Multi-parameter dynamic calibration method and system for measurement while drilling in underground mining

Abstract

The present invention discloses a multi-parameter dynamic calibration method and system for downhole measurement while drilling in a mining well, which relates to the technical field of parameter dynamic calibration, including: dynamically selecting at least one main monitoring parameter from a downhole measurement while drilling parameter set based on a real-time acquired downhole multi-parameter data stream; continuously collecting time-series measurement values ​​of the main monitoring parameter, and calculating the comprehensive confidence impact value of each terminal parameter node; generating a sequence of parameters to be calibrated sorted by calibration priority, and matching the calibration operation type of each parameter; performing calibration on the parameters to be calibrated based on the sequence of parameters to be calibrated and the calibration operation type; and triggering a re-decision process for the main monitoring parameters when a sudden environmental change is detected. The advantages of the present invention are: dynamically selecting the main monitoring parameters for environmental error perception, which not only eliminates the waste of resources and misjudgment risks caused by full parameter scanning, but also ensures the timeliness and reliability of calibration for key drilling conditions by focusing on rapid location of error sources under sudden environmental changes.

Description

Technical Field

[0001] The present invention relates to the technical field of parameter dynamic calibration, and in particular to a multi-parameter dynamic calibration method and system for measurement while drilling in underground mining wells. Background Art

[0002] In the complex downhole environment of measurement while drilling (MWD), real-time multi-parameter calibration presents a significant challenge. Existing technologies generally employ a global, simultaneous parameter monitoring and analysis strategy, resulting in significant computational resource consumption and difficulty meeting the real-time requirements of drilling conditions. Traditional methods require continuous processing of all sensor data, which not only increases the system's computing burden but also can lead to misjudgments and delays due to environmental interference and coupling effects. Especially when downhole conditions change suddenly, the response speed of full-parameter scanning calibration is far slower than the rate of geological change, causing calibration decision lags and cascading errors.

[0003] Recent research has attempted to construct environmental classification models for optimized calibration, but these models still rely on extensive parallel computing resources to build multi-parameter correlation matrices. Such approaches are inefficient in identifying error sources caused by sudden environmental changes and are unable to quickly locate key influencing parameters, resulting in calibration operations that fail to accurately target the core error sources. The mapping relationship between the environment and error is difficult to decouple in real time, forcing the system to maintain a constant high computing load. This not only limits the feasibility of deploying downhole embedded devices but also increases the overall power consumption threshold of while-drilling systems. Summary of the Invention

[0004] To solve the above technical problems, a multi-parameter dynamic calibration method for downhole measurement is provided. This technical solution solves the problem that the response speed of the above-mentioned full-parameter scanning calibration is far lower than the geological change rate, resulting in delayed calibration decisions and causing chain errors.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] The multi-parameter dynamic calibration method for measurement while drilling in underground mining includes:

[0007] Based on the real-time acquired downhole multi-parameter data stream, the current downhole environment type is identified, and according to the preset mapping relationship between the environment type and the main monitoring parameters, at least one main monitoring parameter is dynamically selected from the measurement while drilling parameter set;

[0008] Continuously collect time-series measurement values ​​of the main monitoring parameters and input them into a pre-built error propagation network model. The error propagation network model characterizes the error transmission path through the directed dependency topology and weight coefficients between parameter nodes, activates the corresponding error source nodes according to the abnormal change characteristics of the main monitoring parameters, and calculates the confidence influence value of the activated error source nodes on the associated parameter nodes. The confidence influence values ​​of all activated error source nodes on the associated parameter nodes are summarized, and the comprehensive confidence influence value of each terminal parameter node is calculated;

[0009] The parameter nodes whose comprehensive confidence influence values ​​exceed the dynamic threshold are recorded as parameters to be calibrated, and based on the topological positions of all the parameters to be calibrated in the error propagation network model, a sequence of parameters to be calibrated sorted by calibration priority is generated, and the calibration operation type of each parameter is matched;

[0010] Invoking at least one of a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method according to the sequence of parameters to be calibrated and the type of calibration operation to calibrate the parameters to be calibrated;

[0011] When a sudden change in the environment is detected, the re-decision process of the main monitoring parameters is triggered.

[0012] Preferably, the identifying of the current downhole environment type based on the downhole multi-parameter data stream acquired in real time, and dynamically selecting at least one main monitoring parameter from the measurement while drilling parameter set according to a preset mapping relationship between the environment type and the main monitoring parameter specifically includes:

[0013] Constructing an environmental feature vector based on a multi-parameter downhole data stream, wherein the environmental feature vector includes statistical characteristics of a combination of geological, mechanical, and fluid parameters;

[0014] Outputting an environment type identifier through an environment classification model, wherein the environment classification model is a random forest classifier trained with historical drilling data;

[0015] Based on the output environment type identification, the preset decision matrix is ​​queried according to the label, and the parameter with the highest sensitivity weight is selected as the main monitoring parameter.

