Rock debris type identification method, system and device

Abstract

The invention provides a rock debris type identification method, system and device.The method comprises the steps that real-time rock debris generated in the drilling process of a target well in a target stratum is scanned through CT scanning equipment, and CT scanning data are obtained; extracting internal structure features and external structure features of the real-time rock debris from CT scanning data through a neural network model obtained through pre-training; and through the neural network model obtained through pre-training, based on the internal structure characteristics and the external structure characteristics of the real-time rock debris, carrying out identification so as to obtain the rock debris type of the real-time rock debris. Through the above scheme, efficient, accurate and real-time identification can be performed on the type of the rock debris, especially the rock debris with similar surface features and obvious internal structure difference has higher identification precision, and further, the adaptability of rock debris identification under complex drilling conditions is improved through introduction of the neural network model and fusion of multi-source data.

Description

Technical Field

[0001] This application belongs to the technical field of oil and gas extraction, and particularly relates to a method, system, and device for identifying cuttings types. Background Art

[0002] With the continuous development of oil and gas extraction technologies, cuttings identification has become a very important part of the drilling process. The cuttings generated during drilling are a direct reflection of the downhole formation conditions. Through cuttings, the downhole working conditions and the progress of operations in optimized drilling can be judged. Most existing cuttings identification methods rely on manual on-site observation, or detect and obtain the characteristic information of cuttings through various sensors, cameras and other devices, and finally engineers identify the types of cuttings based on this cuttings information. However, this method cannot accurately capture the morphological characteristics, quantity, volume, and types of cuttings, etc., and consumes a large amount of human resources, with low identification efficiency. Especially when the wellsite conditions change, the existing cuttings identification methods cannot meet the complex and changeable underground conditions.

[0003] Furthermore, during the drilling process, geological personnel directly observe the cuttings and classify and identify different types of cuttings based on experience. Although this implementation method is simple, intuitive, and can quickly obtain results on-site, the identification accuracy depends on the experience of geological personnel, has a large subjectivity, is greatly affected by human factors, and has low identification accuracy and consistency when dealing with complex geological conditions. That is, traditional methods mainly rely on manual experience or microscopic observation. These methods have strong subjectivity and low accuracy. Especially when the surface of the cuttings is covered with mud, it is difficult to accurately identify. If the machine vision identification method is adopted, only the appearance characteristics of the cuttings can be collected, and the internal characteristics under complex geological conditions cannot be identified, and the image quality will be affected by on-site conditions such as lighting and mud. Therefore, the accuracy is low.

[0004] Aiming at the technical problems of low identification accuracy and difficult-to-guarantee real-time performance existing in the existing cuttings identification, no effective solution has been proposed yet. Summary of the Invention

[0005] The purpose of this application is to provide a method, system, and device for identifying cuttings types, which can efficiently, accurately, and real-time identify cuttings types, especially having higher identification accuracy for cuttings with similar surface characteristics but obvious differences in internal structure. Furthermore, by introducing the fusion of neural network models and multi-source data, the cuttings identification results have stronger adaptability to complex drilling conditions.

[0006] The method, system, and device for identifying cuttings types provided by this application are implemented as follows:

[0007] A method for identifying cuttings types includes:

[0008] During the drilling process of the target well in the target formation, the real-time cuttings are scanned by a CT scanning device to obtain CT scan data;

[0009] The internal structure features and external structure features of the real-time cuttings are extracted from the CT scan data through a pre-trained neural network model;

[0010] Based on the internal structure features and external structure features of the real-time cuttings, the cuttings type of the real-time cuttings is identified through the pre-trained neural network model.

[0011] In one embodiment, based on the internal structure features and external structure features of the real-time cuttings, the cuttings type of the real-time cuttings is identified through the pre-trained neural network model, including:

[0012] The environmental parameters of the real-time cuttings are collected by a temperature sensor;

[0013] The electromagnetic characteristic parameters of the real-time cuttings are collected by an electromagnetic scanning device, where the electromagnetic characteristic parameters include: conductivity, permeability;

[0014] Based on the internal structure features, external structure features, environmental parameters and electromagnetic characteristic parameters of the real-time cuttings, the cuttings type of the real-time cuttings is identified through the pre-trained neural network model.

[0015] In one embodiment, the internal structure features and external structure features of the real-time cuttings are extracted from the CT scan data through a pre-trained neural network model, including:

[0016] The external structure features of the real-time cuttings are extracted through the shallow convolution of the neural network model, where the external structure features include: edges, textures, morphological contours;

[0017] The internal structure features of the real-time cuttings are extracted through the deep convolution of the neural network model, where the internal structure features include: porosity, density distribution and internal fractures;

[0018] Correspondingly, based on the internal structure features and external structure features of the real-time cuttings, the cuttings type of the real-time cuttings is identified through the pre-trained neural network model, including:

[0019] Output the cuttings type of the real-time cuttings through the fully connected layer of the output layer of the neural network model, where the fully connected layer includes multiple nodes, the number of nodes is equal to the number of cuttings types, each output node corresponds to a cuttings type, and each output node outputs the probability that the real-time cuttings belong to the corresponding cuttings category. Determine the cuttings category with the largest probability value as the output result of the output layer through the softmax activation function.

[0020] In one embodiment, the pre-trained neural network model is trained as follows:

[0021] Obtain historical cuttings samples of the target well in the target formation;

[0022] Scan the historical cuttings samples through a CT device to obtain CT scan images;

[0023] Perform denoising processing on the CT scan images through Gaussian filtering;

[0024] After the denoising processing, perform normalization processing on the denoised CT scan images through linear normalization;

[0025] Perform image segmentation on the normalized CT scan images, and extract internal structure features and external structure features from the segmented images;

[0026] Label classification labels for the extracted internal structure features and external structure features;

[0027] Use the extracted internal structure features, external structure features, and labeled classification labels as the training data set to train the initial network model to obtain the neural network model.

[0028] In one embodiment, using the extracted internal structure features, external structure features, and labeled classification labels as the training data set to train the initial network model to obtain the neural network model includes:

[0029] Use the extracted internal structure features, external structure features, and labeled classification labels as the first training data set;

[0030] Obtain the geological parameters and historical drilling data of the adjacent wells of the target well, where the historical drilling data includes: the cuttings identification results and formation information of the adjacent wells;

[0031] Use the geological parameters and historical drilling data of the adjacent wells of the target well as the second training data set;

[0032] Train the initial network model through the first training data set and the second training data set to obtain the neural network model.

[0033] In one embodiment, after identifying the type of the real-time cuttings based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model, the method further includes:

[0034] Determining whether the type of the real-time cuttings obtained by identification is accurate;

[0035] When it is determined that the type of the real-time cuttings is inaccurate, obtaining historical adjacent-well cuttings samples of the target well in the target formation;

[0036] Scanning the historical adjacent-well cuttings samples through a CT device to obtain CT scan images of the adjacent-well cuttings;

[0037] Extracting the internal structure features and external structure features of the historical adjacent-well cuttings samples in the CT scan images of the adjacent-well cuttings;

[0038] Optimizing the neural network model with the internal structure features and external structure features of the historical adjacent-well cuttings samples.

[0039] In one embodiment, the pre-trained neural network model is a network model obtained by fusing N deep learning models, where N is a positive integer greater than 1;

[0040] Correspondingly, identifying the type of the real-time cuttings based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model includes:

[0041] Respectively obtaining, by each of the N deep learning models, the probability that the real-time cuttings belong to each of the M cuttings types, where M is a positive integer greater than 1;

[0042] Performing weighted averaging on the M groups of probability values obtained by identification according to the weight values of each of the N deep learning models to obtain M probability values;

[0043] Taking the cuttings type corresponding to the maximum probability value among the M probability values as the type of the real-time cuttings.

