Real-time processing and warehousing method for shield construction timing data based on flow batch integration

Abstract

This invention discloses a real-time processing and database entry method for tunnel boring machine (TBM) construction time-series data based on integrated batch processing. The method includes: acquiring the original time-series data of the TBM and its associated original ring numbers; performing standardization processing based on a pre-built semantic rule base to obtain standardized records; identifying the TBM's operating status and determining the target ring number based on the standardized records and a judgment threshold; performing differentiated data cleaning strategies on the standardized records according to the operating status and evaluating the cleaning results to obtain a comprehensive quality score; and routing the standardized records to matching processing channels based on dynamic diversion conditions including the comprehensive quality score and arrival delay, and writing them into the time-series database using the target ring number as a reference. This invention overcomes the defects of missed physical anomaly detection and reference misalignment, improving the operational condition reproduction accuracy of the entered data.

Description

Technical Field

[0001] This invention relates to the field of intelligent construction and time sequence data processing technology in tunnel engineering, specifically to a method for real-time processing and storage of shield tunneling construction time sequence data based on integrated batch processing and batch processing. Background Technology

[0002] The tunnel boring machine (TBM) process generates high-frequency time-series operational data involving multiple dimensions such as tunneling pressure, torque, and attitude. Performing high-real-time and high-fidelity data cleaning and standardization on this massive, multi-source, heterogeneous data is the underlying technical foundation for achieving dynamic optimization of TBM tunneling parameters, adaptive control in complex geological formations, and accurate identification of construction safety conditions.

[0003] The current shield tunneling data processing architecture mainly adopts a static data routing mode based on data source, which involves "real-time data streaming and historical data batch processing". In the data cleaning and status identification stage, the existing processing system usually uses the passively received programmable logic controller (PLC) hardware feedback loop number as the spatial division benchmark, and uses fixed numerical thresholds for the entire line and independent valid interval judgment logic for a single sensor to perform working condition identification and outlier removal operations.

[0004] However, existing technologies suffer from problems such as spatial reference distortion, poor geological adaptability, and missed detection of physical coupling anomalies when dealing with the complex and ever-changing tunnel boring machine (TBM) construction environment. Specifically:

[0005] The mechanism that passively relies on PLC ring numbers is prone to skipping or missing numbers when there is communication delay or gateway caching, which leads to the failure of the single-ring data statistical benchmark and the lack of low-level autonomous correction capability of the system.

[0006] A fixed global threshold cannot respond to the drastic changes in strength across strata, such as soft soil to hard rock, along the tunnel route, and is prone to misjudging the working condition at the boundary of different strata.

[0007] The parameter-by-parameter independent verification rules completely sever the mechanical balance constraints between thrust, torque and penetration, making it impossible to detect deep coupling anomalies such as single parameter being legal but physical combination being impossible. At the same time, the static diversion mode cannot dynamically perform context reorganization and replenishment according to the degree of data quality degradation, resulting in the final time series data entering the database failing to truly restore the physical working conditions. Summary of the Invention

[0008] Purpose of the invention: To provide a method for real-time processing and storage of shield tunneling construction time sequence data based on integrated batch processing, in order to solve the above-mentioned problems existing in the prior art.

[0009] Technical solution: A method for real-time processing and storage of shield tunneling construction time sequence data based on integrated batch processing, including:

[0010] Obtain the original time-series data generated by the tunnel boring machine during construction, as well as the associated original ring numbers;

[0011] The original time-series data is standardized based on a pre-built semantic rule base to obtain standardized records;

[0012] Based on the standardized records and judgment thresholds, the operating status of the tunnel boring machine is identified and the target ring number is determined;

[0013] Based on the operating status, the corresponding data cleaning strategy is executed on the standardized records, and the cleaning results are evaluated to obtain a comprehensive quality score;

[0014] Based on the triage conditions including the comprehensive quality score and arrival delay, the cleaned standardized records are routed to the matching processing channel and written into the pre-configured time series database with the target ring number as the reference.

[0015] Beneficial effects: A weighted state machine based on multiple physical signals such as thrust drop and travel accumulation was constructed to independently detect ring number boundaries; a spatial mapping logic between tunnel mileage and geological profile dictionary was introduced, and dynamic adaptive scaling of state judgment threshold was realized by combining first-order hysteresis filtering; the dynamics and statics equations of shield machine machinery such as penetration, specific energy and thrust balance were transformed into core data cleaning criteria, and on this basis, the data arrival delay and state confidence were integrated to calculate a comprehensive quality score, thereby driving the dynamic routing of data between stream processing and batch processing channels and the underlying source tracing coverage.

[0016] It solves the problem of spatial statistical benchmark distortion in the traditional passive receiving mode, eliminates the misjudgment of working conditions when the tunnel boring machine crosses the boundary zone of uneven soft and hard strata, and intercepts pseudo-normal dirty data that is legal with single parameters but contradicts the physical coupling logic, thus making up for the shortcoming that static hard diversion cannot repair degraded data.

[0017] This invention not only achieves autonomous correction of the bottom layer of the spatial benchmark for shield tunneling construction and keen perception of deep physical anomalies, but also effectively balances the second-level timeliness of the on-site front-end large screen monitoring and the high-fidelity restoration of the real complex physical conditions of the final stored time-series assets through delayed context recombination and batch replenishment of identifiers and overlay mechanisms. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method for real-time processing and storage of shield tunneling construction sequence data based on the integration of batch processing and batch processing, which is based on the present invention.

[0019] Figure 2 This is a flowchart of the process for constructing standardized records according to the present invention.

[0020] Figure 3 This is a flowchart illustrating the process of identifying the operating status of a tunnel boring machine according to the present invention.

[0021] Figure 4 This is a flowchart of the calculation and determination threshold of the present invention. Detailed Implementation

[0022] Example 1: A method for real-time processing and storage of shield tunneling construction time sequence data based on integrated batch processing is provided, mainly including the following steps:

[0023] Step 101: Obtain the original time-series data generated by the tunnel boring machine during construction, as well as the associated original ring number.

[0024] Specifically, raw time-series data refers to the sequence of physical signals directly acquired from various sensors and programmable logic controllers of the tunnel boring machine (TBM) through a multi-channel acquisition system. In some embodiments, it may include basic physical quantities reflecting the operating conditions of the TBM, such as tunneling pressure, thrust, torque, and attitude parameters. The raw ring number refers to the basic ring number identifier transmitted back by the acquisition system along with the data stream; it typically serves as a discrete spatial marker to distinguish the progress of tunnel excavation. At the TBM construction site, due to the complexity of equipment types and the harsh communication environment, the directly acquired raw time-series data often suffers from problems such as inconsistent field naming, inconsistent physical dimensions, and data communication delays.

[0025] In some alternative implementations, the acquisition of the raw time-series data can be achieved through a dual-channel access architecture. Specifically, for monitoring data with high real-time requirements, real-time access can be achieved through message queue middleware to ensure low-latency transmission; while for non-real-time or historical operational data, it can be temporarily stored in a distributed database to form a traceable time-series data base, providing sufficient data support for context reassembly in subsequent delayed batch processing.

[0026] Step 102: Perform standardization processing on the original time-series data based on the pre-built semantic rule base to obtain standardized records.

[0027] In this embodiment, the pre-built semantic rule base refers to a multi-dimensional cleaning and mapping dictionary pre-configured in a relational database, which includes standardized field names, data type templates, and SI conversion coefficients. Standardization processing mainly involves using the semantic rule base to convert the original time-series data, which are scattered and have varying formats, into a unified structure. Specifically, when the data stream enters, the system calls the semantic rule base to rename the fields of each data source, parse their data types, and unify their physical units. The standardized records output after processing have a unified structural form, eliminating data format differences at the underlying hardware device level, thus laying a data foundation for the stable operation of the subsequent unified algorithm model.

[0028] Furthermore, the standardization process may also include a preliminary identification mechanism for abnormal placeholders. When the system detects missing values ​​or error codes generated by offline sensors during data type format verification, it does not discard the data directly, but retains the record and adds a corresponding type error flag to maintain the continuity of the standardized record over time to the greatest extent possible.

[0029] Step 103: Based on the standardized records and judgment thresholds, identify the operating status of the tunnel boring machine and determine the target ring number.

[0030] In this step, the threshold value refers to the numerical limit used to distinguish the critical points of different operating conditions, which may specifically include the minimum thrust threshold, minimum torque threshold, and minimum speed threshold. Operating status refers to the physical condition of the tunnel boring machine at a specific moment, which can be divided into four states: tunneling, shutdown, maintenance, and abnormal. The target ring number refers to the final ring number to which the record actually belongs after independent confirmation or correction by the system data layer.

[0031] By extracting key monitoring parameters from the standardized records and comparing them with the judgment threshold, the real-time status of the tunnel boring machine can be inferred. Simultaneously, by independently detecting ring number boundary events based on the changing characteristics of physical parameters, the absolutely correct target ring number can be identified.

[0032] As an optional implementation, the judgment threshold can be set not only as a globally fixed constant, but also dynamically scaled based on geological exploration profile data along the tunnel. By incorporating geological information to adjust the judgment threshold, large-scale misjudgments of working conditions can be effectively avoided when the tunnel boring machine traverses uneven strata.

[0033] Step 104: Execute the corresponding data cleaning strategy on the standardized records according to the operating status, and evaluate the cleaning results to obtain a comprehensive quality score.

[0034] Specifically, data cleaning strategies refer to specific cleaning actions dynamically matched based on different operating conditions. For example, strict data range verification is performed during tunneling, while gradual descent tolerance verification is performed on propulsion parameters during shutdown. The comprehensive quality score is a comprehensive indicator that quantifies the reliability of a single cleaned data record. After executing the corresponding data cleaning strategy, characteristic parameters such as field completeness rate (key field completeness rate) and status continuity are extracted from the record, and the comprehensive quality score is calculated through multi-factor weighted calculation.

[0035] Traditional data cleaning methods typically employ static rules, which can easily lead to the accidental deletion of reasonable fluctuations caused by normal operating condition switching. By introducing a state-driven data cleaning strategy, the system can accurately identify false anomalies by combining actual on-site operating conditions, and the calculated comprehensive quality score provides a quantitative basis for subsequent data routing and distribution decisions.