[0016] Preferably, the error propagation network model includes a three-layer structure, wherein the first-layer node of the three-layer structure is the selected main monitoring parameter, the middle node of the three-layer structure is the error source, and the end node of the three-layer structure is other parameters except the main monitoring parameter.

[0017] Preferably, the activation of the corresponding error source node according to the abnormal change characteristics of the main monitoring parameter, and calculating the confidence influence value of the activated error source node on the associated parameter node, summarizing the confidence influence values ​​of all activated error source nodes on the associated parameter nodes, and calculating the comprehensive confidence influence value of each terminal parameter node specifically includes:

[0018] Based on the parameter data of the main monitoring parameters collected in real time, extract the real-time data features of the main monitoring parameters;

[0019] Based on the similarity matching between the real-time data features of the main detection parameters and the typical data features of the main detection parameters caused by each error source, if the similarity is greater than the threshold, the corresponding error source node is activated in the error propagation network model; otherwise, the error source node is not activated;

[0020] The terminal node parameters connected to the activated error source node are recorded as associated parameter nodes, and the confidence influence of the activated error source node on all its associated parameter nodes is calculated;

[0021] Based on the confidence influence of all activated error source nodes connected to the terminal node parameter, the comprehensive confidence influence value of each terminal node parameter is determined.

[0022] Preferably, generating a sequence of parameters to be calibrated sorted by calibration priority based on the topological positions of all parameters to be calibrated in the error propagation network model, and matching the calibration operation type of each parameter specifically includes:

[0023] If the parameter to be calibrated is only affected by a single activated error source node, the parameter to be calibrated is divided into an optional calibration domain;

[0024] If the parameter to be calibrated is affected by at least two activated error source nodes and the ranking of the comprehensive confidence impact value is higher than the global median, it is divided into a high-priority calibration domain;

[0025] If the parameter to be calibrated is affected by at least two activated error source nodes, but the ranking of the comprehensive confidence impact value is lower than the global median, it is divided into the mandatory calibration domain.

[0026] Preferably, calling at least one of a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method based on the sequence of parameters to be calibrated and the calibration operation type, and performing calibration on the parameters to be calibrated specifically includes:

[0027] For parameters to be calibrated that are classified as optional calibration domains, a transmission-level calibration method is used, including signal noise reduction filtering and / or sampling rate adaptive improvement;

[0028] For parameters to be calibrated that are classified as high-priority calibration domains, hardware-level calibration is used, including sensor reference voltage reset and / or multi-channel synchronous sampling verification;

[0029] For the parameters to be calibrated that are divided into the mandatory calibration domain, a model-level calibration method is adopted, including residual compensation based on physical equations and / or deep learning prediction deviation correction.

[0030] Preferably, when a sudden change in the environment is detected, the re-decision process of triggering the main monitoring parameters specifically includes:

[0031] Based on the downhole environment type, determine the corresponding standard environment vector;

[0032] Calculate the cosine similarity between the environmental feature vector constructed based on the downhole multi-parameter data stream and the standard environmental vector;

[0033] If the cosine similarity decrease rate of three consecutive sampling windows exceeds the threshold, it is determined to be an environmental mutation.

[0034] Furthermore, a multi-parameter dynamic calibration system for measurement while drilling in underground mining is proposed, which includes an environmental perception and decision-making module, an error propagation analysis module, and a calibration control and execution module:

[0035] The environmental perception and decision-making module includes a multi-channel data acquisition interface, an environmental classification processor, and a main parameter decision maker:

[0036] The multi-channel data acquisition interface is connected to the multi-parameter data stream of the downhole drilling sensor in real time;

[0037] The environment classification processor runs a random forest classification algorithm to extract an environment feature vector based on a multi-parameter data stream and output an environment type label;

[0038] The main parameter decision maker queries the pre-stored decision matrix according to the environment type label and outputs the main monitoring parameter selection instruction;

[0039] The error propagation analysis module includes a three-layer topology network processor and a confidence calculation unit:

[0040] The three-layer topology network processor internally solidifies the directed dependency topology and weight coefficients between parameter nodes and receives parameter data of the main monitoring parameters in real time;

[0041] The confidence calculation unit activates the error source node according to the abnormal characteristics of the main monitoring parameters and calculates its comprehensive confidence impact value on the terminal node;

[0042] The calibration control and execution module includes a dynamic sequence generator, a multi-mode calibration controller and a feedback detection unit:

[0043] The dynamic sequence generator generates a partition parameter sequence to be calibrated based on the comprehensive confidence impact value and the topological position;

[0044] The multi-mode calibration controller calls a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method to perform parameter calibration according to the partition type;

[0045] The feedback detection unit monitors the similarity between the environmental feature vector and the standard environmental vector, and determines whether to trigger a main parameter decision restart signal based on the similarity within a continuous sampling window.

