A construction environment monitoring system based on digital twins

Abstract

This invention discloses a construction environment monitoring system based on digital twins, belonging to the field of construction environment monitoring technology. After a unified calibration time is established, all environmental sampling records are reorganized onto the same digital twin timeline. The environmental benchmark and interval calculation unit can extract the environmental time series of construction procedures around the master clock and spatial center. This makes the environmental data no longer an independent curve detached from construction actions, but an analysis object bound to specific anchor point groups, specific work locations, and specific time periods. By calculating environmental benchmark values ​​and natural fluctuation bands from the environmental time series, the system forms its own environmental baseline and fluctuation range for each anchor point group and each type of construction procedure, eliminating the need to directly judge all scenarios using a single threshold. This structure addresses the deviations caused by the use of the same environmental judgment caliber for different procedures, spatial locations, and work stages in existing practices, making the interpretation of environmental changes closer to the corresponding construction actions themselves.

Description

Technical Field

[0001] This invention relates to the field of construction environment monitoring technology, specifically a construction environment monitoring system based on digital twins. Background Technology

[0002] In existing construction environment monitoring systems, multi-source data often suffers from upload delays, data re-encoding delays, batch data transmission, and time asynchrony across multiple channels. Existing systems often directly correlate various environmental and construction action data based on reception time, without adequately correcting for time offsets, spatial locations, and multi-channel synchronization characteristics. This results in inaccurate correlations between construction actions and environmental anomalies. Furthermore, for scenarios involving multiple work shifts or nighttime data re-encoding, the system struggles to distinguish between environmental anomalies caused by preceding actions and those triggered by subsequent actions, thus affecting anomaly attribution and construction action tracing.

[0003] The main reason for the above problems lies in the asynchronicity of multi-source data and the lack of unified time and space references: there are discrepancies between the time when cameras, sensors, and process inspection records occur on-site and the time when they are entered into the platform database or transmitted back to the system. Furthermore, the time chain information generated by different devices and systems has inconsistent formats, and there is a lack of unified standards for spatial linkage information. As a result, environmental anomalies cannot be accurately assigned to corresponding construction action anchor points, leading to problems such as confused anomaly attribution, reversed timing, and misjudgment of leading and trailing waves. This, in turn, affects the accuracy of construction environment anomaly analysis, construction action tracing, and safety and quality management decisions. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a construction environment monitoring system based on digital twins, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a construction environment detection system based on digital twin, including an anchor point event acquisition module, a twin master clock construction module, a channel time shift rollback module, a wake wave discrimination and anomaly attribution module, and a twin write-back and verification sorting module; The anchor point event acquisition module collects anchor point event data, environmental sampling data, time chain data, and spatial connection data at the construction site, and fits them into an anchor point record set A; The twin master clock construction module extracts candidate anchor events from the anchor record set A, analyzes the candidate anchor events, and obtains the anchor group number k, master clock tk, and spatial center xk; The channel time shift callback module reads the original time chain field from the multi-source raw data based on the master clock tk and spatial center xk, obtains the inherent time shift of each data channel, calculates and obtains the source-level time shift AEs and discrete bandwidth Ds, and outputs three types of labels; The wake wave discrimination and anomaly attribution module establishes environmental reference values ​​μ(k,c) and natural wave bands S(k,c) for different construction procedures based on source-level time shift AEs and master clock tk, and obtains the legal attribution interval and verifies the attribution results; The twin write-back and review sorting module, based on the legal attribution interval and review attribution results, writes back the anchor master clock, correction time, anomaly attribution, and upload tag to the digital twin timeline to obtain the review priority sequence PD, and manages the correspondence between construction environment anomalies and construction actions.

[0006] Preferably, the anchor event acquisition module includes a multi-source event access unit and a spatiotemporal normalization fitting unit; The multi-source event access unit accesses, identifies, and preliminarily organizes multi-source raw data related to the sequence of construction actions within the construction site, including anchor point event data, environmental sampling data, time chain data, and spatial connection data. Among them, the anchor point event data includes vehicle passing through the gate event, tower crane slewing trigger event, concrete pump start event, cutting machine power-on event, and spraying start event; Environmental sampling data includes dust, noise, temperature, and humidity; Time chain data includes the time of local device formation, the time of gateway release, the time of platform entry into the database, the time of manual inspection and observation, and the time of manual inspection submission; Spatial connection data includes work surface number, floor or section number, and twin unit number; Specifically, vehicle gate passage events originate from access control or gate logs; tower crane slewing trigger events originate from tower crane controller or equipment operation logs; concrete pump start events originate from pump truck control records or construction equipment status; cutting machine power-on events originate from power branch status or equipment power-on signals; and sprinkler start events originate from sprinkler controller start / stop records. Dust, noise, temperature, and humidity data are obtained from periodic sampling by environmental sensors. The timestamps of equipment local formation, gateway release, platform entry, manual inspection observation, and manual inspection submission originate from the corresponding equipment, edge gateway, platform database, and inspection terminal. The work surface number, floor or section number, and twin unit number originate from construction site location identifiers, BIM component area mapping, or digital twin spatial index. The spatiotemporal normalization fitting unit normalizes the acquired multi-source raw data and outputs an anchor point record set A. The normalization process is as follows: a unified field template is established for the multi-source raw data of the construction site; semantic merging is performed on the event names, status names and sampling fields of different systems; relative deviation processing is performed on dust, noise, temperature and humidity according to the historical percentile difference of the same data source and the same spatial unit; a unified reference time rule is established for the local formation time of equipment, gateway release time, platform entry time, manual inspection observation time and manual inspection submission time; and the work surface number, floor or section number is mapped to the unified twin unit number.

[0007] Preferably, the twin master clock construction module includes a candidate anchor point extraction unit and a spatiotemporal grouping clocking unit; The candidate anchor point extraction unit extracts event records representing the start of construction actions from the anchor point record set A, and organizes the extracted event records into a candidate anchor point event set with a unified structure. Each event record contains anchor event data, environmental sampling data, time chain data, and spatial attachment data; Extract event items that serve as action start indicators from the event logs, including vehicle gate passage events, tower crane slewing trigger events, concrete pump start events, cutting machine power-on events, and spray start events, and remove environmental sampling items, state continuation items, and duplicate auxiliary items; The extracted event items are uniformly encoded and organized into candidate anchor events E. All candidate anchor events E are then fitted to form a candidate anchor event set LE.

[0008] Preferably, the spatiotemporal grouping clock unit groups anchor points based on the temporal proximity and spatial adjacency relationships between candidate anchor point events in the candidate anchor point event set LE, generating multiple anchor point groups; The temporal distribution of candidate anchor events within each anchor group is analyzed to determine the master clock tk; the spatial distribution of candidate anchor events within each anchor group is analyzed to determine the spatial center xk. The anchor point group is obtained as follows: For any two candidate anchor events Eu and Ev, calculate the time interval between them |Tu-Tv| and compare it with the allowed grouping time interval of the corresponding event category combination: |Tu-Tv|≤δμ(λu,λv); In the formula, δμ(λu, λv) represents the allowed grouping time interval determined by the combination of candidate anchor event categories, Tu represents the event time of the u-th candidate anchor event, Tv represents the event time of the v-th candidate anchor event, λu represents the event category of the u-th candidate anchor event, λv represents the event category of the v-th candidate anchor event, and μ(λu, λv) represents the event category combination identifier. The spatial locations ξu and ξv corresponding to the candidate anchor point events are determined for adjacency based on the spatial topological relationship in the digital twin model. When the two are located on the same working surface, adjacent twin units, or the same continuous construction section, they are considered to be spatially adjacent. The spatial adjacency relationship is expressed as: adj(ξu, ξv) = 1; adj() represents the spatial adjacency discrimination function for the candidate anchor point events. When two candidate anchor events simultaneously satisfy both the temporal proximity condition and the spatial adjacency condition, they are grouped into the same anchor group; the anchor event set LE is gradually merged according to the above rules to form multiple anchor groups; The master clock tk is obtained by statistically analyzing the time distribution of candidate anchor events within each anchor group. Let the k-th anchor group contain mk candidate anchor events with event times t(k, 1), t(k, 2), ..., t(k, mk). Then the master clock tk of the k-th anchor group is defined as follows: tk=med(t(k,1),t(k,2),…,t(k,mk)); In the formula, med() represents the median operation, tk specifically represents the master clock of the k-th anchor group, and m represents the number of anchor groups; Using the median as the master clock ensures that the master clock position stably falls within the central region of the time distribution within the group, reducing local disturbances caused by upload latency, buffer backhaul, or trigger boundary offsets on a single channel. The resulting master clock represents a unified time reference for the corresponding construction actions on the digital twin timeline. The spatial center xk is obtained as follows: After the master clock is determined, the spatial distribution of candidate anchor events within the k-th anchor group is statistically analyzed. The spatial positions of the candidate anchor events within the anchor group are set as ξ(k, 1), ξ(k, 2), ..., ξ(k, mk). Then, the spatial center xk of the k-th anchor group is defined as: xk=mode(ξ(k,1),ξ(k,2),…,ξ(k,mk)); In the formula, mode() represents the mode operation, and xk specifically represents the spatial center of the k-th anchor point group.

