A Construction Process Monitoring Method Based on BIM and IoT

By constructing a construction spatiotemporal correlation network model and a time synchronization protocol, and combining first-order difference and discrete cross-correlation analysis, the problems of sensor static error and theoretical benchmark adaptability were solved, thereby improving the accuracy of anomaly source tracing and monitoring during the construction process.

CN123117531APending Publication Date: 2026-11-17JINAN CHANGQING IND DEVELOPMENT INVESTMENT HOLDING GROUP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611286458.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing construction monitoring technologies, the accumulation of static errors in sensors leads to false alarms. Static theoretical benchmarks cannot adapt to the actual operating rhythm of machinery, and it is difficult to trace the source of abnormalities in concurrent construction operations.

Method used

Based on BIM and IoT, a construction spatiotemporal correlation network model is constructed. The time reference of the structural monitoring terminal and the mechanical condition acquisition terminal is unified through a time synchronization protocol. Combined with first-order difference calculation and discrete cross-correlation analysis, suspected abnormal source processes are retrieved in reverse and time delay consistency is checked. A dynamic reference response sequence is generated to identify the abnormal source.

Benefits of technology

It enables reverse localization of the source of component status anomalies in multi-device concurrent construction scenarios, improves data time consistency and the accuracy of status anomaly judgment, reduces false alarm rate, and meets the needs of anomaly identification and source tracing analysis in the construction process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117531A_ABST
    Figure CN123117531A_ABST
Patent Text Reader

Abstract

This invention relates to the field of information-based monitoring technology for building engineering, and discloses a method for monitoring the construction process of building engineering based on BIM and the Internet of Things. The method includes the following steps: constructing a construction spatiotemporal correlation network model and presetting a mechanical response delay window; unifying the time benchmark to extract the observed rate of change sequence; establishing an equivalent integral relationship of operation duration to generate a dynamic benchmark response sequence; calculating the cumulative distance of the morphological difference between the observed rate of change sequence and the dynamic benchmark response sequence, and determining component anomalies if the distance exceeds the limit; retrieving a set of suspected anomaly source process nodes, calculating the physical delay time through cross-correlation, performing time delay consistency verification using the mechanical response delay window, calculating a comprehensive matching score, and outputting anomaly source tracing alarms. This invention enables the theoretical response benchmark to be scaled and matched with the on-site construction rhythm over time, and achieves anomaly source process location through time delay consistency verification and comprehensive matching score.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of information monitoring technology for building engineering, specifically a method for monitoring the construction process of building engineering based on BIM and the Internet of Things. Background Technology

[0002] Modern large-scale construction projects are characterized by a large number of components, complex connections, and long construction periods. The construction process involves multiple procedures such as hoisting, support removal, concrete pouring, and equipment transportation, with multiple construction machines operating in parallel during the same construction phase. During the construction phase, building components are affected by the temporary support status, construction loads, component connection status, and changes in the sequence of operations, resulting in continuous changes in boundary conditions and stress states.

[0003] In existing technologies, Building Information Modeling (BIM) and the Internet of Things (IoT) have been applied to construction monitoring systems for building projects. Technicians deploy structural monitoring equipment such as displacement sensors, strain sensors, and tilt sensors on the building structure to collect component status data. The data collected by the sensors is processed by a backend server, establishing a mapping relationship with component nodes in the BIM model, and then displayed in the 3D model or construction management platform. In this way, physical data at the construction site can be linked to spatial component information in the BIM model, allowing project managers to view component status data and spatial distribution during the construction phase, achieving remote monitoring, data retention, and construction status display.

[0004] However, existing construction monitoring solutions mostly compare the absolute values ​​of sensor physical quantities with fixed safety thresholds. Temperature changes at the construction site cause slow thermal expansion and contraction of components, and long-term sensor operation leads to zero-point drift. These slow temperature disturbances and zero-point drift are superimposed on the measurement results, reducing the accuracy of fixed threshold judgments. When finite element static simulation data is used as an auxiliary comparison benchmark, the theoretical response data relies on a fixed time axis and cannot be adjusted for changes in the on-site machinery operation rate, resulting in a time misalignment between theoretical response data and on-site observation data. In scenarios with multiple devices operating concurrently, there is a lack of correlation analysis mechanisms based on force transmission paths, time delay constraints, and topological hierarchy between mechanical condition data and structural monitoring data. When component nodes exhibit abnormal states, existing monitoring solutions struggle to identify the source process from multiple simultaneously operating process nodes, failing to meet the needs of anomaly identification and source tracing analysis during construction. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a construction process monitoring method for building engineering based on BIM and the Internet of Things, which solves the problems in existing technologies such as false alarms caused by the accumulation of static errors of sensors, data comparison misalignment caused by the inability of static theoretical benchmarks to adapt to the actual operation rhythm of machinery, and difficulty in reverse tracing the source process of abnormalities in concurrent construction operations.