[0044] A cuttings type identification system includes:

[0045] A cuttings transmission channel for automatically transmitting real-time cuttings;

[0046] CT devices arranged on both sides of the cuttings transmission channel for performing CT scanning on the real-time cuttings to obtain CT scan images;

[0047] Environmental parameter sensors arranged on both sides of the cuttings transmission channel, for collecting environmental parameters of the real-time cuttings;

[0048] A central processor, connected to the cuttings transmission channel, the CT device and the environmental parameter sensors, for controlling the real-time cuttings to pass through the CT device and the environmental parameter sensors in sequence, and identifying the cutting type of the real-time cuttings according to the CT scan image and environmental parameters.

[0049] A cutting type identification device, comprising:

[0050] A scanning module, for scanning real-time cuttings generated during the drilling process of a target well in a target formation through a CT scanning device to obtain CT scan data;

[0051] An extraction module, for extracting internal structural features and external structural features of the real-time cuttings from the CT scan data through a pre-trained neural network model;

[0052] An identification module, for identifying based on the internal structural features and external structural features of the real-time cuttings through the pre-trained neural network model to obtain the cutting type of the real-time cuttings.

[0053] An electronic device, comprising a processor and a memory for storing processor-executable instructions, and the processor implements the steps of the above method when executing the instructions.

[0054] A computer-readable storage medium, on which a computer program / instructions are stored, and the computer program / instructions implement the steps of the above method when executed by a processor.

[0055] The method for identifying cuttings types provided by this application scans real-time cuttings generated during the drilling process of a target well in a target formation through a CT scanning device to obtain CT scan data; extracts the internal structure features and external structure features of the real-time cuttings from the CT scan data through a pre-trained neural network model; and identifies the cuttings types based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model. That is, the internal and external structure features of the cuttings are obtained through a CT scanning device, and the cuttings types are identified in combination with a neural network model, thereby solving the technical problems of low efficiency and low recognition accuracy existing in the existing cuttings recognition based on manual work or cuttings appearance, and achieving the technical effect of automatically and accurately identifying cuttings types. Through the solution provided in this example, the cuttings types can be identified efficiently, accurately and in real time, especially for cuttings with similar surface features but obvious differences in internal structure. Further, by introducing the fusion of a neural network model and multi-source data, the cuttings recognition results have stronger adaptability to complex drilling conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0057] Figure 1 It is a flowchart of a method for an embodiment of the method for identifying cuttings types based on CT scanning and multi-modal data fusion provided by this application;

[0058] Figure 2 It is a flowchart of the training method of the neural network model provided by this application;

[0059] Figure 3 It is a flowchart of the method for identifying the cuttings types of real-time cuttings provided by this application;

[0060] Figure 4 It is a schematic structural diagram of the cuttings acquisition system provided by this application;

[0061] Figure 5 It is a block diagram of the hardware structure of an electronic device for a method for identifying cuttings types based on CT scanning and multi-modal data fusion provided by this application;

[0062] Figure 6 It is a schematic diagram of the module structure of an embodiment of the device for identifying cuttings types based on CT scanning and multi-modal data fusion provided by this application. Detailed implementation manners

[0063] In order to enable those skilled in the art of the present technology to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the scope of protection of this application.

[0064] It should be noted that in the embodiments of this specification, some industry-existing solutions such as certain software, components, models, etc. may be mentioned, and they should be regarded as exemplary. The purpose is only to illustrate the feasibility in the implementation of the technical solutions of this application, but it does not mean that the applicant has already or necessarily used this solution.

[0065] Considering the identification of cuttings, although the existing machine vision identification technology can meet the intelligent identification requirements, there are still the following problems when used in the actual field:

[0066] 1) Dependence on image quality: Machine vision identification highly depends on the clarity and quality of cuttings images. In the complex environment of the drilling site, the downhole environment is dirty and messy. Although there is a cuttings cleaning link, the cuttings may still be covered by mud or drilling fluid, etc., which will inevitably lead to unclear images, thereby affecting the identification accuracy of the system. And some tiny cuttings cannot be clearly presented through visual images, making it impossible to accurately identify these cuttings.

[0067] 2) Difficulty in capturing internal features of cuttings: Machine vision identification can only process the appearance features of cuttings and cannot obtain the internal structure or physical properties of cuttings. This makes it difficult for the identification system to effectively distinguish cuttings with similar appearances but different compositions and physical properties, restricting the identification ability of machine vision for complex cuttings types.

[0068] 3) Sensitive to the lighting environment: Machine vision has high requirements for the ambient lighting. Factors such as on-site light changes, shadows or reflections will interfere with the effect of image acquisition, resulting in unstable identification results. Especially in the complex drilling operation site, the light conditions are difficult to control, which increases the difficulty of the system to adapt to different working conditions.

[0069] 4) High cost of model training and updating: The core of the machine vision system lies in the training and optimization of image recognition algorithms. This requires a large number of cuttings images of known types as training samples, and as new formations and cuttings types appear, the system needs to continuously update the model to maintain the identification ability. This will result in higher costs for model maintenance and updating, and professional personnel are required to participate.

[0070] 5) Unable to fully replace manual judgment: Although machine vision can automatically process cuttings images, when encountering unseen cuttings or complex geological conditions, it is still unable to make accurate judgments and ultimately requires manual intervention and confirmation. This technology still has limitations in completely replacing manual identification. Compared with traditional manual identification, machine vision identification has significantly improved efficiency and automation, but there are still many problems and challenges when dealing with the complexity, diversity of cuttings and their internal physical properties, lacking flexibility and adaptability.

[0071] Based on this, in this example, it is considered that the three-dimensional geometric shape, internal structure, physical parameters, etc. of cuttings can be obtained through CT scanning (Computed Tomography). By designing and training a convolutional neural network, the relationship between the complex features of cuttings and their categories can be automatically learned, so that the model can be trained with large-scale data, enabling the model to handle cuttings with complex shapes. At the same time, the network structure and hyperparameters are optimized to improve classification accuracy and robustness. That is, most of the existing cuttings identification methods rely on a single data source (such as visual features, density, shape, etc.), without fusing multi-dimensional data, which limits the accuracy of identification. In this example, multi-modal data is fused, combining CT scan data, environmental parameters, electromagnetic characteristic parameters, etc. data, and comprehensive identification of multi-dimensional features is achieved through a deep learning model.

[0072] Figure 1 It is a flowchart of a method of an embodiment of the cuttings type identification method provided by this application. Although this application provides method operation steps or device structures as shown in the following embodiments or drawings, based on routine or non-creative labor, more or fewer operation steps or module units may be included in the method or device. In steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure described in the embodiments of this application and shown in the drawings. When the method or module structure is applied to an actual device or terminal product, it can be executed sequentially or in parallel according to the method or module structure connection shown in the embodiments or drawings (for example, in an environment of parallel processors or multi-threaded processing, or even a distributed processing environment).

[0073] Specifically, as Figure 1 shown, the above-mentioned cuttings type identification method may include the following steps:

[0074] Step 101: Scan the real-time cuttings generated during the drilling process of the target well in the target formation through a CT scanning device to obtain CT scan data;

[0075] Step 102: Extract the internal structural features and external structural features of the real-time cuttings from the CT scan data through a pre-trained neural network model;

[0076] Step 103: Identify the type of the real-time cuttings based on the internal structural features and external structural features of the real-time cuttings through the pre-trained neural network model.

