A shield geological prediction method and system for multiple uncertain working conditions

Abstract

The embodiment of the present application provides a shield geological prediction method and system facing multiple uncertain working conditions, and belongs to the technical field of geological prediction. The method comprises the following steps: acquiring shield operation data and corresponding geological label data formed in a shield construction process, associating the shield operation data with the geological label data based on a tunneling sequence, and constructing a basic data set for shield geological prediction; generating multiple uncertain working condition data sets; under each uncertain working condition, constructing a shield geological prediction model, and outputting a geological prediction result and uncertainty representation information corresponding to the geological prediction result; and outputting applicability evaluation results of the shield geological prediction model under each uncertain working condition. The present application can simultaneously output the shield geological prediction result and its uncertainty representation information under different uncertain working conditions, so as to quantitatively evaluate the reliability degree and applicable range of the prediction result under different geological conditions and operation states.

Description

Technical Field

[0001] This invention relates to the field of geological prediction technology, and more specifically to a method and system for geological prediction of tunnel boring machines under multiple uncertain working conditions. Background Technology

[0002] During shield tunnel construction, geological conditions have a significant impact on construction safety and equipment operation. To understand geological changes during tunneling, engineering practice typically involves analyzing and assessing the geological conditions of the tunneling section by combining shield tunneling operating parameters, geological survey data, and construction records.

[0003] With the development of construction data acquisition methods, geological prediction methods based on tunnel boring machine (TBM) operation data have gradually been applied. Existing methods mostly predict the geological conditions during tunneling by establishing a mapping relationship between TBM operation parameters and geological conditions. Under relatively stable working conditions and when the training data matches the application scenario, these methods can achieve certain results.

[0004] However, in actual shield tunneling, the shield operating parameters are not only affected by geological conditions, but also by the switching of propulsion control strategies, attitude adjustments, and parameter adjustments. When the control strategy changes, the operating parameters may generate abrupt signals that are not directly related to the geological changes. If such control behavior interference is not identified and corrected, non-geological response information can easily be introduced in the sample construction stage, thereby affecting the prediction model's ability to accurately represent the geological state.

[0005] Furthermore, under different construction sections or different line conditions, the coverage of geological types, the proportion of sample composition, and the distribution characteristics of operating parameters may vary. Existing methods usually only conduct experimental verification in different scenarios and lack technical means to uniformly quantify the disturbance intensity between different uncertain working conditions, making it difficult to conduct comparable applicability analysis of prediction results under different working conditions.

[0006] Meanwhile, the composition and characteristics of the data used for modeling may fluctuate due to the incompleteness of data acquisition and noise interference. Under these uncertain conditions, existing prediction methods often still output deterministic prediction results, which are difficult to reflect the reliability of the prediction results and the applicability of the model to the current working conditions. Moreover, existing methods usually output prediction results and their confidence levels for a single tunneling unit, but do not perform segment-level cumulative analysis of the prediction uncertainty changes of multiple consecutive tunneling units. When the prediction confidence level shifts within consecutive segments, there is a lack of a quantitative mechanism to characterize the trend of segment-level risk changes.

[0007] Therefore, it is necessary to provide a geological prediction method for shield tunneling under multiple uncertain working conditions. While predicting the geological state, the method evaluates the uncertainty of the prediction results and the applicability of the model through decoupling of control behavior, quantification of working condition disturbance intensity, and section-level uncertainty accumulation analysis mechanism. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for predicting geological conditions in tunnel boring machines (TBMs) under multiple uncertain working conditions, so as to at least solve the problem of insufficient applicability of existing TBM geological prediction methods under complex construction conditions.

[0009] To achieve the above objectives, the first aspect of the present invention provides a method for geological prediction of tunnel boring machines (TBMs) under multiple uncertain working conditions. The method includes: acquiring TBM operation data and corresponding geological label data generated during TBM construction; performing control behavior identification processing on the TBM operation data to generate control behavior status identifiers; and decoupling and correcting operational parameter mutations caused by control strategy switching based on the control behavior status identifiers to obtain equivalent stratum response operation data; associating the equivalent stratum response operation data with the geological label data based on the tunneling sequence to construct a basic dataset for TBM geological prediction; and performing working condition construction processing based on the basic dataset to generate multiple uncertain working condition datasets, and then performing analysis on each... Uncertainty condition calculation is used to quantify the degree of condition disturbance. Under each uncertainty condition, a shield tunneling geological prediction model is constructed based on the corresponding uncertainty condition dataset. The corresponding equivalent stratum response operation data is input into the shield tunneling geological prediction model, and the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results are output. Based on the uncertainty characterization information of multiple consecutive tunneling units, the cumulative offset of the section is calculated to form a section-level uncertainty cumulative characterization result. Based on the geological prediction results, the section-level uncertainty cumulative characterization result, the condition disturbance intensity vector, and the geological label data, the applicability evaluation result of the shield tunneling geological prediction model under each uncertainty condition is output.

[0010] Optionally, the shield tunneling operation data is processed to identify control behavior to generate a control behavior status identifier. Based on the control behavior status identifier, the sudden changes in operating parameters caused by the control strategy switching are decoupled and corrected to obtain equivalent ground response operation data. This includes: arranging the shield tunneling operation data in a time series according to the tunneling sequence and extracting control feature parameters to characterize the propulsion control behavior; calculating a control change rate sequence based on the variation amplitude of the control feature parameters between consecutive tunneling units, and comparing the control change rate sequence with a preset change rate threshold to identify the control strategy switching time; after identifying the control strategy switching time, generating a control behavior status identifier for the corresponding tunneling unit, and establishing a mapping relationship between the control behavior status identifier and the shield tunneling operation data of the corresponding tunneling unit; based on the control behavior status identifier, dividing the operating parameter sequence within the control strategy switching interval into a transition section, and performing trend separation processing on the operating parameters within the transition section to obtain a residual signal reflecting the ground response, which serves as the equivalent ground response operation data.

[0011] Optionally, the geological label data includes any one or more of geological process labels, stratigraphic state labels, and comprehensive geological category labels; the equivalent stratigraphic response operation data is associated with the geological label data based on the tunneling sequence to construct a basic dataset for shield tunneling geological prediction, including: numbering the equivalent stratigraphic response operation data according to the tunneling sequence to form an operation data sequence arranged by tunneling unit; numbering the geological label data according to the corresponding tunneling unit to form a geological label sequence arranged by tunneling unit; using the equivalent stratigraphic response operation data of the current tunneling unit as input features and the geological label data of the next tunneling unit adjacent to the current tunneling unit as supervision labels to construct prediction unit samples; performing the prediction unit sample construction process on all tunneling units to generate a sample set, and determining the sample set as the basic dataset for shield tunneling geological prediction.

[0012] Optionally, a working condition construction process is performed based on the basic dataset to generate multiple uncertain working condition datasets. For each uncertain working condition, a working condition disturbance intensity vector is calculated to quantify the degree of working condition disturbance. This includes: determining data partitioning conditions for working condition construction based on the correspondence between shield tunneling operation data and geological label data in the basic dataset; selecting different data subsets from the basic dataset according to the data partitioning conditions, respectively serving as working condition datasets under different uncertain working conditions; calculating the statistical offset and label coverage missing rate of the shield tunneling operation data based on the data distribution differences and geological label coverage differences between each uncertain working condition dataset and the basic dataset; calculating the sample ratio offset and operating parameter disturbance amplitude based on the sample quantity ratio between the uncertain working condition dataset and the basic dataset, and the operating parameter disturbance processing rules in the uncertain working condition dataset; and combining the statistical offset, the label coverage missing rate, the sample ratio offset, and the operating parameter disturbance amplitude to form a working condition disturbance intensity vector.

[0013] Optionally, under each uncertain working condition, a shield tunneling geological prediction model is constructed based on the corresponding uncertain working condition dataset, including: for each uncertain working condition dataset, determining the modeling data used for model construction from the corresponding uncertain working condition dataset; based on the modeling data, learning the model parameters of the shield tunneling geological prediction model under the constraints of the corresponding uncertain working condition, so that the shield tunneling geological prediction model forms a representation of the mapping relationship between the shield tunneling operation data and the geological label data under the uncertain working condition, thereby obtaining the shield tunneling geological prediction model corresponding to each uncertain working condition.