[0036] In some preferred embodiments, a physical consistency verification module based on the mechanical mechanism of tunnel boring machines can be introduced when calculating the comprehensive quality score. By verifying the kinematic or static coupling relationship between various physical monitoring parameters, the sensitivity of the comprehensive quality score in identifying deep-seated physical logic anomalies can be further improved.

[0037] The arrival delay is calculated based on the difference between the arrival time and the acquisition time of the cleaned standardized record.

[0038] Step 105: Based on the diversion conditions including the comprehensive quality score and arrival delay, the cleaned standardized records are routed to the matching processing channel and written into the pre-configured time series database with the target ring number as the reference.

[0039] In this embodiment, arrival delay refers to the time difference between the generation of the data and its entry into the processing engine. The diversion condition refers to the logical discriminant that determines the data's destination, the core logic of which is to comprehensively consider the timeliness and reliability of the data. The processing channels are specifically divided into a real-time stream processing channel for fast response, a delayed batch processing channel for context re-judgment, and an exception retention channel for archiving and verification.

[0040] A time-series database is a target data warehouse that ultimately stores cleaned, high-quality data.

[0041] The traffic splitting conditions are logically determined based on pre-configured quality and latency thresholds. High-scoring, low-latency records are directly injected into the streaming processing channel for second-level data entry, while low-scoring or late records are diverted to the batch processing channel for reorganization and correction. Finally, data from all channels is written to the time-series database using the target ring number as the primary key of the storage partition.

[0042] Furthermore, for data entering the delayed batch processing channel, after time window reorganization and state re-judgment are completed, a batch replenishment identifier can be attached to it when writing to the time-series database. This identifier can be used to safely overwrite the corresponding dirty data previously written to the time-series database by the stream processing channel, achieving consistent closed-loop reconciliation of stream batch processing results at the underlying storage end.

[0043] Example 2 describes how to build a low-level rule dictionary containing multi-dimensional semantics and geological features offline before the system runs, and how to use this rule dictionary to resolve data semantic ambiguity, soft labeling of abnormal null values, and quantitative extraction of data quality evaluation indicators during real-time stream processing.

[0044] Step 201: Based on the pre-acquired list of historical engineering measurement points and physical dimension standards for the tunnel boring machine, construct the field semantic mapping table, the data type rule table, and the unit conversion rule table in a pre-configured relational database. The data type rule table is configured with a working condition tolerance strategy.

[0045] Specifically, the historical engineering monitoring point list mainly records the sensor tag names preset by different models and manufacturers of tunnel boring machines. The relational database can be implemented using a general-purpose database system with structured query capabilities.

[0046] When constructing the field semantic mapping table, the system not only maintains the basic key-value mapping relationship between the original field names and the standard field names, but also injects four additional semantic tags into it, specifically including physical quantity category code, measurement point role code, equipment component code, and sampling source code.

[0047] For example, for raw data related to thrust, the physical quantity category code identifies its mechanical classification, the measurement point role code distinguishes whether it is total thrust or single cylinder thrust, and the equipment component code locates its belonging to the propulsion hydraulic cylinder, so as to eliminate the physical semantic conflicts that may occur when merging multi-source heterogeneous data.

[0048] Furthermore, when constructing the data type rule table, a corresponding data type template, legal value range boundary values, and state-related tolerance strategies are configured for each standard field. The state-related tolerance strategies are mainly used to define the permissible behavioral deviations of specific parameters under different shield tunneling operating conditions. For example, setting the propulsion speed to be allowed to be zero in a stopped state without being considered abnormal. Simultaneously, the unit conversion rule table is used to store the multiplication coefficients and offset constants required to convert various heterogeneous data into a unified International System of Units (SI).

[0049] In some optional implementations, to reduce the latency of joint queries across multiple relational data tables, the three separately constructed rule tables can also be optimized and merged into an integrated semantic template library. Specifically, the semantic template library uses a standard field identifier as a unique primary key and stores the set of field aliases, applicable working conditions, missing data compensation methods, and numerical conversion coefficients as nested attributes. When data enters the cleaning process, the system only needs to perform a single memory-level match using the field aliases to obtain the complete processing constraints.

[0050] Step 202: Based on the pre-acquired tunnel geological survey data, construct a geological profile table containing mileage intervals and representative stratigraphic strength parameters in the relational database to obtain the pre-constructed semantic rule base.

[0051] In this embodiment, the geological survey data originates from the on-site geological drilling report prior to tunnel construction. The motivation for constructing the geological profile table is to provide an accurate spatial three-dimensional geological context for subsequent calculation of adaptive state thresholds. The system enters data in ascending order of mileage, with each record strictly defining the mileage start point, mileage end point, stratigraphic physical classification, and the representative stratigraphic strength parameter representing the compressive strength characteristics of that section. For soft soil sections, the representative stratigraphic strength parameter can be set as the standard penetration test blow count; for hard rock sections, it can be set as the unconfined uniaxial compressive strength of the rock. Through this step, the system transforms the static text survey report into a digital geological slice matrix that can be retrieved by the program in real time.

[0052] Step 203: During the processing of the original time-series data, the pre-built semantic rule library is hot-loaded into the memory of the real-time computing node through a state broadcasting mechanism.

[0053] Specifically, the state broadcast mechanism relies on the global state distribution interface at the bottom layer of the distributed stream processing computing engine. Since tunnel boring machine (TBM) construction typically lasts for several months, the need to add new sensors or correct calibration coefficients is inevitable during construction. To avoid restarting the entire real-time computing task every time rules are updated, the system monitors the change logs of the relational database. Once a rule update instruction is detected, the system immediately captures all the latest rules and pushes them as broadcast state data streams to all downstream parallel worker nodes. After receiving the broadcast data, each worker node seamlessly replaces the old rule dictionary in its local memory, thus achieving hot switching of the cleaning logic within milliseconds.

[0054] Step 204: Based on the field semantic mapping table, parse the original time series data to determine the standard field name, physical quantity category code, and measurement point role code corresponding to each data field.

[0055] As data flows into the cleaning pipeline, the system extracts the name of each original field from the raw time-series data and uses it as a search key to perform a lookup comparison in the field semantic mapping table loaded in memory. Upon successful matching, the system extracts the corresponding standard field name and all related semantic attribute codes, expanding the originally flat numerical record into a structured data body containing multi-dimensional identity information. For unknown fields where no match is found in the rule base, the system does not forcibly intercept them but assigns them a default unrecognized semantic code and retains their original data value for downstream transmission, preventing the loss of potentially unknown high-value on-site information.

[0056] Step 205: During the format validation process, when a data field is identified as null or an abnormal placeholder, a null value reservation mark is added to the corresponding data field.

[0057] To address common communication interruptions or device offline issues in industrial field sensors, a retention flag is implemented. Specifically, anomaly placeholder codes are invalid constants automatically filled by the underlying programmable logic controller (PLC) when it cannot obtain a real signal, such as -9999 or a non-numeric character. These types of characteristic data are identified through regular expression scanning.

[0058] Unlike traditional data cleaning processes that directly remove outliers or perform crude mean interpolation, this solution adds a null value retention flag. The purpose is that during manual maintenance shutdowns of the tunnel boring machine, signal loss due to power outages in some sensors is a normal phenomenon expected during construction. Deleting these records prematurely would disrupt the strict equidistant distribution of time-series data, causing subsequent pattern recognition algorithms that rely on fixed-length historical windows to fail.

[0059] Step 206: When the data field type conversion fails, attach a type error flag to the corresponding data field.

[0060] Specifically, the system attempts to perform forced type conversion on each field based on the type templates defined in the rule base. For example, if a pressure field is required to be a high-precision floating-point number, but a garbled string containing unexpected letters is actually received, the conversion process throws an exception. The system catches this exception, stops parsing the field's value, preserves the original erroneous string, and attaches the aforementioned type error flag in parallel. This operation also aims to protect the integrity of the data's appearance framework, delegating the final processing decision to a downstream exception preservation channel that incorporates operational status.

[0061] Step 207: Based on the number of data fields containing the null value retention flag and the type error flag, calculate the key field completeness rate and the total field missing rate of the record respectively, and write the key field completeness rate and the total field missing rate as metadata into the standardized record.

[0062] Previously, anomalous data was only qualitatively characterized by attribute tagging. The system must further convert these tags into continuous floating-point numerical parameters for use in subsequent dynamic routing formulas. The system uses separate denominator bases to calculate these two characteristic indicators separately.

[0063] Specifically, the key field completeness rate c _i =N _valid_key / N _total_key ; where N _valid_keyN represents the number of key monitoring parameter fields in this record that successfully passed format validation and were not tagged. _total_key This refers to the total number of key monitoring parameter fields that are strictly defined in the system configuration file beforehand. These key monitoring parameters typically refer to core control variables that directly affect equipment safety and tunneling attitude, such as propulsion force, cutterhead torque, propulsion speed, and soil chamber pressure.

[0064] Meanwhile, the total field missing rate m _i =N _missing_all / N _total_all ; where N _missing_all N is the total number of fields in the record that have been appended with the null value preservation flag or the type error flag. _total_all This represents the total number of all sensor point fields actually included in this single physical record.

[0065] Two completely independent calculation logics are used because they reflect different levels of quality degradation risk. The total field missing rate is used to macroscopically measure the overall congestion of the current communication link or the degree of large-scale equipment failure; while the critical field completeness rate is used to implement a veto system. Even if the total field missing rate is low, if core parameters such as propulsion are damaged, causing a decrease in the critical field completeness rate, the record is invalid data for the subsequent physical consistency verification model and must be downgraded.

[0066] Step 208: Based on the data type rule table and the unit conversion rule table, perform format verification and unit unification on the parsed data fields, and output the standardized record.

[0067] After completing the above type verification and labeling statistics, for normal data fields that do not have any error or missing markers, the system extracts the corresponding parameters from the unit conversion rule table and performs algebraic conversion operations.

