model, a dynamic rating algorithm, and an anomaly detection engine. The large visual model has the following infrastructure: SAM + CLIP visual language model. The large visual model achieves zero-sample adaptation by using LoRA (Low-Rank Adaptation) fine-tuning technology, updating only 0.3% of the parameters.

[0065] In this embodiment, the Python pseudocode for fine-tuning the visual prediction model can be as follows:

[0066] class CustomSAM(nn.Module):

[0067] def __init__(self):

[0068] super().__init__()

[0069] self.sam = sam_model_registry["vit_h"]()

[0070] self.lora_adapter = LoRA(rank=64, layer_ids=[4, 8, 12]) # Fine-tune only the intermediate layers

[0071] def forward(self, x):

[0072] features = self.sam.image_encoder(x)

[0073] adapted_features = self.lora_adapter(features)

[0074] return adapted_features

[0075] In this embodiment, the dynamic rating algorithm can be the following hybrid rating formula:

[0076] ;

[0077] Among them, V loss The volume loss of the cutting teeth is calculated through point cloud reconstruction; E AE ΔTWF represents the energy integral of the acoustic emission signal in the 15-25kHz frequency band; ΔTWF represents the rate of change of the torque fluctuation coefficient. 0.6 is the initial dynamic weighting coefficient for the volume loss of the cutting teeth, which is subsequently dynamically adjusted according to the calculation formula of α(t) above with the cumulative working time of the target drill bit; 0.3 is the initial dynamic weighting coefficient for the energy integral, which is subsequently dynamically adjusted according to the calculation formula of β(t) above with the cumulative working time of the target drill bit; 0.1 is the initial dynamic weighting coefficient for the rate of change of the torque fluctuation coefficient, which is subsequently dynamically adjusted according to the calculation formula of γ(t) above with the cumulative working time of the target drill bit.

[0078] In this embodiment, the anomaly detection engine can perform wear anomaly detection.

[0079] In this embodiment, the decision output layer includes:

[0080] Visualization interface: 3D wear heat map (grading results are displayed in red-yellow-green grading);

[0081] Prediction module: LSTM-based remaining lifetime prediction (error ±3.2 hours);

[0082] Maintenance suggestion library: a decision tree that links 3000+ historical maintenance cases.

[0083] In addition, the drill bit wear detection and remaining life prediction system also includes dynamic model calibration using digital twin simulation and drilling rig control commands using a ground control console.

[0084] In one embodiment, the method and system of the present invention also provide a remaining lifetime prediction scheme based on Long Short-Term Memory (LSTM) networks (error ±3.2 hours).

[0085] 1. Data Acquisition and Preprocessing:

[0086] (1) Data sources can utilize data from the data acquisition layer, including:

[0087] Sensor data: Real-time parameters such as vibration, torque, rotational speed, temperature, pressure, and drilling depth.

[0088] Operating data: Environmental parameters such as formation hardness, drill bit type, and drilling fluid properties.

[0089] Historical life data: failure records and remaining life tags (RUL) for similar drill bits.

[0090] (2) Data preprocessing:

[0091] Missing value handling: interpolation methods (linear interpolation, spline interpolation) or domain knowledge-based filling (such as fixed thresholds for stratigraphic changes).

[0092] Noise filtering: Wavelet Transform, Filter.

[0093] Outlier detection: Isolation Forest, 3σ principle (dynamically adjusting thresholds based on operating conditions).

[0094] Data alignment: When multiple sensors have different sampling frequencies, resampling or dynamic time warping (DTW) is used.

[0095] (3) Feature engineering:

[0096] Time-domain characteristics: mean, variance, peak factor, kurtosis, waveform factor.

[0097] Frequency domain characteristics: FFT spectrum energy distribution, wavelet packet decomposition.

[0098] Time-series characteristics: sliding window statistics (such as mean within the window, trend slope), lag features.

[0099] Domain characteristics: formation lithology classification (e.g., hardness coefficient of sandstone and shale), and the influence of drilling fluid viscosity on wear.

[0100] 2. Data partitioning and sequence construction:

[0101] (1) Sliding window method:

[0102] Convert the time series data into a supervised learning format, and the window length (T) needs to cover the drill bit wear cycle (e.g., T = 50 time steps).

[0103] The output is the Remaining Life Label (RUL) for the current window, which can be generated by degradation curve modeling or physical failure modeling.

[0104] (2) Data partitioning strategy:

[0105] Time-order partitioning: The training set (70%), validation set (15%), and test set (15%) are divided according to time order to avoid future information leakage.

[0106] Stratified sampling: Sampling is performed in strata according to different working conditions or geological types to ensure balanced data distribution.

[0107] 3. LSTM Model Construction:

[0108] (1) Network structure design:

[0109] Input layer: Window length (T) × Feature dimension (N).

