Deep rock mass mechanical parameter dynamic sensing method based on multi-source while-drilling data fusion

Abstract

The invention relates to the technical field of data processing, discloses a deep rock mass mechanical parameter dynamic sensing method based on multi-source while-drilling data fusion, and aims at solving the problem that an existing method is poor in precision and reliability. According to the scheme, the method mainly comprises the steps of collection and preprocessing of multi-source while-drilling data, multi-source data fusion and parameter inversion under the constraint of physical rules, prediction model construction of an AI driving model, and data integration and storage. According to the method, the precision and reliability of rock mass mechanical parameter sensing are improved, intelligent sensing of deep engineering rock mass safety is achieved, the data storage efficiency and the value mining capacity are optimized, and the method is suitable for deep geological engineering.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a deep rock mass mechanical parameter dynamic perception method based on multi-source while-drilling data fusion. BACKGROUND

[0002] In the process of deep geological exploration, oil and gas exploitation, deep tunnel and large underground cavern group construction, accurately obtaining in-situ rock mass mechanical parameters (such as uniaxial compressive strength, elastic modulus, Poisson's ratio, etc.) is crucial for wellbore stability analysis, drilling parameter optimization, support design decision and resource assessment. The traditional method mainly relies on indoor testing after drilling cores, which is tedious, long-period and high-cost, and it is difficult to obtain the real mechanical response of rock mass at engineering scale, especially for broken and soft rock mass which is difficult to core, the traditional method is almost ineffective.

[0003] In recent years, the while-drilling measurement technology has developed rapidly, which indirectly evaluates the rock mass properties by analyzing the parameters (such as torque, rotational speed, drilling speed) during drilling. However, the existing while-drilling parameter interpretation and rock mass mechanical perception method still has significant limitations. First, the existing method often relies on a single data source (such as drilling pressure, torque) or simple empirical formula for parameter inversion, which fails to fully utilize the internal correlation between multi-source heterogeneous data, resulting in low precision and insufficient reliability of the interpretation results. Second, the mainstream data-driven artificial intelligence model lacks physical law constraints, and its prediction results often violate the basic principles of rock mass mechanics, such as Poisson's ratio exceeding the reasonable range, poor interpretability, and difficult to be trusted and adopted by field engineers. Third, the downhole working conditions are complex and variable, such as sticking, drilling through broken zones, and static models cannot adaptively adjust parameter correlation logic, leading to prediction failure when the working conditions change suddenly.

[0004] Therefore, there is an urgent need in the field for a deep rock mass mechanical parameter dynamic perception method that can fuse multi-source information, embed domain knowledge, and adaptively adjust to dynamic working conditions, to overcome the shortcomings of existing technology and provide reliable technical support for safe and efficient development of deep resources. SUMMARY

[0005] The present application aims to solve the problem of poor precision and reliability in existing while-drilling parameter interpretation and rock mass mechanical perception method, and proposes a deep rock mass mechanical parameter dynamic perception method based on multi-source while-drilling data fusion.

[0006] The technical solution adopted by the present application to solve the above technical problems is: The deep rock mass mechanical parameter dynamic perception method based on multi-source while-drilling data fusion, the method comprises: Real-time acquisition and preprocessing of multi-source while-drilling data to obtain standardized features corresponding to the multi-source while-drilling data; the multi-source while-drilling data includes while-drilling parameters in the drilling process obtained by a drill pipe built-in sensor array and auxiliary formation parameters obtained by a monitoring tool arranged in the downhole or around the well; Based on the standardized features, a graph neural network combining attention mechanism and physical constraints is constructed; the graph neural network defines different depth points of the borehole as nodes, and the node feature vector contains the while-drilling parameters and the auxiliary formation parameters; the edge weight between nodes is dynamically calculated according to the physical constraints, statistical correlation and spatial distance, and the physical knowledge vector is embedded in the node feature update to guide the update of the node feature through the message passing mechanism; the updated high-dimensional node feature is interpreted for physical properties to output rock mass mechanical parameters and calculate the comprehensive confidence; According to the rock mass mechanical parameters and the comprehensive confidence, a physical-data dual-driven mechanism is used to ensure that the prediction of the rock mass mechanical parameters conforms to the rock mass constitutive relation, a dynamic graph topology reconstruction mechanism adaptive to working conditions is used to identify the current drilling working condition in real time, the edge weight and the adjacency matrix of the graph neural network are dynamically adjusted according to the identified working condition, and a multi-scale spatio-temporal alignment federated learning framework is used to realize multi-drilling rig collaborative training and model updating to output risk indicators and decision suggestions; A database index is constructed based on physical indicators, a hierarchical early warning is performed according to the risk indicators, and a differential dynamic storage strategy is implemented according to the comprehensive confidence.

[0007] Further, the while-drilling parameters include the torque, rotational speed and drilling speed of the drill bit measured and obtained; The auxiliary formation parameters include the longitudinal wave velocity, transverse wave velocity, absolute permeability and compressibility coefficient; The rock mass mechanical parameters include uniaxial compressive strength, elastic modulus, Poisson's ratio, rock strength index and drilling condition compliance index.