[0009] Preferably, the channel time-shift callback module includes a time-shift feature calculation unit and a callback correction and tag output unit; The time shift feature calculation unit reads the original time chain field from the multi-source raw data, and aggregates the records within the scope of the same anchor point group based on the master clock tk and the spatial center xk, and calculates the time offset component of each record. The time offset component includes the gateway retention segment length gi of the i-th record, the inbound retention segment length ni of the i-th record, the backfill distance ri of the i-th record, and the clustering back transmission segment length ci of the i-th record; The gateway dwell time gi of the i-th record is obtained by the difference between the gateway release time and the device local creation time of the i-th record; it is used to characterize the dwell time of the record from the device local creation to the gateway release. The storage dwell time ni of the i-th record is obtained by the difference between the platform storage time and the gateway release time of the i-th record; it is used to characterize the dwell time of the record from the gateway release to the platform storage. The backfill distance ri of the i-th record is obtained by the difference between the manual inspection submission time and the manual inspection observation time of the i-th record; it is used to characterize the time difference between the observation time and the submission time in the manual inspection scenario. The group upload segment length ci of the i-th record is obtained by the following formula: The group upload segment length ci of the i-th record is used to characterize the time compression length caused by multiple records being uploaded in batches. ; In the formula, dbti represents the time when the i-th record is entered into the platform database, gwti represents the time when the i-th record is released through the gateway, and bi represents the number of records that arrive at the platform in the same batch as the i-th record. Based on the time offset component, the time offset distribution of multiple historical records within the same data source in the anchor record set A is statistically analyzed to obtain the source-level time offset AEs and discrete bandwidth Ds, as shown in the following formula: ; ; In the formula, medi represents the median operation of the i-th record, and Ωs represents the historical record sample set of the data source s.

[0010] Preferably, the callback correction and tag output unit calls the corresponding source-level time shift AEs according to the data source number si to which each record belongs, and performs a unified callback on the platform entry time dbti of the i-th record to obtain the unified correction time yti; Yti = dbti - AEsi; where yti specifically represents the uniform correction time of the i-th record, and AEsi represents the source-level time shift of the data source to which the i-th record belongs; After calculating the unified correction time, the total time shift length Li of a single record is constructed; the formula is as follows: Li = gi + ni + ri + ci; where Li specifically represents the total time shift length of the i-th record; The total time shift length Li is compared with the source-level time shift AEs and the discrete bandwidth Ds, and three types of labels are output. The process is as follows: When the total backlog of a single record exceeds the sum of the usual time shift and discrete bandwidth of the data source, i.e., Li > AEsi + Dsi, the record is marked as a late record. When the backfill distance is greater than zero, i.e., ri > 0, the record is marked as a supplementary record. When the number of records arriving at the platform in the same batch is greater than 1, i.e., bi > 0, the record is marked as a batch record tag.

[0011] Preferably, the wake wave discrimination and anomaly attribution module includes an environmental benchmark and interval calculation unit and an anomaly attribution determination unit; The environmental benchmark and interval calculation unit are based on the unified correction time yti, combined with the master clock tk and the spatial center xk, and for different construction procedures c and each anchor point group k, the environmental time series yc(t) is extracted based on the source-level time shift AEs. The steps for obtaining the environmental time series yc(t) are as follows: Within a fixed period, select all environmental sampling records belonging to the process from the environmental data output by the anchor event acquisition module, including indicators such as dust, noise, temperature, and humidity. Spatial location is also considered during the screening process; only environmental records located adjacent to the spatial center of the anchor group or within its twin unit are included in the series. The selected environmental records are sorted by time according to the unified correction time yti provided by the channel time shift callback module to form a continuous time series, ensuring that the series reflects the real time change sequence of the construction site. The sorted records are arranged continuously by time, and each time point corresponds to its environmental indicator value, forming an environmental time series yc(t). Based on the environmental time series yc(t), the environmental reference value μ(k,c) and the natural wave band S(k,c) are calculated and obtained, and the leading width Pc and the tail width Wc of the process are obtained. The formula for obtaining the environmental reference value μ(k, c) is: ; In the formula, Q0.25 represents the quartile operation, taking down the quartile, and t(k-1) represents the master clock of the (k-1)th anchor group; The formula for obtaining the natural wave band S(k,c) is: ; The formulas for obtaining the leader width Pc and the wake width Wc are: ; ; In the formula, Θc represents the set of anchor points for construction procedure c, min represents the start time of the leading zone, δ≥0 represents the time elapsed after the master clock, and δ represents the amount of time required for environmental indicators to fall back to the natural fluctuation range starting from the center time. By integrating the leader width Pc and the wake width Wc, the legal home interval Ik = [tk - Pc, tk + Wc] is obtained.

[0012] Preferably, the anomaly attribution determination unit is based on the natural fluctuation zone S(k,c), fits the time segment exceeding the natural fluctuation zone S(k,c) to obtain the environmental anomaly segment Eh, and records the start time thst, end time thed and anomaly spatial location ξh of the environmental anomaly segment Eh. For each environmental anomaly segment Eh, calculate the center time ηh using the formula: ηh = (thst + thed) / 2; For each environmental anomaly segment Eh, detect whether the detection center time ηh falls within the legal attribution interval Ik, and determine the spatial proximity between the spatial position ξh of the anomaly segment and the spatial center xk of the anchor point group. If both time and space conditions are met, then the environmental anomaly segment Eh is assigned to anchor point group k. If the center time ηh of the environmental anomaly segment Eh falls within the legal allocation interval of multiple anchor point groups, the most suitable anchor point group is selected based on the comprehensive distance of time and space. ; In the formula, K+ represents the final anchor group number to which the anomalous segment belongs, and ρ(ξh, xk) represents the distance or topological distance between the spatial location of the anomalous segment and the spatial center of the anchor group. If the environmental anomaly segment Eh carries a late record label, a supplementary record label, or a batch record label, and a unique attribution result cannot be formed, it is marked as pending attribution.

[0013] Preferably, the twin write-back and verification sorting module includes a write-back unit and a temporal clarity calculation and verification sorting unit; The write-back unit writes the anchor group number of the environmental anomaly segment Eh, the center time ηh of the environmental anomaly segment Eh, the unified correction time yti, the master clock tk, and the spatial center xk into the corresponding time axis and spatial unit of the digital twin model. Based on the anchor group number k, the master clock tk and spatial center xk are mapped to the center time ηh of each environmental anomaly segment Eh. The attribution result and label of the environmental anomaly segment Eh are attached to the corresponding time axis position and twin unit position in the digital twin model, and the source information and attribution anchor group number of each record are retained.