[0077] After the type of the cuttings is identified in real time, the system can analyze and give feedback according to the real-time identification result, and adjust the drilling operation in time to ensure the real-time response ability and operation optimization under complex formation conditions.

[0078] In the above example, by introducing a CT scanning device to scan the cuttings, the internal structural features and external structural features of the cuttings can be obtained, and combined with a neural network model for identifying the type of the cuttings, thus solving the technical problems of low efficiency and low recognition accuracy existing in the existing cuttings identification based on manual or appearance of the cuttings, and achieving the technical effect of automatically and accurately identifying the type of the cuttings.

[0079] In order to introduce multi-dimensional feature data to improve the recognition accuracy of the type of the cuttings, in this example, it is considered that environmental parameters, electromagnetic characteristic parameters, etc. can be added as additional dimensional data and used together as feature data for identifying the type of the cuttings. In one embodiment, identifying the type of the real-time cuttings based on the internal structural features and external structural features of the real-time cuttings through the pre-trained neural network model may include:

[0080] S1: Collect the environmental parameters of the real-time cuttings through a temperature sensor;

[0081] S2: Collect the electromagnetic characteristic parameters of the real-time cuttings through an electromagnetic scanning device, where the electromagnetic characteristic parameters include conductivity and permeability;

[0082] S3: Identify the type of the real-time cuttings based on the internal structural features, external structural features, environmental parameters and electromagnetic characteristic parameters of the real-time cuttings through the pre-trained neural network model.

[0083] That is, most of the existing cuttings identifications rely on a single data source (such as visual features, density, morphology, etc.), and multi-dimensional data is not fused, which limits the accuracy of the identification. In this example, multi-modal data is fused, combining data such as CT scan data, environmental parameters, and electromagnetic characteristic parameters, and realizing the comprehensive identification of multi-dimensional features through a deep learning model, so that the cuttings identification result has stronger adaptability to complex drilling conditions.

[0084] For the neural network model in this example, it may include shallow convolution, deep convolution, and an output layer. Specifically, when extracting the internal structural features and external structural features of the real-time cuttings from the CT scan data through the pre-trained neural network model, the external structural features of the real-time cuttings can be extracted through the shallow convolution of the neural network model, where the external structural features may include, but are not limited to, at least one of the following: edges, textures, morphological contours; extracting the internal structural features of the cuttings through the neural network model, including porosity, density distribution, and internal fractures; correspondingly, through the pre-trained neural network model, based on the internal structural features and external structural features of the real-time cuttings for identification to obtain the cutting type of the real-time cuttings, which may include: through the fully connected layer of the output layer of the neural network model, outputting the cutting type of the real-time cuttings, where the fully connected layer includes multiple nodes, the number of nodes is equal to the number of cutting types, each output node corresponds to a cutting type, each output node outputs the probability that the real-time cuttings belong to the corresponding cutting category, and the cutting category with the largest probability value is determined as the output result of the output layer through the softmax activation function.

[0085] That is, the neural network model constructs a system that can automatically extract features and classify from CT scan data. The input layer receives CT scan data, the number of nodes in the output layer is equal to the number of cutting categories, and each node represents the probability of a cutting category. Through the trained convolutional neural network, different layers of the network can extract different levels of features. At each layer, the feature map represents the cutting features at different levels, and the maximum probability is used to determine the cutting category. The distributions of these categories are closely related to the geometric morphology, density, fractures, and other features of the cuttings.

[0086] The above-mentioned neural network model can be trained according to the steps as Figure 2 shown below:

[0087] Step 201: Obtain historical cutting samples of the target well in the target formation;

[0088] Step 202: Scan the historical cutting samples through a CT device to obtain CT scan images;

[0089] Step 203: Denoise the CT scan images through Gaussian filtering;

[0090] Step 204: After denoising, standardize the denoised CT scan images through linear normalization;

[0091] Step 205: Segment the CT scan images after normalization processing, and extract the internal structure features and external structure features from the images after segmentation;

[0092] Step 206: Label classification labels for the extracted internal structure features and external structure features;

[0093] Step 207: Use the extracted internal structure features, external structure features, and the labeled classification labels as a training data set to train the initial network model to obtain the neural network model.

[0094] For a convolutional neural network model, it can enhance the adaptability to new cuttings types and complex layers by continuously learning and updating model parameters, and can link the data of adjacent wells to further enrich the database, thereby improving the accuracy of recognition. Specifically, data from different sources can be independently processed first to extract their respective features, and then these features are concatenated together as input, and z-score normalization is used to ensure the consistency of the scale of the input data. The features extracted from the data of each source can include: one is the features of CT scan data, including geometric shape features (such as structural parameters such as the shape, size, and volume of cuttings); density features, which reflect the density and material of cuttings through the gray value distribution of CT scan images; porosity and fracture features, which identify the pore structure, fracture size, and distribution inside cuttings; texture features, which are used to characterize the local texture changes of the image and reflect the fine characteristics of the cuttings surface. The other is to fuse relevant data of other drilling or geological facies, such as geological parameters, historical drilling data (cuttings recognition results and formation information of adjacent wells), etc.

[0095] Based on this, using the extracted internal structure features, external structure features, and the labeled classification labels as a training data set to train the initial network model to obtain the neural network model can include: using the extracted internal structure features, external structure features, and the labeled classification labels as the first training data set; obtaining the geological parameters and historical drilling data of the adjacent wells of the target well, where the historical drilling data includes: cuttings recognition results and formation information of the adjacent wells; using the geological parameters and historical drilling data of the adjacent wells of the target well as the second training data set; training the initial network model with the first training data set and the second training data set to obtain the neural network model.

[0096] After identifying the cuttings type of the real-time cuttings based on the internal structural features and external structural features of the real-time cuttings through the pre-trained neural network model, considering that the correctness of the identification result can be determined according to the actual identification situation, if it is incorrect, the characteristic data of adjacent wells can be introduced to optimize the neural network model. For example, it can be determined whether the cuttings type of the identified real-time cuttings is accurate; in the case where the determined cuttings type of the real-time cuttings is inaccurate, obtain the historical adjacent-well cuttings samples of the adjacent wells of the target well in the target formation; scan the historical adjacent-well cuttings samples through a CT device to obtain adjacent-well cuttings CT scan images; extract the internal structural features and external structural features of the historical adjacent-well cuttings samples in the adjacent-well cuttings CT scan graphics; optimize the neural network model through the internal structural features and external structural features of the historical adjacent-well cuttings samples.

[0097] To realize the judgment of the cuttings identification result, it is necessary to accurately determine the type of cuttings. For this reason, in this example, a cuttings identification technology is also provided. By analyzing the elemental composition in the cuttings, a quantitative lithology identification and formation division method is provided. Through the analysis and comparison of elemental combinations, the logging process is optimized to improve the lithology identification efficiency and accuracy in drilling operations. Specifically, the following steps can be included: First, determine the elemental combinations that can represent the main lithological features through experimental analysis. Among them, the elements can include Si, Fe, Al, Ca, Mg, etc. These elements account for a relatively high proportion in the cuttings and can reflect the characteristics of different formations. Then, during the drilling process, the cuttings are detected in real time through an X-ray fluorescence device to obtain the elemental spectrogram of the sample. Since the fluorescence intensity and wavelength of each element have specific identification characteristics, the chemical composition of the cuttings can be preliminarily judged based on these characteristics. Then, compare the obtained spectral data with the standardized elemental database, and through mathematical models or algorithms for data processing and analysis in combination with the geological background of the actual well site, determine the lithology of the cuttings and the formation in which it is located, that is, determine the cuttings category of the cuttings. Take the determined cuttings category as the true cuttings category of the real-time cuttings, and compare it with the cuttings category identified by the neural network model to judge whether the cuttings category identified by the neural network model is accurate.