[0006] To achieve the above objectives, this application provides the following technical solution: A construction process monitoring method based on BIM and IoT includes the following steps: A construction spatiotemporal relationship network model containing process nodes and component nodes is constructed based on building information model data, and a mechanical response delay window is preset. Based on the time synchronization protocol, the time base of the structural monitoring terminal and the mechanical condition acquisition terminal is unified, and the observation data corresponding to the component nodes is obtained. The observation rate of change sequence is extracted by first-order difference calculation. Obtain the operating rate parameter from the mechanical operating condition data, establish the equivalent integral relationship of the operating duration by combining it with the theoretical response benchmark data, solve the actual mapping time and resample to generate a dynamic benchmark response sequence; Calculate the cumulative distance of morphological differences between the observed rate of change sequence and the dynamic baseline response sequence. When the cumulative distance of morphological differences is greater than the tolerance threshold, it is determined that the corresponding component node has an abnormal state. When a corresponding component node is determined to have an abnormal state, a set of suspected abnormal source process nodes is retrieved in reverse from the construction spatiotemporal correlation network model; the physical delay time is obtained by cross-correlation calculation of the mechanical condition change rate sequence and the observed change rate sequence corresponding to the suspected abnormal source process node set; based on the mechanical response delay window between the suspected abnormal source process node and the component node with an abnormal state, the physical delay time is checked for consistency; a comprehensive matching score is calculated for the candidate process nodes that pass the consistency check, and an abnormal source tracing alarm is output.

[0007] Preferably, a construction spatiotemporal relationship network model containing process nodes and component nodes is constructed based on building information model data, including: Extract the association information between construction procedures and physical components, and establish an initial node set containing procedure nodes and component nodes; Establish process sequence edges, mechanical transmission edges, and process component association edges in the initial node set; By fusing the process sequence subgraph connected by the process sequence edge with the component mechanics transfer subgraph connected by the mechanics transfer edge, a construction spatiotemporal relationship network model is generated through the process component association edge.

[0008] Preferably, based on a time synchronization protocol, the time base of the structural monitoring terminal and the mechanical condition acquisition terminal is unified to obtain the observation data corresponding to the component nodes, including: The master clock device of the time synchronization protocol sends synchronization messages to the structural monitoring terminal and the mechanical condition acquisition terminal; The structural monitoring terminal and the mechanical condition acquisition terminal send delay request messages to the master clock device; After receiving the delay request message, the master clock device returns a delay response message to the structural monitoring terminal and the mechanical condition acquisition terminal. The structural monitoring terminal and the mechanical condition acquisition terminal calculate the network link transmission delay and local clock deviation based on the timestamps corresponding to the synchronization message, delay request message and delay response message; The structural monitoring terminal and the mechanical condition acquisition terminal use local clock deviation to perform phase compensation calibration on the local timing unit, generate a unified time reference, and add an absolute timestamp when acquiring observation data and mechanical condition data.

[0009] Preferably, an equivalent integral relationship for the operation duration is established by combining theoretical response benchmark data to solve for the actual mapping time, including: Obtain the theoretical static time axis from the theoretical response baseline data; Convert the operation rate parameter into a continuous actual operation rate function; Based on the preset theoretical standard operating rate, and with the theoretical time point on the theoretical static time axis as the target quantity, a normalized cumulative operating quantity equation for the continuous actual operating rate function is constructed. The normalized workload cumulative equation is discretized and summed for solution. When the cumulative normalized workload equals the workload corresponding to the theoretical time point, the upper limit of the integral is taken as the corresponding actual mapping time.

[0010] Preferably, the process of solving for the actual mapping time and resampling to generate a dynamic benchmark response sequence includes: Obtain the discrete sequence of the rate of change of the theoretical response from the theoretical response benchmark data; A discrete scatter set is constructed using the actual mapped time series as the x-axis and the data points of the discrete sequence of the theoretical response rate of change as the y-axis. A baseline rate of change interpolation curve is generated by fitting a discrete scatter point set using a cubic spline interpolation algorithm. Resampling is performed on the benchmark rate of change interpolation curve with a uniform sampling period as the step size to generate a dynamic benchmark response sequence.