[0068] Q _SI_i =k _U *Q _src_i ; where Q _SI_i k is the converted output value in International Standard Units (SI). _U Q is a pre-configured conversion factor loaded into memory via a broadcast mechanism. _src_i The system parses and extracts the raw collected values. For example, it multiplies imperial torque values ​​by a fixed constant to convert them to Newton-meters. After performing the above operations on all normal fields, the system encapsulates and packages the mapped names, the converted uniform values, and numerous evaluation metadata, including completeness and missing rates, to formally generate and output the standardized record to downstream nodes.

[0069] In some embodiments, an on-machine offset constant, such as Q, can also be used. _SI_i =k_U *Q _src_i + b _U For sensors requiring zero bias correction, a non-zero b can be configured. _U accomplish.

[0070] Example 3: Based on the above examples, this example describes how to combine the three-dimensional spatial geological information of the tunnel to dynamically calculate and smooth the threshold for identifying the working condition of the tunnel boring machine, and then combine the historical context to determine the actual operating condition.

[0071] It should be noted that, within each data processing cycle, the target ring number on which the dynamic determination threshold is calculated is a historical target ring number that has been confirmed and stored in the memory state machine at the end of the previous calculation cycle, rather than a new ring number that has just been detected in this step within the current cycle.

[0072] When the system starts up for the first time, the initial target ring number is preloaded into the memory state machine using the starting ring number of the shield tunneling shaft (usually the first ring) as the initial value. This time-series decoupling mechanism, which uses the state of the previous cycle to drive the calculation of the current cycle, can effectively avoid the circular dependency between the target ring number and the decision threshold, and ensure the deterministic sequential execution of each processing step within each calculation cycle.

[0073] Step 301: Calculate the current mileage of the tunnel boring machine based on the target ring number corresponding to the current record, the pre-configured tunnel starting mileage, and the pre-configured design ring width.

[0074] Specifically, shield tunnel construction often spans several kilometers. In the time-series data stream, a single record does not contain absolute spatial coordinates. The system must establish a mapping relationship between time, space, and geology through logical deduction. The pre-configured tunnel starting mileage refers to the absolute station mileage value of the shield launching shaft. The pre-configured design ring width refers to the theoretical design length of a single ring of precast concrete segments along the tunnel axis. By reading the target ring number in each record, which has been independently confirmed or corrected by the data layer, and combining it with the aforementioned engineering constants, the system converts the discrete ring number identifiers into continuous absolute spatial mileage coordinates.

[0075] In this step, the current mileage χ _i =χ _0 +ring _id_i *L _ring ;

[0076] Where, χ _0 For the pre-configured tunnel starting mileage, ring _id_i Let L be the target ring number. _ringThe system uses a pre-configured design ring width. For example, if the starting mileage of the tunnel is set to 1 kilometer and the design ring width is 1.2 meters, then when the target ring number of the data record is ring 50, the calculated current mileage is 1060 meters. Through this mapping mechanism, the system successfully assigns spatial geometric attributes to purely time-dimensional data.

[0077] Step 302: Based on the current mileage, query the corresponding representative stratigraphic strength parameters in the pre-constructed geological profile table.

[0078] In this embodiment, the pre-constructed geological profile table is a spatial lookup dictionary obtained by discretizing exploration borehole data. It records the geotechnical mechanical characteristics of different mileage intervals along the tunnel. Using the calculated current mileage as the retrieval key, interval matching is performed in the geological profile table to extract the representative stratum strength parameters corresponding to that point. The selection rules for representative stratum strength parameters depend on the main geological medium. For soft soil strata, the standard penetration test blow count is typically used as the characterization value; while for hard rock strata, the unconfined uniaxial compressive strength of the rock is preferred.

[0079] Step 303: Based on the ratio of the representative formation strength parameter to the pre-configured reference formation strength value, the pre-configured benchmark threshold is linearly scaled to calculate the initial judgment threshold.

[0080] Furthermore, traditional state recognition algorithms often employ globally fixed thrust and torque thresholds. However, the normal thrust required for a tunnel boring machine (TBM) to excavate in soft clay may be deemed an abnormal shutdown state in dense gravel strata, and vice versa. To address this challenge, an adaptive threshold mechanism that dynamically changes with geological conditions is introduced. The system uses a pre-configured reference stratum strength value as a base, calculates the proportionality coefficient of the current representative stratum strength parameter relative to the base, and directly applies this proportionality coefficient to the pre-configured benchmark threshold, thereby dynamically raising or lowering the threshold water level.

[0081] F _min_initial_i =F _min_ref *(σ _geo_i / σ _ref );

[0082] Among them, F _min_initial_i As the initial threshold for determining propulsion, F _min_ref σ is the pre-configured reference threshold for propulsion. _geo_i σ represents the representative stratigraphic strength parameter at the current mileage. _refThis refers to a pre-configured reference stratum strength value. For example, if the standard blow count for the reference stratum is set to ten blows, the corresponding baseline thrust threshold is 5000 kN. When the tunnel boring machine (TBM) traverses a soft clay stratum with a standard blow count of four blows, the calculated initial judgment threshold will automatically and proportionally decrease to 2000 kN; however, when the TBM subsequently enters a dense gravel stratum with a standard blow count as high as thirty-five blows, the initial judgment threshold will instantly increase to 17500 kN. This adaptive scaling mechanism enhances the robustness of the state determination logic to complex and composite strata.

[0083] Step 304: Based on the pre-configured smoothing coefficient, perform time-series smoothing weighted calculation on the initial judgment threshold and the pre-stored historical judgment threshold of the previous moment, and output the smoothed judgment threshold.

[0084] It should be noted that while the stratigraphic boundary is represented as a sudden step function in the theoretical model, the massive physical inertia of the tunnel boring machine (TBM) means that its actual operating parameters cannot undergo instantaneous changes. If the initial threshold calculated above is used directly, it can easily cause drastic fluctuations in the threshold at the interface between soft and hard strata, leading to high-frequency oscillations in the state recognition engine between tunneling and anomalies. To suppress the algorithmic instability caused by spatial abrupt changes, a first-order hysteresis filtering stage is introduced. The system utilizes a pre-configured smoothing coefficient to fuse and weight the newly calculated initial threshold with the historical threshold from the previous moment, pre-stored in memory.

[0085] F _min_i =λ*F _min_initial_i +(1-λ)*F _min_prev ;

[0086] Among them, F _min_i The threshold value for smoothed propulsion, λ is the pre-configured smoothing coefficient, and F is the threshold value for smoothed propulsion. _min_initial_i F is the initial judgment threshold. _min_prev This is the pre-stored historical judgment threshold from the previous moment. By setting a reasonable smoothing coefficient, such as 0.1, the system can force the jump process of the initial judgment threshold to be transformed into a smooth-transitioning exponential decay curve, which conforms to the objective engineering law of the large inertia of the tunnel boring machine's physical system.

[0087] The process of identifying the operating status of the tunnel boring machine based on the standardized records and judgment thresholds mainly includes the following steps:

[0088] Step 305: Extract key monitoring parameters from the standardized records. The key monitoring parameters include at least propulsion force, cutterhead torque, and propulsion speed.

[0089] After dynamically generating the adaptive judgment threshold, the system proceeds to analyze the time-series data itself. The system parses the standardized records, which have been cleaned by the front end, and extracts key monitoring parameters that most directly reflect the tunnel boring machine's excavation and forward movement status. The thrust reflects the total power to overcome the frontal earth pressure and shield friction; the cutterhead torque represents the rotational resistance to cutting the ground; and the thrust speed is the actual motion response of the equipment as it cuts into the ground. The combination of these three parameters constitutes the necessary and sufficient conditions for judging any physical working condition.

[0090] Step 306: Based on the comparison results between the key monitoring parameters and the judgment threshold, generate an instantaneous status judgment result.

[0091] Specifically, the system constructs a set of multi-dimensional logical threshold rules. The system compares the extracted real-time values ​​of propulsion force, cutterhead torque, and propulsion speed with the smoothed propulsion force threshold, torque threshold, and speed threshold output from previous steps. When the propulsion force is greater than or equal to the propulsion force threshold, the cutterhead torque is greater than or equal to the torque threshold, and the propulsion speed is strictly greater than the speed threshold, the system determines that the equipment is in a continuous working and advancing state, thus generating an instantaneous state determination result for the tunneling state. If all parameters fall significantly below a specific lower limit, a shutdown state determination result is generated. Any parameter combination that does not conform to the known normal physical coupling mode is directly intercepted and an abnormal state determination result is generated.

[0092] Step 307: Perform a consistency comparison calculation between the instantaneous state determination result and the pre-stored state history sequence to obtain the state continuity confidence level.

[0093] In practical engineering, sensor bus communication is highly susceptible to electromagnetic interference from high-power electrical appliances such as frequency converters, causing some key monitoring parameters to drop to zero or plummet within a single sampling period. If absolute judgments are made based solely on single-point data, numerous false shutdowns or abnormal alarms will occur. Therefore, the system maintains a fixed-length sliding time window in the memory backend to store historical operating condition tags over a continuous period. The system inputs the currently generated real-time status judgment result into this sliding window and calculates the frequency of its consistency with historical states, using this as a basis for evaluating the reliability of the current judgment.

[0094] State continuity confidence s _i =(1 / w)*SUM(I _func (π _raw ,π _hist ));

[0095] Where w is the total length of the sliding history window, SUM is the summation operation for all elements within the history window, and I... _funcA binary indicator function π for determining whether the values ​​of two internal variables are exactly equal. _raw For the immediate state determination result, π _hist This score is the label extracted for each historical state when traversing a pre-stored sequence of state histories. It visually reflects the stability of the current physical conditions over recent time.

[0096] Step 308: When the confidence level of the state continuity meets the preset conditions, the instantaneous state determination result is confirmed as the running state.