[0008] Further, the preprocessing includes: According to the real-time drilling working condition, a noise reduction algorithm is adaptively switched; the real-time drilling working condition includes sticking, broken zone and normal drilling; through depth alignment and outlier rejection, the multi-source while-drilling data is aligned in time series and spatial depth; for different rock types, the corresponding compensation algorithm is called for automatic correction.

[0009] Further, the calculation formula of the edge weight between nodes is: ; Wherein, represents the edge weight between adjacent nodes and node , represents the weight coefficient of the physical constraint term, a weight coefficient representing a statistical correlation term, a weight coefficient representing a spatial distance term, a Young's modulus of a node , a Young's modulus of a node , a tolerance threshold of a Young's modulus, an eigenvector of a node , a Pearson correlation coefficient between an eigenvector of a node , an eigenvector of a node , a spatial coordinate of a node , a spatial coordinate of a node , a maximum neighborhood distance.

[0010] Further, the node feature update adopts a physically guided residual connection mechanism, and the update formula is: ; wherein, a feature vector of a node in a neural network of the th layer, a feature vector of a node in a neural network of the th layer, an edge weight between a node and a node , a feature vector of a node in a neural network of the th layer, respectively represent a learnable weight matrix, a set of adjacent nodes of a node , a GELU activation function, a physical knowledge vector function.

[0011] Further, the output rock mass mechanical parameters are calculated and the comprehensive confidence thereof is calculated, including: The output rock mass mechanical parameters are interpreted by a physical property interpretation formula, and the physical property interpretation formula is as follows: ; wherein, a predicted physical property vector output by a node , including an output vector of multiple rock mass mechanical parameters, This represents the learnable decoding weight matrix. Represents a node The final layer of the graph neural network outputs a high-dimensional feature vector that incorporates multi-source data and physical constraints. Indicates the bias term; After interpreting the physical property vectors, calculate the node-level residuals: ; in, Indicates the first Node-level residuals under each physical rule Indicates based on the predicted physical property vector The theoretical value is obtained by calculation using physical rules and formulas from the rock mechanics knowledge base. This represents the actual theoretical value corresponding to the physical rule. Indicates taking the absolute value; Gaussian noise was added to the predicted physical property vector to perform a perturbation test, and the perturbation sensitivity coefficient was calculated: ; in, Represents the first element in the predicted physical property vector. Disturbance sensitivity coefficients of individual rock mass mechanical parameters. This indicates that the original input rock mechanics parameters are subjected to Gaussian noise and then interpreted by a graph neural network to obtain the perturbed rock mechanics parameters. The L2 norm of a vector; The overall confidence level is calculated using the following formula: ; in, Represents a node The overall confidence level, Represents a node The node-combined residual vector, Represents a node The node-wide integrated disturbance sensitivity coefficient. Represents a node The consistency error between the predicted values ​​of its neighboring nodes and the predicted values ​​of its neighboring nodes. Represents the natural constant.

[0012] Furthermore, the physical-data dual-drive mechanism includes: The core laws of rock mechanics are encoded as graphical structural a priori constraints, and the range of differences in the elastic modulus of adjacent nodes is clearly defined: ; in, Represents a node The elastic modulus, representing a node a modulus of elasticity, representing a configurable tolerance threshold, representing taking an absolute value; designing an edge weight calculation function of physical constraints: wherein, representing an edge weight between adjacent nodes and nodes , representing the association strength of physical properties between nodes and nodes , representing the Pearson correlation coefficient between the permeability of nodes and nodes , representing a weight coefficient.

[0013] Further, the federated learning framework comprises: adopting a loss function with a fusion of physical rule residuals for training, and the expression is: wherein, representing a federated learning global loss function, representing a standard data error term representing a physical rule residual vector, representing a physical constraint weight, representing the modulus of a vector; After each drilling rig node is trained based on local data, only encrypted model gradients are uploaded to the cloud server; The cloud server allocates aggregation weights according to the data volume and reliability of each drilling rig node, and obtains a global model through weighted average aggregation and issues an update.

[0014] Further, the calculation formula of the risk index is as follows: wherein, representing a risk index, representing a uniaxial compressive strength prediction value, representing a maximum horizontal principal stress; The graded early warning comprises: When the risk index is in a first preset range, it is determined as high risk, and an emergency drilling stop command is triggered; When the risk index is in a second preset range, it is determined as medium risk, and a decision suggestion of reducing the rotating speed and starting a support plan is automatically issued; ​​​When the risk indicator is in the third preset range, it is determined as low risk, the current torque, rotating speed and drilling speed are maintained, and continuous monitoring is carried out.

[0015] Further, a differentiated dynamic storage strategy is implemented according to the comprehensive confidence, including: For the rock mass mechanical parameters corresponding to the comprehensive confidence in the fourth preset range, lossless compression storage is adopted, and complete original physical parameters and characteristic vectors are reserved; For the rock mass mechanical parameters corresponding to the comprehensive confidence in the fifth preset range, moderate compression storage is adopted, and only core physical parameters and residual information are reserved; For the rock mass mechanical parameters corresponding to the comprehensive confidence in the sixth preset range, only abnormal identifiers and key error parameters are stored.