[0014] Optionally, the equivalent stratum response operation data is input into the shield tunneling geological prediction model, and the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results are output. This includes: inputting the equivalent stratum response operation data into the corresponding uncertainty condition prediction model in sequence according to the tunneling order to obtain geological prediction results corresponding to each tunneling unit; based on the prediction confidence distribution information output by the shield tunneling geological prediction model in the process of generating the geological prediction results, calculating the confidence index and distribution dispersion index corresponding to each geological prediction result, and determining the confidence index and distribution dispersion index as the uncertainty characterization information corresponding to each geological prediction result.

[0015] Optionally, the cumulative offset of a section is calculated based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result, including: arranging the uncertainty characterization information according to the tunneling sequence and constructing a sliding window with a preset number of tunneling units; in each sliding window, the confidence index and distribution dispersion index in the uncertainty characterization information are accumulated and calculated to obtain the section cumulative offset of the corresponding window; and arranging the section cumulative offsets corresponding to each sliding window according to the tunneling sequence to form a section-level uncertainty cumulative characterization result.

[0016] Optionally, based on the geological prediction results, the segment-level uncertainty accumulation characterization results, the working condition disturbance intensity vector, and the geological label data, the applicability evaluation results of the shield tunneling geological prediction model under various uncertain working conditions are output, including: calculating a prediction consistency index based on the correspondence between the geological prediction results and the geological label data; performing a correlation calculation between the prediction consistency index and the segment-level uncertainty accumulation characterization results to obtain a segment stability index; performing a weighted combination calculation between the segment stability index and the working condition disturbance intensity vector to generate a working condition fit index; and outputting the applicability evaluation results of the shield tunneling geological prediction model under various uncertain working conditions based on the working condition fit index.

[0017] A second aspect of this invention provides a shield tunneling geological prediction system for multiple uncertain working conditions. The system includes: a data acquisition unit, configured to acquire shield tunneling operation data and corresponding geological tag data generated during shield tunneling construction; perform control behavior identification processing on the shield tunneling operation data to generate control behavior status identifiers; and decouple and correct sudden changes in operating parameters caused by control strategy switching based on the control behavior status identifiers to obtain equivalent stratum response operation data; associate the equivalent stratum response operation data with the geological tag data based on the tunneling sequence to construct a basic dataset for shield tunneling geological prediction; and a processing unit, configured to perform working condition construction processing based on the basic dataset to generate multiple uncertain working condition datasets, and address each uncertainty... The working condition calculation is used to quantify the degree of working condition disturbance by a working condition disturbance intensity vector; the prediction unit is used to construct a shield tunneling geological prediction model based on the corresponding uncertain working condition dataset under various uncertain working conditions, input the corresponding equivalent stratum response operation data into the shield tunneling geological prediction model, output the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results, and calculate the cumulative offset of the section based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result; the output unit is used to output the applicability evaluation result of the shield tunneling geological prediction model under various uncertain working conditions based on the geological prediction results, the section-level uncertainty cumulative characterization result, the working condition disturbance intensity vector and the geological label data.

[0018] On the other hand, the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described shield tunneling geological prediction method for multiple uncertain working conditions.

[0019] Through the above technical solution, this invention systematically correlates operational data and geological labeling data generated during shield tunneling construction. Furthermore, during the sample construction phase, it identifies and decouples the impact of control strategy switching on the operational data, enabling the data used in modeling to more accurately reflect the geological response characteristics and reducing the influence of control behavior interference on the prediction results. Based on this, it constructs datasets for multiple uncertain working conditions, freeing shield tunneling geological prediction from the single assumption that training data and application scenarios are identically distributed.

[0020] By quantifying the degree of disturbance under different uncertain working conditions, a comparable working condition disturbance intensity vector is formed. Prediction models are constructed under different uncertain working conditions, and corresponding geological prediction results and uncertainty characterization information are output. This invention can simultaneously reflect the model's understanding of the current working conditions during the prediction process.

[0021] Furthermore, by performing segment-level cumulative analysis of uncertainty characterization information from multiple consecutive tunneling units, the predicted risk is extended from a single tunneling unit to the segment scale, enabling the identification of persistent prediction deviation trends. By combining prediction results, uncertainty characterization information, and real geological label data, the applicability of the model under various uncertain conditions is systematically evaluated, thus avoiding the bias caused by measuring model performance solely based on a single prediction accuracy index. This technical solution expands the shield tunneling geological prediction results from a single deterministic output to an evaluation result containing applicability information, providing a basis for model deployment and working condition selection, and better meeting the engineering needs of complex and variable geological conditions during actual shield tunneling construction.

[0022] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0023] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the steps of a shield tunneling geological prediction method for multiple uncertain working conditions provided by one embodiment of the present invention; Figure 2 This is a detailed flowchart of step S10 of the shield tunneling geological prediction method for multiple uncertain working conditions provided by one embodiment of the present invention. Figure 3 This is a schematic diagram of the uncertainty distribution and decomposition results of shield tunneling geological prediction under different geological conditions provided by one embodiment of the present invention; Figure 4 This is a schematic diagram of the uncertainty distribution and decomposition results of shield tunnel geological prediction under changing operating conditions provided by one embodiment of the present invention; Figure 5 This is a system structure diagram of a shield tunneling geological prediction system for multiple uncertain working conditions provided by one embodiment of the present invention; Figure 6 This is an internal structural diagram of a computer device provided in one embodiment of the present invention. Detailed Implementation

[0024] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0025] like Figure 1 As shown, embodiments of the present invention provide a method for geological prediction of tunnel boring machines under multiple uncertain conditions, the method comprising: Step S10: Obtain shield operation data and corresponding geological label data generated during shield tunneling construction; perform control behavior identification processing on the shield operation data to generate control behavior status identifiers; and decouple and correct the sudden changes in operation parameters caused by control strategy switching based on the control behavior status identifiers to obtain equivalent stratum response operation data; associate the equivalent stratum response operation data with the geological label data based on the tunneling sequence to construct a basic dataset for shield tunneling geological prediction.

[0026] Specifically, by acquiring shield operation data generated during shield tunneling and combining it with corresponding geological tag data, control behavior identification processing is performed on the shield operation data. This identifies abnormal fluctuations in operation parameters caused by changes in control strategies such as propulsion parameter adjustments and attitude control switching. Based on the generated control behavior status identifiers, these abnormal fluctuations are decoupled and corrected to obtain equivalent ground response operation data that characterizes the actual ground response features. Correlation processing of the two types of data according to the shield tunneling sequence establishes a correspondence between the shield operation status and the geological status at both temporal and spatial levels.

[0027] Data organization based on the tunneling sequence ensures consistency between shield tunneling operation data and geological labeling data at the tunneling ring scale. A clear input-output mapping relationship is established through prediction unit construction rules between the current and next tunneling units. This facilitates accurate characterization of the relationship between shield tunneling behavior and subsequent geological changes. The foundational dataset built upon this basis can uniformly carry shield tunneling operation information and geological state information, providing a stable and consistent data foundation for subsequent working condition construction, model building, and prediction evaluation. This avoids analytical biases caused by data misalignment or unclear correlations, thereby ensuring the feasibility and consistency of subsequent shield tunneling geological prediction processes. Specifically, for example... Figure 2 Step S10 includes the following steps: Step S101: Arrange the shield tunneling operation data into a time series according to the tunneling sequence, and extract control feature parameters to characterize the propulsion control behavior.

[0028] Specifically, arranging the tunnel boring machine (TBM) operation data in a time series according to the tunneling sequence means using the tunneling unit as the basic organizational unit, and numbering and sorting the propulsion parameters, attitude parameters, cutterhead operation parameters, and auxiliary control parameters corresponding to each tunneling unit according to the actual construction sequence to form a continuous time series data structure. This time series arrangement uses the tunneling unit number as the primary index, ensuring that the operation data between adjacent tunneling units maintains a logical correspondence, providing continuity constraints for identifying subsequent changes in control behavior.