[0097] In this step, the preset condition refers to the minimum confidence threshold that the system can tolerate for state transitions. When the calculated confidence level of state continuity is greater than or equal to the specific confidence threshold, the system considers the current state identification not to be caused by occasional noise, but to indicate a confirmed change or continuation of operating conditions, and then formally solidifies and outputs the instantaneous state determination result as the final operating state. Conversely, if the confidence level does not meet the threshold, the system will discard the instantaneous determination result and force the current operating state to inherit the steady-state label from the previous moment. This secondary verification mechanism based on contextual confidence is equivalent to adding a digital anti-shake gimbal to the state identification engine, eliminating glitches and fluctuations in state labels in discrete time-series data, and providing an absolutely stable and reliable source of scheduling instructions for the differentiated cleaning algorithm that depends on operating conditions in subsequent embodiments.

[0098] The confidence threshold can be determined by those skilled in the art based on actual engineering needs by analyzing the state transition frequency and sensor interference statistical characteristics in historical shield tunneling construction data.

[0099] Example 4: This example describes how, during system operation, the absolute dependence on the ring number returned by the programmable logic controller is eliminated, and a weighted voting state machine is constructed using multi-channel physical signals to independently detect boundary events of shield tunnel segment assembly and realize autonomous ring number correction.

[0100] Step 401: Extract the physical monitoring parameters for ring number detection from the standardized record. The physical monitoring parameters for ring number detection include at least the propulsion force, propulsion speed, propulsion stroke, and grouting pressure.

[0101] Specifically, in actual tunnel boring machine (TBM) construction environments, the original ring numbers returned by the programmable logic controller (PLC) often exhibit systematic deviations. Common anomalies include delayed reports where tunneling has begun but the ring number has not yet incremented, and premature reports where the system has skipped a ring number before the tunnel segments have been assembled.

[0102] To establish an independent calibration mechanism at the data processing layer, it is necessary to bypass the limitations of directly relying on network-returned tags and seek the absolute truth from the physical motion trajectory of the equipment at its core. In this step, the system specifically extracts four core monitoring data: propulsion force, propulsion speed, propulsion stroke, and grouting pressure. Propulsion stroke represents the physical elongation of the propulsion cylinder, while grouting pressure reflects the synchronous grouting condition behind the shield tail after the segment exits. These parameters are the golden indicators for reconstructing the complete operation cycle of the tunnel boring machine.

[0103] Step 402: Calculate the propulsion stroke increment between adjacent sampling points, and accumulate the propulsion stroke increment to obtain the cumulative stroke amount.

[0104] In this embodiment, the advance stroke value at a single time point only represents the current absolute extension length of the hydraulic cylinder. Due to differences in the initial jacking positions of different rings caused by manual intervention, directly using the absolute position cannot accurately measure the advance of a single ring. Therefore, the system introduces a differential accumulation mechanism. Specifically, the system reads the difference in hydraulic cylinder displacement between the current sampling point and the previous sampling point as the advance stroke increment between adjacent sampling points. Subsequently, the system continuously integrates and accumulates the above increment in a strictly isolated memory state machine.

[0105] The above difference accumulation process is as follows:

[0106] d_acc _i =d_acc _i -1+Δd _i ;

[0107] Among them, d_acc _i d_acc represents the cumulative travel amount at the current moment. _i -1 represents the cumulative amount of the process stored in memory at the previous moment, Δd _i This represents the incremental travel distance between adjacent sampling points. This design enables the system to continuously memorize long-period spatial displacements, successfully transforming discrete single-point sensor jitter into a macroscopically smooth distance measurement scale.

[0108] Step 403: The matching feature between the propulsion stroke and the pre-configured design ring width is determined by the comparison result between the cumulative stroke and the pre-configured design ring width.

[0109] Further, the pre-configured design ring width refers to the standard ex-factory geometric width of a single-ring precast reinforced concrete segment, which can usually be set to 1.2 meters or 1.5 meters. After clarifying the above-mentioned continuous memory distance scale, the system will continuously monitor the climbing level of the stroke accumulation. When the stroke accumulation gradually approaches the pre-configured design ring width, it means that the actual displacement of the equipment in the physical space is sufficient to accommodate the assembly size of a complete ring of segments. The system calculates the absolute difference between the two in real time. Once this difference falls within the preset millimeter-level tolerance range, it can be highly confirmed from the geometric kinematics level that the tunneling task for this ring is coming to an end.

[0110] Step 404: Based on the sudden drop characteristic of the propulsion force, the zeroing characteristic of the propulsion speed, the matching characteristic between the propulsion stroke and the pre-configured design ring width, the switching characteristic of the operating state, and the pulse characteristic of the grouting pressure, construct a ring number boundary characteristic signal set.

[0111] In this step, the system completely quantifies the on-site experience of human engineering experts into five independent mathematical judgment thresholds, and thus constructs a full-dimensional characteristic matrix covering mechanics, kinematics, geometry, and technology. When the tunneling of a ring is completed and the segment assembly is about to start, the operator will stop the machine and retract the cylinder, thereby triggering synchronous grouting. This standard operating procedure leaves a physical mark on the sensor curve.

[0112] Regarding the five physical marks, the specific process of extracting the characteristic signals is as follows:

[0113] b _1_i =I _func ((F _i-1 -F _i ) / F _i-1 >γ _F );

[0114] b _2_i =I _func (v _i <v_zero and v _i-1 >=v_min);

[0115] b _3_i =I _func (abs(d _acc_i -L _ring )<ε_d);

[0116] b _4_i =I _func (π _i-1 equals tunneling and π _i equals shutdown);

[0117] b_ 5_i =I _func(There exists j belonging to the time interval [i, i+w_g] satisfying p) _g_j >p _g_threshold And p _g_j-1 <p _g_threshold );

[0118] Among them, b _1_i For the characteristic of a sudden drop in propulsion, I _func F is an indicator function that outputs one if the condition is true and zero otherwise. _i For current propulsion, F _i-1 For the propulsion force of the previous moment, γ _F b is the preset thrust drop ratio threshold. _2_i To propel the velocity to zero, v _i Given the current propulsion speed, v _zero For the lower limit of velocity approaching zero, v _min b is the lower limit of normal tunneling speed. _3_i To improve the matching characteristics between the travel and the pre-configured design loop width, abs is an absolute value function, d _acc_i For the cumulative trip amount, L _ring For the pre-configured design ring width, ε _d For travel tolerance, b _4_i π represents the switching characteristics of the running state. _i The current running state is π. _i-1 b represents the running state at the previous moment. _5_i The pulse characteristics of the grouting pressure are given by j, where j is the traversal index and w is the pulse characteristic of the grouting pressure. _g p is the width of the observation time window. _g_j The grouting pressure at the observation point, p _g_threshold This is the threshold pressure for triggering the grouting pulse.

[0119] The pulse characteristic b_ of the grouting pressure 5_i The observation time window of length wg is used. In the stream processing engine, caching and delay calculation can be realized through the event time sliding window mechanism. This mechanism is a common method in distributed stream processing technology in this field. The width wg of the observation time window can be adaptively configured by those skilled in the art according to the typical time interval from the shutdown of the tunnel boring machine to the triggering of synchronous grouting.

[0120] Step 405: Perform ring number boundary detection based on the ring number boundary feature signal set to obtain the independently detected ring number.

[0121] Specifically, at each time step, the Boolean states of the five characteristic signals mentioned above are calculated concurrently, assembling a high-dimensional signal vector reflecting the current critical characteristics of the device. The independent detection ring number is an incrementing integer variable independently maintained by the data processing platform, detached from the underlying hardware. Its initial value can be consistent with the first ring of the originating well. The system performs pattern recognition on the input high-dimensional signal vector using a specialized discriminant model. Once a signal combination that highly fits the ring-changing characteristics is detected, the system autonomously increments this independent variable at the data level.

[0122] The step of performing ring number boundary detection based on the ring number boundary feature signal set to obtain independently detected ring numbers includes:

[0123] Step 406: The weighted sum of each feature signal in the ring number boundary feature signal set and the pre-configured feature signal weights is calculated to obtain the weighted voting value.

[0124] In this embodiment, different sensors exhibit significant differences in their anti-interference capabilities and reliability under extreme operating conditions. For example, the speed zeroing signal is highly susceptible to false positives due to short-term shutdowns for slag removal, while features based on cumulative stroke are relatively robust. To enhance the robustness of the discrimination system, pre-configured feature signal weights are assigned to the five features, and weighted majority voting engineering logic is used instead of absolute decisions based on a single condition.

[0125] B _i =∑ k=1 5 (w _k *b _k_i );

[0126] Among them, B _i Let ∑ be the weighted voting value at the current moment. k=1 5 w is a summation function for five feature indices. _k b is the pre-configured feature signal weight corresponding to the k-th feature. _k_i This is the specific extracted value of the k-th feature signal at the current time.

[0127] Step 407: When the weighted voting value is greater than or equal to the pre-configured voting threshold, a ring number boundary event is determined to have occurred, and the ring number counter is incremented based on the ring number boundary event to update the independent detection ring number.

[0128] In continuous real-time streaming computation, the weighted voting value typically hovers at a low level. The system closely monitors this numerical sequence, and a pre-configured voting threshold can be set to 70% of the full weight of all features. When the weighted voting value exceeds this pre-configured threshold, a high-priority system interrupt is triggered in the background, determining that a genuine ring number boundary event has occurred at the current timestamp. Immediately afterwards, the system sends an increment signal to the ring number counter in memory, incrementing the original independently detected ring number by one.

[0129] Step 408: When the ring number boundary event is determined to occur, the travel accumulation is reset to zero to start a new round of travel accumulation for the target ring number.

[0130] Within the same millisecond that the system confirms the occurrence of the ring number boundary event, not only is the ring number counter incremented, but the spatial memory of the previous work cycle must also be immediately cleared. A zeroing operation is performed, erasing the accumulated travel value of 1.2 meters or 1.5 meters from memory. This ensures that in subsequent newly generated time-series data streams, all propulsion travel increments will restart from a clean zero position. Through this cyclical accumulation and zeroing mechanism, the physical process of the tunnel boring machine's cylinder extension and retraction is simulated.

[0131] Step 409: Compare the independent detection ring number with the original ring number, and determine the target ring number based on the comparison result.