[0016] The beneficial effects of the present application are: the deep rock mass mechanical parameter dynamic perception method based on multi-source drilling data fusion provided by the present application overcomes the uncertainty and limitations of single data source inversion models by collecting multi-source drilling data and combining information fusion with an attention mechanism and a physical constraint graph neural network (AGNN), while embedding the rock mass mechanical law deeply into the data fusion process, forcing the fusion and inversion process to strictly comply with physical criteria such as Hooke's law, effectively avoiding unreasonable output of pure data-driven models, and greatly improving the accuracy and reliability of rock mass mechanical parameter inversion; through the working condition adaptive module, sudden conditions such as sticking and broken zones are identified in real time, and the topological structure and parameter association logic of the graph neural network are dynamically adjusted, so that the model can flexibly respond to the complex and variable geological environment underground, further improving the accuracy and reliability of parameter perception; the comprehensive confidence scoring mechanism is introduced, the prediction results of each data point are quantitatively evaluated in multiple dimensions by comprehensively considering the physical rule residual, disturbance sensitivity and spatial consistency error, and a quantifiable confidence evaluation and decision basis is provided; by converting the rock mass mechanical parameters into risk indicators in real time and grading the warning, the intelligent perception of deep engineering rock mass safety is realized, and major safety accidents such as wellbore collapse are effectively avoided; the residual priority dynamic storage strategy based on confidence realizes the best balance between the integrity of key data and storage efficiency under limited storage resources, and optimizes the data storage efficiency and value mining capability. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the deep rock mass mechanical parameter dynamic perception method based on multi-source drilling data fusion provided for the embodiment is shown. DETAILED DESCRIPTION

[0018] Since the current while-drilling parameter interpretation and rock mass mechanics perception method rely on a single data source for parameter inversion prediction, lack of physical law constraints, and are prone to prediction failure due to sudden changes in working conditions, the rock mass mechanics parameters have the problems of poor precision and reliability.

[0019] Based on this, the technical scheme of the present application is proposed. In the present application, first, the while-drilling parameters directly reflecting the mutual feedback between drilling and rock are obtained in real time by the sensor array built in the drill pipe; at the same time, auxiliary parameters characterizing the formation characteristics are obtained by the monitoring tool arranged in the well, realizing multi-dimensional perception of the downhole environment, and the multi-source, heterogeneous, noisy raw data collected are preprocessed to provide high-quality standardized feature data input for subsequent analysis. Then, the borehole space is discretized into a graph structure, where each node represents a depth point, and the feature vector of the node is all the preprocessed parameters of the point; the physical law is deeply embedded in the graph neural network, and the physical constraint edge weight calculation and physical guided feature update force the fusion and inversion process to strictly comply with the physical criteria such as Hooke's law, thereby ensuring the precision and reliability of the rock mass mechanics parameter interpretation; after calculating the edge weight between nodes, the graph neural network starts message passing and node feature updating, and then translates the high-dimensional features output by the AGNN into rock mass mechanics parameters that can be directly applied to engineering through the decoding layer, and evaluates the comprehensive confidence of these parameters. Then, dynamically perceive the changes in working conditions and adjust the connection logic of the graph network accordingly, so that the model can adapt to complex geological environments; at the same time, through the federated learning framework, each drilling machine only shares encrypted model gradients after local training, and aggregates and updates in the cloud, thereby fusing multiple knowledge and improving the generalization ability of the model on the premise of protecting data privacy. Finally, according to its comprehensive confidence, the data is stored differently, giving priority to ensuring the integrity of high-value data, while compressing or only storing summaries of low-value or low-confidence data, thereby intelligently optimizing the utilization rate of storage resources.

[0020] The technical solutions in the embodiments will be clearly and completely described below with reference to the drawings in the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0021] Figure 1 A flowchart of a deep rock mass mechanics parameter dynamic perception method based on multi-source while-drilling data fusion is shown, please refer to Figure 1 The method comprises the following steps: Step 1, acquisition and preprocessing of multi-source while-drilling data: Real-time acquisition of multi-source while-drilling data and preprocessing thereof to obtain standardized features corresponding to the multi-source while-drilling data; the multi-source while-drilling data includes while-drilling parameters obtained by the sensor array built in the drill pipe during drilling, and auxiliary formation parameters obtained by the monitoring tool arranged in the well or around the well.

[0022] In the embodiment, the while-drilling parameters include measured torque, rotation speed and drilling speed of the drill bit; the auxiliary formation parameters include rock density, longitudinal wave speed, transverse wave speed, absolute permeability, compressibility and maximum horizontal principal stress; and the rock mass mechanical parameters include uniaxial compressive strength, elastic modulus, Poisson's ratio, rock strength index and drilling condition compliance index.

[0023] In actual application, the built-in sensor array is arranged in the TBM (Tunnel Boring Machine) cutter drive part and the main bearing, mainly including torque sensors, rotation speed sensors and the like to collect the while-drilling parameters in the drilling process, and the auxiliary formation parameters are obtained through the monitoring tools arranged in the well or around the well, such as acoustic logging sensors, resistivity sensors and the like.