[0098] Furthermore, according to the identification result of the cuttings category, real-time lithology discrimination suggestions can be provided for drilling, prompting the on-site personnel whether it is necessary to adjust the drilling parameters or optimize the completion plan to adapt to the complexity of different geological layers. Furthermore, through continuous elemental data feedback, the elemental combination and lithology identification model can be automatically adjusted and calibrated to ensure accuracy and consistency under different well sites and geological conditions, and historical drilling data can also be combined to continuously optimize the lithology identification accuracy.

[0099] Specifically, by accurately identifying and classifying the characteristics of cuttings, such as three-dimensional shape, porosity, volume, fracture degree, mineral composition, etc., real-time feedback and optimization suggestions are provided during the drilling operation based on the recognition results of cutting types, achieving more efficient drilling operations and resource utilization rates. The system can analyze and provide feedback based on real-time data, timely adjust the drilling operation, and ensure real-time response capabilities and operation optimization under complex formation conditions. By replacing the existing manual identification process with an intelligent system, the risk of human misjudgment is reduced, and the identification speed is significantly improved, making the drilling operation more efficient and reliable. The system can operate independently and automatically generate identification results, reducing the dependence on manual experience while improving identification standardization and consistency.

[0100] When optimizing the model, in order to cope with the complex environments of different downhole formations, an optimization strategy of multi-model integration can be adopted. That is, by integrating multiple deep learning models, comparing and analyzing the recognition effects of each model in different environments, and selecting the optimal model for real-time cutting identification. For example, different deep learning models such as CNN and U-Net can be integrated for model fusion to give full play to the respective advantages of each model, thereby reducing the overfitting or underfitting problems that may occur in a single model. Specifically, for the neural network of multi-model fusion, different models can be fused through weighted voting, and the class probabilities output by each model are weighted and averaged, and the class with the highest probability is selected as the final prediction.

[0101] Specifically, the pre-trained neural network model described above can be a network model obtained by fusing N deep learning models, where N is a positive integer greater than 1; correspondingly, through the pre-trained neural network model, based on the internal structure characteristics and external structure characteristics of the real-time cuttings, the cutting type of the real-time cuttings can be identified, as Figure 3 shown, including the following steps:

[0102] Step 301: Respectively identify the probabilities that the real-time cuttings belong to each of the M cutting types through each of the N deep learning models, where M is a positive integer greater than 1;

[0103] Step 302: According to the weight values of each of the N deep learning models, perform weighted averaging on the M groups of probability values obtained by identification to obtain M probability values;

[0104] Step 303: Take the cutting type corresponding to the maximum probability value among the M probability values as the cutting type of the real-time cuttings.

[0105] In this example, a cutting acquisition system is also provided. The cutting acquisition system includes:

[0106] A cuttings transmission channel for automatically transmitting real-time cuttings;

[0107] CT devices arranged on both sides of the cuttings transmission channel for performing CT scans on the real-time cuttings to obtain CT scan images;

[0108] Environmental parameter sensors arranged on both sides of the cuttings transmission channel for collecting the environmental parameters of the real-time cuttings;

[0109] A central processor connected to the cuttings transmission channel, the CT devices, and the environmental parameter sensors for controlling the real-time cuttings to sequentially pass through the CT devices and the environmental parameter sensors, and identifying the type of the real-time cuttings based on the CT scan images and environmental parameters.

[0110] That is, in this example, a cuttings collection system is provided, including: a cavity 1 for loading and equipped with a high-resolution CT machine to perform three-dimensional imaging on cuttings samples to obtain the internal structural characteristics and physical parameters of the cuttings. Through CT scanning, key information such as the volume, density, and fracture distribution of the cuttings can be accurately analyzed, providing basic data for cuttings classification and modeling. Cavity 2 and cavity 3 can load the integration of multiple types of sensors, such as: temperature sensors, etc. for real-time collection of the environmental parameters of the samples. Combining these data with the CT scan results for multi-dimensional analysis of cuttings characteristics, thereby improving the recognition accuracy. Cavity 4 is a cuttings transmission channel. Further, positions for an electromagnetic scanning device or other modules can be reserved in the cuttings collection system. Through the electromagnetic scanning device, the electromagnetic characteristics (such as: conductivity, permeability, etc.) of the cuttings samples can be measured and analyzed to further enrich the acquisition dimension of the physical information of the cuttings. The above-mentioned cuttings transmission channel is used to realize the automatic transmission of samples, enabling the cuttings to sequentially pass through each functional module to complete CT scanning, sensor data collection, and other test operations, thereby avoiding manual intervention and improving work efficiency and recognition accuracy.

[0111] In the above example, CT scan image segmentation technology is used to extract the geometric morphology, porosity, and physical parameters of the cuttings. The model is trained based on the CT scan data of the cuttings, and non-linear features and noise interference are solved through data preprocessing. By combining the CT scan data with real-time data such as drilling information, a dynamically updated intelligent recognition system is formed. The output result of the recognition model provides specific decision-making suggestions for drilling operations, ensuring accurate classification and enabling on-site application.

[0112] The above method will be described below in conjunction with a specific embodiment. However, it should be noted that this specific embodiment is only for better explaining the present application and does not constitute an improper limitation of the present application.

[0113] In the existing manual cuttings identification, it relies on manual experience and visual judgment, with limited accuracy and being easily affected by subjective factors. In diverse formation environments, existing machine vision cannot capture the subtle features of cuttings, resulting in large identification errors. Based on this, in this example, a cuttings identification method combining high-resolution three-dimensional imaging by CT scanning and a deep learning model is proposed to enhance the adaptability and robustness of the cuttings identification system in complex geological environments. That is, by introducing CT scanning and neural network technology, high-precision and automated cuttings identification is achieved, thereby improving the overall identification efficiency.

[0114] Therefore, in this example, an integrated device and method capable of realizing intelligent cuttings identification and analysis are provided to improve the accuracy and real-time performance of cuttings detection, and optimize the drilling process based on the identified cuttings types, and make judgments and operation decisions on the drilling conditions. Specifically, the collected cuttings data and historical data in the database can be integrated to establish a cuttings identification model, and through accurate identification and classification of the characteristics of the cuttings such as three-dimensional shape, porosity, volume, fracture degree, and mineral composition, real-time feedback and optimization suggestions can be provided during the drilling operation, realizing more efficient drilling operations and resource utilization.

[0115] To solve the problems of efficiency and accuracy existing in the existing cuttings identification technology, especially the poor real-time performance and automated processing ability of the existing cuttings identification methods for cuttings identification under complex formation conditions, in this example, a cuttings identification system based on CT scanning and neural network is provided. This system can not only improve the accuracy and efficiency of identification, but also has strong environmental adaptability and real-time feedback ability, thus achieving the technical effect of accurately identifying cuttings.

[0116] Specifically, the intelligent cuttings identification technology based on CT scanning and neural network provided in this example realizes the automation and intelligence of cuttings identification through multi-level feature extraction and deep learning models, combines the accurate physical data collected by CT scanning with the learning ability of neural networks, and thus optimizes the cuttings classification. It can include the following steps:

[0117] Step 1: Data collection and data preprocessing.

[0118] Data collection: First, use CT to scan the three-dimensional internal structure imaging of cuttings, capture the subtle physical characteristics of cuttings (such as porosity, density distribution, and internal fractures, etc.) through high-resolution images. At the same time, build a three-dimensional solid model of cuttings. Through multi-dimensional data extraction, the system can comprehensively obtain the external and internal information of cuttings, providing data support for subsequent intelligent identification.