[0132] At this point, two coordinate systems are simultaneously available. One is the original ring number sequence, which is prone to jumps or lags due to communication delays. The other is an independently detected ring number sequence based on a weighted reconstruction of massive amounts of underlying physical sensor features. The system loads these two integer variables in parallel and executes rigorous cross-validation comparison logic to eliminate inferior labels and extract the unique and valid spatial benchmark that can guide subsequent accurate statistical analysis, namely the target ring number.

[0133] The process of comparing the independent detection ring number with the original ring number and determining the target ring number based on the comparison result includes:

[0134] Step 410: When the independent detection ring number is consistent with the original ring number, the original ring number is confirmed as the target ring number.

[0135] Specifically, within most normal operating ranges where communication is unimpeded and the industrial control computer is functioning well, the ring-switching time calculated independently by the weighted state machine is highly synchronized with the ring-switching time automatically reported by the programmable logic controller. Under this ideal operating condition, the result of subtracting two integer variables is always zero. The system determines that the external environment is reliable and readily assigns the current value of the original ring number to the target ring number, while simultaneously preserving the original flow path attributes in the metadata of the data record.

[0136] Step 411: When the independent detection ring number is inconsistent with the original ring number, the independent detection ring number is confirmed as the target ring number, and a ring number correction mark is generated.

[0137] Furthermore, when network congestion causes tag delays, or when sensor malfunctions on the segment assembly machine causes premature number skipping, the comparison results of the two integer variables mentioned above will show a significant non-zero difference. For example, the system's underlying physical detection may indicate that the travel distance is full and the speed has returned to zero, and the independent detection ring number has jumped to the 100th ring, but the network packet is still stuck on the 99th ring.

[0138] The system seizes calibration authority, taking the independent detection ring number derived from the physical world as absolute truth, overwriting and confirming it as the final target ring number. Simultaneously, to provide troubleshooting clues for downstream big data backtesting analysis, the system adds a dedicated ring number correction marker to the corrected data entry, recording in detail the specific difference in value that occurred.

[0139] In some alternative implementations, for offline retransmission scenarios involving large volumes of data after communication interruption, since single-point streaming judgment cannot take the overall picture into account, the aforementioned ring number correction logic can also be optimized globally using a dynamic time planning algorithm in batch processing mode. By loading the historical waveforms of the entire work shift at once, the ring switching breakpoint with the most consistent energy consumption characteristics can be found from a global perspective, which can further improve the accuracy of ring number correction mark generation under complex electromagnetic interference environments.

[0140] Example 5: In this example, we mainly explain how to implement targeted and differentiated cleaning strategies for time-series data based on different operating states, and how to construct a hierarchical decision tree model to assign a comprehensive data quality level to the cleaned records.

[0141] Step 501: When the running state is the tunneling state, perform strict range verification based on the pre-configured legal range for each data field in the standardized record, and mark outliers that exceed the legal range.

[0142] Specifically, the tunneling state refers to the full-load operation condition where the tunnel boring machine's cutterhead continuously rotates and cuts the soil, and the propulsion cylinders continuously extend to perform work. Under this high-load physical environment, the readings of various sensors must strictly conform to the equipment design parameters and the bearing capacity limits of the current geology. The system extracts all data fields from the standardized records and cross-validates them against the absolute physical upper and lower limits pre-configured in the rule base. For example, the legal range of propulsion force for a certain type of tunnel boring machine is set to 2000KN to 15000KN. Once the received propulsion force value reaches 20000KN, the system immediately determines that the value exceeds the laws of physics or the hardware range, and labels the specific field as an outlier exceeding the legal range. Retaining the outlier value and its label instead of directly removing it is to provide real negative sample data for the neural network in subsequent model training, while ensuring the structural integrity of the time series matrix in the time dimension.

[0143] In some alternative implementations, the specific threshold for the strict range verification can be based not only on the absolute factory rating of the device, but also on a dynamic statistical boundary based on a time sliding window. Specifically, the system can calculate the mean and standard deviation of the physical parameter over the past one hundred sampling periods in real time, and dynamically tighten the valid interval to a range of three standard deviations above and below the mean, thereby achieving sensitive detection of minute sensor drift.

[0144] Step 502: When the operating state is the shutdown state, perform a slow descent tolerance check on the propulsion force and cutterhead torque in the standardized record based on a pre-configured working condition tolerance strategy.

[0145] In this embodiment, the shutdown state typically occurs during the segment assembly stage after the tunnel boring machine (TBM) has completed one ring of excavation. Due to the significant physical inertia and stress release lag between the massive mechanical equipment and the surrounding soil, when the operator issues a shutdown command, the thrust and cutterhead torque do not instantly drop to zero like electronic signals, but rather exhibit an exponentially decaying residual stress curve. If the strict range verification logic under the aforementioned tunneling state is continued, these normal residual readings in the decay process are easily misjudged as exceeding limits. Therefore, the system specifically activates the pre-configured working condition tolerance strategy for the thrust and cutterhead torque.

[0146] Specifically, after the operating state is detected to have switched to a shutdown state, non-zero residual values ​​of propulsion force and cutterhead torque are allowed within a certain tolerable attenuation range. As long as these residual values ​​show a monotonically decreasing trend over time and do not exhibit an abnormal secondary surge, they are considered to have passed the gradual descent tolerance check, and no abnormality markers are applied. This more closely reflects the actual mechanical release process of tunnel boring machine construction and reduces the false alarm rate of on-site data cleaning.

[0147] The observation time window length and the threshold of the second sudden increase in the descent tolerance test can be determined by those skilled in the art through statistical analysis of the residual stress curves in historical shutdown data, based on the inertial characteristics of the hydraulic system of the specific tunnel boring machine model. In an optional implementation, least squares linear fitting can be used to regress the residual value sequence within the window. If the slope is negative and the residuals conform to a monotonic trend, then the descent tolerance test is deemed to have passed.

[0148] Step 503: When the running state is the maintenance state, add expected missing markers to the data fields that are missing in the standardized records.

[0149] Furthermore, the maintenance status indicates that the equipment has entered a planned shutdown for maintenance or a long-term standby phase. During this period, to ensure the safety of operators or conserve energy, the power supply to high-power actuators such as the main drive of the cutter head and the screw conveyor, as well as the associated sensors, is usually proactively cut off on-site. This causes the programmable logic controller (PLC) to obtain a large number of null values ​​or communication error codes during polling. After confirming through the status engine that the current period is a legitimate maintenance period, these missing fields are no longer considered as communication faults or hardware damage, but are instead uniformly assigned a specially designed expected missing flag. This flexible handling strategy clarifies the business rationale for data missingness and prevents an alarm storm caused by misidentification in the underlying communication alarm system.

[0150] Step 504: When the running state is the abnormal state, retain all the original field values ​​in the standardized record and attach the abnormal reason code.

[0151] In this step, the abnormal state refers to a dangerous situation where the tunnel boring machine encounters extreme engineering accidents such as cutterhead jamming, main bearing overheating, or severe segment rupture. In such critical moments, every set of extreme out-of-limit values, or even erratic pulse jumps, recorded by sensors are invaluable resources for subsequent accident tracing, accountability determination, and safety review by engineering experts. Therefore, all data filtering and smoothing logic is eliminated, and the highest level of full data retention strategy is implemented. No interpolation or correction is performed on any garbled data exceeding physical limits; all original field values ​​in the standardized records are locked intact, and a unique anomaly cause code is added based on the characteristic dimension that triggered the abnormal situation. This improves the fidelity of digital evidence at the disaster site.

[0152] Step 505: Based on the verification pass rate, missing data rate, and abnormal retention rate of key and non-key data fields in the standardized records, assign a comprehensive data quality level to the cleaned standardized records.

[0153] After completing the aforementioned state-driven customized cleaning process, each data record is filled with various verification conclusions and traceability tags. To facilitate rapid data usability identification by downstream data mining models, visualization dashboards, and long-term storage components, the system must converge these scattered micro-tags into a macro-level quantitative rating. The system parses a pre-configured dictionary table, strictly dividing all monitoring variables into critical data fields for evaluating core operational safety and non-critical data fields for auxiliary monitoring. Subsequently, the system extracts the final verification results of these two types of fields and feeds them into a hierarchical decision tree logic, ultimately outputting a clear comprehensive data quality level for the entire time-series record.

[0154] Step 506: The comprehensive data quality level is divided into four levels: Level 1: all key data fields pass strict range verification; Level 2: non-key data fields are missing but key data fields pass verification; Level 3: key data fields are missing, abnormal, or trigger slack tolerance; Level 4: the running state is the abnormal state and all data fields are fully retained.

[0155] Specifically, these four levels correspond to perfect data, usable data, degraded data, and archived data in engineering semantics, forming a strict quality ladder. The first level has the most stringent requirements. Only when the record is in a normal tunneling state and all vital parameters, including propulsion and torque, fall flawlessly within the physical boundaries can it be awarded this highest rating, which can be directly used for training high-precision artificial intelligence prediction models.

[0156] Level 2 represents a common scenario in real-world operations involving defects. For example, an occasional disconnection of a pressure sensor at an auxiliary grouting hole at the tail of the shield could cause the loss of a non-critical data field, while the critical data fields of the main drive system remained completely normal and passed verification. The system classifies this as Level 2, indicating that although the data has a localized flaw, the overall structure remains healthy and is fully capable of supporting routine progress statistics and energy consumption analysis.

[0157] The third level is the concentrated embodiment of the fault tolerance mechanism. When the equipment is in a shutdown state and the thrust reduction tolerance is triggered, or when a core thrust sensor in the tunneling state shows an out-of-limit anomaly, the system will decisively downgrade its quality to the third level. This type of data contains residual physical inertia or local single-point faults and cannot be directly used for precise machine learning fitting. It must be sent to a delayed batch processing channel in a later stage for context re-judgment over a wider time window.

[0158] The fourth level is a special channel reserved for extreme incidents. Once the engine determines that the current state is abnormal, regardless of the field missing rate within the record, the system will bypass the regular quality bonuses and directly lock it to the lowest level, the fourth level. Data at this level is strictly sealed, warning downstream stream processing services not to access it, and is only available for in-depth incident reverse analysis in the offline data lake. Through this four-level classification mechanism, this solution successfully transforms the ambiguous industrial field data stream into standardized digital assets with clear contractual constraints.