[0011] Preferably, calculating the cumulative morphological difference distance between the observed rate of change sequence and the dynamic baseline response sequence includes: Calculate the squared Euclidean distance between each eigenvalue in the observed rate of change sequence and each eigenvalue in the dynamic baseline response sequence, and construct a two-dimensional local distance matrix; Find regular paths in the two-dimensional local distance matrix and recursively calculate the cumulative cost function corresponding to each element in the two-dimensional local distance matrix; During the recursive calculation process, a window width parameter is introduced to limit the deviation of the regularized path from the diagonal of the two-dimensional local distance matrix. The cumulative cost function value corresponding to the endpoint of the two-dimensional local distance matrix is ​​used as the cumulative distance of morphological difference.

[0012] Preferably, the set of suspected anomaly source process nodes is retrieved in reverse from the construction spatiotemporal correlation network model, including: Starting from the component node that is determined to have an abnormal state, the preceding component node and the process node associated with the preceding component node are retrieved layer by layer in the opposite direction of the directed edge in the construction spatiotemporal association network model. Determine whether the retrieved process node is in a working state; Within the maximum search depth, aggregate the process nodes that are in operation to generate a set of process nodes suspected of being sources of anomalies.

[0013] Preferably, the physical delay time is obtained by cross-correlation calculation of the mechanical condition change rate sequence and the observed change rate sequence corresponding to the suspected anomaly source process node set, including: Calculate the discrete cross-correlation function between the mechanical operating condition rate of change sequence and the observed rate of change sequence at different time offsets; Traverse the discrete cross-correlation function within the preset delay step range to determine the delay step corresponding to the maximum positive correlation peak value of the discrete cross-correlation function. The physical delay time is obtained by multiplying the delay step size by the uniform sampling period.

[0014] Preferably, based on the mechanical response delay window between the suspected anomaly source process node and the component node with an abnormal state, a delay consistency check is performed on the physical delay time, including: The path-level mechanical response delay window between the suspected abnormal source process node and the component node with an abnormal state is used as the mechanical response delay window; Compare the physical delay time with the mechanical response delay window values; When the physical delay time is within the mechanical response delay window, the corresponding suspected anomaly source process node is retained as a candidate process node. When the physical delay time is not within the mechanical response delay window, the corresponding suspected abnormal source process node is deleted from the suspected abnormal source process node set.

[0015] Preferably, a comprehensive matching score is calculated for candidate process nodes that pass the latency consistency check, including: Obtain the normalized cross-correlation peak value, latency matching degree, and topology path cost of candidate process nodes; The normalized cross-correlation peak value is calculated by dividing the original cross-correlation peak value by the geometric mean of the zero-delay autocorrelation peak values ​​of the mechanical condition rate of change sequence and the observed rate of change sequence. The delay matching degree is determined based on the absolute value of the difference between the physical delay time and the midpoint of the mechanical response delay window; The construction spatiotemporal network model has a preset topology path cost weight; the topology path cost is determined based on the transmission path with the minimum cumulative topology path cost weight between the candidate process node and the component node with an abnormal state. The normalized cross-correlation peak, time delay matching degree, and topology path cost are weighted according to the preset weight coefficients to obtain the comprehensive matching score, in which the topology path cost is deducted.

[0016] Compared with the prior art, this application has at least the following beneficial effects: 1. This invention employs a technical solution that combines actual mechanical operating rates to establish an equivalent integral relationship of operating time to generate a dynamic benchmark response sequence. This achieves the technical effect of allowing the theoretical response benchmark to be scaled and matched with changes in the on-site construction rhythm over time. Compared to existing technologies that rely on static finite metadata or fixed-time-axis theoretical response data for judgment and comparison, this invention overcomes the shortcomings of these technologies, such as difficulty in adapting to fluctuations in the operating rate of construction machinery, and the tendency to cause time misalignment between theoretical response data and on-site observation data, leading to an increased false alarm rate.

[0017] 2. This invention employs a reverse retrieval based on a construction spatiotemporal correlation network model, combined with discrete cross-correlation calculation, time delay consistency verification, and multi-feature comprehensive matching scoring to determine the abnormal source process. This achieves the technical effect of reverse locating the source of abnormal component status in multi-device concurrent construction scenarios. Compared to existing technologies that only provide isolated alarms for over-limit sensors at the component's end, this invention overcomes the deficiency of these technologies in tracing the abnormal source process from abnormal component phenomena when facing complex force transmission paths and multiple concurrent operations.