[0029] After arranging the time series, control feature parameters characterizing the propulsion control behavior are extracted from the tunnel boring machine (TBM) operation data. These control feature parameters describe the adjustment behavior of the propulsion system across different tunneling units and may include propulsion speed setpoints, thrust control commands, rotational speed adjustments, attitude corrections, or other parameters reflecting changes in the control strategy. The extraction of these control feature parameters is based on the original record fields in the TBM operation data, without altering the original data source; only the data undergoes structured filtering and organization.

[0030] It should be noted that the control characteristic parameters in this step are not limited to a specific type or a fixed number of parameters. Any operational data that can characterize the changing trend of the control strategy can be included in the processing scope as control characteristic parameters. The implementation method of this step is not limited to a specific construction line or specific geological conditions. Any technical solution that organizes shield tunneling operation data in a time sequence and extracts control characteristic parameters based on the tunneling sequence falls within the protection scope of this invention.

[0031] Step S102: Based on the variation amplitude of the control characteristic parameters between continuous tunneling units, calculate the control change rate sequence and compare the control change rate sequence with a preset change rate threshold to identify the control strategy switching time.

[0032] Specifically, calculating the control change rate sequence based on the variation amplitude of the control characteristic parameters between consecutive tunneling units involves performing a differential operation on the control characteristic parameters of adjacent tunneling units according to the tunneling sequence after the time series arrangement is completed. This yields the parameter variation amount between each tunneling unit, and the variation amount is then normalized based on the corresponding tunneling unit interval to form the control change rate sequence. The control change rate sequence characterizes the intensity of change in propulsion control behavior between adjacent tunneling units, and its numerical value reflects the degree of change in the adjustment amplitude of the control strategy.

[0033] After obtaining the control change rate sequence, it is compared with a preset change rate threshold to identify the control strategy switching moment. Specifically, when the control change rate between consecutive tunneling units exceeds the preset change rate threshold, it is determined that a control strategy switching behavior exists at the corresponding tunneling unit, and the tunneling unit is marked as the control strategy switching moment. The preset change rate threshold can be obtained based on historical construction data statistics or can be preset according to the equipment operating characteristics; its value is used to distinguish between normal fine-tuning behavior and strategy-level adjustment behavior.

[0034] It should be noted that the calculation method of the control rate of change sequence is not limited to the simple difference form, but can also adopt the sliding window average difference, weighted difference or multi-parameter joint rate of change calculation method; the preset rate of change threshold can also be set separately for different control characteristic parameters.

[0035] Step S103: After identifying the control strategy switching time, generate the control behavior status identifier of the corresponding tunneling unit, and establish a mapping relationship between the control behavior status identifier and the shield operation data of the corresponding tunneling unit.

[0036] Specifically, after identifying the control strategy switching moment, a control behavior state identifier is generated for the tunneling unit corresponding to the control strategy switching moment. This control behavior state identifier is used to distinguish the control behavior states of different tunneling units. It can adopt a discrete state encoding form, for example, dividing the tunneling unit into categories such as control stable state, control transition state, or control adjustment state, and assigning a corresponding state identifier value to each category. The generation of the control behavior state identifier is based on the comparison result between the aforementioned control change rate sequence and a preset change rate threshold. Tunneling units that meet the switching judgment conditions are marked, and the states of several adjacent tunneling units are extended according to the principle of continuity, thereby forming a complete control behavior state distribution sequence.

[0037] After generating the control behavior status identifier, a mapping relationship is established between the control behavior status identifier and the shield tunneling operation data of the corresponding tunneling unit. Specifically, in the shield tunneling operation data structure arranged in the tunneling sequence, a corresponding control behavior status identifier field is added to each tunneling unit, so that the control behavior status identifier and the operation parameters of the tunneling unit form a one-to-one correspondence. By establishing this mapping relationship, when performing segmented processing or decoupling correction of the operation parameters, the operation data under different states can be distinguished and selected based on the control behavior status identifier.

[0038] Step S104: Based on the control behavior status identifier, the sequence of operating parameters in the control strategy switching interval is divided into transition segments, and trend separation processing is performed on the operating parameters in the transition segments to obtain residual signals reflecting the formation response, which are used as equivalent formation response operating data.

[0039] Specifically, the tunneling unit marked as being in a control strategy switching state and its adjacent consecutive tunneling units are defined as the control strategy switching interval, and the sequence of operating parameters within the control strategy switching interval is divided into transition sections. These transition sections characterize the range of non-formation response changes in operating parameters caused by the control strategy adjustment, thus distinguishing them from the control stability interval.

[0040] After the transition zone is defined, trend separation processing is performed on the operating parameters within the transition zone. This trend separation processing involves decomposing the operating parameter sequence within the transition zone into trend and fluctuation terms, separating the trend variation components caused by control strategy adjustments from the original operating parameters, and retaining the residual variation components related to the formation response. Specific implementation methods may include linear trend fitting and subtraction processing based on a sliding window, trend estimation processing based on polynomial fitting, or low-frequency component separation processing based on filtering algorithms. Through these processes, a residual signal reflecting the formation response characteristics is obtained.

[0041] The residual signal, as equivalent formation response operational data, is used to replace the original shield tunneling operational data in the construction of the basic dataset and subsequent model building process. It should be noted that the specific algorithm for trend separation processing is not limited to a single mathematical model, as long as it can distinguish and separate the control strategy adjustment trend from the formation response variation components.

[0042] Step S105: Based on the tunneling sequence, associate the equivalent stratum response operation data with the geological label data to construct a basic dataset for shield tunneling geological prediction.

[0043] Specifically, the geological label data includes any one or more of geological process labels, stratigraphic state labels, and comprehensive geological category labels.

[0044] Furthermore, the equivalent stratum response operation data is numbered according to the tunneling sequence to form an operation data sequence arranged by tunneling unit; the geological label data is numbered according to the corresponding tunneling unit to form a geological label sequence arranged by tunneling unit; using the equivalent stratum response operation data of the current tunneling unit as input features and the geological label data of the next tunneling unit adjacent to the current tunneling unit as supervision labels, a prediction unit sample is constructed; the prediction unit sample construction process is performed on all tunneling units to generate a sample set, and the sample set is determined as the basic dataset for shield tunneling geological prediction.

[0045] In this embodiment of the invention, after obtaining the equivalent formation response operational data, the equivalent formation response operational data is associated with the geological label data based on the tunneling sequence to construct a basic dataset for shield tunneling geological prediction. The association process uses the tunneling unit as the basic organizational unit, establishing a temporal and spatial correspondence between the operational data and the geological labels through the tunneling sequence, thus forming a one-to-one matching structure between the two types of data at the same tunneling scale.

[0046] Specifically, the geological tagging data may include any one or more of geological process tags, stratigraphic state tags, and comprehensive geological category tags. Geological process tags characterize the attributes of geological change stages during tunneling, stratigraphic state tags describe the stratigraphic structure characteristics corresponding to the tunneling section, and comprehensive geological category tags are used to categorize geological conditions. The aforementioned tagging data may originate from construction records, geological survey data, or on-site assessment results, and be stored accordingly according to tunneling units.

[0047] During the data organization phase, the equivalent formation response operational data are numbered according to the tunneling sequence, forming an operational data sequence arranged by tunneling unit; simultaneously, the geological label data are numbered according to the corresponding tunneling unit, forming a geological label sequence arranged by tunneling unit. This numbering mechanism ensures that each tunneling unit corresponds to a unique operational data and geological label record, guaranteeing clear data correlation.

[0048] When constructing prediction unit samples, the equivalent geological response data of the current tunneling unit is used as input features, and the geological label data of the next adjacent tunneling unit is used as supervision labels, thereby establishing a clear prediction relationship between preceding and subsequent units. This method can characterize the mapping relationship between the operational state and subsequent geological changes, ensuring that the sample construction conforms to the actual law of geological state changes along the tunneling direction during shield tunneling. The above prediction unit sample construction process is performed on all tunneling units to form a sample set, which is then determined as the basic dataset for shield tunneling geological prediction.