[0046] Optionally, the physical architecture of the three-layer topology network processor is a three-layer pipeline structure implemented by an FPGA chip, specifically including:

[0047] First-layer processing unit: receives the time series data of the main monitoring parameters and performs wavelet packet decomposition to extract high-frequency features;

[0048] Middle layer processing unit: performs similarity matching based on high-frequency features and the stored error source feature library, and activates error source nodes whose similarity exceeds the threshold;

[0049] Terminal layer processing unit: calculates the comprehensive confidence impact value of the activated error source node on the associated terminal parameters and outputs it to the dynamic sequence generator.

[0050] Optionally, the multi-mode calibration controller includes three independent hardware channels;

[0051] Hardware calibration channel: connects to the sensor's reference voltage adjustment circuit and multiplexing switch to perform reference voltage reset or synchronous sampling;

[0052] Transmission optimization channel: embedded with programmable noise reduction filter and bandwidth switching circuit to achieve signal noise reduction and sampling rate improvement;

[0053] Model compensation channel: deploys physical equation residual compensators and neural network predictors to output parameter deviation corrections.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention significantly reduces the computing power consumption brought by traditional multi-parameter global analysis by dynamically selecting the main monitoring parameters as the core nodes of environmental error perception. The changing characteristics of the main monitoring parameters drive the rapid error tracing mechanism, avoiding the real-time calculation requirements of complex correlation models, and enabling the downhole embedded system to achieve rapid calibration response with lower power consumption. At the same time, the targeted calibration strategy based on the main parameter feature identification only performs precise corrections on the associated parameter sub-measurement system affected by error propagation, eliminating the waste of resources and misjudgment risks caused by full parameter scanning, and ensuring the timeliness and reliability of calibration under key drilling conditions by focusing on breakthroughs in the rapid positioning of error sources under sudden environmental changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of the multi-parameter dynamic calibration method for downhole measurement while drilling proposed in Example 1;

[0057] Figure 2 This is a flow chart of the method for dynamically selecting at least one main monitoring parameter from a set of measurement-while-drilling parameters proposed in Example 1;

[0058] Figure 3 This is a flow chart of the method for summarizing and calculating the comprehensive confidence influence value of each terminal parameter node proposed in Example 2;

[0059] Figure 4 This is a flow chart of the method proposed in Example 3 for triggering the re-decision process of the main monitoring parameters when a sudden change in the environment is monitored. DETAILED DESCRIPTION

[0060] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0061] Example 1

[0062] Reference Figure 1 As shown in FIG, a multi-parameter dynamic calibration method for measurement while drilling in underground mining includes:

[0063] Based on the real-time acquired downhole multi-parameter data stream, the current downhole environment type is identified, and according to the preset mapping relationship between the environment type and the main monitoring parameters, at least one main monitoring parameter is dynamically selected from the measurement while drilling parameter set;

[0064] By replacing parallel monitoring of all parameters with targeted focusing on key parameters, the computing resources of the downhole embedded system can be concentrated on the analysis of core sensitive parameters, breaking through the computing power bottleneck of traditional global scanning.

[0065] Continuously collect the time series measurement values ​​of the main monitoring parameters and input them into a pre-built error propagation network model. The error propagation network model characterizes the error transmission path through the directed dependency topology and weight coefficients between parameter nodes, activates the corresponding error source nodes according to the abnormal change characteristics of the main monitoring parameters, and calculates the confidence influence value of the activated error source nodes on the associated parameter nodes. The confidence influence values ​​of all activated error source nodes on the associated parameter nodes are summarized, and the comprehensive confidence influence value of each terminal parameter node is calculated. The error propagation network model includes a three-layer structure, wherein the first-layer nodes are the selected main monitoring parameters, the middle nodes are the error sources, and the terminal nodes are other parameters except the main monitoring parameters;

[0066] Rapidly locate error sources based on a single parameter, avoiding the complex matrix operations of traditional multi-parameter coupling analysis, improving error tracing speed, and quantifying the urgency of calibration through comprehensive confidence impact values ​​to ensure the timeliness of key parameter corrections;

[0067] The parameter nodes whose comprehensive confidence influence values ​​exceed the dynamic threshold are recorded as parameters to be calibrated, and based on the topological positions of all the parameters to be calibrated in the error propagation network model, a sequence of parameters to be calibrated sorted by calibration priority is generated, and the calibration operation type of each parameter is matched;

[0068] Calibration resources are dynamically allocated based on the principle of network hub priority, source-level calibration is implemented for high-error risk parameters, and low-error risk parameters are delayed on demand to achieve a global optimal balance between computing power and calibration accuracy.