[0014] Preferably, the timing clarity calculation and review sorting unit receives the sorted anchor group's abnormal information and tags, and counts abnormal records for each anchor group k: including the number of abnormal records belonging to the current anchor group curNk, the number of abnormal records belonging to the wake wave taNk, the number of abnormal records with tags laNk, and the number of records with reversed order reNk. The temporal clarity Ck is calculated based on the anomaly records, using the following formula: ; In the formula, Ck specifically represents the timing clarity of anchor group k; the closer the timing clarity Ck is to 1, the higher the consistency between the anomaly of the anchor group and the master clock sequence; the lower it is, the more tail waves, delays or sequence conflicts there are. All anchor point groups are arranged in ascending order of temporal clarity to obtain the review priority order PD. The anchor point groups with low clarity are reviewed first. The environmental anomaly segment Eh information after review and confirmation, including the environmental anomaly attribution result, unified correction time, master clock tk and spatial center xk, as well as late record label, supplementary record label and batch record label, are written back to the corresponding time axis and spatial unit of the digital twin model. At the same time, the timing clarity of the corrected anchor point group is recalculated, and an updated review priority sequence is formed based on the timing clarity, so as to trace, verify and manage the correspondence between construction environment anomalies and construction actions; The formula for obtaining the review priority order PD is: PD = sortasc(C1, C2, ..., CKs); In the formula, CKs represents the total number of anchor groups, and sortasc represents the ascending sort operation.

[0015] This invention provides a construction environment monitoring system based on digital twins, which has the following advantages: (1) When the system is running, the channel time shift callback module collects the records within the range of the same anchor point group based on the master clock and the spatial center xk, and then calculates the gateway delay length, the warehouse delay length, the backfill distance and the batch back transmission length respectively, thereby splitting the originally general "data late arrival" into four sources: gateway delay, platform warehouse delay, manual supplementation and batch back transmission. This processing method addresses the problem of asynchronous arrival of multi-source data in the background but the reasons are mixed and unclear, so that the subsequent callback correction has a clearer basis.

[0016] (2) By constructing the total time shift length and comparing it with the source-level time shift and discrete bandwidth, late record labels, supplementary record labels and batch record labels are output respectively, which can separate and identify different arrival mechanisms that were originally mixed in the same time axis. For the situation in the existing system where it is difficult to distinguish between late reporting, manual supplementary recording and centralized cache return, this processing method makes each type of record have a clear category in the subsequent attribution analysis, which makes it easier to distinguish which records belong to normal link fluctuations and which records belong to special upload forms.

[0017] (3) After the unified correction time is formed, the environmental sampling records are reorganized onto the same digital twin time axis. The environmental benchmark and interval calculation unit can extract the environmental time series of the construction process around the master clock and spatial center. This makes the environmental data no longer an independent curve detached from the construction action, but an analysis object bound to a specific anchor group, specific work location, and specific time period. By calculating the environmental benchmark value and natural fluctuation band of the environmental time series, the system forms its own environmental baseline value and fluctuation range for each anchor group and each type of construction process, and no longer uses a single threshold to directly judge all scenarios. For the deviation caused by the use of the same environmental judgment caliber for different processes, different spatial locations, and different work stages in the existing practice, this structure makes the interpretation of environmental changes closer to the corresponding construction action itself.

[0018] (4) By writing back the anchor group number, center time, unified correction time, master clock and spatial center of the environmental anomaly segment into the time axis and spatial unit of the digital twin model, the construction action timeline, environmental anomaly timeline and data upload label are written back into the same twin interface.

[0019] By retaining the source information, anchor group number, and anomaly label of each record in the digital twin model, the correspondence between environmental anomalies and construction actions is no longer limited to simple alarms, but can be traced back to the specific data source, the specific construction action group, and the specific time period of the anomaly. This matches the inadequacy of explaining "who caused it, when it caused it, and where it was located" after the anomaly occurred in the background, making the review and management of construction environmental anomalies more organized. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the block flow of a construction environment monitoring system based on digital twins according to the present invention. Figure 2 This is a schematic diagram of the anchor point grouping judgment process of the present invention; Figure 3 This is a schematic diagram of the abnormal attribution legality judgment process of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Example 1 This invention provides a construction environment monitoring system based on digital twins. Please refer to [link / reference]. Figure 1It includes an anchor point event acquisition module, a twin master clock construction module, a channel time shift rollback module, a wake-wave discrimination and anomaly attribution module, and a twin write-back and verification sorting module; The anchor point event acquisition module collects anchor point event data, environmental sampling data, time chain data, and spatial connection data at the construction site, and fits them into an anchor point record set A; The twin master clock construction module extracts candidate anchor events from the anchor record set A, analyzes the candidate anchor events, and obtains the anchor group number k, master clock tk, and spatial center xk; The channel time shift callback module reads the original time chain field from the multi-source raw data based on the master clock tk and spatial center xk, obtains the inherent time shift of each data channel, calculates and obtains the source-level time shift AEs and discrete bandwidth Ds, and outputs three types of labels; The wake wave discrimination and anomaly attribution module establishes environmental reference values ​​μ(k,c) and natural wave bands S(k,c) for different construction procedures based on source-level time shift AEs and master clock tk, and obtains the legal attribution interval and verifies the attribution results; The twin write-back and review sorting module, based on the legal attribution interval and review attribution results, writes back the anchor master clock, correction time, anomaly attribution, and upload tag to the digital twin timeline to obtain the review priority sequence PD, and manages the correspondence between construction environment anomalies and construction actions.

[0023] For example, in the E3 work area on the east side of B1 of a two-story underground commercial complex, concrete pouring, rebar cutting, and vehicle transportation operations occur simultaneously from 20:00 to 22:00. Two dust sensors, one noise sensor, one set of road gate, one tower crane control recording point, and one set of pump start / stop recording points are deployed on site. The digital twin model divides this area into six twin units, DU-021 to DU-026. This system uses this work area as the anchor point for grouping and assigning environmental anomalies, and performs a linkage analysis on construction actions and environmental changes.

[0024] In this embodiment, the anchor event acquisition module and the channel time-shift rollback module are used to normalize and roll back construction action data, environmental sampling data, and process inspection data from different sources to a unified correction time, enabling various types of data to be compared within the same digital twin time axis and spatial unit. This solves the problem of inaccurate correspondence between construction actions and environmental anomalies caused by asynchronous multi-source data and upload delays in existing systems.

[0025] The twin master clock construction module extracts candidate anchor points and performs spatiotemporal grouping analysis to generate a master clock time and spatial center for each construction action, serving as a reference benchmark for attributing various environmental anomalies. This avoids misjudging the wake waves of preceding actions as anomalies in subsequent actions, improving the accuracy of anomaly attribution and the clarity of time sequence.

[0026] The channel time-shift callback module statistically analyzes the gateway delay, database delay, backfilling, and batch backhaul segment length of historical records for each data source, calculates the source-level time shift and discrete bandwidth, and labels each record as late, supplemented, or batch-recorded. This resolves the attribution confusion caused by multi-channel upload delays and supplemented records, providing a reliable foundation for subsequent wake wave identification.

[0027] The wake wave discrimination and anomaly attribution module combines environmental benchmark values ​​and natural wave bands to extract the leader width and wake wave width, generate a valid attribution interval, and accurately assign anomaly segments to anchor point groups based on temporal and spatial proximity. This effectively distinguishes between preceding wake waves and anomaly segments caused by subsequent construction actions, resolving the wake wave misjudgment problem in existing systems.