[0049] The basic dataset constructed through the above steps structurally includes both operational information decoupled from control behavior and corresponding geological state information, and maintains a continuous tunneling sequence relationship among the samples. This provides a unified and coherent data foundation for subsequent working condition construction, prediction model training, and applicability evaluation. Any technical solution that uses tunneling sequence to number and associate equivalent stratum response operational data and geological label data to construct prediction unit samples falls within the protection scope of this invention.

[0050] Step S20: Perform working condition construction processing based on the basic dataset to generate multiple uncertain working condition datasets, and calculate the working condition disturbance intensity vector for each uncertain working condition to quantify the degree of working condition disturbance.

[0051] Specifically, based on the correspondence between shield tunneling operation data and geological label data in the basic dataset, data partitioning conditions for constructing working conditions are determined. Based on these conditions, different data subsets are selected from the basic dataset as working condition datasets for different uncertain working conditions. Based on the differences in data distribution and geological label coverage between each uncertain working condition dataset and the basic dataset, the statistical offset and label coverage missing rate of the shield tunneling operation data are calculated. Based on the sample quantity ratio between the uncertain working condition dataset and the basic dataset, and the rules for handling operational parameter disturbances in the uncertain working condition dataset, the sample ratio offset and operational parameter disturbance amplitude are calculated. The statistical offset, the label coverage missing rate, the sample ratio offset, and the operational parameter disturbance amplitude are combined to form a working condition disturbance intensity vector.

[0052] In this embodiment of the invention, by determining the data partitioning conditions based on the correspondence between shield tunneling operation data and geological label data in the basic dataset, shield tunneling construction data can be reorganized at the working condition level without changing the semantics of the original data. The data partitioning conditions characterize the basic dataset from multiple perspectives, including geological features, operational status, and data attributes, ensuring that data partitioning no longer relies on a single dimension but comprehensively reflects various uncertainties that may occur during shield tunneling construction. Based on these data partitioning conditions, different data subsets are selected from the basic dataset and used as working condition datasets for different uncertainty conditions. This allows each working condition dataset to have clear differences in data composition and characteristics, thus corresponding to different construction scenarios and data states.

[0053] Specifically, the data partitioning conditions include: 1) The division conditions used to characterize the differences in geological label coverage include: based on the value range or category composition of the geological label data used to characterize the geological state in the basic dataset, data containing different combinations of geological labels are distinguished to form a working condition dataset with differences in geological label coverage.

[0054] In this embodiment of the invention, the differences in the coverage of geological labels are used to reflect the completeness and diversity of geological states contained in the basic dataset under different working conditions. Because geological conditions exhibit significant spatial variations along the tunneling direction during shield tunneling, the combinations of geological types covered by different construction sections or different tunneling stages may differ considerably. To characterize these differences, this embodiment differentiates the data based on the value range or category composition of the geological label data used to represent geological states in the basic dataset.

[0055] Specifically, by statistically analyzing the occurrence of geological labels in different subsets of the basic dataset, the types and combinations of geological labels contained in different subsets are identified. When a subset contains a relatively concentrated range of geological labels or has a narrow coverage, it can be classified as a working condition dataset with relatively limited geological label coverage; conversely, when another subset contains a large number of geological labels or has a wide coverage, it can be classified as a working condition dataset with rich geological label coverage. Through this method, different working condition datasets exhibit differences in geological label coverage, thereby reflecting the model's applicability under conditions of varying geological coverage.

[0056] 2) The classification conditions used to characterize the differences in shield tunneling operation status include: based on the construction line, construction section or operation stage identifier corresponding to the shield tunneling operation data in the basic dataset, the data reflecting different shield tunneling operation statuses are distinguished to form a working condition dataset with differences in shield tunneling operation status.

[0057] In this embodiment of the invention, the differences in shield tunneling operation status are used to reflect the changes in the operating characteristics of the shield tunneling machine under different construction routes, different construction sections, or different operating stages. In actual construction, even if the geological conditions are similar, the construction organization methods, operating strategies, and equipment status of different routes or sections may differ, resulting in different statistical characteristics and trends in shield tunneling operation data.

[0058] Based on the above considerations, this embodiment differentiates the data according to the construction line, construction section, or operation stage identifier corresponding to the shield tunneling operation data in the basic dataset. Specifically, based on the division of shield tunneling tasks, data from different construction lines in the basic dataset can be categorized into different data subsets; or based on the stage division of the shield tunneling process, data at different operation stages can be distinguished. In this way, different data subsets create differences at the shield tunneling operation status level, thereby constructing an uncertainty condition dataset reflecting different operation conditions, which can be used for subsequent predictive modeling and evaluation for different operation conditions.

[0059] 3) The division conditions used to characterize the differences in the proportion of data composition include: combining data with different proportions based on the distribution relationship of the number of samples in the basic dataset to form a working condition dataset with different proportions of data composition.

[0060] In this embodiment of the invention, the difference in data composition ratio is used to reflect the changes in the quantity distribution of different types of samples in the basic dataset. Due to factors such as construction progress, data collection completeness, and uneven geological distribution, the data used for model construction often exhibits an imbalance in the number of samples at different stages or in different sections. To characterize these differences, this embodiment recombines the data based on the distribution relationship of the number of samples in the basic dataset.

[0061] Specifically, by controlling the proportion of different samples in the data subsets, multiple work condition datasets with varying sample composition ratios are constructed. For example, in some work condition datasets, the number of samples corresponding to a certain geological label is relatively high, while in other work condition datasets, the number of samples of this type is relatively low. In this way, different work condition datasets can be made to have significant differences in sample composition structure, thereby enabling the analysis of the applicability of the prediction model under conditions of changing sample distribution.

[0062] 4) The division conditions used to characterize the differences in data disturbance states include: based on applying different disturbance states to the shield tunneling operation data in the basic dataset, the original data and the disturbed data are distinguished to form a working condition dataset with differences in data disturbance states.

[0063] In this embodiment of the invention, the difference in data disturbance state is used to simulate the noise interference or abnormal fluctuations that shield tunneling operation data may experience during actual acquisition and transmission. In actual engineering, shield tunneling operation data may experience varying degrees of data disturbance due to factors such as changes in sensor accuracy, fluctuations in equipment status, or construction disturbances. To reflect the above situation, this embodiment distinguishes between the original data and the disturbed data by applying different disturbance states to the shield tunneling operation data in the basic dataset.

[0064] Specifically, without altering the overall structure of the tunnel boring machine (TBM) operation data, a preset perturbation process is applied to a portion of the TBM operation data, and the perturbed data and the original data are then categorized into different data subsets. This method creates a dataset of operating conditions with varying perturbation states, which can then be used to analyze the stability and applicability of the prediction model under conditions of changing data quality.

[0065] In one specific implementation, after constructing a basic dataset based on the actual construction and operation of the tunnel boring machine (TBM), a working condition construction process is performed to generate multiple uncertainty working condition datasets. This embodiment simulates different uncertainty working conditions that may occur during TBM construction by adjusting the data state used for model construction and the selection method of test data.

[0066] In this embodiment, shield tunneling operation data and its matching geological label data corresponding to a specific construction route are selected as the basic data source, with geological label values ​​ranging from 0 to 3 used as training data for model building. Based on this, an uncertain working condition dataset with varying geological label coverage is constructed by adjusting the range of geological labels included in the test data. Specifically, a geological label 4, which was not present during the training phase, is introduced into the test data, making the test data simultaneously contain geological labels 0, 1, 2, 3, and 4. This creates a data state containing unknown geological conditions, used to simulate working conditions encountered by the shield tunneling process.

[0067] Furthermore, by adjusting the source of the test data, an uncertain working condition dataset with varying shield tunneling operation states is constructed. In this embodiment, the data used for model construction comes from one construction route, while the test data comes from another construction route, thus forming a data configuration that simulates the changes in data distribution corresponding to the shield tunneling operation under different construction routes or different operating conditions.

[0068] Furthermore, by adjusting the sample size ratio between the data used for model building and the test data, an uncertain working condition dataset with varying data composition ratios is constructed. In this embodiment, several different data ratio configurations are set, including combinations of training and test data ratios such as 8:2, 7:3, 6:4, 5:5, 4:6, and 3:7. Under these ratio configurations, both the training and test data include geological labels from 0 to 3 to ensure a consistent range of geological labels, thereby simulating only the uncertainties introduced by changes in the sample size ratio.