[0024] In the embodiment, the preprocessing includes: adaptively switching the noise reduction algorithm according to the real-time drilling conditions, the real-time drilling conditions including sticking, broken zone and normal drilling; through depth alignment and abnormal value elimination, the multi-source while-drilling data is aligned in time sequence and spatial depth; and for different rock types, the corresponding compensation algorithm is called for automatic correction.

[0025] Specifically, the preprocessing part adopts a three-layer architecture design, including: Signal layer processing: through the working condition adaptive noise reduction, the calculation accuracy of the uniaxial compressive strength, the rock strength index and the drilling condition compliance index is directly improved, the Kalman filter eliminates the torque jump when sticking, thereby reducing the uniaxial compressive strength calculation error; the notch filter preserves the acoustic wave main frequency when in the broken zone, ensuring more accurate Poisson's ratio inversion; and the wavelet noise reduction optimizes the vibration signal when in normal drilling, which is beneficial to more stable calculation of the drilling condition compliance index; Data layer cleaning: through depth alignment and abnormal value elimination, the spatio-temporal consistency of the parameters is ensured; Engineering layer compensation: for different rock types, the parameter deviation caused by environmental interference is automatically corrected.

[0026] Step 2, multi-source data fusion and parameter inversion under physical rule constraints: Based on the standardized features, a graph neural network combining attention mechanism and physical constraints is constructed; the graph neural network defines different depth points of the borehole as nodes, and the node feature vector contains the while-drilling parameters and the auxiliary formation parameters; the edge weight between nodes is dynamically calculated according to the physical constraints, statistical correlation and spatial distance, and the physical knowledge vector is embedded in the node feature update for guidance, and the node features are updated through the message passing mechanism; the updated high-dimensional node features are interpreted for physical properties, and the rock mass mechanical parameters are output, and the comprehensive confidence is calculated.

[0027] This step employs an Aggregate Graph Neural Network (AGNN) that combines attention mechanisms with physical constraints. It dynamically calculates edge weights between nodes using a rock mechanics knowledge base and embeds physical rule residuals into feature updates, forcing the fusion process to strictly adhere to physical laws such as Hooke's Law. Ultimately, it interprets and outputs highly reliable rock mechanics parameters and real-time risk indicators, achieving accurate and interpretable dynamic perception of rock mechanics parameters from multi-source noisy data.

[0028] Specifically, in this embodiment, the borehole space is discretized into a graph structure, where each node represents a depth point, and the node feature vector is all the preprocessed parameters of that point.

[0029] Next, the edge weights between nodes are calculated using the following formula: ; in, Indicates adjacent nodes With nodes Edge weights between them This represents the weighting coefficient of the physical constraint term. The weighting coefficients of the statistical correlation terms. This represents the weighting coefficient of the spatial distance term. Represents a node The elastic modulus, Represents a node The elastic modulus, This represents the tolerance threshold for the elastic modulus. Represents a node eigenvectors With nodes eigenvectors The Pearson correlation coefficient between them Represents a node spatial coordinates and nodes spatial coordinates The Euclidean distance between them This represents the maximum neighborhood distance.

[0030] The larger the edge weight, the more "similar" or "related" the two nodes are in a physical and statistical sense, and the stronger the information transmission. In other words, the larger the result of the formula, the higher the reasonable connection strength between the two nodes.

[0031] In the above formula for calculating edge weights: This represents the physical constraint term, which has the highest percentage and is assigned the highest weight. =0.6; where, and The "intrinsic rigidity" of the batholith, which directly represents the integrity of the rock mass, is the most representative physical constraint, and the difference in elastic modulus between the two nodes helps to avoid the model treating the nodes on both sides of the fault as continuous rock mass; in normal rock mass without faults and fracture zones, the change in elastic modulus within a 5m range usually does not exceed 2GPa, so it can be taken as = 2GPa.

[0032] The statistical correlation term is represented, and the weight is lower than the physical constraint, which can be taken as = 0.3; Quantifies the degree of linear correlation of node observation data in statistics, the value range is [-1, 1], positive value represents positive correlation, negative value represents negative correlation, and the absolute value is greater, the correlation is stronger.

[0033] The spatial distance penalty term is represented, and the weight is the lowest, which can be taken as = 0.1; In drilling, the rock mass properties change little within a 5m range, and beyond 5m, the continuity of the rock mass will be significantly weakened, so it can be taken as = 5m.

[0034] In the calculation formula of the above edge weight, the three rules of "physical constraint (elastic modulus continuity), statistical correlation (permeability correlation), and spatial continuity (distance)" are integrated, and each parameter has a clear rock mechanics basis and engineering experience support, avoiding unreasonable connections driven by pure data from the source. The "isolated data points" in step 1 are translated into "structured relationship network consistent with geological laws", which provides a "high-quality data basis" for subsequent feature fusion, parameter interpretation, and AI prediction.