[0028] Through the collaboration of the aforementioned modules, this system can achieve multi-source data synchronization, accurate anomaly attribution, precise wake wave discrimination, clear and quantifiable time series analysis, and priority sequence generation for verification without increasing the amount of hardware acquisition. It solves problems such as asynchronous multi-source data, misjudgment of preceding wake waves, and the impact of supplementary / batch recording on attribution in existing construction monitoring systems, significantly improving the reliability of construction environment anomaly analysis and the level of construction action traceability management.

[0029] Example 2 Please refer to Figure 1 Specifically: the anchor event acquisition module includes a multi-source event access unit and a spatiotemporal normalization fitting unit; The multi-source event access unit accesses, identifies, and preliminarily organizes multi-source raw data related to the sequence of construction actions within the construction site, including anchor point event data, environmental sampling data, time chain data, and spatial connection data. Among them, the anchor point event data includes vehicle passing through the gate event, tower crane slewing trigger event, concrete pump start event, cutting machine power-on event, and spraying start event; Environmental sampling data includes dust, noise, temperature, and humidity; Time chain data includes the time of local device formation, the time of gateway release, the time of platform entry into the database, the time of manual inspection and observation, and the time of manual inspection submission; Spatial connection data includes work surface number, floor or section number, and twin unit number; The spatiotemporal normalization fitting unit normalizes the acquired multi-source raw data and outputs an anchor point record set A. The normalization process is as follows: a unified field template is established for the multi-source raw data of the construction site; semantic merging is performed on the event names, status names and sampling fields of different systems; relative deviation processing is performed on dust, noise, temperature and humidity according to the historical percentile difference of the same data source and the same spatial unit; a unified reference time rule is established for the local formation time of equipment, gateway release time, platform entry time, manual inspection observation time and manual inspection submission time; and the work surface number, floor or section number is mapped to the unified twin unit number.

[0030] The twin master clock construction module includes a candidate anchor point extraction unit and a spatiotemporal grouping clocking unit; The candidate anchor point extraction unit extracts event records representing the start of construction actions from the anchor point record set A, and organizes the extracted event records into a candidate anchor point event set with a unified structure. Each event record contains anchor event data, environmental sampling data, time chain data, and spatial attachment data; Extract event items that serve as action start indicators from the event logs, including vehicle gate passage events, tower crane slewing trigger events, concrete pump start events, cutting machine power-on events, and spray start events, and remove environmental sampling items, state continuation items, and duplicate auxiliary items; The extracted event items are uniformly encoded and organized into candidate anchor events E. All candidate anchor events E are then fitted to form a candidate anchor event set LE.

[0031] In this embodiment, the multi-source event access unit first performs unified access to anchor event data, environmental sampling data, time chain data, and spatially linked data. Then, the spatiotemporal normalization fitting unit completes field template unification, semantic merging, relative deviation processing, time rule unification, and spatial unit mapping, so that the heterogeneous data originally scattered in access control, equipment, environmental sensors, verification terminals, and digital twin models enters the same data framework. For the problem of data being difficult to directly correlate due to inconsistent event names, different sources of time fields, and inconsistent spatial numbering standards in the existing system, this structure can first organize the data into a set A of comparable, mergeable, and linkable anchor records, establishing a unified input foundation for subsequent analysis.

[0032] By using the candidate anchor point extraction unit to filter out event items that serve as the starting identifier of construction actions from the anchor point record set A, and removing environmental sampling items, state continuation items, and duplicate ancillary items, it is possible to distinguish between "records that truly represent the starting point of the action" and "accompanying records generated around the action". In response to the problem in the existing system where environmental sampling fluctuations, state continuation signals and action trigger signals are mixed together, making it difficult to identify the action boundary, this processing method makes the identification of the start of construction actions more focused and reduces the situation where irrelevant records are merged into the same action chain during subsequent grouping.

[0033] By uniformly encoding and fitting candidate anchor point events through spatiotemporal grouping clock units, and further forming a candidate anchor point event set LE, multiple record traces left by the same construction action in different sensing channels can be converged into the same analysis object. This corresponds to the problem of "asynchronous multi-source data and inconsistent time and position references in different systems" mentioned in the background section, so that the subsequently generated anchor point group number, master clock and spatial center have a unified reference meaning, which makes it easier to put construction actions and environmental changes into the same time axis and the same spatial unit for corresponding analysis.

[0034] Because the environmental sampling data is processed using the historical quantile difference of the same data source and the same spatial unit for relative deviation, dust, noise, temperature and humidity of different dimensions and fluctuation scales are no longer directly compared by mixing the original values, but are transformed into deviation expressions around their respective spatial norms. For situations where various environmental quantities are difficult to place in the same analysis link due to different units and numerical ranges in existing monitoring methods, this processing method makes the relationship between environmental fluctuations and construction actions easier to identify and compare in the future.

[0035] Because the time chain data retains the device local formation time, gateway release time, platform entry time, manual inspection observation time, and manual inspection submission time during the access phase, and establishes a unified reference time rule during the normalization phase, subsequent modules can continue to process the differences between the "on-site occurrence time" and the "platform reception time". This directly corresponds to the upload delay, supplementary recording delay, and multi-channel asynchrony issues mentioned in the background, enabling the system to no longer simply understand the sequence of construction actions according to the order of platform reception, but to organize data around the actual action chain.

[0036] Because spatially attached data is mapped to a unified twin unit number during the normalization stage, and candidate anchor point events will be further formed into a master clock and spatial center in the future, construction actions are no longer just abstract points in time, but spatiotemporal events with clear spatial landing points. For situations in the existing system where "time can be seen but location cannot be attached" or "multiple sets of numbers coexist on the same work surface", this structure provides a clear spatial basis for subsequent construction action tracing, environmental anomaly location, and digital twin write-back.

[0037] Example 3 Please refer to Figure 2 Specifically: the spatiotemporal grouping clock unit groups anchor points based on the temporal proximity and spatial adjacency between each candidate anchor point event in the candidate anchor point event set LE, generating multiple anchor point groups; The temporal distribution of candidate anchor events within each anchor group is analyzed to determine the master clock tk; the spatial distribution of candidate anchor events within each anchor group is analyzed to determine the spatial center xk. The master clock tk is obtained by statistically analyzing the time distribution of candidate anchor events within each anchor group. Let the k-th anchor group contain mk candidate anchor events with event times t(k, 1), t(k, 2), ..., t(k, mk). Then the master clock tk of the k-th anchor group is defined as follows: tk=med(t(k,1),t(k,2),…,t(k,mk)); In the formula, med() represents the median operation, tk specifically represents the master clock of the k-th anchor group, and m represents the number of anchor groups; The spatial center xk is obtained by statistically analyzing the spatial distribution of candidate anchor point events within the k-th anchor point group after the master clock is determined. The spatial location ξ(k,j) is represented by a ternary location code of “floor number - section number - twin unit number” and is bound to the three-dimensional coordinates (X, Y, Z) in the digital twin model; among them, the spatial location of the vehicle passing through the gate event is obtained from the entrance and exit coordinates bound to the gate controller, the spatial location of the tower crane slewing trigger event is calculated from the coordinates of the tower crane slewing center and the landing point of the boom projection, the spatial location of the concrete pumping start event is obtained from the coordinates of the pouring zone where the pump pipe end is located, the location of the cutting machine power-on event is obtained from the working surface positioning code bound to the equipment distribution box, and the location of the sprinkler start event is obtained from the sprinkler zone coordinates bound to the sprinkler branch control point; Let the spatial positions of candidate anchor events within an anchor group be ξ(k, 1), ξ(k, 2), ..., ξ(k, mk), then the spatial center xk of the k-th anchor group is defined as: xk=mode(ξ(k,1),ξ(k,2),…,ξ(k,mk)); In the formula, mode() represents the mode operation, and xk specifically represents the spatial center of the k-th anchor point group.

[0038] In a specific embodiment, the spatial location ξ(k,j) is represented by a ternary location code consisting of “floor number - section number - twin unit number”, and is bound one-to-one with the three-dimensional coordinates (X, Y, Z) in the digital twin model.