[0069] Meanwhile, to simulate the acquisition status of shield tunneling operation data under different noise environments, different degrees of perturbation processing were applied to the input parameters of the shield tunneling operation data in the basic dataset, thereby constructing an uncertain working condition dataset with different data perturbation states. Specifically, for multiple input parameters in the shield tunneling operation data, perturbation noise with different standard deviation ratios was added. The standard deviation ratios included 0, 0.05, 0.1, 0.2, 0.5, and 1.0 to form various data perturbation states, which were used to simulate the changes in data quality caused by factors such as sensor noise and environmental interference during actual construction.

[0070] It should be noted that in the above embodiments, the shield tunneling geological prediction process aims at prediction, that is, predicting the geological state of the next tunneling ring based on the shield operation data corresponding to the current tunneling ring, rather than modeling based on the regression relationship between input and output at the same moment. The uncertainty condition dataset constructed in the above manner can be used to construct and evaluate shield tunneling geological prediction models under different uncertainty conditions.

[0071] Furthermore, in order to achieve quantitative comparison between different uncertain operating conditions, after the uncertain operating condition dataset is constructed, a quantitative index for characterizing the intensity of operating condition disturbance is calculated based on the degree of difference between each uncertain operating condition dataset and the basic dataset.

[0072] To address the data distribution differences between each uncertain working condition dataset and the basic dataset, key statistical features in the tunnel boring machine (TBM) operation data are compared. Specifically, the mean, variance, quantiles, or other statistical features of the operating parameters can be calculated, and the difference between the uncertain working condition dataset and the basic dataset in the corresponding statistical features is used as a statistical offset. This statistical offset characterizes the overall degree of shift in the distribution pattern of the operating data, thus reflecting the magnitude of change in the current working condition relative to the basic working condition. Simultaneously, regarding differences in geological label coverage, geological label categories that are not included or have decreased coverage in the uncertain working condition dataset are identified, and the label coverage missing rate is calculated to quantify the changes in the coverage of geological categories.

[0073] Furthermore, based on the ratio of sample size between the uncertain operating condition dataset and the basic dataset, a sample ratio offset is calculated. This offset reflects the quantitative difference between the sample composition under the current operating condition and the basic dataset, thus characterizing the degree of change in the data structure. Simultaneously, for the operational parameter disturbance processing rules implemented in the uncertain operating condition dataset, such as parameter noise superposition, data truncation, or sampling changes, the magnitude of the operational parameter disturbance is calculated to quantify the intensity of data disturbance introduced by human or environmental factors.

[0074] After calculating the statistical offset, label coverage missing rate, sample proportion offset, and operational parameter disturbance amplitude, the aforementioned quantitative indicators are combined in a preset order to form a working condition disturbance intensity vector. This working condition disturbance intensity vector comprehensively represents the overall offset of the current uncertain working condition relative to the basic dataset in a multidimensional form, providing a quantitative basis for subsequent model applicability evaluation. This vectorized representation makes the disturbance degrees between different uncertain working conditions comparable, thereby supporting systematic evaluation under multiple working conditions. The implementation of this step does not depend on a specific geological type or data scale; any technical solution that calculates multidimensional quantitative indicators based on the differences between the basic dataset and the uncertain working condition dataset and combines them into a vector falls within the protection scope of this invention.

[0075] Step S30: Under each uncertain working condition, construct a shield tunneling geological prediction model based on the corresponding uncertain working condition dataset, input the corresponding equivalent stratum response operation data into the shield tunneling geological prediction model, output the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results, and calculate the cumulative offset of the section based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result.

[0076] Specifically, under each uncertain working condition, a shield tunneling geological prediction model is constructed based on the corresponding uncertain working condition dataset. This includes: for each uncertain working condition dataset, determining the modeling data used for model construction from the corresponding uncertain working condition dataset; based on the modeling data, learning the model parameters of the shield tunneling geological prediction model under the constraints of the corresponding uncertain working condition, so that the shield tunneling geological prediction model can represent the mapping relationship between the shield tunneling operation data and the geological label data under the uncertain working condition, thereby obtaining the shield tunneling geological prediction model corresponding to each uncertain working condition.

[0077] In this embodiment of the invention, after the construction of the uncertainty working condition dataset is completed, a shield tunneling geological prediction model is constructed based on the corresponding uncertainty working condition dataset under each uncertainty working condition, and the corresponding shield tunneling operation data is input into the shield tunneling geological prediction model to obtain the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results.

[0078] Specifically, under each uncertainty condition, modeling data for model construction is first determined from the corresponding uncertainty condition dataset. This modeling data consists of tunnel boring machine (TBM) operation data and its corresponding geological label data. The TBM operation data characterizes the operational state of the TBM during tunneling, while the geological label data characterizes the geological state corresponding to that operational state. By determining modeling data separately for each uncertainty condition, it is ensured that the data state used for model construction remains consistent with the current uncertainty condition, enabling the model's learning process to be tailored to specific conditions.

[0079] After determining the modeling data, the model parameters of the shield tunneling geological prediction model are learned under the constraints of the corresponding uncertainty conditions. The model parameter learning process is based on the correspondence between shield tunneling operation data and geological label data in the modeling data. By adjusting the model parameters, the shield tunneling geological prediction model gradually forms a representation of the mapping relationship between shield tunneling operation data and geological label data under the uncertainty conditions. Since the geological coverage, operational characteristics, or data disturbance states of the data differ under different uncertainty conditions, performing model parameter learning separately under each uncertainty condition helps the obtained shield tunneling geological prediction model adapt to the data distribution characteristics under the corresponding condition, thus obtaining a shield tunneling geological prediction model that corresponds one-to-one with each uncertainty condition.

[0080] Furthermore, the corresponding equivalent stratum response operation data is input into the shield tunneling geological prediction model, and the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results are output. This includes: inputting the equivalent stratum response operation data into the corresponding uncertainty condition prediction model in sequence according to the tunneling order to obtain geological prediction results corresponding to each tunneling unit; based on the prediction confidence distribution information output by the shield tunneling geological prediction model in the process of generating the geological prediction results, calculating the confidence index and distribution dispersion index corresponding to each geological prediction result, and determining the confidence index and distribution dispersion index as the uncertainty characterization information corresponding to each geological prediction result.

[0081] Furthermore, the cumulative offset of a section is calculated based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result. This includes: arranging the uncertainty characterization information according to the tunneling sequence and constructing a sliding window with a preset number of tunneling units; within each sliding window, the confidence index and distribution dispersion index in the uncertainty characterization information are accumulated and calculated to obtain the section cumulative offset of the corresponding window; and arranging the section cumulative offsets corresponding to each sliding window according to the tunneling sequence to form a section-level uncertainty cumulative characterization result.

[0082] In this embodiment of the invention, after constructing the shield tunneling geological prediction model corresponding to each uncertain working condition, the corresponding equivalent stratum response operation data is input into the shield tunneling geological prediction model to obtain the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results. Specifically, the equivalent stratum response operation data is sequentially input into the corresponding uncertain working condition prediction model according to the tunneling sequence, so that the model outputs the corresponding geological prediction results for each tunneling unit while maintaining the consistency of the tunneling unit sequence. Through the sequential input mechanism, the prediction results are continuously distributed in the tunneling direction, thereby maintaining consistency with the actual construction progress.

[0083] While generating the geological prediction results, the shield tunneling geological prediction model outputs prediction confidence distribution information. This prediction confidence distribution information may include probability distribution values ​​or confidence level distribution structures for different geological categories. Based on this prediction confidence distribution information, a confidence index and a distribution dispersion index are calculated respectively. The confidence index characterizes the concentration of the model's prediction results for the current tunneling unit, such as the maximum probability value of the prediction category; the distribution dispersion index characterizes the dispersion of the prediction distribution, such as the variance or information entropy value of the probability distribution. By simultaneously calculating the confidence index and the distribution dispersion index, the uncertainty characterization information can reflect the stability and distribution diffusion degree of the model's predictions. The confidence index and the distribution dispersion index correspond one-to-one with the geological prediction results of each tunneling unit, forming a structured sequence of uncertainty characterization information.