[0035] In this embodiment, the deterministic rules of rock mechanics (such as wave velocity-elastic modulus relationship, Poisson's ratio value range) are converted into model recognizable vector signals through node feature updating, that is, "physical knowledge vectors" are used to guide model learning, avoiding deviation from mechanical criteria. The node feature update adopts a physical-guided residual connection mechanism, and the update formula is: ; Wherein, represents the feature vector of the node in the first layer neural network, represents the feature vector of the node in the first layer neural network, represents the edge weight between the node and the node , represents the feature vector of the node in the first feature vector of a layer neural network, respectively represent a learnable weight matrix, represent a node adjacent node set of the node, represent a GELU activation function, represent a physical knowledge vector function.

[0036] It should be noted that a purely data-driven model may learn features that violate the laws of mechanics, such as a Poisson's ratio greater than 0.5, while will directly embed the correct physical parameters into the feature update, guide the model learning, and ensure that each layer of feature update does not deviate from the laws of rock mass mechanics.

[0037] It can be understood that the high-dimensional feature vector obtained through node feature update is not directly applicable, and the updated high-dimensional node features need to be interpreted for physical properties, output rock mass mechanical parameters, and calculate their comprehensive confidence, that is, abstract features are mapped to physical property vectors containing rock mass mechanical parameters such as uniaxial compressive strength and elastic modulus, providing directly applicable mechanical parameters for subsequent rock mass stability prediction, parameter inversion and other downstream tasks, and realizing the conversion from fused features to practical parameters. Specifically, it includes: The rock mass mechanical parameters are interpreted and output through the physical property interpretation formula, and the physical property interpretation formula is as follows: ; wherein, represent a node output prediction physical property vector, including an output vector of multiple rock mass mechanical parameters, represent a learnable decoding weight matrix, represent a node high-dimensional feature vector fused with multi-source data and physical constraints output by the final layer of the graph neural network, represent a bias term; After the physical property vector is interpreted, the node-level residual error is calculated: ; wherein, represent the node-level residual error under the physical rule, represent the theoretical value calculated according to the predicted physical property vector through the physical rule formula in the rock mass mechanics knowledge base, represent the true theoretical value corresponding to the physical rule, represent the absolute value; Gaussian noise is added to the predicted physical property vector for perturbation test, and the perturbation sensitivity coefficient is calculated: ; wherein, represents the perturbation sensitivity coefficient of the i-th rock mass mechanical parameter in the predicted physical property vector, represents the perturbed rock mass mechanical parameter obtained by interpreting the original input rock mass mechanical parameter added with Gaussian noise through the graph neural network, represents the L2 norm of a vector; The comprehensive confidence is calculated by the following formula: ; wherein, represents the comprehensive confidence of the node , wherein, represents the node comprehensive residual vector of the node , wherein, represents the node comprehensive perturbation sensitivity coefficient of the node , wherein, represents the consistency error of the node with the predicted values of its neighbor nodes, wherein, represents a natural constant.

[0038] The comprehensive confidence is a comprehensive score of the reliability of the rock mass mechanical parameter: ≥ 0.8 indicates that the parameter is reliable and can be directly used for decisions such as drilling pressure optimization; 0.6≤ <0.8 indicates low confidence, which needs to be verified in combination with downhole photography; <0.6 indicates that it is unreliable and needs to suspend operation and reacquire data.

[0039] Step 3, prediction model construction of the AI-driven model: According to the rock mass mechanical parameters and the comprehensive confidence, a physical-data dual-driven mechanism is used to ensure that the prediction of the rock mass mechanical parameters conforms to the rock mass constitutive relationship, a dynamic graph topology reconstruction mechanism adaptive to working conditions is used to identify the current drilling working condition in real time, the edge weight and the adjacency matrix of the graph neural network are dynamically adjusted according to the identified working condition in real time, and multi-drilling machine collaborative training and model updating are realized through a multi-scale spatio-temporal alignment federated learning framework, and risk indicators and decision suggestions are output.

[0040] It can be understood that step 3 receives the rock mass mechanical parameters and their comprehensive confidence output by step 2, ensures that the parameters conform to the mechanical criteria through a physical-data dual-driven mechanism, responds to drilling mutations through a working condition adaptive dynamic graph topology reconstruction, realizes multi-drilling machine safety collaboration through a multi-scale spatio-temporal alignment federated learning, and forms a closed loop from perception to decision.

[0041] ​The physics-data dual-drive mechanism primarily addresses the problem of traditional pure data-driven models neglecting rock mass constitutive relationships, leading to parameter inversion results that violate fundamental mechanical principles. By deeply integrating physical rules with data learning, the reliability of predicted parameters is ensured. The physics-data dual-drive mechanism includes: The core laws of rock mechanics are encoded as graphical structural a priori constraints, and the range of differences in the elastic modulus of adjacent nodes is clearly defined: ; in, Represents a node The elastic modulus, Represents a node The elastic modulus, This indicates the configurable tolerance threshold. Indicates taking the absolute value; Design the edge weight calculation function for physical constraints: ; in, Indicates adjacent nodes With nodes Edge weights between them Represents a node With nodes The strength of the correlation between their physical properties. Represents a node With nodes The Pearson correlation coefficient between permeability and This represents the weighting coefficient.