[0119] Data preprocessing: After obtaining the initial data, denoising can be performed through Gaussian filtering, image segmentation can be carried out through deep learning, dimensionality reduction using PCA (Principal Component Analysis), and 3D reconstruction, etc., so as to improve the quality and recognition accuracy of CT scan data. By standardizing these data, the quality of the data input into the neural network can be ensured.

[0120] Step 2: Feature extraction and deep learning.

[0121] Feature extraction: Based on the CT scan data, key physical features of the cuttings are extracted. Among them, the key physical features can include: geometric morphology, internal structure, and physical parameters. These feature data are input into the trained neural network model.

[0122] CT image preprocessing: First, denoise the image through Gaussian filtering to smooth the image; standardize the CT image through linear normalization. Then, perform image segmentation and then feature extraction to extract geometric morphology features, volume and density features, and texture features. Finally, model training is carried out based on these feature data.

[0123] Deep learning: The deep CNN (Convolutional Neural Networks) gradually learns the correlation between the complex morphology and categories of cuttings through the training of a large-scale cuttings dataset, forming an adaptive classification model.

[0124] Specifically, first collect the 3D structure data of the cuttings through CT scanning technology, including: features such as the geometric morphology, internal structure, density distribution, porosity, and fracture degree of the cuttings. After preprocessing these 3D structure data (such as denoising, standardization, dimensionality reduction, etc.), high-quality input data is formed. Then, key physical features (such as volume, density, geometric morphology, texture, etc.) are extracted from the CT scan data. These features are the core data input into the neural network and provide a direct classification basis for model training. For the training of the neural network, a large-scale labeled cuttings dataset is required, which contains CT scan data of various different types of cuttings and corresponding classification labels. The neural network learns the relationship between the cuttings features and categories through these data to achieve the classification and recognition of new cuttings data.

[0125] The neural network model is to build a system that can automatically extract features from CT scan data and classify them. The input layer receives CT scan data, and the number of nodes in the output layer is equal to the number of rock cuttings categories, and each node represents the probability of a rock cutting category. Through the trained convolutional neural network, different layers of the network can extract features at different levels. In each layer, the feature map represents the rock cutting features at different levels. The maximum probability is used to determine the category of the rock cuttings. The distribution of these categories is closely related to the geometric morphology, density, cracks and other characteristics of the rock cuttings. Among them, the shallow convolution of the convolutional neural network is used to extract basic features, such as simple features such as edges, textures, and morphological contours. Deep convolution is used to extract complex high-order features, such as abstract features such as the geometric morphology, internal cracks, and density distribution of rock cuttings. The number of nodes in the output layer is determined according to the number of rock cutting categories. Specifically, the output layer can be a fully connected layer with the number of nodes equal to the number of categories of rock cuttings to be identified. Each output node corresponds to a rock cutting category, and the output value represents the probability that the input sample belongs to that category. Then, the value of the output node is normalized to a probability distribution through the softmax activation function, and the category with the highest probability is taken as the classification result of the model.

[0126] Step 3: Model optimization and linkage identification.

[0127] Model optimization: In order to cope with the complex environment of different underground formations, a multi-model integrated optimization strategy can be adopted. By integrating multiple deep learning models, comparing and analyzing the recognition effects of each model in different environments, the optimal model is selected for real-time cuttings recognition.

[0128] By integrating different deep learning models such as CNN and U-Net, we can give full play to their respective advantages, thereby reducing the overfitting or underfitting problems that may occur in a single model. Specifically, we can fuse different models through weighted voting, take a weighted average of the class probabilities output by each model, and select the class with the highest probability as the final prediction.

[0129] For the convolutional neural network model, the adaptability to new cuttings types and complex layers can be enhanced by continuously learning and updating model parameters, and the data of adjacent wells can be linked to further enrich the database and improve the accuracy of recognition. Using a model trained in one geological environment, and then fine-tuning it, it can be adapted to new cuttings types or new geological environments.

[0130] Among them, data fusion means that data from different sources are first processed independently to extract their respective features, and then these features are concatenated together as input, and z-score normalization is used to ensure the consistency of the scale of the input data. The features extracted respectively can include: one is the feature of CT scan data, including geometric shape features (such as structural parameters such as the shape, size, and volume of cuttings); density features, which reflect the density and material of cuttings through the gray value distribution of CT scan images; porosity and fracture features, which identify the pore structure, fracture size and distribution inside cuttings, etc.; texture features, which are used to characterize the local texture changes of the image and reflect the subtle characteristics of the surface of cuttings. The other is to fuse relevant data of other drilling or geological facies, such as geological parameters, historical drilling data (cuttings identification results of adjacent wells, formation information), etc.

[0131] Step 4: Real-time identification and feedback optimization.

[0132] Real-time identification: The real-time identification module combines CT scan data with a neural network model to quickly output the cuttings type and corresponding feature information. The system optimizes the identification model by continuously collecting and analyzing operation data, enabling it to quickly adapt to formation changes and achieve accurate identification.

[0133] Feedback mechanism: The drilling operation provides real-time feedback, and the neural network model is adjusted and optimized according to the feedback to achieve continuous learning and improvement of the model.

[0134] Among them, the adjustment of the feedback mechanism can include: misjudgment feedback, that is, the system compares the prediction result with the actual sample category. If the identification result does not match the actual category, the system will mark the misjudgment and use the error data for subsequent model optimization. Adjacent sample adjustment, that is, by jointly analyzing the features of adjacent samples (such as density, fractures, porosity), the system will re-evaluate the classification result of the current sample and adjust the identification strategy using the principle of continuity.

[0135] Specifically, the category feedback result can be obtained by comparing the cuttings types predicted by the model with the cuttings samples collected during the actual drilling process. Then, based on the category feedback result, it can be determined whether the cuttings category identified based on the model is accurate. For adjacent cuttings samples, if the model makes a misjudgment, the system can adjust the model's recognition strategy by tracking the relationship between adjacent samples. The influence of certain features (such as porosity, fractures, etc.) on the recognition result under specific formation conditions will also be fed back to the model. According to the feedback result, it can be judged which cuttings types have a lower recognition accuracy and whether it is due to the data imbalance problem (such as fewer samples of certain cuttings types). The generalization ability of the model can also be optimized by augmenting the samples of the minority classes (such as image rotation, flipping, transformation, etc.). When implementing, the network weights can be adjusted according to the feedback result to make the model pay more attention to the learning of specific category features; by fine-tuning the network layer structure, the performance of recognizing different features can be optimized. For example, increasing the depth of the network to capture complex features and enhancing the sensitivity of the activation function to continuous values such as density and porosity.

[0136] In the above example, through the deep integration of CT scanning and neural network, the transformation of cuttings recognition from the traditional mode to intelligent automation is realized, which can improve the recognition efficiency and significantly enhance the recognition accuracy under complex geological conditions, with broad application prospects.

[0137] In this example, a cuttings recognition method based on CT scanning and neural network is provided to achieve efficient and accurate cuttings recognition. The method may include the following steps:

[0138] Step 1: Data acquisition and preprocessing:

[0139] Use a CT scanning device to collect the three-dimensional data of the cuttings. Among them, the three-dimensional data may include physical parameters such as morphology, density, and porosity. Then, preprocess these original three-dimensional data, where the preprocessing includes: denoising, image enhancement, and standardization processing of the data format, so as to ensure that the data quality meets the training requirements of the subsequent deep learning model.