[0018] 3. This invention employs a time synchronization protocol to unify the time reference between the structural monitoring terminal and the mechanical condition acquisition terminal, and utilizes a first-order difference extraction technique to extract the observed rate of change sequence. This achieves the technical effect of improving the time consistency of multi-source acquired data and extracting the dynamic response change characteristics of components. Compared with the existing technology that directly compares the absolute values ​​of sensor physical quantities with fixed thresholds, this invention overcomes the shortcomings of the previous technology, which is easily affected by slow environmental temperature disturbances, long-term zero-point drift of sensors, and time asynchrony of multi-source data, thus reducing the accuracy of abnormal state judgment. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the system architecture according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 3A flowchart illustrating the principle of constructing a construction spatiotemporal correlation network model according to an embodiment of the present invention; Figure 4 This is a flowchart of the multi-terminal time synchronization and observation rate of change feature extraction according to an embodiment of the present invention; Figure 5 This is a flowchart illustrating the theoretical time axis dynamic mapping principle of an embodiment of the present invention. Figure 6 This is a flowchart of the structural state anomaly detection and regularization path according to an embodiment of the present invention; Figure 7 This is a flowchart of concurrent process anomaly tracing and multi-feature comprehensive matching scoring according to an embodiment of the present invention; Figure 8 The following is a waveform diagram of the dynamic monitoring and anomaly identification signal of the steel main beam in an embodiment of the present invention, wherein (a) is a comparison diagram of the observed rate of change sequence and the dynamic reference response sequence, and (b) is a diagram of the calculation results of the dynamic time warping path; Figure 9 The diagram shows the cross-correlation calculation and feature comparison of anomaly tracing in this embodiment of the invention, where (a) is the cross-correlation function calculation result of candidate processes, and (b) is the comparison diagram of multi-feature comprehensive matching score. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 This invention provides a construction process monitoring system for building engineering based on BIM and the Internet of Things, including an edge computing server, a BIM server, a structural monitoring terminal, and a mechanical condition acquisition terminal.

[0022] Edge computing servers are deployed at the construction site to receive, aggregate, and process structural monitoring data and mechanical condition data generated on-site. They also serve as the master clock for the time synchronization protocol, providing a unified time reference for all data acquisition terminals. The edge computing servers establish communication connections with the BIM server, structural monitoring terminals, and mechanical condition acquisition terminals, respectively.

[0023] The BIM server stores 4D building information model data and finite element simulation analysis data for building projects, and provides edge computing servers with the basic data and theoretical response benchmark data required to construct the construction spatiotemporal network model. The theoretical response benchmark data includes at least one of the following: process node number, associated component node number, construction stage condition number, theoretical response change rate sequence, theoretical static time axis, theoretical response physical quantity type, and standard operating rate.

[0024] The structural monitoring terminal is deployed on the physical components of the building structure and integrates physical state sensors. Configured as a slave clock device with a time synchronization protocol, the terminal collects at least one physical state quantity from the components, including displacement and strain, adds an absolute timestamp based on a unified time reference to the collected observation data, and then uploads the data to an edge computing server.

[0025] The machinery condition acquisition terminal is deployed in the control system of the construction machinery, acquiring operation rate parameters by reading the machinery control bus. The machinery condition acquisition terminal is configured as a slave clock device using a time synchronization protocol, used to add absolute timestamps to the machinery condition data based on a unified time base, aligned with the structural monitoring terminal, and to upload the machinery condition data to the edge computing server.

[0026] Reference Figure 2 The construction process monitoring method based on BIM and IoT includes the following steps: S100 constructs a construction spatiotemporal correlation network model based on building information model data, and presets the mechanical response delay window and topology path cost weights. S200 uses a unified time reference between the unified structure monitoring terminal and the mechanical condition acquisition terminal based on a time synchronization protocol to acquire observation data with absolute timestamps and extract the observation rate of change sequence. S300: Obtain the working rate parameter from the mechanical working condition data, combine it with the theoretical response benchmark data to establish the equivalent integral relationship of the working duration, and dynamically map to generate a dynamic benchmark response sequence. S400: Calculate the cumulative distance of the morphological difference between the observed rate of change sequence and the dynamic baseline response sequence. When the cumulative distance exceeds the tolerance threshold, it is determined that there is an abnormal state of the component node. S500 extracts a set of suspected anomaly source process nodes from the construction spatiotemporal correlation network model, and outputs anomaly source tracing alarm results through discrete cross-correlation calculation, time delay consistency verification, and multi-feature comprehensive matching score.