[0042] Furthermore, the module employs a physics-guided residual learning mechanism, explicitly embedding physical knowledge such as the conversion formula between wave velocity and elastic parameters during each layer update process. The following update rules are used to bring the prediction results closer to the theoretical values: ; in, Represents a node In the Feature vectors of a layered neural network Represents a node In the Feature vectors of a layered neural network Represents a node With nodes Edge weights between them Represents a node In the Feature vectors of a layered neural network Let represent the learnable weight matrices, Represents a node The set of adjacent nodes, represents the GELU activation function, represents the physical knowledge vector function.

[0043] The working condition adaptive dynamic graph topology reconstruction is used to address the limitations of static models in dealing with sudden working conditions in the drilling process (such as sticking, entering broken zones, etc.). By real-time sensing of working condition changes, dynamically adjusting parameter correlation logic and reconstructing graph topology structure, the model can adapt to complex geological environments.

[0044] Specifically, first, based on the vertical acceleration, torque rate of change and horizontal principal stress gradient data collected by the IMU sensor on the drill rod, a working condition classifier is constructed to realize real-time identification of the current drilling state (normal drilling, sticking, broken zone, etc.). Then, according to the identified working condition label, the parameter correlation logic is dynamically adjusted, for example, in the sticking working condition, to strengthen the correlation between uniaxial compressive strength and related parameters, the weight coefficient of the statistical correlation term in the edge weight formula is increased from 0.3 to 0.7; in the broken zone working condition, to enhance the spatial continuity of rock mass parameters, the weight coefficient of the spatial distance term is reduced from 0.1 to 0.01; and in the normal working condition, the default parameters are maintained. At the same time, the adjacency matrix of the graph is updated in real time according to the working condition label, allowing connections to be established across fault nodes to adapt to the discontinuity of broken zones, ensuring that the model still maintains stable prediction performance when the working condition changes.

[0045] The multi-scale spatio-temporal alignment federated learning framework is used to solve the problem of difficult safe sharing of downhole and cloud data and limited generalization ability of single-drill model when multiple drills are working together. This module realizes safe and efficient collaboration of multiple devices through hierarchical feature compression, physical constraint federated learning and distributed model updating.

[0046] Specifically, first, a rock mass physical feature extractor is designed to compress the features output by the final layer of AGNN, reducing the high-dimensional feature vector to 30% of the original dimension. This not only preserves the core physical information such as elastic modulus and permeability, but also reduces data transmission and ensures data security.

[0047] In the federated learning process, a loss function that combines physical rule residuals is used for training, and its expression is: ; wherein, represents the global loss function of federated learning, represents the standard data error term represents the physical rule residual vector, represents the physical constraint weight (default 0.5), which ensures that the distributed learning process does not violate the rock mass strength theory, represents the norm of the vector.​​

[0048] In terms of model updating, each drilling node trains based on local data and only uploads encrypted model gradients to the cloud server. The cloud server allocates aggregation weights according to the data volume and reliability of each drilling node, and obtains a global model by weighted average aggregation and issues updates. This mechanism protects data privacy while integrating the geological experience of multiple drilling rigs, significantly improving model generalization ability and reducing generalization error.

[0049] The above three mechanisms work together. The physical-data dual driving mechanism ensures the physical reasonableness of parameter prediction, the working condition self-adaptive reconstruction ensures the stable output of the model in complex environment, and the federated learning framework improves the cooperation efficiency of multiple devices, forming a complete closed loop from perception to decision-making, and providing strong support for safe and efficient construction of deep rock mass engineering.

[0050] Step 4, data integration and storage: Based on the physical index, a database index is constructed, a graded early warning is performed according to the risk index, and a differentiated dynamic storage strategy is implemented according to the comprehensive confidence.

[0051] It can be understood that after obtaining the rock mass mechanical parameters, risk indicators and decision recommendations, a physical constraint driven space-time database architecture is constructed to realize the whole process data management from while-drilling perception to intelligent decision-making, and provide data support and real-time response capability for safe and intelligent management of deep engineering rock mass.

[0052] Specifically, to realize efficient data retrieval and intuitive presentation of three-dimensional geological state, a hierarchical index tree structure based on physical indicators is adopted, breaking through the index limitations of traditional single dimension of time or space. The first level index takes wave velocity ratio as the core indicator, and divides the lithology category according to the wave velocity ratio characteristics of different lithology, to quickly locate the lithology attribute of the target stratum; the second level index is based on elastic modulus curvature, and identifies stress concentration area by calculating the second order derivative of elastic modulus at adjacent depth.

[0053] In this embodiment, the calculation formula of the risk index is as follows: ; Wherein, represents the risk index, represents the uniaxial compressive strength prediction value, represents the maximum horizontal principal stress; The graded early warning includes: When the risk index is in the first preset range (such as ), it is determined as high risk, and an emergency stop drilling instruction is triggered to avoid safety accidents such as well wall collapse; When the risk index is in the second preset range (such as When the risk level is determined to be medium, a decision to reduce the rotation speed and activate the support plan will be automatically issued. When the risk indicator is within the third preset range (e.g.) If the risk is low, maintain the current torque, rotation speed, and drilling speed, and continue monitoring.