[0140] For example: Such as Figure 4The described cuttings collection system includes: a cavity 1 for loading a high-resolution CT scanner. Through this CT scanner, three-dimensional imaging of the cuttings samples is performed to obtain the internal structural characteristics and physical parameters of the cuttings. Through CT scanning, key information such as the volume, density, and fracture distribution of the cuttings can be accurately analyzed, providing basic data for cuttings classification and modeling. Cavity 2 and cavity 3 can load the integration of multiple types of sensors, such as temperature sensors, etc., for real-time collection of the environmental parameters of the samples. Combining these data with the CT scanning results, multi-dimensional analysis of the cuttings characteristics is carried out to improve the recognition accuracy. Cavity 4 is a cuttings transmission channel. Further, in this cuttings collection system, positions for electromagnetic scanning or other modules can also be reserved for measuring and analyzing the electromagnetic characteristics (such as conductivity, permeability, etc.) of the cuttings samples to further enrich the acquisition dimension of the physical information of the cuttings. The above-mentioned cuttings transmission channel is used to achieve the automatic transmission of the samples, enabling the cuttings to pass through each functional module in sequence, completing CT scanning, sensor data collection, and other test operations, thereby avoiding manual intervention and improving work efficiency and recognition accuracy.

[0141] That is, a short section as shown in Figure 4 is set and connected behind the cuttings cleaning tank to achieve automatic scanning and detection of the cuttings, thereby realizing the recognition of the cuttings type.

[0142] Step 2: Three-dimensional model construction and feature extraction:

[0143] Based on the slice data generated by CT scanning, a three-dimensional model of the cuttings is reconstructed and its features are extracted. The extracted features can include: the geometric shape, internal structure, and material distribution of the cuttings, etc. The extracted feature data will be used as the input for the subsequent intelligent recognition model.

[0144] Step 3: Neural network model design and training:

[0145] According to the characteristics of the cuttings, a deep neural network model is designed and trained. Through supervised learning with a large amount of labeled data, the convolutional neural network learns the mapping relationship between different cuttings types and their characteristics. During the model training process, cross-validation and parameter tuning techniques are used to ensure that the model has good generalization ability and high-precision recognition ability.

[0146] Step 4: Real-time recognition and feedback system:

[0147] After the model training is completed, the model is deployed in the real-time system to instantly recognize the cuttings type using the CT scanning data. Specifically, an independent system can be developed, which can provide intelligent feedback based on the current recognition result and make dynamic adjustments during the drilling process to ensure the real-time and accuracy of recognition and classification.

[0148] Step 5: Result Monitoring and Model Optimization:

[0149] Based on the real-time monitoring of the operation effect of the recognition system, analyze the performance of the model and determine the accuracy of the recognition result. According to the feedback in actual applications, regularly update the model and further optimize the model parameters to improve the overall performance and recognition efficiency of the system.

[0150] Step 6: System Iteration and Upgrade:

[0151] With the accumulation of data, the model can be continuously iteratively optimized. Further, by introducing more sample data and improving deep learning algorithms, the intelligence level and accuracy of the recognition system can be gradually enhanced to adapt to the changing on-site environment and formation conditions.

[0152] Specifically, in the above example, the three-dimensional geometric shape, internal structure, and physical parameters of cuttings are obtained through CT scanning, and these data are efficiently processed. Through image processing algorithms and data augmentation methods, it is ensured that the extracted features are accurate and comprehensive, providing a solid data foundation for subsequent classification. By designing and training a convolutional neural network, automatically learn the relationship between the complex features of cuttings and their categories, and train the model with a large amount of data, so that the recognition model can handle cuttings with complex shapes. At the same time, optimize the network structure and hyperparameters to improve the classification accuracy and robustness. By combining the CT scanning results with other relevant data (such as drilling records, geological information, etc.), the fusion processing of multi-source data is realized, and the model can be updated in real time according to the new scanning results. Through the classification results of the convolutional neural network, the system can analyze in real time and provide an optimized solution for cuttings recognition and classification, ensuring the provision of optimal drilling decision support in complex geological environments.

[0153] That is, through the intelligent cuttings recognition method based on CT scanning and deep learning provided above, it can be ensured that the cuttings recognition process can be automated and real-time, and meet the recognition requirements in complex geological environments. Specifically, by integrating CT scanning technology, deep neural network, real-time recognition and feedback system, an efficient and intelligent cuttings recognition solution is provided. In this way, the accuracy and efficiency of cuttings recognition can be improved. The physical characteristics of cuttings are obtained through CT scanning, and a deep learning model is used for intelligent recognition, which can quickly and accurately identify different types of cuttings, greatly improving the recognition efficiency and accuracy. The system can analyze and give feedback based on real-time data, timely adjust drilling operations, and ensure real-time response ability and operation optimization in complex formation conditions. By replacing the existing manual recognition process with an intelligent system, the risk of human misjudgment is reduced, and the recognition speed is significantly increased, making drilling operations more efficient and reliable. The system can operate independently and automatically generate recognition results, reducing the dependence on human experience, while improving recognition standardization and consistency.

[0154] Furthermore, the deep learning model can learn from a large number of different types of cuttings data and provide accurate identification results in complex and variable geological environments, adapting to diverse cuttings types under different formation conditions. The system can be continuously upgraded and iterated with the development of data and technology to handle more complex identification tasks and higher identification requirements. The real-time feedback system can provide operation suggestions while identifying cuttings, reducing unnecessary downtime and drilling adjustments, thus improving the overall operation efficiency and reducing the operation cost. Through accurate cuttings identification, it can help the drilling operation team optimize resource allocation and utilization, avoid unnecessary waste, and further enhance economic benefits. By continuously accumulating identification experience and updating the model, the system can achieve continuous optimization of the identification algorithm and possess self-adaptive capabilities, promoting the development and innovation of cuttings identification technology. The identification method provided in this example is not only applicable to the existing drilling environment but also can be extended to other resource development fields, such as mineral exploration, geological research, etc., with broad application potential. Through the above solutions, significant improvements can be provided in terms of the accuracy, efficiency, automation, and environmental adaptability of cuttings identification. At the same time, through real-time feedback and intelligent optimization, the operation decision-making ability during drilling is greatly enhanced, providing strong support for the intelligent development of the cuttings identification field.

[0155] The method embodiment provided in the above embodiment of this application can be executed in a mobile terminal, a computer terminal, or a similar computing device. Taking running on an electronic device as an example, Figure 5 is a hardware structure block diagram of an electronic device for a cuttings type identification method provided by this application. As Figure 5 shown, the electronic device 10 may include one or more (only one is shown in the figure) processors 02 (the processor 02 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 04 for storing data, and a transmission module 06 for communication functions. Those of ordinary skill in the art can understand that Figure 5 the structure shown is only schematic and does not limit the structure of the above electronic device. For example, the electronic device 10 may further include more or fewer components than those shown in Figure 5 , or have a different configuration from that shown in Figure 5 .

[0156] The memory 04 can be used to store software programs and modules of application software, such as the program instructions / modules corresponding to the cuttings type recognition method in the embodiments of the present application. The processor 02 executes various functional applications and data processing by running the software programs and modules stored in the memory 04, that is, realizes the cuttings type recognition method of the above application program. The memory 04 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 04 may further include a memory remotely provided with respect to the processor 02, and these remote memories can be connected to the electronic device 10 through a network. Examples of the above network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.