[0069] Invoking at least one of a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method according to the sequence of parameters to be calibrated and the type of calibration operation to calibrate the parameters to be calibrated;

[0070] By decoupling the calibration method from the error type, hardware-level calibration directly targets sensor drift, model-level compensation suppresses system error accumulation, and transmission-level optimization blocks signal interference, forming a multi-dimensional defense chain and reducing the overall calibration error rate in complex environments.

[0071] When a sudden change in the environment is detected, the re-decision process of the main monitoring parameters is triggered.

[0072] Establish an environment-parameter dynamic binding mechanism to immediately switch the main monitoring parameters when the lithology changes suddenly or the drilling tool is abnormal, avoid errors caused by environmental mismatch and ensure the real-time reliability of the calibration strategy throughout the drilling process.

[0073] Reference Figure 2 As shown, in this embodiment, based on the real-time acquired downhole multi-parameter data stream, the current downhole environment type is identified, and according to the preset mapping relationship between the environment type and the main monitoring parameter, at least one main monitoring parameter is dynamically selected from the measurement while drilling parameter set, specifically including:

[0074] Constructing environmental feature vectors based on downhole multi-parameter data streams. The environmental feature vectors include statistical characteristics of geological, mechanical, and fluid parameter combinations.

[0075] Outputting the environmental type identifier through the environmental classification model, which is a random forest classifier trained with historical drilling data;

[0076] Based on the output environment type identification, the preset decision matrix is ​​queried according to the label, and the parameter with the highest sensitivity weight is selected as the main monitoring parameter.

[0077] The preset decision matrix is ​​pre-set based on the typical error sources under different environmental types and the sensitivity of various parameters to the typical error sources. By setting the parameters that are highly sensitive to the typical error sources under the environmental type as the main monitoring parameters, the error source information in the downhole environment can be obtained by analyzing the parameter change characteristics of the main detection parameters, and then targeted parameter calibration can be carried out in combination with the error source information.

[0078] By integrating multi-dimensional environmental features with a collaborative mechanism based on a preset decision matrix, an intelligent correlation mapping between environmental types and error sources is achieved, providing an accurate basis for the dynamic selection of primary monitoring parameters: a highly discriminative environmental signature is formed based on the composite statistical characteristics of geological, mechanical, and fluid parameters, enabling the random forest classifier to accurately identify complex downhole working conditions; at the same time, the preset decision matrix anchors the characteristics of typical error sources and parameter sensitivity weights in different environments, ensuring that the selected primary monitoring parameters have the optimal characterization capability for the current core error source. This design allows the system to reversely infer the dominant downhole error source by simply analyzing the changing characteristics of a single primary parameter, significantly reducing the computational complexity of the error tracing process and significantly saving computing power resources on the embedded platform. It also directly triggers targeted parameter calibration based on error source information, effectively eliminating redundant operations caused by traditional global analysis.

[0079] Example 2

[0080] Reference Figure 3 As shown, based on the first embodiment, this embodiment further points out that the corresponding error source node is activated according to the abnormal change characteristics of the main monitoring parameters, and the confidence influence value of the activated error source node on the associated parameter node is calculated, the confidence influence values ​​of all activated error source nodes on the associated parameter nodes are summarized, and the comprehensive confidence influence value of each terminal parameter node is calculated. Specifically, the following steps are included:

[0081] Based on the parameter data of the main monitoring parameters collected in real time, extract the real-time data features of the main monitoring parameters;

[0082] Based on the similarity matching between the real-time data features of the main detection parameters and the typical data features of the main detection parameters caused by each error source, if the similarity is greater than the threshold, the corresponding error source node is activated in the error propagation network model; otherwise, the error source node is not activated;

[0083] The terminal node parameters connected to the activated error source node are recorded as associated parameter nodes, and the confidence influence of the activated error source node on all its associated parameter nodes is calculated;

[0084] Based on the confidence influence of all activated error source nodes connected to the terminal node parameter, the comprehensive confidence influence value of each terminal node parameter is determined.