[0159] Example 6 provides a physical consistency verification method based on mechanical mechanisms. Specifically, it addresses how to identify hidden coupling anomalies between various physical monitoring parameters.

[0160] It should be noted that the static calculation process involving thrust and earth pressure balance in this embodiment is preferably applicable to the specific model of earth pressure balance shield tunneling machine. For slurry balance shield tunneling machine or hard rock tunnel boring machine, it can be adapted to replace it with its corresponding air cushion chamber pressure model or rock breaking thrust model.

[0161] Step 601: When the operating state is the tunneling state, extract the physical monitoring parameters for mechanical verification from the cleaned standardized records. The physical monitoring parameters for mechanical verification include at least the thrust, cutterhead torque, thrust speed, cutterhead rotation speed, and soil chamber pressure.

[0162] Specifically, the physical coupling law only strictly holds when the equipment is cutting the soil at full load. Once the system confirms through the state engine that the current time-series record belongs to the tunneling state, it activates the physical consistency verification logic. The system specifically extracts five core variables from the cleaned, standardized records that can fully describe the closed-loop energy interaction at the excavation face. Extracting propulsion force and propulsion speed quantifies the longitudinal translational work done by the equipment; extracting cutterhead torque and cutterhead speed quantifies the rotational cutting work done by the equipment; and extracting soil chamber pressure establishes the mechanical boundary conditions for the excavation face support pressure.

[0163] Step 602: Based on the kinematic ratio of the propulsion speed to the cutterhead rotation speed, calculate the calculated penetration depth and determine the consistency deviation between the calculated penetration depth and the pre-configured reasonable range of penetration depth.

[0164] In this embodiment, penetration is defined as the absolute physical distance the tunnel boring machine (TBM) advances along the tunnel axis per revolution of the cutterhead. This is a rigid kinematic constraint independent of external geological conditions. The system uses the extracted propulsion speed and cutterhead rotation speed to derive the instantaneous theoretical value of this parameter through division.

[0165] p_r _i =v _i / n _i;where p_r _i To calculate penetration, v _i To increase speed, n _i This represents the rotational speed of the cutter head.

[0166] Furthermore, in engineering practice, to facilitate on-site personnel's intuitive verification of parameters, the unit for feed speed is usually retained as millimeters per minute (mm / min) and the unit for cutterhead rotation speed is revolutions per minute (RPM). With this non-international standard unit input, the system automatically performs a division operation internally to directly cancel out the time dimension, thus smoothly outputting the calculated penetration value in the engineering-customary unit of millimeters per revolution. Subsequently, the system subtracts the calculated penetration from the median of the pre-configured reasonable penetration range, takes the absolute value, and then divides it by the median to calculate the normalized penetration consistency deviation. If, at a certain moment, the feed speed is extremely high while the cutterhead rotation speed is extremely slow, causing the calculated penetration to deviate significantly from the reasonable range, the system can determine that one of the two independent sensors must be drifting.

[0167] Step 603: Based on the propulsion force, the cutterhead torque, the propulsion speed, the cutterhead rotation speed, and the pre-configured excavation surface area, perform power conversion to obtain the calculated specific energy, and determine the consistency deviation between the calculated specific energy and the pre-configured specific energy within a reasonable range.

[0168] Specifically, the system calculates the difference between the calculated specific energy and the median of the pre-configured reasonable range of specific energy, takes the absolute value, and then divides it by the median to calculate the normalized specific energy consistency deviation δSEi. The calculation formula is: δSEi =|SEi - SEmid| / SEmid; where SEmid is the median value of the pre-configured reasonable range of specific energy.

[0169] In this step, specific energy is defined as the total mechanical energy consumed by the tunnel boring machine (TBM) cutterhead in excavating a unit volume of stratum. The overall tunneling energy efficiency of the system is measured by summing the rotary cutting power and the longitudinal propulsion power and dividing by the volume of soil excavated per unit time.

[0170] SE _i =(2*π*M _i *n _i +F _i *v _i ) / (A _face *v _i );

[0171] Among them, SE _i To calculate the specific energy, π is the constant of pi, and M _i n is the torque of the cutter head. _i F is the rotational speed of the cutter head. _i For propulsion, v _i To increase speed, A_face This refers to the pre-configured excavation area.

[0172] This explanation uses an earth pressure balance tunnel boring machine (TBM) with a cutterhead diameter of 6.28 meters, applied to typical soft soil strata, as an example. The pre-configured excavation face area is calculated to be approximately 30.97 square meters using geometric formulas. Assume that at a certain sampling moment, the system extracts a cutterhead torque of 800 kNm, a cutterhead rotation speed of 1.5 rpm, a thrust of 12,000 kN, and a thrust speed of 40 mm / min. Before performing the calculation, the system converts these parameters to SI units and substitutes them into the formula, ultimately obtaining a calculated specific energy of approximately 6.47 MJ / m³. The system determines that this value falls perfectly within the reasonable range of 3 to 15 MJ / m³ for typical soft soil, therefore the consistency deviation of the output specific energy is close to zero.

[0173] Step 604: Calculate the theoretical thrust based on the earth chamber pressure, the pre-configured shield friction coefficient, the pre-configured lateral earth pressure, the pre-configured outer area of ​​the shield, and the excavation face area, and determine the thrust balance deviation between the theoretical thrust and the thrust.

[0174] Specifically, the system takes the absolute value of the difference between the theoretical thrust and the actual thrust, divides it by a pre-configured reference thrust benchmark value, and normalizes the result to calculate the dimensionless thrust balance deviation δF. _i , i.e. δF _i = |F_theory _i - F _i | / F _ref Among them, F _ref The pre-configured reference thrust baseline value can be taken as the pre-configured reference threshold F_min. _ref Or other reference quantities with the same dimensions as propulsion force.

[0175] Specifically, this is a static verification barrier tailored for earth pressure balance tunnel boring machines (TBMs). Under uniform speed tunneling conditions, the total forward thrust of the TBM's main hydraulic cylinders must be macroscopically equal to the sum of the earth pressure reaction force at the excavation face and the frictional resistance of the outer surface of the shield.

[0176] Theoretical propulsion F _theory_i =p _i *A _face +μ*p _lateral *A _shield ;

[0177] Where, p _i For the pressure of the earthwork, A _face Here, μ is the pre-configured excavation face area, and p is the pre-configured shield friction coefficient. _lateral For the pre-configured lateral earth pressure, A_shield The area of ​​the pre-configured outer side of the shield shell.

[0178] The system calculates the theoretical thrust required to maintain balance under the current operating conditions through the aforementioned mechanical modeling, and compares it with the actual thrust collected by the sensor. If the difference between the two is too large, resulting in a highly significant thrust balance deviation, it indicates that the system has encountered an unknown surge in external resistance or that the soil pressure sensor has experienced a severe mud clogging failure.

[0179] Step 605: Calculate the physical consistency score based on the penetration consistency deviation, the specific energy consistency deviation, and the thrust balance deviation.

[0180] After completing the cross-validation of the three mechanical dimensions, the system needs to converge the three discrete deviation indicators into a unified quality control variable. The system sets weighting coefficients to linearly combine the three deviation values ​​and deduct them from the full score benchmark.

[0181] φ _i =1-(β _1 *δ _pr_i +β _2 *δ _SE_i +β _3 *δ _F_i );

[0182] Where, φ _i For the physical consistency score, β _1 For the pre-configured penetration weight, δ _pr_i β represents the penetration consistency deviation. _2 For the pre-configured specific energy weight, δ _SE_i β represents the specific energy consistency deviation. _3 For the pre-configured thrust balance weights, δ _F_i This refers to the thrust balance deviation.

[0183] In some alternative implementations, when the input data is extremely disordered, resulting in a boundary case where the weighted sum of the three deviation values ​​is greater than one, directly applying the above formula will calculate a negative physical consistency score.

[0184] To avoid unpredictable mathematical disturbances caused by negative values ​​in subsequent routing and distribution formulas, the system supplements this with a set of negative value truncation rules. Specifically, when the calculated physical consistency score is less than zero, the system forcibly truncates its value to zero and immediately triggers the highest-level physical coupling collapse warning signal in the metadata of that record, thereby ensuring the stability of the security domain for subsequent quality score calculations.

[0185] Step 606: Based on the physical consistency score, the comprehensive quality score is constructed by combining the field completeness rate of the standardized records.

[0186] The field completeness rate derived from missing and error statistics in the previous embodiments is dimensionally integrated with the physical consistency score derived from deep mechanical mechanisms in this embodiment. The field completeness rate represents the health of the data in terms of surface format, while the physical consistency score represents the credibility of the data in terms of its underlying logic.

[0187] When the operating state is not the tunneling state, the system does not activate the above-mentioned physical consistency verification logic, and the physical consistency score φ is then... _i The pre-configured default values ​​are used. Specifically, for records of shutdown and maintenance states, φ _i The default setting is a pre-configured non-tunneling baseline score (e.g., 0.5) to reflect the objective situation where such records lack effective evaluation criteria in the physical and logical dimensions; for records in abnormal states, since the system forcibly routes them to the abnormal retention channel without performing regular quality scoring calculations, φ _i Assigning a value does not affect the routing result and can be set to zero.

[0188] It should be noted that the above weighted calculation formulas primarily use numerical values, and in some embodiments, standardization and normalization can be performed. Common dimensionless methods include absorbing dimensions with coefficients, dividing by a reference value or unit 1. For example, dimensionless transformation of time parameters can be achieved using t / t0, t / t... ref When the denominator is 1, numerically, it represents the parameter itself. Normalization methods include various approaches such as maximum / minimum value normalization. These methods are common techniques in this field and will not be detailed here.