[0157] The transmission module 06 is used to receive or send data via a network. Specific examples of the above network may include a wireless network provided by a communication provider of the electronic device 10. In one instance, the transmission module 06 includes a network interface controller (NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission module 06 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0158] At the software level, the above cuttings type recognition device can be as Figure 6 shown, including:

[0159] A scanning module 601, configured to scan real-time cuttings generated during the drilling process of a target well in a target formation through a CT scanning device to obtain CT scanning data;

[0160] An extraction module 602, configured to extract internal structure features and external structure features of the real-time cuttings from the CT scanning data through a pre-trained neural network model;

[0161] An identification module 603, configured to identify based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model to obtain the cuttings type of the real-time cuttings.

[0162] In one embodiment, the above-mentioned identification module 603 is specifically configured to collect the environmental parameters of the real-time cuttings through a temperature sensor; collect the electromagnetic characteristic parameters of the real-time cuttings through an electromagnetic scanning device, where the electromagnetic characteristic parameters include: conductivity, permeability; and perform identification based on the internal structure features, external structure features, environmental parameters, and electromagnetic characteristic parameters of the real-time cuttings through the pre-trained neural network model to obtain the cutting type of the real-time cuttings.

[0163] In one embodiment, the above-mentioned extraction module 602 is specifically configured to extract the external structure features of the real-time cuttings through the shallow convolution of the neural network model, where the external structure features include: edges, textures, morphological contours; extract the internal structure features of the real-time cuttings through the deep convolution of the neural network model, where the internal structure features include: porosity, density distribution, and internal fractures; correspondingly, perform identification based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model to obtain the cutting type of the real-time cuttings, including: outputting the cutting type of the real-time cuttings through the fully connected layer of the output layer of the neural network model, where the fully connected layer includes multiple nodes, the number of nodes is equal to the number of cutting types, each output node corresponds to a cutting type, each output node outputs the probability that the real-time cuttings belong to the corresponding cutting category, and the cutting category with the largest probability value is determined as the output result of the output layer through the softmax activation function.

[0164] In one embodiment, the pre-trained neural network model can be trained in the following manner: obtain historical cutting samples of a target well in a target formation; scan the historical cutting samples through a CT device to obtain CT scan images; perform denoising processing on the CT scan images through Gaussian filtering; after the denoising processing, perform normalization processing on the denoised CT scan images through linear normalization; perform image segmentation on the normalized CT scan images, and extract the internal structure features and external structure features of the images after graphic segmentation; label classification labels for the extracted internal structure features and external structure features; use the extracted internal structure features, external structure features, and labeled classification labels as a training data set to train an initial network model to obtain the neural network model.

[0165] In one embodiment, the extracted internal structure features, external structure features, and labeled classification labels are used as a training data set to train an initial network model to obtain the neural network model, which may include: using the extracted internal structure features, external structure features, and labeled classification labels as a first training data set; obtaining the geological parameters and historical drilling data of the adjacent wells of the target well, where the historical drilling data includes: the cuttings identification results and formation information of the adjacent wells; using the geological parameters and historical drilling data of the adjacent wells of the target well as a second training data set; training the initial network model with the first training data set and the second training data set to obtain the neural network model.

[0166] In one embodiment, after identifying the cuttings type of the real-time cuttings based on the internal structure features and external structure features of the real-time cuttings by using the pre-trained neural network model, it is also possible to determine whether the identified cuttings type of the real-time cuttings is accurate; in the case where the determined cuttings type of the real-time cuttings is inaccurate, obtaining the historical adjacent well cuttings samples of the target well in the target formation; scanning the historical adjacent well cuttings samples by a CT device to obtain adjacent well cuttings CT scan images; extracting the internal structure features and external structure features of the historical adjacent well cuttings samples in the adjacent well cuttings CT scan graphics; optimizing the neural network model by using the internal structure features and external structure features of the historical adjacent well cuttings samples.

[0167] In one embodiment, the above-mentioned pre-trained neural network model is a network model obtained by fusing N deep learning models, where N is a positive integer greater than 1; correspondingly, identifying the cuttings type of the real-time cuttings based on the internal structure features and external structure features of the real-time cuttings by using the pre-trained neural network model may include: respectively identifying the probabilities that the real-time cuttings belong to each of the M cuttings types by each of the N deep learning models, where M is a positive integer greater than 1; performing weighted averaging on the identified M sets of probability values according to the weight values of each of the N deep learning models to obtain M probability values; taking the cuttings type corresponding to the maximum probability value among the M probability values as the cuttings type of the real-time cuttings.

[0168] Embodiments of the present application also provide a specific implementation manner of an electronic device that can implement all steps in the cuttings type recognition method in the above embodiments. The electronic device specifically includes the following: a processor, a memory, a communications interface, and a bus; wherein, the processor, the memory, and the communications interface complete communication with each other through the bus; the processor is used to call the computer program in the memory, and when the processor executes the computer program, all steps in the cuttings type recognition method in the above embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0169] Step 1: Scan the real-time cuttings generated during the drilling process of the target well in the target formation by a CT scanning device to obtain CT scan data;

[0170] Step 2: Extract the internal structure features and external structure features of the real-time cuttings from the CT scan data through a pre-trained neural network model;

[0171] Step 3: Identify based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model to obtain the cuttings type of the real-time cuttings.

[0172] Embodiments of the present application also provide a computer-readable storage medium that can implement all steps in the cuttings type recognition method in the above embodiments. A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, all steps in the cuttings type recognition method in the above embodiments are implemented. For example, when the processor executes the computer program, the following steps are implemented:

[0173] Step 1: Scan the real-time cuttings generated during the drilling process of the target well in the target formation by a CT scanning device to obtain CT scan data;

[0174] Step 2: Extract the internal structure features and external structure features of the real-time cuttings from the CT scan data through a pre-trained neural network model;

[0175] Step 3: Identify based on the internal structure features and external structure features of the real-time cuttings through the pre-trained neural network model to obtain the cuttings type of the real-time cuttings.

[0176] As can be seen from the above description, in the embodiment of the present application, a CT scanning device is used to scan the real-time cuttings generated during the drilling process of a target well in a target formation to obtain CT scan data; an internal structure feature and an external structure feature of the real-time cuttings are extracted from the CT scan data through a pre-trained neural network model; and the type of the real-time cuttings is identified based on the internal structure feature and the external structure feature of the real-time cuttings through the pre-trained neural network model. That is, the internal and external structure features of the cuttings are obtained through a CT scanning device, and the type of the cuttings is identified in combination with a neural network model, thereby solving the technical problems of low efficiency and low recognition accuracy existing in the existing cuttings recognition based on manual work or the appearance of cuttings, and achieving the technical effect of automatically and accurately identifying the type of cuttings. Through the solution provided in this example, the type of cuttings can be efficiently, accurately and real-time identified, especially for cuttings with similar surface features but obvious differences in internal structure. Further, by introducing the neural network model and the fusion of multi-source data, the recognition result of the cuttings has stronger adaptability to complex drilling conditions.

[0177] The embodiments in this specification are all described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the hardware + program type embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0178] The specific embodiments of this specification are described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be executed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the figures do not necessarily require the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0179] Although the present application provides method operation steps as described in the embodiments or flowcharts, based on routine or non-creative labor, there may be more or fewer operation steps. The order of steps listed in the embodiments is only one way among many step execution orders and does not represent the only execution order. When actually executed by a device or client product, it can be executed in the order shown in the embodiments or the figures or in parallel (such as in an environment of parallel processors or multi-threaded processing).