[0039] For example, the pumping landing point of the E3 working face on the east side of basement B1 is recorded as [B1, E3, DU-021], with corresponding coordinates of (36.5m, 18.0m, -5.2m); the location of the vehicle passing through the gate event is obtained from the entrance and exit coordinates bound to the gate controller; the location of the tower crane slewing trigger event is obtained by calculating the coordinates of the tower crane slewing center and the landing point of the boom projection; the location of the concrete pumping start event is obtained from the coordinates of the pouring zone where the pump pipe ends; the location of the cutting machine power-on event is obtained from the working face positioning code bound to the equipment distribution box; and the location of the sprinkler start event is obtained from the sprinkler zone coordinates bound to the sprinkler branch control point. The channel time-shift callback module includes a time-shift feature calculation unit and a callback correction and tag output unit; The time shift feature calculation unit reads the original time chain field from the multi-source raw data, and aggregates the records within the scope of the same anchor point group based on the master clock tk and the spatial center xk, and calculates the time offset component of each record. The time offset component includes the gateway retention segment length gi of the i-th record, the inbound retention segment length ni of the i-th record, the backfill distance ri of the i-th record, and the clustering back transmission segment length ci of the i-th record; The gateway demurrage length gi of the i-th record is obtained by the difference between the gateway release time of the i-th record and the local device formation time. The length ni of the inbound storage period of the i-th record is obtained by the difference between the platform inbound time and the gateway release time of the i-th record; The backfill distance ri of the i-th record is obtained by the difference between the manual inspection submission time and the manual inspection observation time of the i-th record; The group return segment length ci of the i-th record is obtained using the following formula: ; In the formula, dbti represents the time when the i-th record is entered into the platform database, gwti represents the time when the i-th record is released through the gateway, and bi represents the number of records that arrive at the platform in the same batch as the i-th record. Based on the time offset component, the time offset distribution of multiple historical records within the same data source in the anchor record set A is statistically analyzed to obtain the source-level time offset AEs and discrete bandwidth Ds, as shown in the following formula: ; ; In the formula, medi represents the median operation of the i-th record, and Ωs represents the historical record sample set of the data source s.

[0040] For example, in the aforementioned work area E3 on the east side of B1, a vehicle passing through the gate event was collected at 20:13:08, with the location [B1, E3, DU-022]; a concrete pump start event was collected at 20:13:26, with the location [B1, E3, DU-021]; and a tower crane slewing trigger event was collected at 20:13:31, with the location [B1, E3, DU-021]. These three events satisfy the conditions of temporal proximity and spatial adjacency, and are grouped into the same anchor point group k=12, with its master clock t12 set to 20:13:26 and its spatial center x12 set to DU-021. In this embodiment, the events in the candidate anchor point event set LE are grouped according to temporal proximity and spatial adjacency by the spatiotemporal grouping clock unit. This allows multiple event traces formed by the same construction action in different sensing channels to be merged into the same anchor point group, avoiding the mixing of events that are close in time but separate in space into the same action chain, and also avoiding the erroneous merging of spatially adjacent events with large time spans. This transforms the construction action recognition from single-point recording judgment to spatiotemporal joint judgment, correspondingly solving the problems of scattered multi-source data and blurred action boundaries in the background.

[0041] By statistically analyzing the temporal distribution of candidate anchor point events within each anchor point group and using the median method to determine the master clock tk, the master clock no longer depends on the upload time of a single device or channel, but reflects the central time position of the construction actions in that group. When there are situations on site such as inconsistent device upload speeds, gateway cache release, and platform data entry delays, this approach helps to reduce the interference of local abnormal moments on the reference time of construction actions, making the time correspondence between subsequent environmental anomalies and construction actions clearer.

[0042] By statistically analyzing the spatial distribution of candidate anchor point events within each anchor point group and determining the spatial center xk using the mode method, construction actions have a clear spatial affiliation unit in the digital twin model. This spatial center can serve as a unified landing point for subsequent environmental data collection and anomaly attribution, addressing the issues of inconsistent spatial numbering of different systems in the background and the difficulty in linking the work surface with environmental monitoring points, making it easier to trace the spatial relationship between construction actions and environmental changes.

[0043] Based on the master clock and spatial center xk, the channel time-shift callback module aggregates records within the same anchor point group's range and then calculates the gateway delay length, warehouse entry delay length, backfill distance, and batch transmission length. This breaks down the previously vague "late data arrival" into four sources: gateway delay, platform warehouse entry delay, manual supplementation, and batch transmission. This approach addresses the issue of asynchronous arrival of multiple data sources with unclear and mixed causes, providing a clearer basis for subsequent callback correction.

[0044] By statistically analyzing the time offset distribution of multiple historical records within the same data source, the source-level time shift and discrete bandwidth Ds are obtained. This ensures that each data source has a time shift reference value and fluctuation range corresponding to its own upload characteristics, eliminating the need to process all channels with a uniform fixed time difference. For different data sources such as cameras, sensors, inspection terminals, and controllers, this method can process them separately according to their respective link characteristics, thus solving the deviation problem caused by the asynchronous processing of multiple channel time chains in the background.

[0045] By combining source-level time shift and discrete bandwidth Ds, the system can distinguish between late records, supplementary records, and batch records in subsequent modules, so that the platform's entry time is no longer directly equivalent to the on-site occurrence time. For scenarios such as nighttime supplementary recording, link fluctuations, and centralized cache uploads, this processing method helps to prevent supplementary information from being mistaken for real-time construction actions and batch transmissions from being mistaken for anomalies occurring at the same time, thereby mitigating the situation in the background where "the tail wave of the previous process is counted as an anomaly caused by the next process".

[0046] In this embodiment, the spatiotemporal grouping clock unit first forms the anchor point group, master clock, and spatial center. Then, the channel time shift callback module extracts the inherent time shift of each channel and constructs the source-level time shift and discrete bandwidth, so that the three parts of construction action identification, time correction, and spatial connection form a continuous processing framework. This corresponds to the current situation of "multi-source asynchronous, inconsistent time, inconsistent spatial reference, and chaotic anomaly attribution" in the background, so that the anomalies in the construction environment have clearer time and spatial references in the subsequent wake wave discrimination, anomaly attribution, and digital twin write-back process, which facilitates the tracing and verification of the correspondence between construction environment anomalies and construction actions.

[0047] Example 4 Please refer to Figure 3 Specifically: The callback correction and tag output unit calls the corresponding source-level time shift AEs according to the data source number si to which each record belongs, and performs a unified callback on the platform entry time dbti of the i-th record to obtain the unified correction time yti. Yti = dbti - AEsi; where yti specifically represents the uniform correction time of the i-th record, and AEsi represents the source-level time shift of the data source to which the i-th record belongs; After calculating the unified correction time, the total time shift length Li of a single record is constructed; the formula is as follows: Li = gi + ni + ri + ci; where Li specifically represents the total time shift length of the i-th record; The total time shift length Li is compared with the source-level time shift AEs and the discrete bandwidth Ds, and three types of labels are output. The process is as follows: When the total backlog of a single record exceeds the sum of the usual time shift and discrete bandwidth of the data source, i.e., Li > AEsi + Dsi, the record is marked as a late record. When the backfill distance is greater than zero, i.e., ri > 0, the record is marked as a supplementary record. When the number of records arriving at the platform in the same batch is greater than 1, i.e., bi > 0, the records are marked as batch record tags.