[0189] It should be noted that the aforementioned physical consistency check is only activated when the tunnel boring machine is in the tunneling state. When the operating state is stopped or under maintenance, since the cutterhead is not continuously performing work, the mechanical coupling relationship is not strictly valid, and the system sets the physical consistency score to the pre-configured default value. In one implementation, this default value is set to one, indicating that no additional penalty is imposed on the physical consistency dimension, and the comprehensive quality score is calculated based solely on three factors: field completeness rate, total field missing rate, and state continuity confidence.

[0190] Example 7 describes a multidimensional quality scoring and dynamic distribution method, namely, how to comprehensively consider the time delay and multidimensional quality characteristics of the data, dynamically distribute the cleaned time-series data to the stream processing channel, batch processing channel or abnormal channel, and realize the process of stream-batch integrated replenishment, coverage and reconciliation in the underlying time-series database.

[0191] Step 701: Calculate the arrival delay based on the difference between the arrival time and the acquisition time recorded in the cleaned standardized record.

[0192] Specifically, at the tunnel boring machine (TBM) construction site, data typically undergoes forwarding through multiple gateway nodes from its generation by the underlying programmable logic controller (PLC) to its reception by the data processing engine. The acquisition time refers to the absolute timestamp generated by the sensor's physical slice, and the arrival time refers to the absolute timestamp when the message queue middleware receives the data packet. Due to the complex wireless communication environment inside the tunnel or temporary network outages, the arrival time of some data packets may significantly lag behind the acquisition time. The system extracts these two time metadata elements from the cleaned, standardized records and performs a subtraction operation.

[0193] Δt _i =t _arr_i -t_event _i ; where Δt _i For arrival delay, t _arr_i For arrival time, t_event _i The acquisition time is the key parameter. This parameter precisely quantifies the latency of a single data record on the transmission link and is a core veto indicator that determines whether it is valuable for participating in real-time stream computing.

[0194] Step 702: Extract the completeness rate of the key fields and the missing rate of all fields recorded in the cleaned standardized records.

[0195] In this embodiment, the system directly loads the statistical features generated in the preceding cleaning and labeling step from the metadata area of ​​the standardized record. The field completeness rate represents the health of the core parameters in the record, while the total field missing rate reflects the macroscopic proportion of auxiliary sensor disconnections. These two parameters are numerically distributed between 0 and 1 and are physically independent, together forming a two-dimensional coordinate system for evaluating data format quality.

[0196] Step 703: Based on pre-configured weighting coefficients, the field completeness rate, the total field missing rate, the state continuity confidence score, and the physical consistency score are weighted and calculated to obtain the comprehensive quality score.

[0197] In this step, the system merges four evaluation metrics distributed across different dimensions into a single scalar score. These four metrics represent format health, overall missingness, temporal stability, and the truthfulness of underlying mechanical logic, respectively. The system balances the influence of each dimension using pre-configured weighting coefficients.

[0198] q _i =α _1 *c _i +α _2*s _i +α _3 *φ _i -α _4 *m _i ; where q _i For the overall quality score, α _1 For the pre-configured completeness weight, c _i α represents the field completeness rate. _2 For the pre-configured state continuity weights, s _i α represents the confidence level for state continuity. _3 For the pre-configured physical consistency weights, φ _i For the physical consistency score, α _4 For the pre-configured missing rate penalty weights, m _i This represents the total field missing rate.

[0199] The aforementioned weighting coefficients are subject to strict normalization constraints. Specifically, the sum of each positive gain weight and negative penalty weight must satisfy the normalization condition, i.e., α. _1 +α _2 +α _3 +α _4 =1. By adopting the above normalization constraint, the score can reach a definite upper limit of α1+α2+α3 when all positive indicators are full and the missing rate is zero, and a definite lower limit of negative α4 when the negative penalty is maximum. The physical boundary of the score interval is therefore determined by the weight parameters.

[0200] For example, α_1 can be set to 0.4, α_2 to 0.2, α_3 to 0.3, and α_4 to 0.1. When a record in the tunneling state has all fields complete and a perfect physical score (i.e., ci=1, si=1, φi=1, mi=0), its final calculated comprehensive quality score will approach the theoretical upper limit of α1+α2+α3 (i.e., 1-α4), reaching 0.9 with the example weights. Conversely, if the physical consistency score drops to zero due to sensor anomalies and some fields are missing, the comprehensive quality score will be quickly lowered below the pre-configured quality threshold, thus preventing it from being mixed into the high-fidelity data stream. In actual deployment, the pre-configured quality threshold should be adjusted accordingly based on the selected weight coefficients to ensure that high-fidelity data can be correctly routed to the real-time stream processing channel.

[0201] Step 704: When the running state is not the abnormal state, the arrival delay is less than or equal to the pre-configured delay threshold, and the comprehensive quality score is greater than or equal to the pre-configured quality threshold, the cleaned standardized record is routed to the real-time stream processing channel.

[0202] Furthermore, the system performs a rigorous three-way branch routing decision. This is the first preferred channel among the three branches. The system sets a pre-configured latency threshold, for example, five seconds, and a pre-configured quality threshold, for example, 0.85. When a record simultaneously meets the three stringent conditions of non-accident operation, extremely low transmission latency, and extremely high overall quality, the system determines that the record is completely reliable and has excellent timeliness. It is then directly injected into the real-time streaming processing channel. This channel skips the cumbersome context reorganization logic, pursuing the ultimate processing throughput and extremely low end-to-end latency, providing second-level parameter feedback to the on-site construction command screen.

[0203] Step 705: When the operating state is the abnormal state, the cleaned standardized record is routed to the abnormal retention channel.

[0204] This is the second isolation channel in the three-way branch. When the system's front-end state recognition engine determines that the current record occurred during an extreme accident or disaster downtime, the system will trigger a forced circuit breaker mechanism regardless of its arrival latency or field integrity status. The system bypasses the regular quality scoring comparison logic, packages the entire original data, and routes it to the anomaly preservation channel. This strategy aims to protect valuable on-site physical snapshots that, although formatted incorrectly, contain clues to the accident, preventing them from being mistakenly discarded as noise by the stream processing model.

[0205] Step 706: When the above-mentioned stream processing and exception retention conditions are not met, the cleaned standardized records are routed to the delayed batch processing channel.

[0206] This is the third compensation channel in the three-way branch. For high-latency records arriving late due to network congestion, or low-quality records whose physical scores have decreased due to electromagnetic interference, the system will not discard them but will instead redirect them to the delay batch processing channel. Data entering this channel means that its single-point information is insufficient to prove its innocence; it must wait for time to pass and obtain corroboration from more adjacent records before final cleaning and evaluation can be completed. Through this dynamic routing mechanism, this solution completely breaks through the bottleneck of the traditional solution's rigid data source-based traffic distribution.

[0207] Step 707: In the real-time stream processing channel, a streaming first-write identifier is added to the received standardized record, and the record is written to the time series database based on the target ring number.

[0208] Specifically, for data that successfully enters the real-time stream processing channel, the system performs rapid, lightweight encapsulation. The system forcibly appends a specific string called a streaming first-write identifier to the record's tag metadata set, indicating that the record is the fastest decision-making output based on single-point local information. Subsequently, based on the target ring number it is attached to, the system locates the corresponding project library and ring number table partition in the time-series database and performs direct insert / write operations.

[0209] Step 708: In the delayed batch processing channel, the context window is reorganized based on the pre-stored adjacent historical records of the same ring number in the time series database.

[0210] Meanwhile, within the delayed batch processing channel, the system does not rush to calculate upon receiving a low-scoring record. Instead, based on the record's target ring number and acquisition timestamp, the system proactively initiates a wide-ranging search request to the underlying time-series database, retrieving all pre-stored historical records with adjacent ring numbers within several minutes before and after the record's occurrence. The system uses these historical records to construct a broad temporal context window, re-embedding isolated, single-point dirty data back into a continuous physical timeline.

[0211] Step 709: Perform cross-record consistency checks and status re-judgments on the received standardized records based on the context window, and generate batch processing replenishment records with batch replenishment identifiers.

[0212] Building upon this, the system activates a high-performance global analysis model in the delayed batch processing channel. By comparing the smooth evolution trend of parameters within the context window, the system performs cross-record consistency checks to identify whether the dirty data is a transient sensor out-of-order issue, transient noise, or a genuine operational condition switch. Using a global perspective, the system performs a secondary reassessment of the record's operational status and executes final repair / filling or removal operations based on the reassessment results. After repair, the system attaches a new batch replenishment identifier to the record, thereby generating a highly reliable batch replenishment record.

[0213] Step 710: Write the batch processing backfill records into the time series database, and use the batch backfill identifier to overwrite the records in the time series database that have the corresponding data identifier and the streaming first write identifier.

[0214] The system addresses the technical challenge of dual-version conflicts arising from data at the same physical moment in a time-series database. When writing batch processing backfill records, the system extracts the globally unique record identifier as the primary key for overwrite verification. The system's search engine searches the corresponding target ring number table for existing older records of the same origin. If an earlier version with a streaming first-write identifier is found, the system immediately uses the current record carrying the batch backfill identifier to physically overwrite or perform a version update operation. Through this closed-loop overwrite design, the system ensures both the rapid display requirements of the streaming channel at the front end and the global data consistency and final fidelity of the batch processing channel at the back end, achieving integrated streaming and batch processing collaboration in the field of data governance.

[0215] According to one aspect of this application, the original time-series data is parsed based on the field semantic mapping table to determine the standard field name, physical quantity category code, and measurement point role code corresponding to each data field, including:

[0216] In the above embodiments, the system employs a multi-table separated relational database structure to maintain rules for naming, type, and unit, etc. As an equivalent alternative to the above scheme, in some optional embodiments, the system can use an integrated semantic template library method to reconstruct the rule base. Specifically, the system constructs a global tree-structured hash structure in an in-memory or document-oriented database, using a standard field identifier as the unique primary key. This tree structure tightly encapsulates all field aliases, physical quantity categories, measurement point roles, equipment components, applicable operating conditions, and conversion coefficients corresponding to the same physical measurement point within a single independent data object.