[0180] Although the embodiments of this specification provide method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative means. The step sequences listed in the embodiments are only one way among the execution sequences of numerous steps and do not represent the only execution sequence. When the actual device or terminal product executes, it can be executed in the method sequence shown in the embodiments or the drawings or executed in parallel (for example, in an environment of parallel processors or multi-threaded processing, or even in a distributed data processing environment). The term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, product or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, product or device. Without more limitations, it does not exclude the existence of additional identical or equivalent elements in the process, method, product or device comprising the said elements.

[0181] For convenience of description, the above device is described by dividing it into various modules according to functions. Of course, when implementing the embodiments of this specification, the functions of each module can be implemented in the same or multiple software and / or hardware, or the modules implementing the same function can be realized by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0182] Those skilled in the art also know that in addition to implementing the controller in the form of pure computer-readable program code, the method steps can be logically programmed to enable the controller to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. to achieve the same function. Therefore, such a controller can be regarded as a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0183] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0184] Those skilled in the art should understand that the embodiments of the present specification can be provided as a method, system, or computer program product. Therefore, the embodiments of the present specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0185] The embodiments of the present specification can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The embodiments of the present specification can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0186] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the key point of each embodiment is to illustrate the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the description of the method embodiment. In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of this specification. In this specification, the schematic expression of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0187] The above is only the embodiment of the embodiments of this specification and is not used to limit the embodiments of this specification. For those skilled in the art, various changes and modifications can be made to the embodiments of this specification. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of this specification shall be included within the scope of the claims of the embodiments of this specification.

Claims

1. A method for identifying rock cuttings types, characterized in that: include: Scanning the real-time rock cuttings generated during the drilling process of the target well in the target formation by means of a CT scanning device to obtain CT scanning data; Extracting the internal and external structural features of the real-time rock cuttings from the CT scan data through a pre-trained neural network model; The neural network model obtained by the pre-training is used to identify the rock cuttings based on the internal structural features and external structural features of the real-time rock cuttings to obtain the rock cuttings type of the real-time rock cuttings.

2. The method according to claim 1, characterized in that Using the pre-trained neural network model, identification is performed based on the internal structural features and external structural features of the real-time rock cuttings to obtain the rock cutting type of the real-time rock cuttings, including: Collecting environmental parameters of the real-time cuttings through a temperature sensor; Collecting electromagnetic characteristic parameters of the real-time rock cuttings by an electromagnetic scanning device, wherein the electromagnetic characteristic parameters include at least one of the following: electrical conductivity and magnetic permeability; The pre-trained neural network model is used to identify the real-time rock cuttings based on their internal structural features, external structural features, environmental parameters and electromagnetic characteristic parameters, so as to obtain the rock cuttings type of the real-time rock cuttings.

3. The method according to claim 1, characterized in that The internal and external structural features of the real-time rock cuttings are extracted from the CT scan data by using a pre-trained neural network model, including: Extracting external structural features of the real-time rock cuttings through shallow convolution of the neural network model, wherein the external structural features include at least one of the following: edge, texture, and morphological contour; Extracting internal structural features of the real-time rock cuttings through deep convolution of the neural network model, wherein the internal structural features include at least one of the following: porosity, density distribution, and internal fractures; Accordingly, the pre-trained neural network model is used to identify the real-time rock cuttings based on their internal and external structural features to obtain the rock cuttings type of the real-time rock cuttings, including: The rock cuttings type of the real-time rock cuttings is output through the fully connected layer of the output layer of the neural network model, wherein the fully connected layer includes multiple nodes, the number of nodes is equal to the number of rock cuttings types, each output node corresponds to a rock cutting type, each output node outputs the probability that the real-time rock cuttings belong to the corresponding rock cutting category, and the rock cutting category with the largest probability value is determined by the softmax activation function as the output result of the output layer.

4. The method according to claim 1, characterized in that: The pre-trained neural network model is trained in the following manner: Obtain historical rock cuttings samples from the target well in the target formation; Scanning the historical rock cuttings sample by a CT device to obtain a CT scan image; Performing denoising on the CT scan image by Gaussian filtering; After denoising, the denoised CT scan images were standardized by linear normalization; Perform image segmentation on the standardized CT scan image, and extract internal and external structural features of the segmented image; Label the extracted internal and external structural features with classification labels; The extracted internal structure features and external structure features, as well as the annotated classification labels, are used as training data sets to train the initial network model to obtain the neural network model.

5. The method according to claim 4, characterized in that The extracted internal structure features and external structure features, as well as the annotated classification labels, are used as training data sets to train the initial network model to obtain the neural network model, including: The extracted internal structure features and external structure features, as well as the annotated classification labels, are used as the first training data set; Acquire geological parameters and historical drilling data of neighboring wells of the target well, wherein the historical drilling data includes: cuttings identification results and formation information of neighboring wells; Using the geological parameters and historical drilling data of the neighboring wells of the target well as a second training data set; The initial network model is trained using the first training data set and the second training data set to obtain the neural network model.

6. The method according to claim 1, characterized in that After the pre-trained neural network model is used to identify the real-time rock cuttings based on their internal and external structural features to obtain the rock cuttings type of the real-time rock cuttings, the method further includes: Determine whether the rock cutting type of the real-time rock cutting identified is accurate; When it is determined that the cuttings type of the obtained real-time cuttings is inaccurate, obtaining cuttings samples of historical adjacent wells of the target well in the target formation; Scanning the historical adjacent well cuttings samples by CT equipment to obtain adjacent well cuttings CT scan images; Extracting internal and external structural features of the historical adjacent well cuttings samples from the adjacent well cuttings CT scanning image; The neural network model is optimized by using the internal and external structural characteristics of historical adjacent well cuttings samples.

7. The method according to claim 1, characterized in that The pre-trained neural network model is a network model obtained by fusing N deep learning models, where N is a positive integer greater than 1; Accordingly, the pre-trained neural network model is used to identify the real-time rock cuttings based on their internal and external structural features to obtain the rock cuttings type of the real-time rock cuttings, including: Using each of the N deep learning models, respectively identify the probability that the real-time rock cuttings belong to each of the M rock cuttings types, where M is a positive integer greater than 1; According to the weight value of each deep learning model in the N deep learning models, weighted average is performed on the M groups of probability values ​​obtained by identification to obtain M probability values; The rock cutting type corresponding to the maximum probability value among the M probability values ​​is used as the rock cutting type of the real-time rock cutting.

8. A rock cuttings type identification system, characterized in that: include: Cuttings transmission channel, used for automatic transmission of real-time cuttings; CT devices arranged on both sides of the cuttings transmission channel are used to perform CT scanning on the real-time cuttings to obtain CT scanning images; Environmental parameter sensors arranged on both sides of the cuttings transmission channel are used to collect the environmental parameters of the cuttings in real time; The central processing unit is connected to the rock cuttings transmission channel, the CT device and the environmental parameter sensor, and is used to control the real-time rock cuttings to pass through the CT device and the environmental parameter sensor in sequence, and identify the rock cuttings type of the real-time rock cuttings according to the CT scan image and the environmental parameters.

9. A rock cuttings type identification device, characterized in that: include: A scanning module, used to scan the real-time rock cuttings generated during the drilling process of the target well in the target formation by means of a CT scanning device to obtain CT scanning data; An extraction module, used to extract the internal structure features and external structure features of the real-time rock cuttings from the CT scanning data through a pre-trained neural network model; The identification module is used to identify the rock cuttings type of the real-time rock cuttings based on the internal structure characteristics and external structure characteristics of the real-time rock cuttings through the pre-trained neural network model.

10. An electronic device comprising a processor and a memory for storing instructions executable by the processor, characterized in that: When the processor executes the instructions, the steps of the method according to any one of claims 1 to 7 are implemented.

Images