[0084] Furthermore, to identify the changing trend of predicted stability within continuous sections, the cumulative offset of the section is calculated based on the uncertainty characterization information of multiple consecutive tunneling units. The uncertainty characterization information is arranged according to the tunneling sequence, and a sliding window is constructed with a preset number of tunneling units, so that each window covers a continuous tunneling unit section. The preset number of tunneling units can be set according to the length of the construction section or risk assessment requirements. Within each sliding window, the confidence index and distribution dispersion index in the uncertainty characterization information are accumulated and calculated to obtain the cumulative offset of the corresponding window. The cumulative offset of the section is used to characterize the overall changing trend of the predicted confidence and the degree of uncertainty accumulation within that section.

[0085] The cumulative offsets of each sliding window corresponding to a segment are arranged according to the tunneling sequence to form a segment-level uncertainty accumulation characterization result. This method expands the instantaneous uncertainty characterization originally targeting a single tunneling unit into a segment-scale cumulative risk measure, providing segment-level quantitative basis for subsequent model applicability assessment. This step does not depend on a specific prediction model structure; any technical solution that can calculate uncertainty indices and perform segment accumulation processing based on prediction confidence distribution information falls within the protection scope of this invention.

[0086] In another possible implementation, by combining geological survey data or existing stratigraphic zoning information corresponding to the construction section, the continuous tunneling unit is divided into several stratigraphic structure sub-segments. Within the same stratigraphic structure sub-segment, a structural consistency analysis is performed on the cumulative offset of the segment. When the geological prediction category of adjacent tunneling units remains consistent while the uncertainty index suddenly increases, it is identified as a local prediction fluctuation event, and a smoothing correction is performed on the corresponding segment's cumulative offset. When both the prediction category and the uncertainty index experience persistent shifts within the same sub-segment, the segment is marked as a potential stratigraphic change risk segment. By introducing stratigraphic structure continuity constraints, the segment-level uncertainty accumulation characterization results not only reflect the fluctuation of the model output but also undergo reasonable correction based on the actual stratigraphic spatial continuity characteristics, thereby improving the engineering adaptability of the segment risk identification results.

[0087] In one specific implementation, based on the aforementioned uncertain working condition dataset, the prediction results and uncertainty characterization information of the shield tunneling geological prediction model are analyzed under different working conditions to evaluate the applicability of the model under different working conditions.

[0088] First, under the condition that the geological conditions are within a known range, the data in the basic dataset belonging to the known geological conditions are analyzed. Under this condition, both the training and testing data are derived from sample sequences with geological labels 0 to 3, and are arranged according to the tunneling sequence to form a shield tunneling operation data sequence for prediction. In this embodiment, the prediction results under the corresponding working condition are indexed by the ring number, and the corresponding geological prediction results are output for each tunneling ring. By comparing and analyzing the prediction results with the corresponding geological label data, it can be found that the prediction results for various geological conditions are generally within a reasonable range. Among them, the prediction results for geological label 0 are the most stable, with no obvious misjudgments. There are a small number of mutual misjudgments between geological labels 1 and 3, and the prediction error of geological label 2 is slightly higher than that of labels 1 and 3, but the overall number of misjudgments is still within an acceptable range. This phenomenon is closely related to the difference in the distribution of the number of samples for different geological labels in the basic dataset; the prediction accuracy is relatively lower for geological labels with fewer samples.

[0089] Furthermore, geological conditions not included in the training data were introduced into the test data, forming an uncertain working condition dataset containing unknown geological conditions. Under this working condition, since the training phase did not include corresponding geological labels, the prediction model could not directly form a stable mapping relationship when facing unknown geological conditions. Analysis of the distribution of prediction results revealed that different prediction models exhibited significant misjudgments under this working condition, with some prediction results showing an over-biased bias towards known geological conditions. Compared to prediction methods that only output a single prediction result, the uncertainty characterization information formed based on the model's prediction output features can reflect the decreasing trend in the reliability of the prediction results.

[0090] Based on the above, statistical analysis is performed on the uncertainty characterization information corresponding to each tunneling ring to obtain the uncertainty distribution characteristics under different working conditions. In this embodiment, the corresponding uncertainty value is calculated for each tunneling ring, and the uncertainty values ​​are summarized and statistically analyzed. The statistical results are as follows: Figure 3 As shown, where Figure 3 (a) shows the distribution of uncertainty as a function of the tunneling ring under known and unknown geological conditions. Figure 3 (b) shows a comparison of the statistical distribution of uncertainty values ​​under the two operating conditions.

[0091] Furthermore, the uncertainty characterization information was decomposed and analyzed, breaking down the total uncertainty into cognitive uncertainty and random uncertainty. Statistical results show that, under known geological conditions, the overall uncertainty level is relatively low, with cognitive uncertainty remaining close to zero. Random uncertainty plays a major role in the total uncertainty, primarily stemming from fluctuations and noise interference in the tunnel boring machine's operational data. The corresponding uncertainty decomposition results are as follows: Figure 3 As shown in (c).

[0092] Under uncertain conditions including unknown geological features, the overall uncertainty level increases significantly. Compared to known geological conditions, the total uncertainty distribution shifts to a higher value range, increasing by approximately 95%. Under this condition, both cognitive uncertainty and random uncertainty increase to varying degrees. The peak value of cognitive uncertainty changes significantly, while the peak value of random uncertainty changes relatively less, indicating that unknown geological conditions have a more significant impact on the model's cognitive ability. The corresponding uncertainty decomposition results are as follows: Figure 3 As shown in (d).

[0093] Furthermore, the uncertainty of the shield tunneling geological prediction model under complex working conditions is further analyzed in detail by combining different operating state changes, and the corresponding analysis results are introduced into the specification as supplementary embodiments.

[0094] In this embodiment, the training data consists of operational data and geological label data generated during the left-line shield tunneling process, with the geological label range still limited to 0 to 3. The data order is shuffled during the training phase. Correspondingly, the test data consists of operational data generated during the right-line shield tunneling process, and the test phase also only includes geological labels 0 to 3. This working condition is used to simulate the uncertainty of the operational state changes caused by differences in the construction conditions of the left and right lines of the shield.

[0095] Under this working condition, shield tunneling geological prediction models were constructed based on the aforementioned uncertain working condition dataset. The right-line shield tunneling operation data was input into the corresponding prediction models according to the tunneling ring number order, and the geological prediction results and their uncertainty characterization information were output. Statistical analysis of the prediction results revealed that the prediction accuracy of different prediction models under this working condition decreased significantly. Furthermore, some prediction results corresponding to tunneling rings exhibited geological label shift, reflecting the impact of differences in the geological mapping relationship between the left and right lines on the model's predictive ability.

[0096] Furthermore, the uncertainty characterization information corresponding to each tunneling ring is statistically summarized to obtain the following results: Figure 4 The results are shown. Figure 4 (a) shows the distribution of uncertainty as the tunneling ring changes under the changing operating conditions. It can be seen that the overall uncertainty level is significantly higher than that of the known geological conditions, but there is still a certain distribution in the low value range. Figure 4 (b) shows a comparison of the statistical distribution of uncertainty values ​​under this working condition with those under known geological working conditions, where the uncertainty distribution under the working condition with changing operating status generally shifts towards the higher value range.

[0097] Further decomposition analysis of the total uncertainty yields the distribution of cognitive uncertainty and random uncertainty under this operating condition, such as... Figure 4 (c) and Figure 4 As shown in (d), the analysis results indicate that under changing operating conditions, random uncertainty is lower than under known geological conditions, while cognitive uncertainty increases significantly, becoming the main component of total uncertainty. This result suggests that when the shield tunneling operation status and the geological mapping relationship between the left and right lines change, the model is more likely to treat the test data as unseen data, leading to a significant increase in cognitive uncertainty.

[0098] By introducing this embodiment, it can be further illustrated that the method of the present invention can not only output geological prediction results under different uncertain working conditions, but also effectively reflect the differences in the applicability of the model under changing operating conditions through uncertainty characterization information, providing a basis for model selection and risk assessment under different construction sections and operating conditions in engineering practice.