[0110] Hidden layers: multi-layer stacked LSTM (e.g., 2-3 layers), number of neurons (64-256), activation function is `tanh`.

[0111] Regularization: Dropout layer (0.2-0.5), L2 regularization.

[0112] Output layer: Fully connected layer (Dense), outputs the remaining lifetime label RUL value (regression task).

[0113] (2) Model optimization:

[0114] Loss function: Weighted mean squared error (WMSE), which assigns higher weights to time points that are close to failure.

[0115] Optimizer: Adam (initial learning rate 1e-3) with learning rate decay (ReduceLROnPlateau).

[0116] Early Stopping: Monitor the loss on the validation set, and set the patience value to 10-20 epochs.

[0117] 4. Model Training and Optimization:

[0118] (1) Hyperparameter optimization:

[0119] Grid search / Bayesian optimization: Adjust window length, number of LSTM layers, Dropout rate, and learning rate.

[0120] Attention mechanism: Introduce Transformer or Self-Attention layers to capture feature contributions at key time points.

[0121] (2) Model fusion:

[0122] Hybrid models: Combining LSTM with convolutional neural networks (CNN) (such as ConvLSTM) to extract spatiotemporal features.

[0123] Physical information embedding: Drilling mechanics equations (such as drill bit wear rate formulas) are added as constraints to the loss function.

[0124] 5. Model Evaluation and Validation:

[0125] (1) Evaluation indicators:

[0126] Regression indicators (RMSE, MAE, R², etc.).

[0127] Classification metrics: Discretize RUL into health statuses (e.g., normal / warning / dangerous) and calculate the F1-Score.

[0128] Engineering metrics: False Alarm Rate and Missed Detection Rate.

[0129] (2) Interpretability analysis:

[0130] SHAP value / LIME: Explains the impact of features on prediction results.

[0131] Gradient analysis: Visualizing the sensitivity of the LSTM hidden state to input features.

[0132] 6. Deployment and Online Forecasting:

[0133] (1) Real-time reasoning:

[0134] Edge computing: Lightweight model deployment (TensorFlow Lite, ONNX) to edge devices on drilling platforms.

[0135] Streaming data processing: Apache Kafka or Flink for real-time processing of sensor data streams.

[0136] 2) Dynamic updates:

[0137] Online learning: Incremental updates of model parameters (e.g., using Elastic Weight Consolidation to prevent catastrophic forgetting).

[0138] Digital twin: Combining drill bit physical model simulation data to improve prediction robustness.

[0139] This invention also proposes a drill bit wear detection and remaining life prediction device, the principle of which is similar to the drill bit wear detection and remaining life prediction method, and will not be described in detail here.

[0140] Figure 3This is a schematic diagram of a drill bit wear detection and remaining life prediction device provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the drill bit wear detection and remaining life prediction device may include:

[0141] The acquisition module 301 is used to acquire drill bit multimodal data; the drill bit multimodal data includes point cloud data of the target drill bit, sensor data and historical life data of drill bits of the same type; the historical life data of drill bits of the same type is the historical life data of drill bits of the same type as the target drill bit;

[0142] The timing alignment module 302 is used to perform timing alignment of drill bit multimodal data using a dynamic time warping algorithm;

[0143] Output module 303 is used to input time-aligned drill bit multimodal data into the intelligent analysis center and output the remaining life prediction result of the target drill bit. The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The visual prediction model is a large-scale visual model combining the CLIP model and the SAM model. The visual prediction model is used to segment the wear area of ​​the target drill bit and generate wear area segmentation results. The dynamic rating algorithm is used to classify the wear area segmentation results and generate classification results. The digital twin simulation model is a multi-scale simulation system driven by digital twin simulation technology using wear area segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data.

[0144] In one embodiment, the drill bit wear detection and remaining life prediction device may further include: a preprocessing module, used for:

[0145] The point cloud data of the target drill bit is denoised using an improved moving least squares filtering algorithm with an adaptive parameter adjustment mechanism and a weighting function.

[0146] The point cloud data of the denoised target drill bit is registered using the path planning-based NDT point cloud registration algorithm.

[0147] Timing alignment module 302 is specifically used for:

[0148] The dynamic time warping algorithm is used to perform time-series alignment of sensor data, historical life data of similar drill bits, and point cloud data of the registered target drill bit.

[0149] In one embodiment, the sensor data includes torque data acquired by a mechanical sensor and acoustic emission signals acquired by an acoustic emission array;

[0150] The dynamic rating algorithm includes:

[0151] Based on the point cloud data of the target drill bit, determine the volume loss of the cutting teeth at each location in the wear area;

[0152] Based on the acoustic emission signal, determine the energy integral of the acoustic emission signal at each location in the wear area within a preset frequency band;

[0153] Based on the torque data, determine the rate of change of the torque fluctuation coefficient at each location in the wear zone;

[0154] The classification result of each position in the wear zone is determined based on the volume loss of the cutting teeth, energy integral, torque fluctuation coefficient change rate, and dynamic weighting coefficient; the dynamic weighting coefficient is adaptively adjusted according to the cumulative working time of the target drill bit.