[0217] When raw data flows into the system, the system extracts its original field names and performs a single deep traversal search within the global tree-structured hash table. Once a corresponding alias node is matched, the system can directly extract all the standard semantics and cleaning strategies encapsulated within the data object. Using an integrated semantic template library to replace join queries across multiple data tables effectively eliminates the state network overhead caused by frequent cross-table join operations in distributed stream processing engines, making it particularly suitable for large-scale tunnel engineering monitoring networks with extremely high concurrency throughput.

[0218] According to one aspect of this application, based on a triage condition including the comprehensive quality score and arrival delay, the cleaned standardized records are routed to a matching processing channel and written to a pre-configured time-series database with the target ring number as a reference.

[0219] Previous embodiments disclosed an optimal scheme for dynamic routing based on a multidimensional quality scoring formula. However, in conventional tunnel boring machine (TBM) projects where the computing power of some edge computing nodes is severely limited, or where there is a high tolerance for data physical consistency, the system provides a more lightweight alternative implementation method: the static source-driven routing method. Specifically, instead of extracting indicators one by one and calculating the quality score for each record, the system directly extracts the source marker of the data packet at the network communication layer for hard routing.

[0220] In this static source-driven traffic splitting method, as long as the system detects that a data packet originates from a real-time message queue middleware, it is assumed to have excellent timeliness, and the system will directly redirect it to the real-time stream processing channel to ensure the rapid refresh of the front-end monitoring dashboard. Conversely, if the system detects that the data packet is historical late data retrieved in batches from the local physical gateway cache via a scheduled task, it will directly package all of it and push it into the delayed batch processing channel. Although this method abandons the ability to perform micro-correction based on mechanical characteristics, its internal computing architecture is extremely streamlined, which can significantly reduce the memory resource consumption of the stream processing engine, making it an effective engineering compromise solution for balancing computing costs and basic business requirements.

[0221] According to one aspect of this application, the batch processing backfill records are written into the time series database, and the batch backfill identifier is used to overwrite the records in the time series database that have the corresponding data identifier and the streaming first-write identifier.

[0222] To ensure the absolute security and advanced retrieval performance of the aforementioned integrated batch and flow replenishment and coverage mechanism, this embodiment has undergone in-depth engineering optimization of the underlying data organization structure and index design of the time-series database. Specifically, the system physically partitions the time-series database according to the isolation principle of one project corresponding to one database space and one physical ring number corresponding to one data table. More importantly, the system configures an extremely rich multi-dimensional tag index system for each measurement point data table. In addition to including project identifier, device identifier, and target ring number, the tag index is also forcibly written with flow tracking feature attributes by the system.

[0223] Specifically, when data is first entered into the database through the real-time streaming channel, the system configures its tag with a write mode attribute of "streaming first write" and sets the boolean value tag for the backup status to false. After the delayed batch processing channel completes the status re-judgment and generates a more accurate record, the system uses the data's timestamp and unique auto-incrementing sequence code to accurately locate the corresponding old record in the time-series database. During version updates, the system not only overwrites the original values ​​with the new accurate values ​​but also simultaneously updates the record's write mode attribute to "batch backup" and flips the backup status boolean value tag to true. Through this multi-dimensional tag index design, downstream data miners can use the time-series database's native query language to instantly filter out all high-quality data that has undergone deep batch processing correction, or reverse-engineer sets of abnormal data fragments that have triggered stream-batch conflicts, thus providing a highly intuitive quantitative view for evaluating the communication and acquisition quality of the entire tunnel boring machine IoT system.

[0224] Furthermore, it should be noted that throughout the entire lifecycle of the aforementioned judgment, routing, overwriting, and database insertion operations, the system deeply integrates the underlying distributed snapshot mechanism of the stream processing computing engine. At set time intervals, the system periodically serializes the state information of internal operators that are halfway through processing and writes it to a persistent key-value pair storage backend. In the event of a physical power outage or a core computing node failure, the system can automatically and accurately restore the complete computing progress from the most recently verified snapshot file after restarting. This ensures that all batch processing recovery records are neither lost in a disaster nor repeatedly written to the time-series database after a restart, achieving end-to-end fault-tolerant semantics in the harsh and complex industrial construction site environment.

[0225] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for real-time processing and storage of shield tunneling construction time sequence data based on integrated batch processing, characterized in that, include: Obtain the original time-series data generated by the tunnel boring machine during construction, as well as the associated original ring numbers; Standardized records are obtained by performing standardization processing on the original time-series data based on a pre-built semantic rule base; Based on standardized records and judgment thresholds, the operating status of the tunnel boring machine is identified and the target ring number is determined; Based on the operational status, corresponding data cleaning strategies are executed on standardized records, and the cleaning results are evaluated to obtain a comprehensive quality score; Based on the triage conditions that include comprehensive quality scores and arrival delays, the cleaned and standardized records are routed to the matching processing channels and written to the pre-configured time series database with the target ring number as the reference.

2. The method according to claim 1, characterized in that, The pre-built semantic rule base includes a field semantic mapping table, a data type rule table, and a unit conversion rule table. The field semantic mapping table is configured with physical quantity category codes, measurement point role codes, and working condition tolerance strategies. Standardization processing is performed on the original time-series data based on a pre-built semantic rule base to obtain standardized records, including: The original time series data is parsed based on the field semantic mapping table to determine the standard field name, physical quantity category code and measurement point role code corresponding to each data field; Based on the data type rule table and unit conversion rule table, the parsed data fields are formatted and their units are standardized, and standardized records are output.

3. The method according to claim 1, characterized in that, Operating status includes tunneling status, shutdown status, maintenance status, and abnormal status; Based on standardized records and judgment thresholds, the operating status of the tunnel boring machine is identified, including: Key monitoring parameters are extracted from standardized records. These key monitoring parameters include at least propulsion force, cutterhead torque, and propulsion speed. Based on the comparison results between key monitoring parameters and judgment thresholds, an instantaneous status judgment result is generated. The consistency comparison between the instantaneous state determination result and the pre-stored state history sequence is calculated to obtain the state continuity confidence. When the confidence level of state continuity meets the preset conditions, the instantaneous state determination result is confirmed as the running state.

4. The method according to claim 3, characterized in that, The decision threshold is dynamically calculated through the following steps: Calculate the current mileage of the tunnel boring machine based on the target ring number, the pre-configured starting mileage of the tunnel, and the pre-configured design ring width; Based on the current mileage, query the corresponding representative stratigraphic strength parameters in the pre-constructed geological profile table; The judgment threshold is obtained by linearly scaling the pre-configured benchmark threshold based on the ratio of the representative formation strength parameter to the pre-configured reference formation strength value.

5. The method according to claim 1, characterized in that, Determine the target ring number, including: Extract the ring number detection physical monitoring parameters from the standardized records. The physical monitoring parameters include at least the propulsion force, propulsion speed, propulsion stroke, and grouting pressure. Based on the characteristics of sudden drop in propulsion force, zeroing of propulsion speed, matching characteristics of propulsion stroke and pre-configured design ring width, switching characteristics of operating state, and pulse characteristics of grouting pressure, a set of ring number boundary characteristic signals is constructed. Perform ring number boundary detection based on the ring number boundary feature signal set to obtain independently detected ring numbers; The independent detection ring number is compared with the original ring number, and the target ring number is determined based on the comparison result.

6. The method according to claim 5, characterized in that, Ring number boundary detection is performed based on the ring number boundary feature signal set to obtain independently detected ring numbers, including: The weighted voting value is calculated by weighting and summing each feature signal in the ring number boundary feature signal set with the pre-configured feature signal weights. When the weighted voting value is greater than or equal to the pre-configured voting threshold, a ring number boundary event is determined to have occurred, and the ring number counter is incremented based on the ring number boundary event to update the independent detection ring number; The independent detection ring number is compared with the original ring number, and the target ring number is determined based on the comparison result, including: When the independently detected ring number matches the original ring number, the original ring number is confirmed as the target ring number; When the independent detection ring number is inconsistent with the original ring number, the independent detection ring number is confirmed as the target ring number, and a ring number correction mark is generated.

7. The method according to claim 3, characterized in that, Based on the operational status, perform corresponding data cleaning strategies on the standardized records, including: When the running status is tunneling, strict range validation is performed on each data field in the standardized record, and outliers that exceed the legal range are marked. When the operating status is a shutdown state, perform a slow descent tolerance check on the propulsion force and cutterhead torque in the standardized records based on the pre-configured working condition tolerance strategy; When the running status is maintenance, add expected missing tags to the data fields that are missing in the standardized records; When the running status is abnormal, retain all the original field values ​​in the standardized record and attach the exception reason code.

8. The method according to claim 7, characterized in that, The cleaning results are evaluated to obtain a comprehensive quality score, including: When the operation is in the tunneling state, mechanical verification physical monitoring parameters are extracted from the cleaned standardized records. The physical monitoring parameters include at least the thrust, cutterhead torque, thrust speed, cutterhead speed and soil chamber pressure. Based on physical monitoring parameters and pre-configured physical constants of the tunnel boring machine, the penetration consistency deviation, specific energy consistency deviation and thrust balance deviation are calculated respectively. The physical consistency score is calculated based on the penetration consistency deviation, specific energy consistency deviation, and thrust balance deviation. A comprehensive quality score is constructed based on the physical consistency score and the field completeness rate of standardized records.

9. The method according to claim 8, characterized in that, Based on physical monitoring parameters and pre-configured physical constants of the tunnel boring machine, the penetration consistency deviation, specific energy consistency deviation, and thrust balance deviation are calculated respectively, including: Based on the kinematic ratio of the feed rate to the cutterhead rotation speed, the calculated penetration is obtained, and the consistency deviation between the calculated penetration and the pre-configured penetration within a reasonable range is determined. Power conversion is performed based on propulsion force, cutterhead torque, propulsion speed, cutterhead rotation speed and pre-configured excavation area to obtain calculated specific energy, and the consistency deviation between the calculated specific energy and the pre-configured specific energy within a reasonable range is determined. The theoretical thrust is calculated based on the earth pressure, the pre-configured shield friction coefficient, the pre-configured lateral earth pressure, and the excavation area, and the thrust balance deviation between the theoretical thrust and the actual thrust is determined.

Images