[0085] Specifically, the influence of an activated error source node on the confidence of all its associated parameter nodes is calculated as follows: in, is the comprehensive confidence impact of the i-th activated error source node on its j-th associated parameter node, is the similarity between the typical data characteristics of the main detection parameters caused by the i-th activated error source node and the real-time data characteristics of the main detection parameters, is the correlation coefficient of the i-th activated error source node to its j-th associated parameter node. In some embodiments, the correlation coefficient is expressed as the error rate of the associated parameter caused by the error;

[0086] Specifically, the calculation formula for the comprehensive confidence impact value of each terminal node parameter is: in, is the comprehensive confidence impact value of the j-th terminal node parameter, is the set of all activated error source nodes connected to the j-th terminal node parameter, For the elements, is the influence of e on the confidence of the j-th associated parameter node.

[0087] Specifically, the calculation process of the comprehensive confidence influence value in this embodiment is illustrated by taking an example;

[0088] If similarity matching is performed based on the real-time data features of the main detection parameters and the typical data features of the main detection parameters caused by each error source, it is concluded that the activated error source nodes are A and B, and the similarity of A is 0.85, and the similarity of B is 0.75. The error rate of parameter t caused by error source A is 0.3, and the error rate of parameter t caused by error source B is 0.5. It can be obtained that error source A will cause the parameter t to have an unreliability of 0.85×0.3=0.255, and error source B will cause the parameter t to have an unreliability of 0.75×0.5=0.375. Without considering the coupling relationship between error sources A and B, under the joint action of error sources A and B, the comprehensive confidence impact value of parameter t is 1-(1-0.255)×(1-0.375)=0.53. That is, under the dual action of the activated error sources A and B, the credibility of parameter t is 1-0.53=0.47.

[0089] This embodiment establishes a dynamic matching mechanism between the main monitoring parameter characteristics and the influence of the error source, accurately activates the associated error nodes through similarity comparison, and objectively quantifies the comprehensive influence of multi-source errors on the terminal parameters based on the probabilistic synthesis model. This design realizes the intelligent deconstruction of the error propagation chain in a complex downhole environment, and transforms the traditional multi-parameter coupling analysis that relies on manual experience into a data-driven mathematical calculation process. With the help of the probabilistic synthesis mechanism of the confidence influence value, the system can deeply restore the nonlinear effect of the superposition and interaction of multiple errors, significantly improve the accuracy of the parameter credibility assessment, and provide a scientific basis for the targeted calibration strategy. At the same time, the activation operation is only performed on the high-similarity associated error sources, which effectively avoids the computing power consumption caused by the traversal of the entire network nodes, so that the limited computing resources are concentrated on the real-time tracking of the key propagation paths, and strengthens the decision-making timeliness and resource utilization efficiency of the embedded platform from the source.

[0090] Specifically, in this embodiment, based on the topological position of all parameters to be calibrated in the error propagation network model, a sequence of parameters to be calibrated sorted by calibration priority is generated, and the calibration operation type of each parameter is matched specifically including;

[0091] If the terminal parameter node is only affected by a single activated error source node, the terminal parameter node is divided into the optional calibration domain;

[0092] If the terminal parameter node is affected by at least two activated error source nodes and the comprehensive confidence impact value is ranked higher than the global median, it is divided into a high-priority calibration domain;

[0093] If the terminal parameter node is affected by at least two activated error source nodes, but the ranking of the comprehensive confidence impact value is lower than the global median, it is divided into the forced calibration domain.

[0094] Specifically, this embodiment further proposes that a transmission-level calibration method be used for the optional calibration domain parameters, including signal noise reduction filtering and / or sampling rate adaptive improvement;

[0095] For high-priority calibration domain parameters, hardware-level calibration is used, including sensor reference voltage reset and / or multi-channel synchronous sampling verification;

[0096] For the mandatory calibration domain parameters, a model-level calibration method is adopted, including residual compensation based on physical equations and / or deep learning prediction bias correction.

[0097] This embodiment constructs a three-layer dynamic calibration domain partitioning mechanism based on confidence influence and topological coupling strength. The terminal parameter node's calibration urgency is intelligently determined by the complexity of its error source and the comprehensive confidence level. Parameters affected by a single error source are placed in the optional calibration domain, allowing the system to defer processing of non-urgent interference; parameters with multi-source coupling and prominent confidence risk are placed in the high-priority calibration domain to ensure timely correction of key parameters; and parameters with multi-source effects but a gentle confidence decay are allocated to the mandatory calibration domain to prevent potential error propagation. The layered calibration strategy at the transmission level, hardware level, and model level precisely matches the characteristics of each partition, significantly improving the intelligent level of system resource allocation. This design not only avoids the resource waste problem caused by traditional uniform calibration, but also achieves the optimal allocation of limited computing power underground through gradient mapping of error propagation depth and calibration strength, making targeted maintenance with millisecond-level response possible, and building a dynamic protection barrier for drilling safety under complex working conditions.