[0048] The wake wave discrimination and anomaly attribution module includes an environmental baseline and interval calculation unit and an anomaly attribution determination unit; The environmental benchmark and interval calculation unit are based on the unified correction time yti, combined with the master clock tk and the spatial center xk, and for different construction procedures c and each anchor point group k, the environmental time series yc(t) is extracted based on the source-level time shift AEs. Based on the environmental time series yc(t), the environmental baseline value μ(k,c) and the natural fluctuation band S(k,c) are calculated and obtained, and the leading width Pc and tail width Wc of the process are obtained. The formula for obtaining the environmental reference value μ(k, c) is: ; In the formula, Q0.25 represents the quartile operation, taking down the quartile, and t(k-1) represents the master clock of the (k-1)th anchor group; The formula for obtaining the natural wave band S(k,c) is: ; The formulas for obtaining the leader width Pc and the wake width Wc are: ; ; In the formula, Θc represents the set of anchor points for construction procedure c, min represents the start time of the leading zone, δ≥0 represents the time elapsed after the master clock, and δ represents the amount of time required for environmental indicators to fall back to the natural fluctuation range starting from the center time. By integrating the leader width Pc and the wake width Wc, the legal home interval Ik = [tk - Pc, tk + Wc] is obtained.

[0049] For example, in the pouring operation corresponding to anchor point group k=12, the local device formation time of dust sensor S03 is 20:13:42, the gateway release time is 20:13:47, and the platform entry time is 20:13:59. Then gi=5s, ni=12s, and bi=1, ri=0, so Li=17s. According to the statistics of 480 historical records of the dust data source in the past 5 days, AEs=24s, Ds=8s, and the unified correction time after the record is returned is 20:13:35.

[0050] In the same scenario, the observation time of the manual inspection record is 20:14:00 and the submission time is 20:26:00, so ri=720s and it is marked as a supplementary record; the number of video records uploaded in the same batch is bi=6, and the time difference between the gateway and the platform is 18s, so ci=(6-1)×18 / 6=15s. When Li=41s is greater than the sum of AESi and Dsi, which is 32s, the record is marked as a late record and a batch record.

[0051] Taking the cutting process as an example, within the time period [t11, t12) corresponding to anchor point group k=12, the dust time series yc(t) is calculated to obtain μ(12, cutting)=0.18mg / m³, S(12, cutting)=0.05mg / m³, Pc=42s, Wc=185s. Therefore, the legal assignment interval I12=[20:12:44, 20:16:31]; In this embodiment, by calling the corresponding source-level time shift according to the data source number, the platform's entry time is uniformly rolled back to form a unified correction time. This ensures that video, access control, environmental sensors, inspection terminals, and equipment control records no longer directly use the order of platform reception as the basis for the action sequence, but instead return to a position closer to the order of occurrence on site.

[0052] By constructing the total time shift length and comparing it with the source-level time shift and discrete bandwidth, late record tags, supplementary record tags, and batch record tags are output respectively. This allows for the separation and identification of different arrival mechanisms that were originally mixed in the same time axis. For situations where it is difficult to distinguish between late reporting, manual supplementary recording, and centralized cached backhaul in the existing system, this processing method gives each type of record a clear category in subsequent attribution analysis, making it easier to distinguish which records belong to normal link fluctuations and which records belong to special upload forms.

[0053] The formation of the legal attribution range means that subsequent anomaly attribution no longer depends solely on the instantaneous high value at a certain moment, but is analyzed based on the entire link of "unified time after callback - process baseline fluctuation - leading and tail wave intervals". For scenarios such as multi-shift cross-operation, continuous transportation, nighttime supplementary recording, and edge gateway fluctuations, this structure can view "anomalies caused by actual actions" and "apparent anomalies caused by the link mechanism" in a hierarchical manner, making the basis for attributing construction environment anomalies more complete.

[0054] This embodiment first uses source-level time shift to perform unified rollback and tag splitting on multi-source records, then extracts the environmental time series based on the unified correction time, and then calculates the environmental baseline, natural fluctuation band, leading width and tail width around the master clock, finally forming a legal attribution interval. This link puts several core deficiencies in the background part, namely, inconsistent time, mixed upload formats, difficulty in corresponding environmental fluctuations and action boundaries, and confusion between tail waves and new anomalies, into the same processing framework, so that the construction action timeline and environmental fluctuation timeline in the digital twin model can be compared with each other.

[0055] Example 5 Please refer to Figure 2 Specifically: The anomaly attribution determination unit is based on the natural fluctuation zone S(k,c). It fits the time segments that exceed the natural fluctuation zone S(k,c) to obtain the environmental anomaly segment Eh, and records the start time thst, end time thed, and anomaly spatial location ξh of the environmental anomaly segment Eh. For each environmental anomaly segment Eh, calculate the center time ηh using the formula: ηh = (thst + thed) / 2; For each environmental anomaly segment Eh, detect whether the detection center time ηh falls within the legal attribution interval Ik, and determine the spatial proximity between the spatial position ξh of the anomaly segment and the spatial center xk of the anchor point group. If both time and space conditions are met, then the environmental anomaly segment Eh is assigned to anchor point group k. If the center time ηh of the environmental anomaly segment Eh falls within the legal allocation interval of multiple anchor point groups, the most suitable anchor point group is selected based on the comprehensive distance of time and space. ; In the formula, K+ represents the final anchor group number to which the anomalous segment belongs, and ρ(ξh, xk) represents the distance or topological distance between the spatial location of the anomalous segment and the spatial center of the anchor group. If the environmental anomaly segment Eh carries a late record label, a supplementary record label, or a batch record label, and a unique attribution result cannot be formed, it is marked as pending attribution.

[0056] The twin write-back and review sorting module includes a write-back unit and a timing clarity calculation and review sorting unit; The write-back unit writes the anchor group number of the environmental anomaly segment Eh, the center time ηh of the environmental anomaly segment Eh, the unified correction time yti, the master clock tk, and the spatial center xk into the corresponding time axis and spatial unit of the digital twin model. Based on the anchor group number k, the master clock tk and spatial center xk are mapped to the center time ηh of each environmental anomaly segment Eh. The attribution result and label of the environmental anomaly segment Eh are attached to the corresponding time axis position and twin unit position in the digital twin model, and the source information and attribution anchor group number of each record are retained.

[0057] For example, in the E3 working face corresponding to anchor point group k=12, the start time of the continuous dust abnormality segment Eh1 is 20:13:40, the end time is 20:14:50, and the abnormal spatial location is DU-021. Then its center time ηh1=20:14:15. Since ηh1 falls into I12 and the abnormal spatial location is consistent with x12, Eh1 is assigned to anchor point group 12.

[0058] Another abnormal segment, Eh2, has a start time of 20:16:05, an end time of 20:17:20, and a center time of 20:16:42, falling into both I12 and I13. After normalizing the time and spatial distances, the comprehensive distance of anchor group 12 is 109, and the comprehensive distance of anchor group 13 is 46. Therefore, Eh2 is assigned to anchor group 13. If Eh2 also has a supplementary recording tag and its spatial location is missing, it is first marked as pending review and assignment. The timing clarity calculation and review sorting unit receives the sorted anchor group's abnormal information and tags, and counts the abnormal records for each anchor group k: including the number of abnormal records belonging to the current anchor group curNk, the number of abnormal records belonging to the wake wave taNk, the number of abnormal records with tags laNk, and the number of records with reversed order reNk. The temporal clarity Ck is calculated based on the anomaly records, using the following formula: ; In the formula, Ck specifically represents the temporal clarity of anchor group k; All anchor point groups are arranged in ascending order of temporal clarity to obtain the review priority order PD. The anchor point groups with lower clarity are reviewed first. The environmental anomaly segment Eh information after review and confirmation is written back to the corresponding time axis and spatial unit of the digital twin model. At the same time, the timing clarity of the corrected anchor point group is recalculated, and an updated review priority sequence is formed based on the timing clarity, so as to trace, verify and manage the correspondence between construction environment anomalies and construction actions; The formula for obtaining the review priority order PD is: PD = sortasc(C1, C2, ..., CKs); In the formula, CKs represents the total number of anchor groups, and sortasc represents the ascending sort operation.