[0155] In one embodiment, the dynamic weighting coefficient is adaptively adjusted according to the following formula:

[0156] ;

[0157] ;

[0158] ;

[0159] Where α(t) is the dynamic weighting coefficient of the volume loss of the cutting teeth; β(t) is the dynamic weighting coefficient of the energy integral; γ(t) is the dynamic weighting coefficient of the rate of change of the torque fluctuation coefficient; and t is the cumulative working time of the target drill bit.

[0160] In one embodiment, the dynamic rating algorithm specifically includes:

[0161] ;

[0162] Among them, V loss E represents the volume loss of the cutting teeth. AE ΔTWF is the energy integral of the acoustic emission signal in the preset frequency band; V0 is the initial volume of the cutting teeth of the target drill bit type in a brand new state; E0 is the initial reference value of the energy integral of the acoustic emission signal measured in the preset frequency band of the target drill bit type in a brand new state; α(t) is the dynamic weighting coefficient of the volume loss of the cutting teeth; β(t) is the dynamic weighting coefficient of the energy integral; γ(t) is the dynamic weighting coefficient of the rate of change of the torque fluctuation coefficient; and t is the cumulative working time of the target drill bit.

[0163] In one embodiment, the CLIP model extracts visual features from point cloud data using a first image encoder and semantically annotates these features using preset text prompts; the SAM model performs zero-sample segmentation using the visual features extracted by the CLIP model using a second image encoder and outputs the wear area of ​​the target drill bit through the SAM model's decoder; the first and second image encoders share weights; the visual prediction model fine-tunes its parameters using LoRA fine-tuning technology, drill bit multimodal data, and user-input prompts; the visual prediction model is trained using drill bit multimodal data and preset text prompts during fine-tuning.

[0164] Figure 4 This is a schematic diagram of a specific example of a drill bit wear detection and remaining life prediction device provided in an embodiment of the present invention, such as... Figure 4 As shown, in one embodiment, the drill bit wear detection and remaining life prediction device may further include: a maintenance suggestion output module 401, used for:

[0165] Based on the predicted remaining life of the target drill bit and the decision tree generated from historical drill bit maintenance cases, maintenance recommendations are output.

[0166] Compared with existing technologies for drill bit wear detection and remaining life prediction, this invention acquires multimodal data of the drill bit. This multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of similar drill bits. The historical life data of similar drill bits refers to the historical life data of drill bits of the same type as the target drill bit. A dynamic time warping algorithm is used to perform time-series alignment on the multimodal data. The time-series aligned multimodal data is then input into an intelligent analysis center, which outputs the remaining life prediction result of the target drill bit. The intelligent analysis center includes a visual prediction model and a pre-established dynamic evaluation system. The system includes a multi-scale algorithm and a digital twin simulation model; a visual prediction model that combines the CLIP and SAM models; a visual prediction model used to segment the wear area of ​​the target drill bit and generate wear area segmentation results; a dynamic rating algorithm used to classify the wear area segmentation results and generate classification results; and a digital twin simulation model that is a multi-scale simulation system driven by digital twin simulation technology using wear area segmentation results, classification results, real-time point cloud data of the target drill bit, and real-time sensor data, which can improve the detection accuracy, real-time performance, model generalization, and life prediction capabilities of drill bit wear.

[0167] The innovative points and core problems solved by the embodiments of the present invention are shown in Table 1 below:

[0168] Table 1

[0169]

[0170] The technical effects of the embodiments of the present invention compared with those of the prior art are shown in Table 2 below:

[0171] Table 2

[0172]

[0173] This invention also provides a computer device. Figure 5 This is a schematic diagram of a computer device in an embodiment of the present invention. The computer device 500 includes a memory 510, a processor 520, and a computer program 530 stored in the memory 510 and executable on the processor 520. When the processor 520 executes the computer program 530, it implements the above-mentioned drill bit wear detection and remaining life prediction method.

[0174] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described drill bit wear detection and remaining life prediction method.

[0175] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described drill bit wear detection and remaining life prediction method.

[0176] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0177] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0178] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0179] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0180] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting drill bit wear and predicting remaining life, characterized in that, include: Acquire drill bit multimodal data; the drill bit multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of drill bits of the same type; the historical life data of drill bits of the same type refers to the historical life data of drill bits of the same type as the target drill bit. A dynamic time warping algorithm is used to perform time-series alignment on drill bit multimodal data. The time-aligned multimodal data of the drill bit is input into the intelligent analysis center, which outputs the remaining life prediction result of the target drill bit. The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The visual prediction model is a large-scale visual model combining the CLIP model and the SAM model. The visual prediction model is used to segment the wear area of ​​the target drill bit and generate wear area segmentation results. The dynamic rating algorithm is used to classify the wear area segmentation results and generate classification results. The digital twin simulation model is a multi-scale simulation system driven by digital twin simulation technology, which utilizes the wear area segmentation results, the grading results, the real-time point cloud data of the target drill bit, and the real-time data of the sensors.