[0098] Example 3

[0099] Reference Figure 4 As shown, based on the first embodiment, this embodiment further points out that when a sudden change in the environment is detected, the re-decision process of triggering the main monitoring parameters specifically includes:

[0100] Based on the downhole environment type, determine the corresponding standard environment vector;

[0101] Calculate the cosine similarity between the environmental feature vector constructed based on the downhole multi-parameter data stream and the standard environmental vector;

[0102] If the similarity decrease rate of three consecutive sampling windows exceeds the threshold, it is determined to be an environmental mutation.

[0103] This embodiment constructs a dynamic tracking and trend analysis mechanism for environmental feature vectors, and intelligently perceives qualitative changes in downhole working conditions through the attenuation trend of the similarity between environmental feature vectors and standard vectors in continuous cycles. This design breaks through the threshold triggering limitations of traditional environmental mutations, and at key turning points of working conditions such as lithologic conversion and drill tool anomalies, it determines the environmental reconstruction node based on the feature mutation rate, and achieves near-zero-delay reconfiguration of the main monitoring parameters. The instant refresh of the environmental mapping relationship effectively blocks the error tracing distortion caused by parameter environment mismatch, establishes an environmentally adaptive elastic decision chain for the downhole drilling system, ensures that the calibration strategy during dynamic drilling is always accurately synchronized with the actual working conditions, and essentially improves the safety redundancy and measurement reliability of drilling in complex formations.

[0104] Example 4

[0105] This embodiment, based on the same inventive concept as embodiments 1 to 3, proposes a multi-parameter dynamic calibration system for measurement while drilling in underground mining, including an environmental perception and decision module, an error propagation analysis module, and a calibration control and execution module:

[0106] Environmental perception and decision-making module, including multi-channel data acquisition interface, environmental classification processor and main parameter decision maker:

[0107] Multi-channel data acquisition interface accesses multi-parameter data streams of downhole drilling sensors in real time;

[0108] The environment classification processor runs a random forest classification algorithm to extract environment feature vectors based on multi-parameter data streams and output environment type labels;

[0109] The main parameter decision maker queries the pre-stored decision matrix according to the environment type label and outputs the main monitoring parameter selection instruction;

[0110] The error propagation analysis module includes a three-layer topology network processor and a confidence calculation unit:

[0111] The three-layer topology network processor internally solidifies the directed dependency topology and weight coefficients between parameter nodes and receives parameter data of the main monitoring parameters in real time;

[0112] The confidence calculation unit activates the error source node according to the real-time data characteristics of the main monitoring parameters, and calculates the comprehensive confidence impact value of the activated error source node on the terminal node;

[0113] Calibration control and execution module, including dynamic sequence generator, multi-mode calibration controller and feedback detection unit:

[0114] The dynamic sequence generator generates a partition parameter sequence to be calibrated based on the comprehensive confidence impact value and topological position;

[0115] The multi-mode calibration controller calls the hardware-level calibration method, the transmission-level calibration method or the model-level calibration method to perform parameter calibration according to the partition type;

[0116] The feedback detection unit monitors the similarity between the environmental feature vector and the standard environmental vector, and determines whether to trigger a main parameter decision restart signal based on the similarity within a continuous sampling window.

[0117] Optionally, the physical architecture of the three-layer topology network processor is a three-layer pipeline structure implemented by an FPGA chip;

[0118] First-layer processing unit: receives the time series data of the main monitoring parameters and performs wavelet packet decomposition to extract high-frequency features;

[0119] Middle layer processing unit: performs similarity matching based on high-frequency features and the stored error source feature library, and activates error source nodes whose similarity exceeds the threshold;

[0120] Terminal layer processing unit: calculates the comprehensive confidence impact value of the activated error source node on the associated terminal parameters and outputs it to the dynamic sequence generator.

[0121] Optionally, the multi-mode calibration controller includes three independent hardware channels;

[0122] Hardware calibration channel: connects to the sensor's reference voltage adjustment circuit and multiplexing switch to perform reference voltage reset or synchronous sampling;

[0123] Transmission optimization channel: embedded with programmable noise reduction filter and bandwidth switching circuit to achieve signal noise reduction and sampling rate improvement;

[0124] Model compensation channel: deploys the physical equation residual compensator and neural network predictor to output parameter deviation correction.