[0059] Before the review, the anchor group 12 was statistically analyzed and found to have curN12=6, taN12=1, laN12=1, and reN12=0. Therefore, C12=6 / (6+1+1+0)=0.75. The anchor group 13 was statistically analyzed and found to have C13=0.43. The anchor group 14 was statistically analyzed and found to have C14=0.86. Therefore, the review priority order PD=[13, 12, 14]. First, the nighttime cutting and vehicle cross-operation corresponding to the anchor group 13 was manually reviewed, and then the correction results were written back to the digital twin time axis. In this embodiment, by using natural fluctuation bands as a basis, time segments that continuously exceed the natural fluctuation bands are fitted as environmental anomaly segments, and their start time, end time, and anomaly spatial location are recorded. This transforms environmental anomalies from scattered single-point exceedance records into anomaly objects with duration and spatial location. This corresponds to the problem mentioned in the background section that existing systems can only see discrete fluctuations and are difficult to reconstruct complete anomaly processes, allowing subsequent anomaly attribution to be based on continuous anomaly behavior rather than isolated sampling points.

[0060] By calculating the center time for each environmental anomaly segment and combining it with the legal attribution interval and spatial center for attribution judgment, the correspondence between environmental anomaly segments and construction actions has a unified temporal and spatial basis for judgment. In the context of inconsistent time and unclear action boundaries of multi-source data, this method can place the anomaly segment into the spatiotemporal framework of a specific anchor point group for analysis, reducing the mismatch caused by simply attaching based on the platform's receiving time.

[0061] When the center moment of an environmental anomaly segment falls within the legal assignment interval of multiple anchor point groups simultaneously, the most suitable anchor point group number K+ is selected by a comprehensive judgment method of time distance and spatial distance, so that the anomaly assignment no longer depends on a single time condition or a single spatial proximity condition.

[0062] When an abnormal environmental segment carries a late record tag, a supplementary record tag, or a batch record tag, and a unique attribution result cannot be formed, it is marked as pending attribution. This avoids directly attributing apparent anomalies caused by upload link anomalies, manual supplementary records, or centralized backhauls to a certain construction action. This approach addresses the deficiency in the background where "late data, supplementary data, and batch backhaul data are mistakenly regarded as real-time action results," allowing the system to retain a review entry point in addition to automatic attribution, facilitating subsequent verification.

[0063] By writing back the unit, the anchor point group number, center time, unified correction time, master clock and spatial center of the environmental anomaly segment are uniformly written into the time axis and spatial unit of the digital twin model, so that the construction action timeline, environmental anomaly timeline and data upload label form a corresponding relationship in the same twin interface.

[0064] By retaining the source information, anchor group number, and anomaly label of each record in the digital twin model, the correspondence between environmental anomalies and construction actions is no longer limited to simple alarms, but can be traced back to the specific data source, the specific construction action group, and the specific time period of the anomaly. This matches the inadequacy of explaining "who caused it, when it caused it, and where it was located" after the anomaly occurred in the background, making the review and management of construction environmental anomalies more organized.

[0065] By sorting all anchor point groups according to their temporal clarity from smallest to largest, a review priority order is obtained. The system can prioritize handling anchor point groups with disordered temporal order, more tail interference, and concentrated tagged records, rather than processing all anomalies equally. By rewriting the digital twin model of the environmental anomaly segments after review and confirmation, and recalculating the temporal clarity of the corrected anchor point groups, an updated review priority sequence is formed. The system forms a continuous closed loop of "anomaly segment identification - attribution judgment - twin rewriting - review sorting - correction rewriting". Corresponding to the problem in the background that the existing system lacks the ability to conduct subsequent verification and iterative correction after anomaly attribution, this structure enables the correspondence between construction environment anomalies and construction actions to continue to converge after review.

[0066] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended technical solutions and their equivalents.

Claims

1. A construction environment monitoring system based on digital twins, characterized in that: It includes an anchor point event acquisition module, a twin master clock construction module, a channel time shift rollback module, a wake wave discrimination and anomaly attribution module, and a twin write-back and verification sorting module; The anchor point event acquisition module collects anchor point event data, environmental sampling data, time chain data, and spatial connection data at the construction site, and fits them into an anchor point record set A; The twin master clock construction module extracts candidate anchor events from the anchor record set A, analyzes the candidate anchor events, and obtains the anchor group number k, master clock tk, and spatial center xk; The channel time shift callback module reads the original time chain field from the multi-source raw data based on the master clock tk and spatial center xk, obtains the inherent time shift of each data channel, calculates and obtains the source-level time shift AEs and discrete bandwidth Ds, and outputs three types of labels; The wake wave discrimination and anomaly attribution module establishes environmental reference values ​​μ(k,c) and natural wave bands S(k,c) for different construction procedures based on source-level time shift AEs and master clock tk, and obtains the legal attribution interval and verifies the attribution results; The twin write-back and review sorting module, based on the legal attribution interval and review attribution results, writes back the anchor master clock, correction time, anomaly attribution, and upload tag to the digital twin timeline to obtain the review priority sequence PD, and manages the correspondence between construction environment anomalies and construction actions.

2. The construction environment monitoring system based on digital twin according to claim 1, characterized in that: The anchor point event acquisition module includes a multi-source event access unit and a spatiotemporal normalization fitting unit; The multi-source event access unit accesses, identifies, and preliminarily organizes multi-source raw data related to the sequence of construction actions within the construction site, including anchor point event data, environmental sampling data, time chain data, and spatial connection data. Among them, the anchor point event data includes vehicle passing through the gate event, tower crane slewing trigger event, concrete pump start event, cutting machine power-on event, and spraying start event; Environmental sampling data includes dust, noise, temperature, and humidity; Time chain data includes the time of local device formation, the time of gateway release, the time of platform entry into the database, the time of manual inspection and observation, and the time of manual inspection submission; Spatial connection data includes work surface number, floor or section number, and twin unit number; The spatiotemporal normalization fitting unit normalizes the acquired multi-source raw data and outputs an anchor point record set A. The normalization process is as follows: a unified field template is established for the multi-source raw data of the construction site; semantic merging is performed on the event names, status names and sampling fields of different systems; relative deviation processing is performed on dust, noise, temperature and humidity according to the historical percentile difference of the same data source and the same spatial unit; a unified reference time rule is established for the local formation time of equipment, gateway release time, platform entry time, manual inspection observation time and manual inspection submission time; and the work surface number, floor or section number is mapped to the unified twin unit number.

3. The construction environment monitoring system based on digital twin according to claim 2, characterized in that: The twin master clock construction module includes a candidate anchor point extraction unit and a spatiotemporal grouping clocking unit; The candidate anchor point extraction unit extracts event records representing the start of construction actions from the anchor point record set A, and organizes the extracted event records into a candidate anchor point event set with a unified structure. Each event record contains anchor event data, environmental sampling data, time chain data, and spatial attachment data; Extract event items that serve as action start indicators from the event logs, including vehicle gate passage events, tower crane slewing trigger events, concrete pump start events, cutting machine power-on events, and spray start events, and remove environmental sampling items, state continuation items, and duplicate auxiliary items; The extracted event items are uniformly encoded and organized into candidate anchor events E. All candidate anchor events E are then fitted to form a candidate anchor event set LE.