[0099] As can be seen from the above examples, under different uncertain conditions, the shield tunneling geological prediction model not only exhibits differences in prediction results, but also displays significantly different distribution characteristics in its corresponding uncertainty characterization information. Joint analysis based on prediction results and uncertainty characterization information can provide a basis for subsequent applicability assessment of the shield tunneling geological prediction model under different construction conditions.

[0100] Step S40: Based on the geological prediction results, the segment-level uncertainty accumulation characterization results, the working condition disturbance intensity vector, and the geological label data, output the applicability evaluation results of the shield tunneling geological prediction model under various uncertainty working conditions.

[0101] Specifically, based on the correspondence between the geological prediction results and the geological label data, a prediction consistency index is calculated; the prediction consistency index is correlated with the cumulative characterization results of the segment-level uncertainty to obtain a segment stability index; the segment stability index is weighted and combined with the working condition disturbance intensity vector to generate a working condition fit index; and the applicability evaluation results of the shield tunneling geological prediction model under various uncertain working conditions are output according to the working condition fit index.

[0102] In this embodiment of the invention, after obtaining the geological prediction results, the cumulative uncertainty characterization results at the section level, and the disturbance intensity vector under various working conditions, the applicability of the shield tunneling geological prediction model under various uncertainty conditions is systematically evaluated based on the above information. This evaluation process no longer relies solely on a single prediction accuracy index, but instead constructs a comprehensive adaptability judgment mechanism through multi-level indices, enabling the model's application reliability under different uncertainty conditions to be quantitatively expressed.

[0103] Based on the correspondence between the geological prediction results and the geological label data, a prediction consistency index is calculated. Specifically, the geological prediction results of each tunneling unit are compared with the corresponding real geological label data unit by unit. Quantitative statistics can be performed using methods such as accuracy, classification consistency rate, or error rate to form a prediction consistency index for the current uncertain working condition. This prediction consistency index reflects the model's overall predictive ability of the geological state under this working condition and is a fundamental quantitative indicator for evaluating model performance.

[0104] Furthermore, the prediction consistency index is correlated with the cumulative uncertainty characterization result at the segment level to obtain the segment stability index. The cumulative uncertainty characterization result at the segment level reflects the cumulative trend of prediction confidence and distribution dispersion within multiple consecutive tunneling units. When the prediction consistency index is high but the cumulative uncertainty characterization result at the segment level fluctuates significantly, it indicates that although the model has a certain degree of accuracy, it suffers from insufficient prediction stability at the segment scale. The segment stability index can be obtained by correlating the prediction consistency index with the cumulative segment offset, for example, by constructing a functional relationship or weighted fusion relationship between the two. The segment stability index is used to characterize the prediction stability of the model within continuous tunneling segments.

[0105] Furthermore, the segment stability index and the operating condition disturbance intensity vector are weighted and combined to generate an operating condition fit index. The operating condition disturbance intensity vector is used to characterize the overall deviation of the current uncertain operating condition relative to the basic dataset, including factors such as data distribution deviation, label coverage difference, sample ratio change, and parameter disturbance amplitude. By weighting and combining the components in the segment stability index and the operating condition disturbance intensity vector, an operating condition fit index that comprehensively reflects the matching degree between model performance and operating condition difficulty can be formed. The weighting combination method can be based on linear weighting based on preset weights, or the weights can be adjusted based on historical operating condition performance. The purpose is to establish a quantifiable correlation between prediction stability and operating condition complexity.

[0106] The applicability assessment results of the shield tunneling geological prediction model under various uncertain working conditions are output based on the working condition adaptability index. These applicability assessment results can be presented in a graded format, such as classifying the working condition adaptability index into high, medium, or low adaptability levels, or they can be output as continuous numerical values ​​to guide model deployment decisions. In practical applications, when the working condition adaptability index falls below a preset threshold, a model retraining, model switching, or manual review mechanism can be triggered, thereby improving risk control capabilities during construction.

[0107] Through the above steps, this invention structurally integrates three types of information: prediction accuracy, segment-level stability, and working condition disturbance intensity. This allows the applicability assessment results to not only reflect the model's prediction accuracy but also comprehensively demonstrate the model's stability and fit under current uncertain working conditions. Compared to evaluations based solely on a single accuracy index, this technical solution more comprehensively reflects the model's adaptability in complex construction environments, providing quantifiable and comparable decision-making basis for the practical application of shield tunneling geological prediction models. Any technical solution that constructs multi-level adaptability indices based on prediction results, segment-level uncertainty accumulation information, and working condition disturbance intensity vectors, and outputs applicability assessment results, falls within the protection scope of this invention.

[0108] In another possible implementation, without changing the construction method of the uncertainty condition dataset and the structure of the shield tunneling geological prediction model, a condition assessment strategy based on tunneling continuity is introduced into the applicability assessment process. This implementation takes into account that the geological conditions during shield tunneling usually have a certain continuity between adjacent tunneling rings, and the prediction results and uncertainty characterization information corresponding to a single tunneling ring may be affected by local disturbances, thus causing instantaneous deviations in the applicability assessment results.

[0109] In this implementation, for each uncertain working condition, after calculating the predictive evaluation index and the working condition uncertainty index, the applicability evaluation result is not directly output based on a single tunneling ring. Instead, a tunneling segment consisting of multiple consecutive tunneling rings is used as the evaluation unit to perform a continuous analysis on the predictive evaluation index and the working condition uncertainty index. Specifically, the working condition evaluation characteristics corresponding to adjacent tunneling rings are sequentially organized to determine their changing trend and stability within a preset continuous tunneling range.

[0110] When the performance evaluation characteristics within a continuous tunneling segment remain within a relatively stable range, the tunnel geological prediction model under this uncertain condition is deemed to have high applicability at this construction stage. Conversely, when the performance evaluation characteristics exhibit frequent fluctuations or abrupt changes within the continuous tunneling segment, the model's applicability under this uncertain condition is deemed to be at risk of decline. By introducing tunneling continuity constraints into the applicability assessment process, the assessment results are made to better reflect the actual operation of tunnel boring machine (TBM) construction.

[0111] like Figure 5As shown, this invention provides a shield tunneling geological prediction system for multiple uncertain working conditions. The system includes: a data acquisition unit, used to acquire shield tunneling operation data and corresponding geological tag data generated during shield tunneling construction; perform control behavior identification processing on the shield tunneling operation data to generate control behavior status identifiers; and decouple and correct sudden changes in operating parameters caused by control strategy switching based on the control behavior status identifiers to obtain equivalent stratum response operation data; associate the equivalent stratum response operation data with the geological tag data based on the tunneling sequence to construct a basic dataset for shield tunneling geological prediction; and a processing unit, used to perform working condition construction processing based on the basic dataset to generate multiple uncertain working condition datasets, and to address each uncertain condition... The system includes a working condition calculation unit for calculating the working condition disturbance intensity vector to quantify the degree of working condition disturbance; a prediction unit for constructing a shield tunneling geological prediction model based on the corresponding uncertain working condition dataset under various uncertain working conditions, inputting the corresponding equivalent stratum response operation data into the shield tunneling geological prediction model, outputting the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results, and calculating the cumulative offset of the section based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result; and an output unit for outputting the applicability evaluation results of the shield tunneling geological prediction model under various uncertain working conditions based on the geological prediction results, the section-level uncertainty cumulative characterization results, the working condition disturbance intensity vector, and the geological label data.

[0112] The present invention also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described shield tunneling geological prediction method for multiple uncertain working conditions.

[0113] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor A01, a network interface A02, a memory (not shown), and a database (not shown) connected via a system bus. The processor A01 provides computing and control capabilities. The memory includes internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01, a computer program B02, and a database (not shown). The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 stored in the non-volatile storage medium A06. The network interface A02 is used for communication with external terminals via a network connection. When the computer program B02 is executed by the processor A01, it implements a shield tunneling geological prediction method for multi-uncertain working conditions.

[0114] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0115] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details described above. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not further describe the various possible combinations.

[0116] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the embodiments of the present invention, they should also be regarded as the content disclosed by the embodiments of the present invention.