[0125] In summary, the advantages of the present invention are: by dynamically selecting the main monitoring parameters as the core nodes of environmental error perception, the computing power consumption brought by traditional multi-parameter global analysis is significantly reduced, and the changing characteristics of the main monitoring parameters drive the rapid error tracing mechanism, avoiding the real-time calculation requirements of complex correlation models, so that the downhole embedded system can achieve millisecond-level calibration response with lower power consumption; at the same time, the targeted calibration strategy based on the main parameter feature identification only performs precise corrections on the associated parameter subsystems affected by error propagation, which not only eliminates the waste of resources and misjudgment risks caused by full parameter scanning, but also ensures the timeliness and reliability of calibration of key drilling conditions by focusing on breakthroughs in the rapid positioning of error sources under sudden environmental changes.

[0126] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-parameter dynamic calibration method for measurement while drilling in underground mining, characterized in that: include: Based on the real-time acquired downhole multi-parameter data stream, the current downhole environment type is identified, and according to the preset mapping relationship between the environment type and the main monitoring parameters, at least one main monitoring parameter is dynamically selected from the measurement while drilling parameter set; Continuously collect time-series measurement values ​​of the main monitoring parameters and input them into a pre-built error propagation network model. The error propagation network model characterizes the error transmission path through the directed dependency topology and weight coefficients between parameter nodes, activates the corresponding error source nodes according to the abnormal change characteristics of the main monitoring parameters, and calculates the confidence influence value of the activated error source nodes on the associated parameter nodes. The confidence influence values ​​of all activated error source nodes on the associated parameter nodes are summarized, and the comprehensive confidence influence value of each terminal parameter node is calculated; The parameter nodes whose comprehensive confidence influence values ​​exceed the dynamic threshold are recorded as parameters to be calibrated, and based on the topological positions of all the parameters to be calibrated in the error propagation network model, a sequence of parameters to be calibrated sorted by calibration priority is generated, and the calibration operation type of each parameter is matched; Invoking at least one of a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method according to the sequence of parameters to be calibrated and the type of calibration operation to calibrate the parameters to be calibrated; When a sudden change in the environment is detected, the re-decision process of the main monitoring parameters is triggered.

2. The multi-parameter dynamic calibration method for measurement while drilling in a mining well according to claim 1, characterized in that: The method of identifying the current downhole environment type based on the real-time acquired downhole multi-parameter data stream and dynamically selecting at least one main monitoring parameter from the measurement while drilling parameter set according to a preset mapping relationship between the environment type and the main monitoring parameter specifically includes: Constructing an environmental feature vector based on a multi-parameter downhole data stream, wherein the environmental feature vector includes statistical characteristics of a combination of geological, mechanical, and fluid parameters; Outputting an environment type identifier through an environment classification model, wherein the environment classification model is a random forest classifier trained with historical drilling data; Based on the output environment type identification, the preset decision matrix is ​​queried according to the label, and the parameter with the highest sensitivity weight is selected as the main monitoring parameter.

3. The multi-parameter dynamic calibration method for measurement while drilling in a mining well according to claim 2, characterized in that: The error propagation network model includes a three-layer structure, wherein the first-layer nodes of the three-layer structure are selected main monitoring parameters, the middle nodes of the three-layer structure are error sources, and the end nodes of the three-layer structure are other parameters except the main monitoring parameters.

4. The multi-parameter dynamic calibration method for measurement while drilling in a mining well according to claim 3, characterized in that: The steps of activating the corresponding error source node according to the abnormal change characteristics of the main monitoring parameters, calculating the confidence influence value of the activated error source node on the associated parameter node, summarizing the confidence influence values ​​of all activated error source nodes on the associated parameter nodes, and calculating the comprehensive confidence influence value of each terminal parameter node specifically include: Based on the parameter data of the main monitoring parameters collected in real time, extract the real-time data features of the main monitoring parameters; Based on the similarity matching between the real-time data features of the main detection parameters and the typical data features of the main detection parameters caused by each error source, if the similarity is greater than the threshold, the corresponding error source node is activated in the error propagation network model; otherwise, the error source node is not activated; The terminal node parameters connected to the activated error source node are recorded as associated parameter nodes, and the confidence influence of the activated error source node on all its associated parameter nodes is calculated; Based on the confidence influence of all activated error source nodes connected to the terminal node parameter, the comprehensive confidence influence value of each terminal node parameter is determined.

5. The multi-parameter dynamic calibration method for measurement while drilling in a mining well according to claim 4, characterized in that: The generating of a sequence of parameters to be calibrated sorted by calibration priority based on the topological positions of all parameters to be calibrated in the error propagation network model and matching the calibration operation type of each parameter specifically includes: If the parameter to be calibrated is only affected by a single activated error source node, the parameter to be calibrated is divided into an optional calibration domain; If the parameter to be calibrated is affected by at least two activated error source nodes and the ranking of the comprehensive confidence impact value is higher than the global median, the parameter to be calibrated is divided into a high-priority calibration domain; If a parameter to be calibrated is affected by at least two activated error source nodes, but the ranking of its comprehensive confidence impact value is lower than the global median, the parameter to be calibrated is divided into the mandatory calibration domain.