[0054] It should be noted that the risk indicators are not fixed values, but are dynamically updated based on the statistical results of multi-drilling rig collaborative operations. By integrating geological data from different boreholes, the threshold range is continuously optimized to improve the adaptability and accuracy of risk assessment.

[0055] In this embodiment, implementing a differentiated dynamic storage strategy based on the comprehensive confidence level includes: For the comprehensive confidence level within the fourth preset range (e.g.) The rock mass mechanical parameters corresponding to ≥0.8 are stored using lossless compression, preserving the complete original physical parameters and feature vectors; For the comprehensive confidence level within the fifth preset range (e.g., 0.6≤...), The rock mechanics parameters corresponding to <0.8) are stored using moderate compression, retaining only the core physical parameters and residual information; For the comprehensive confidence level within the sixth preset range (e.g.) The rock mechanics parameters corresponding to <0.6 are stored only for anomaly identifiers and key error parameters.

[0056] Specifically, to ensure the integrity of critical data under limited storage resources, a differentiated storage strategy is implemented based on the overall confidence level: For high confidence data ( For data with a confidence level ≥ 0.8, lossless compression storage is used to preserve the original physical parameters and feature vectors for model iteration and geological pattern analysis; for data with a confidence level ≤ 0.6, lossless compression storage is employed. For data with a confidence level <0.8, moderate compression is used for storage, preserving core parameters and residual information to support subsequent verification and correction; for low-confidence data ( <0.6), only anomaly identifiers and key error parameters are stored, reducing the space occupied by redundant data.

[0057] This dynamic storage strategy ensures the complete retention of high-risk, high-value data while reducing storage costs through targeted compression, achieving a balance between data storage efficiency and security.

[0058] To sum up, the deep rock mass mechanical parameter dynamic perception method based on multi-source drilling data fusion provided by the embodiment overcomes the uncertainty and limitations of single data source inversion model by collecting multi-source drilling data and combining AGNN for information fusion, at the same time, the depth of rock mass mechanical law is embedded in the data fusion process, forcing the fusion and inversion process to strictly comply with physical criteria such as Hooke's law, effectively avoiding the unreasonable output of pure data-driven models, and greatly improving the accuracy and reliability of rock mass mechanical parameter inversion; through the working condition adaptive module, the sudden working conditions such as sticking and broken zone are identified in real time, and the topological structure and parameter correlation logic of the graph neural network are dynamically adjusted, so that the model can flexibly respond to the complex and changeable geological environment underground, further improving the accuracy and reliability of parameter perception; the comprehensive confidence score mechanism is introduced, the prediction results of each data point are quantitatively evaluated in multiple dimensions by comprehensively considering the physical rule residual, disturbance sensitivity and spatial consistency error, and a quantifiable confidence evaluation and decision basis is provided; by converting the rock mass mechanical parameters into risk indicators in real time and grading the warning, the intelligent perception of the safety of deep engineering rock mass is realized, and major safety accidents such as well wall collapse are effectively avoided; based on the confidence residual priority dynamic storage strategy, the best balance between the integrity and storage efficiency of key data under limited storage resources is realized, and the data storage efficiency and value mining ability are optimized.

Claims

1. A method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion, characterized in that, The method includes: Real-time acquisition of multi-source drilling data and preprocessing of the data to obtain standardized features corresponding to the multi-source drilling data; the multi-source drilling data includes drilling parameters acquired by the drill pipe built-in sensor array during the drilling process, and auxiliary formation parameters acquired by monitoring tools deployed downhole or around the well. Based on the standardized features, a graph neural network combining attention mechanism and physical constraints is constructed. The graph neural network defines different depth points of the borehole as nodes, and the node feature vector includes the drilling parameters and auxiliary formation parameters. The edge weights between nodes are dynamically calculated according to physical constraints, statistical correlation and spatial distance, and physical knowledge vectors are embedded in the node feature update for guidance. The node features are updated through a message passing mechanism. The updated high-dimensional node features are interpreted for physical properties, rock mechanics parameters are output and their comprehensive confidence is calculated. Based on the rock mass mechanics parameters and comprehensive confidence level, the prediction of rock mass mechanics parameters is ensured to conform to the constitutive relationship of the rock mass based on the physical-data dual-drive mechanism. Based on the dynamic graph topology reconstruction mechanism of working condition adaptation, the current drilling working condition is identified in real time. The edge weights and adjacency matrix of the graph neural network are dynamically adjusted according to the real-time identified working condition. Multi-scale spatiotemporal aligned federated learning framework is used to realize multi-drilling rig collaborative training and model update, and output risk indicators and decision suggestions. A database index is built based on physical indicators, and graded early warnings are carried out according to the risk indicators. A differentiated dynamic storage strategy is implemented based on the comprehensive confidence level.

2. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The drilling parameters include the measured torque, rotational speed, and drilling rate of the drill bit; The auxiliary formation parameters include P-wave velocity, S-wave velocity, absolute permeability, and compressibility. The rock mass mechanical parameters include uniaxial compressive strength, elastic modulus, Poisson's ratio, rock strength index, and drilling condition compliance index.

3. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The preprocessing includes: Based on the real-time drilling condition adaptive switching noise reduction algorithm, the real-time drilling conditions include stuck pipe, fracture zone and normal drilling; through depth alignment and outlier removal, the multi-source drilling data is aligned in terms of time series and spatial depth; for different rock types, the corresponding compensation algorithm is called for automatic correction.

4. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The formula for calculating the edge weights between the nodes is: ； in, Indicates adjacent nodes With nodes Edge weights between them This represents the weighting coefficient of the physical constraint term. The weighting coefficients of the statistical correlation terms. This represents the weighting coefficient of the spatial distance term. Represents a node The elastic modulus, Represents a node The elastic modulus, This represents the tolerance threshold for the elastic modulus. Represents a node eigenvectors With nodes eigenvectors The Pearson correlation coefficient between them Represents a node spatial coordinates and nodes spatial coordinates The Euclidean distance between them This represents the maximum neighborhood distance.

5. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The node feature update adopts a physically guided residual connection mechanism, and the update formula is: ； in, Represents a node In the Feature vectors of a layered neural network Represents a node In the Feature vectors of a layered neural network Represents a node With nodes Edge weights between them Represents a node In the Feature vectors of a layered neural network Let represent the learnable weight matrices, Represents a node The set of adjacent nodes, This represents the GELU activation function. A vector function representing physical knowledge.

6. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The output rock mass mechanical parameters, and the calculation of their overall confidence level, include: Rock mass mechanical parameters are output by interpreting physical properties using the following formulas: ； in, Represents a node The output predicted physical property vector includes output vectors of multiple rock mass mechanical parameters. This represents the learnable decoding weight matrix. Represents a node The final layer of the graph neural network outputs a high-dimensional feature vector that incorporates multi-source data and physical constraints. Indicates the bias term; After interpreting the physical property vectors, calculate the node-level residuals: ； in, Indicates the first Node-level residuals under each physical rule Indicates based on the predicted physical property vector The theoretical value is obtained by calculating using physical rules and formulas from the rock mechanics knowledge base. This represents the actual theoretical value corresponding to the physical rule. Indicates taking the absolute value; Gaussian noise was added to the predicted physical property vector to perform a perturbation test, and the perturbation sensitivity coefficient was calculated: ； in, Represents the first element in the predicted physical property vector. Disturbance sensitivity coefficients of individual rock mass mechanical parameters. This indicates that the original input rock mechanics parameters are subjected to Gaussian noise and then interpreted by a graph neural network to obtain the perturbed rock mechanics parameters. The L2 norm of a vector; The overall confidence level is calculated using the following formula: ； in, Represents a node The overall confidence level, Represents a node The node-combined residual vector, Represents a node The node-wide integrated disturbance sensitivity coefficient. Represents a node The consistency error between the predicted values ​​of its neighboring nodes and the predicted values ​​of its neighboring nodes. Represents the natural constant.

7. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The physical-data dual-drive mechanism includes: The core laws of rock mechanics are encoded as graphical structural a priori constraints, and the range of differences in the elastic modulus of adjacent nodes is clearly defined: ； in, Represents a node The elastic modulus, Represents a node The elastic modulus, This indicates the configurable tolerance threshold. Indicates taking the absolute value; Design the edge weight calculation function for physical constraints: ； in, Indicates adjacent nodes With nodes Edge weights between them Represents a node With nodes The strength of the correlation between their physical properties. Represents a node With nodes The Pearson correlation coefficient between permeability and This represents the weighting coefficient.

8. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The federated learning framework includes: The training is performed using a loss function that incorporates the residuals of physical rules, and its expression is as follows: ； in, This represents the global loss function of federated learning. Indicates standard data error term Represents the physical rule residual vector. Represents the physical constraint weights. Represents the magnitude of a vector; Each drilling rig node, after training on local data, only uploads encrypted model gradients to the cloud server; The cloud server allocates aggregation weights based on the data volume and reliability of each drilling rig node, and obtains a global model through weighted average aggregation and then distributes updates.

9. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, The formula for calculating the risk indicator is as follows: ； in, Indicates risk indicators, This represents the predicted value of uniaxial compressive strength. Indicates the maximum horizontal principal stress; The tiered early warning system includes: When the risk indicator is within the first preset range, it is judged as high risk and an emergency drilling stop command is triggered. When the risk indicator is within the second preset range, it is judged as medium risk, and a decision suggestion to reduce the rotation speed and activate the support plan is automatically issued. When the risk indicators are within the third preset range, the risk is determined to be low, and the current torque, rotation speed and drilling speed are maintained, while continuous monitoring continues.

10. The method for dynamic sensing of deep rock mass mechanical parameters based on multi-source drilling data fusion according to claim 1, characterized in that, Implement differentiated dynamic storage strategies based on the comprehensive confidence level, including: For the rock mass mechanics parameters corresponding to the comprehensive confidence level within the fourth preset range, lossless compression storage is adopted to retain the complete original physical parameters and feature vectors; For the rock mechanics parameters corresponding to the comprehensive confidence level within the fifth preset range, medium-compression storage is adopted, retaining only the core physical parameters and residual information; For rock mechanics parameters corresponding to a comprehensive confidence level within the sixth preset range, only anomaly identifiers and key error parameters are stored.

Images