4. The construction environment monitoring system based on digital twin according to claim 3, characterized in that: The spatiotemporal grouping clock unit groups anchor points based on the temporal proximity and spatial adjacency relationships between candidate anchor point events in the candidate anchor point event set LE, generating multiple anchor point groups; The temporal distribution of candidate anchor events within each anchor group is analyzed to determine the master clock tk. Analyze the spatial distribution of candidate anchor point events within each anchor point group to determine the spatial center xk; The master clock tk is obtained by statistically analyzing the time distribution of candidate anchor events within each anchor group. Let the k-th anchor group contain mk candidate anchor events with event times t(k, 1), t(k, 2), ..., t(k, mk). Then the master clock tk of the k-th anchor group is defined as follows: tk=med(t(k,1),t(k,2),…,t(k,mk)); In the formula, med() represents the median operation, tk specifically represents the master clock of the k-th anchor group, and m represents the number of anchor groups; The spatial center xk is obtained as follows: After the master clock is determined, the spatial distribution of candidate anchor events within the k-th anchor group is statistically analyzed. The spatial positions of the candidate anchor events within the anchor group are set as ξ(k, 1), ξ(k, 2), ..., ξ(k, mk). Then, the spatial center xk of the k-th anchor group is defined as: xk=mode(ξ(k,1),ξ(k,2),…,ξ(k,mk)); In the formula, mode() represents the mode operation, and xk specifically represents the spatial center of the k-th anchor point group.

5. A construction environment monitoring system based on digital twins according to claim 4, characterized in that: The channel time-shift callback module includes a time-shift feature calculation unit and a callback correction and tag output unit; The time shift feature calculation unit reads the original time chain field from the multi-source raw data, and aggregates the records within the scope of the same anchor point group based on the master clock tk and the spatial center xk, and calculates the time offset component of each record. The time offset component includes the gateway retention segment length gi of the i-th record, the inbound retention segment length ni of the i-th record, the backfill distance ri of the i-th record, and the clustering back transmission segment length ci of the i-th record; The gateway demurrage length gi of the i-th record is obtained by the difference between the gateway release time of the i-th record and the local device formation time. The length ni of the inbound storage period of the i-th record is obtained by the difference between the platform inbound time and the gateway release time of the i-th record; The backfill distance ri of the i-th record is obtained by the difference between the manual inspection submission time and the manual inspection observation time of the i-th record; The group return segment length ci of the i-th record is obtained using the following formula: ； In the formula, dbti represents the time when the i-th record is entered into the platform database, gwti represents the time when the i-th record is released through the gateway, and bi represents the number of records that arrive at the platform in the same batch as the i-th record. Based on the time offset component, the time offset distribution of multiple historical records within the same data source in the anchor record set A is statistically analyzed to obtain the source-level time offset AEs and discrete bandwidth Ds, as shown in the following formula: ； ； In the formula, medi represents the median operation of the i-th record, and Ωs represents the historical record sample set of the data source s.

6. A construction environment monitoring system based on digital twins according to claim 5, characterized in that: The callback correction and tag output unit calls the corresponding source-level time shift AEs according to the data source number si to which each record belongs, and performs a unified callback on the platform entry time dbti of the i-th record to obtain the unified correction time yti. Yti = dbti - AEsi; where yti specifically represents the uniform correction time of the i-th record, and AEsi represents the source-level time shift of the data source to which the i-th record belongs; After calculating the unified correction time, the total time shift length Li of a single record is constructed; the formula is as follows: Li = gi + ni + ri + ci; where Li specifically represents the total time shift length of the i-th record; The total time shift length Li is compared with the source-level time shift AEs and the discrete bandwidth Ds, and three types of labels are output. The process is as follows: When the total backlog of a single record exceeds the sum of the usual time shift and discrete bandwidth of the data source, i.e., Li > AEsi + Dsi, the record is marked as a late record. When the backfill distance is greater than zero, i.e., ri > 0, the record is marked as a supplementary record. When the number of records arriving at the platform in the same batch is greater than 1, i.e., bi > 0, the records are marked as batch record tags.

7. A construction environment monitoring system based on digital twins according to claim 6, characterized in that: The wake wave discrimination and anomaly attribution module includes an environmental baseline and interval calculation unit and an anomaly attribution determination unit; The environmental benchmark and interval calculation unit are based on the unified correction time yti, combined with the master clock tk and the spatial center xk, and for different construction procedures c and each anchor point group k, the environmental time series yc(t) is extracted based on the source-level time shift AEs. Based on the environmental time series yc(t), the environmental baseline value μ(k,c) and the natural fluctuation band S(k,c) are calculated and obtained, and the leading width Pc and tail width Wc of the process are obtained. The formula for obtaining the environmental reference value μ(k, c) is: ; In the formula, Q0.25 represents the quartile operation, taking down the quartile, and t(k-1) represents the master clock of the (k-1)th anchor group; The formula for obtaining the natural wave band S(k,c) is: ; The formulas for obtaining the leader width Pc and the wake width Wc are: ； ； In the formula, Θc represents the set of anchor points for construction procedure c, min represents the start time of the leading zone, δ≥0 represents the time elapsed after the master clock, and δ represents the amount of time required for environmental indicators to fall back to the natural fluctuation range starting from the center time. By integrating the leader width Pc and the wake width Wc, the legal home interval Ik = [tk - Pc, tk + Wc] is obtained.

8. A construction environment monitoring system based on digital twins according to claim 7, characterized in that: The anomaly attribution determination unit is based on the natural wave zone S(k,c). It fits time segments that exceed the natural wave zone S(k,c) to obtain the environmental anomaly segment Eh, and records the start time thst, end time thed, and anomaly spatial location ξh of the environmental anomaly segment Eh. For each environmental anomaly segment Eh, calculate the center time ηh using the formula: ηh = (thst + thed) / 2; For each environmental anomaly segment Eh, detect whether the detection center time ηh falls within the legal attribution interval Ik, and determine the spatial proximity between the spatial position ξh of the anomaly segment and the spatial center xk of the anchor point group. If both time and space conditions are met, then the environmental anomaly segment Eh is assigned to anchor point group k. If the center time ηh of the environmental anomaly segment Eh falls within the legal allocation interval of multiple anchor point groups, the most suitable anchor point group is selected based on the comprehensive distance of time and space. ； In the formula, K+ represents the final anchor group number to which the anomalous segment belongs, and ρ(ξh, xk) represents the distance or topological distance between the spatial location of the anomalous segment and the spatial center of the anchor group. If the environmental anomaly segment Eh carries a late record label, a supplementary record label, or a batch record label, and a unique attribution result cannot be formed, it is marked as pending attribution.

9. A construction environment monitoring system based on digital twins according to claim 8, characterized in that: The twin write-back and review sorting module includes a write-back unit and a timing clarity calculation and review sorting unit; The write-back unit writes the anchor group number of the environmental anomaly segment Eh, the center time ηh of the environmental anomaly segment Eh, the unified correction time yti, the master clock tk, and the spatial center xk into the corresponding time axis and spatial unit of the digital twin model. Based on the anchor group number k, the master clock tk and spatial center xk are mapped to the center time ηh of each environmental anomaly segment Eh. The attribution result and label of the environmental anomaly segment Eh are attached to the corresponding time axis position and twin unit position in the digital twin model, and the source information and attribution anchor group number of each record are retained.

10. A construction environment monitoring system based on digital twins according to claim 9, characterized in that: The timing clarity calculation and review sorting unit receives the sorted anchor group's abnormal information and tags, and counts the abnormal records for each anchor group k: including the number of abnormal records belonging to the current anchor group curNk, the number of abnormal records belonging to the wake wave taNk, the number of abnormal records with tags laNk, and the number of records with reversed order reNk. The temporal clarity Ck is calculated based on the anomaly records, using the following formula: ； In the formula, Ck specifically represents the temporal clarity of anchor group k; All anchor point groups are arranged in ascending order of temporal clarity to obtain the review priority order PD. The anchor point groups with lower clarity are reviewed first. The environmental anomaly segment Eh information after review and confirmation is written back to the corresponding time axis and spatial unit of the digital twin model. At the same time, the timing clarity of the corrected anchor point group is recalculated, and an updated review priority sequence is formed based on the timing clarity, so as to trace, verify and manage the correspondence between construction environment anomalies and construction actions; The formula for obtaining the review priority order PD is: PD = sortasc(C1, C2, ..., CKs); In the formula, CKs represents the total number of anchor groups, and sortasc represents the ascending sort operation.

Images