6. The multi-parameter dynamic calibration method for measurement while drilling in underground mining according to claim 5, characterized in that: The step of calling at least one of a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method based on the sequence of parameters to be calibrated and the calibration operation type to calibrate the parameters to be calibrated specifically includes: For parameters to be calibrated that are classified as optional calibration domains, a transmission-level calibration method is used, including signal noise reduction filtering and / or sampling rate adaptive improvement; For parameters to be calibrated that are classified as high-priority calibration domains, hardware-level calibration is used, including sensor reference voltage reset and / or multi-channel synchronous sampling verification; For the parameters to be calibrated that are divided into the mandatory calibration domain, a model-level calibration method is adopted, including residual compensation based on physical equations and / or deep learning prediction deviation correction.

7. The multi-parameter dynamic calibration method for measurement while drilling in a mining well according to claim 6, characterized in that: When a sudden change in the environment is detected, the re-decision process of triggering the main monitoring parameters specifically includes: Based on the downhole environment type, determine the corresponding standard environment vector; Calculate the cosine similarity between the environmental feature vector constructed based on the downhole multi-parameter data stream and the standard environmental vector; If the cosine similarity decrease rate of three consecutive sampling windows exceeds the threshold, it is determined to be an environmental mutation.

8. A multi-parameter dynamic calibration system for measurement while drilling in underground mining, characterized by: The method for implementing a multi-parameter dynamic calibration of measurement while drilling in a mining well according to any one of claims 1 to 7 comprises an environment perception and decision module, an error propagation analysis module, and a calibration control and execution module, wherein: The environmental perception and decision-making module includes a multi-channel data acquisition interface, an environmental classification processor and a main parameter decision maker: The multi-channel data acquisition interface is connected to the multi-parameter data stream of the downhole drilling sensor in real time; The environment classification processor runs a random forest classification algorithm to extract an environment feature vector based on a multi-parameter data stream and output an environment type label; The main parameter decision maker queries the pre-stored decision matrix according to the environment type label and outputs the main monitoring parameter selection instruction; The error propagation analysis module includes a three-layer topology network processor and a confidence calculation unit: The three-layer topology network processor internally solidifies the directed dependency topology and weight coefficients between parameter nodes and receives parameter data of the main monitoring parameters in real time; The confidence calculation unit activates the error source node according to the real-time data characteristics of the main monitoring parameters, and calculates the comprehensive confidence impact value of the activated error source node on the terminal node; The calibration control and execution module includes a dynamic sequence generator, a multi-mode calibration controller and a feedback detection unit: The dynamic sequence generator generates a partition parameter sequence to be calibrated based on the comprehensive confidence impact value and the topological position; The multi-mode calibration controller calls a hardware-level calibration method, a transmission-level calibration method, or a model-level calibration method to perform parameter calibration according to the partition type of the parameter sequence to be calibrated; The feedback detection unit monitors the similarity between the environmental feature vector and the standard environmental vector, and determines whether to trigger a main parameter decision restart signal based on the similarity within a continuous sampling window.

9. The multi-parameter dynamic calibration system for measurement while drilling in underground mining according to claim 8, characterized in that: The physical architecture of the three-layer topology network processor is a three-layer pipeline structure implemented by an FPGA chip, specifically including: First-layer processing unit: receives the time series data of the main monitoring parameters and performs wavelet packet decomposition to extract high-frequency features; Middle layer processing unit: performs similarity matching based on high-frequency features and the stored error source feature library, and activates error source nodes whose similarity exceeds the threshold; Terminal layer processing unit: calculates the comprehensive confidence impact value of the activated error source node on the associated terminal parameters and outputs it to the dynamic sequence generator.

10. The multi-parameter dynamic calibration system for measurement while drilling in underground mining according to claim 8, characterized in that: The multi-mode calibration controller includes three independent hardware channels, specifically: Hardware calibration channel: The hardware calibration channel is connected to the sensor's reference voltage adjustment circuit and multiplexing switch to perform reference voltage reset or multi-channel synchronous sampling verification; Transmission optimization channel: The transmission optimization channel embeds a programmable noise reduction filter and bandwidth switching circuit to achieve signal noise reduction filtering and adaptive sampling rate improvement; Model compensation channel: The model compensation channel deploys the physical equation residual compensator and the neural network predictor to output the deviation correction of the parameters to be calibrated.

Images