2. The method as described in claim 1, characterized in that, After acquiring the drill bit multimodal data, the following is also included: The point cloud data of the target drill bit is denoised using an improved moving least squares filtering algorithm with an adaptive parameter adjustment mechanism and a weighting function. The point cloud data of the denoised target drill bit is registered using the path planning-based NDT point cloud registration algorithm. The dynamic time warping algorithm is used to perform time-series alignment of drill bit multimodal data, including: The dynamic time warping algorithm is used to perform time-series alignment of sensor data, historical life data of similar drill bits, and point cloud data of the registered target drill bit.

3. The method as described in claim 1, characterized in that, The sensor data includes torque data collected by the mechanical sensor and acoustic emission signals collected by the acoustic emission array; The dynamic rating algorithm includes: Based on the point cloud data of the target drill bit, determine the volume loss of the cutting teeth at each location in the wear area; Based on the acoustic emission signal, determine the energy integral of the acoustic emission signal at each location in the wear area within a preset frequency band; Based on the torque data, determine the rate of change of the torque fluctuation coefficient at each location in the wear zone; The classification result of each position in the wear area is determined based on the volume loss of the cutting teeth, energy integral, torque fluctuation coefficient change rate, and dynamic weighting coefficient; the dynamic weighting coefficient is adaptively adjusted according to the cumulative working time of the target drill bit.

4. The method as described in claim 3, characterized in that, The dynamic weighting coefficients are adaptively adjusted according to the following formula: ； ； ； Where α(t) is the dynamic weighting coefficient of the volume loss of the cutting teeth; β(t) is the dynamic weighting coefficient of the energy integral; γ(t) is the dynamic weighting coefficient of the rate of change of the torque fluctuation coefficient; and t is the cumulative working time of the target drill bit.

5. The method as described in claim 4, characterized in that, The dynamic rating algorithm specifically includes: ； Among them, V loss E represents the volume loss of the cutting teeth. AE ΔTWF is the energy integral of the acoustic emission signal in the preset frequency band; V0 is the initial volume of the cutting teeth of the target drill bit type in a brand new state; E0 is the initial reference value of the energy integral of the acoustic emission signal measured in the preset frequency band of the target drill bit type in a brand new state; α(t) is the dynamic weighting coefficient of the cutting tooth volume loss; β(t) is the dynamic weighting coefficient of the energy integral; γ(t) is the dynamic weighting coefficient of the torque fluctuation coefficient change rate; and t is the cumulative working time of the target drill bit.

6. The method as described in claim 1, characterized in that, The CLIP model extracts visual features from point cloud data using a first image encoder and semantically annotates these features using preset text prompts. The SAM model performs zero-sample segmentation using the visual features extracted by the CLIP model using a second image encoder and outputs the wear area of ​​the target drill bit through the SAM model's decoder. The first and second image encoders share weights. The visual prediction model fine-tunes its parameters using LoRA fine-tuning technology, drill bit multimodal data, and user-input prompts. The visual prediction model is trained using drill bit multimodal data and the preset text prompts during fine-tuning.

7. The method as described in claim 1, characterized in that, Also includes: Based on the predicted remaining life of the target drill bit and the decision tree generated from historical drill bit maintenance cases, maintenance recommendations are output.

8. A drill bit wear detection and remaining life prediction device, characterized in that, include: The acquisition module is used to acquire drill bit multimodal data; the drill bit multimodal data includes point cloud data of the target drill bit, sensor data, and historical life data of drill bits of the same type; the historical life data of drill bits of the same type refers to the historical life data of drill bits of the same type as the target drill bit. The timing alignment module is used to perform timing alignment of drill bit multimodal data using a dynamic time warping algorithm. The output module is used to input the time-aligned drill bit multimodal data into the intelligent analysis center and output the remaining life prediction result of the target drill bit. The intelligent analysis center includes a visual prediction model, a pre-established dynamic rating algorithm, and a digital twin simulation model. The visual prediction model is a large-scale visual model combining the CLIP model and the SAM model. The visual prediction model is used to segment the wear area of ​​the target drill bit and generate wear area segmentation results. The dynamic rating algorithm is used to classify the wear area segmentation results and generate classification results. The digital twin simulation model is a multi-scale simulation system driven by digital twin simulation technology, which utilizes the wear area segmentation results, the grading results, the real-time point cloud data of the target drill bit, and the real-time data of the sensors.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

11. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 7.

Images