Claims

1. A method for geological prediction in shield tunneling under multiple uncertain working conditions, characterized in that, The method includes: The shield tunneling operation data and corresponding geological label data generated during the shield tunneling construction process are acquired. The shield tunneling operation data is processed to identify control behavior status identifiers, and the sudden changes in operation parameters caused by control strategy switching are decoupled and corrected based on the control behavior status identifiers to obtain equivalent stratum response operation data. The equivalent stratum response operation data is associated with the geological label data based on the tunneling sequence to construct a basic dataset for shield tunneling geological prediction. Based on the aforementioned basic dataset, a working condition construction process is performed to generate multiple uncertain working condition datasets, and a working condition disturbance intensity vector is calculated for each uncertain working condition to quantify the degree of working condition disturbance. Under various uncertain working conditions, a shield tunneling geological prediction model is constructed based on the corresponding uncertain working condition dataset. The corresponding equivalent stratum response operation data is input into the shield tunneling geological prediction model, and the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results are output. Based on the uncertainty characterization information of multiple consecutive tunneling units, the cumulative offset of the section is calculated to form the section-level uncertainty cumulative characterization result. Based on the geological prediction results, the cumulative uncertainty characterization results at the section level, the working condition disturbance intensity vector, and the geological label data, the applicability evaluation results of the shield tunneling geological prediction model under various uncertain working conditions are output.

2. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 1, characterized in that, The shield tunneling operation data is processed to identify control behavior to generate a control behavior status identifier. Based on the control behavior status identifier, the sudden changes in operating parameters caused by control strategy switching are decoupled and corrected to obtain equivalent ground response operation data, including: The shield tunneling operation data is arranged in a time series according to the tunneling sequence, and control feature parameters used to characterize the propulsion control behavior are extracted; Based on the variation of the control characteristic parameters between continuous tunneling units, a control change rate sequence is calculated, and the control change rate sequence is compared with a preset change rate threshold to identify the control strategy switching time. After identifying the control strategy switching moment, a control behavior status identifier for the corresponding tunneling unit is generated, and a mapping relationship is established between the control behavior status identifier and the shield operation data of the corresponding tunneling unit. Based on the control behavior status identifier, the sequence of operating parameters within the control strategy switching interval is divided into transition segments, and trend separation processing is performed on the operating parameters within the transition segments to obtain residual signals reflecting the formation response, which serve as equivalent formation response operating data.

3. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 2, characterized in that, The geological label data includes any one or more of the following: geological process labels, stratigraphic state labels, and comprehensive geological category labels; Based on the tunneling sequence, the equivalent stratum response operational data is associated with the geological label data to construct a basic dataset for shield tunneling geological prediction, including: The equivalent formation response operation data are numbered according to the tunneling sequence to form an operation data sequence arranged by tunneling unit; The geological tag data is numbered according to the corresponding tunneling unit to form a geological tag sequence arranged by tunneling unit; Using the equivalent formation response operation data of the current tunneling unit as input features and the geological label data of the next tunneling unit adjacent to the current tunneling unit as supervision labels, a prediction unit sample is constructed. The prediction unit sample construction process is performed on all tunneling units to generate a sample set, which is then used as the basic dataset for shield tunneling geological prediction.

4. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 1, characterized in that, Based on the aforementioned base dataset, a working condition construction process is performed to generate multiple uncertain working condition datasets. For each uncertain working condition, a working condition disturbance intensity vector is calculated to quantify the degree of working condition disturbance, including: Based on the correspondence between shield tunneling operation data and geological label data in the basic dataset, the data partitioning conditions for constructing working conditions are determined, and according to the data partitioning conditions, different data subsets are selected from the basic dataset as working condition datasets under different uncertain working conditions. Based on the differences in data distribution and geological label coverage between each uncertain working condition dataset and the basic dataset, the statistical offset and label coverage missing rate of the shield tunneling operation data are calculated respectively. Based on the ratio of sample size between the uncertain operating condition dataset and the basic dataset, and the rules for handling operational parameter disturbances in the uncertain operating condition dataset, the sample ratio offset and the operational parameter disturbance magnitude are calculated respectively. The statistical offset, the label coverage missing rate, the sample proportion offset, and the perturbation amplitude of the operating parameters are combined to form the operating condition perturbation intensity vector.

5. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 4, characterized in that, Under various uncertain conditions, a shield tunneling geological prediction model is constructed based on the corresponding uncertain condition dataset, including: For each uncertain working condition dataset, modeling data for model building is determined from the corresponding uncertain working condition dataset; Based on the modeling data, the model parameters of the shield tunneling geological prediction model are learned under the constraints of the corresponding uncertainty conditions. This enables the shield tunneling geological prediction model to characterize the mapping relationship between the shield tunneling operation data and the geological label data under the uncertainty conditions, thus obtaining the shield tunneling geological prediction model corresponding to each uncertainty condition.

6. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 5, characterized in that, The corresponding equivalent geological response operation data is input into the shield tunneling geological prediction model, and the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results are output, including: The equivalent formation response operation data is sequentially input into the corresponding uncertainty condition prediction model according to the tunneling sequence to obtain geological prediction results corresponding to each tunneling unit. Based on the prediction confidence distribution information output by the shield tunneling geological prediction model during the generation of the geological prediction results, the confidence index and distribution dispersion index corresponding to each of the geological prediction results are calculated, and the confidence index and distribution dispersion index are determined as uncertainty characterization information corresponding one-to-one with the geological prediction results.

7. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 6, characterized in that, The cumulative offset of a section is calculated based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result, including: The uncertainty characterization information is arranged according to the tunneling sequence, and a sliding window is constructed with a preset number of tunneling units; Within each sliding window, the confidence index and the distribution dispersion index in the uncertainty characterization information are accumulated and calculated to obtain the cumulative offset of the corresponding window segment. The cumulative offset of each sliding window segment is arranged according to the tunneling sequence to form a segment-level uncertainty accumulation characterization result.

8. The shield tunneling geological prediction method for multiple uncertain working conditions according to claim 1, characterized in that, Based on the geological prediction results, the cumulative uncertainty characterization results at the section level, the working condition disturbance intensity vector, and the geological label data, the applicability evaluation results of the shield tunneling geological prediction model under various uncertainty conditions are output, including: Based on the correspondence between the geological prediction results and the geological label data, a prediction consistency index is calculated; The prediction consistency index is correlated with the cumulative characterization result of the segment-level uncertainty to obtain the segment stability index. The section stability index and the working condition disturbance intensity vector are weighted and combined to generate the working condition adaptability index. The applicability assessment results of the shield tunneling geological prediction model under various uncertain working conditions are output based on the working condition adaptability index.

9. A shield tunneling geological prediction system for multiple uncertain working conditions, characterized in that, The system includes: The acquisition unit is used to acquire shield operation data and corresponding geological label data generated during shield tunneling construction. It performs control behavior identification processing on the shield operation data to generate control behavior status identifiers, and decouples and corrects the sudden changes in operation parameters caused by control strategy switching based on the control behavior status identifiers to obtain equivalent stratum response operation data. Based on the tunneling sequence, the equivalent stratum response operation data is associated with the geological label data to construct a basic dataset for shield tunneling geological prediction. The processing unit is used to perform working condition construction processing based on the basic dataset, generate multiple uncertain working condition datasets, and calculate a working condition disturbance intensity vector for each uncertain working condition to quantify the degree of working condition disturbance. The prediction unit is used to construct a shield tunneling geological prediction model based on the corresponding uncertainty condition dataset under various uncertainty conditions. It inputs the corresponding equivalent stratum response operation data into the shield tunneling geological prediction model, outputs the geological prediction results and the uncertainty characterization information corresponding to the geological prediction results, and calculates the cumulative offset of the section based on the uncertainty characterization information of multiple consecutive tunneling units to form a section-level uncertainty cumulative characterization result. The output unit is used to output the applicability evaluation results of the shield tunneling geological prediction model under various uncertain working conditions based on the geological prediction results, the section-level uncertainty accumulation characterization results, the working condition disturbance intensity vector, and the geological label data.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the shield tunneling geological prediction method for multiple uncertain working conditions as described in any one of claims 1-8.

Images