[0027] The following describes the step-by-step method for monitoring the construction process of building engineering based on BIM and the Internet of Things, with specific implementation methods.

[0028] Reference Figure 3In this embodiment, step S100 is executed by the edge computing server, which converts the construction progress information, component connection information, and interaction relationships between processes and components in the building information model data into a searchable graph structure, and provides mechanical response delay constraints and path cost parameters for the reverse traversal search after anomaly detection. Step S100 includes the following sub-steps: S101, the edge computing server obtains four-dimensional building information model files, structured attribute library and construction schedule data from the BIM server, reads the entity component identification, component geometric coordinates, component material properties, component connection relationship, construction process number and process schedule time information, and extracts the association information between construction process and entity component from the reading results.

[0029] Edge computing server establishes initial node set and define the node set . This represents a set of construction process nodes, corresponding to construction operation items in the construction schedule, such as steel structure hoisting operations and foundation pit support dismantling operations. It represents a set of physical component nodes, corresponding to the physical structural units that bear force in the physical space of a building project, such as steel main beams and concrete columns.

[0030] During the system initialization phase, the edge computing server establishes a binding relationship between the structural monitoring terminal and the physical component nodes. The binding relationship includes at least one of the following: terminal number, component node number, installation coordinates, measurement direction, and type of physical quantity acquired, so that subsequent observation data can be mapped to the corresponding component node.

[0031] set up For any specific process node, This refers to any specific component node. The parsing of 4D Building Information Model (BIM) files can be completed through the Industrial Foundation Class (IFC) data interface or the data interface provided by the BIM platform. The parsed fields include at least the component's unique identifier, the component's spatial location, the component type, the schedule task identifier, and the association identifier between the component and the work process.

[0032] S102, the edge computing server establishes a set of directed edges of multiple types in the initial node set based on the extracted association information. and define the edge set .

[0033] Edge computing servers parse the temporal relationships between preceding and subsequent tasks in the construction schedule and generate process sequence edges connecting different process nodes. Process timing edge Constraints used to characterize the sequence of actions performed. Because the task sequence in the construction plan progresses along the time axis, ... and The process sequence subgraph formed by the connected process nodes is a directed acyclic graph.

[0034] Edge computing servers parse the geometric coordinates, connection methods, contact relationships, and force transmission relationships of physical components in physical space, and generate mechanical transmission edges connecting different component nodes. Mechanical transmission edge Used to characterize the load transfer path or mechanical coupling relationship between solid components. and The component mechanics transfer subgraph formed by the connected component nodes is a weighted directed graph.

[0035] The edge computing server parses the construction resource configuration table, identifies the correspondence between construction operations and the components that directly bear the construction operations, and generates process-component association edges connecting process nodes and component nodes. Process component related edge The construction resource allocation table is used to characterize the direct external forces exerted on the target physical component during the construction process. It includes at least one field: construction machinery number, work sequence number, target component number, work type, and planned work time period.

[0036] S103, after extracting the node structure and edge structure, the edge computing server performs graph structure fusion of the process sequence subgraph and the component mechanical transfer subgraph to generate a construction spatiotemporal correlation network model. .

[0037] In one implementation, the edge computing server associates edges with process components. As a connecting channel, the process progress logic and structural force transmission logic are associated with the same graph structure. To reduce the risk of invalid searches and loop traversals, the edge computing server traverses the initial node set and edge set during the fusion process, deleting isolated nodes with both zero out-degree and zero in-degree.

[0038] After data association, the construction schedule and building structure geometric model are transformed into a construction spatiotemporal relational network model that simultaneously includes the temporal relationships of work processes and the spatial force transmission relationships of components. In the construction spatiotemporal relational network model, the changing data of work process nodes with associated edges or force transmission paths can be transmitted along the directed edges to the mechanical response calculation path of the corresponding component nodes, thus providing a topological basis for subsequent reverse traversal search.

[0039] S104 After completing the construction of the construction spatiotemporal network model, the edge computing server assigns physical constraint parameters to the directed edges involved in mechanical transmission and completes the network parameter calibration.

[0040] For the mechanical transmission edge and process component related edge Edge computing servers preset mechanical response delay window and topology path cost weights The mechanical response delay window is used to characterize the range of physical time delay required for the mechanical action from the source node to be transmitted along the structural transmission path to the target node and cause an observable physical response.

[0041] In practical applications, the edge computing server reads the finite element simulation analysis results of the construction phase from the BIM server and extracts the initial baseline values ​​for the mechanical response delay window and the topology path cost weight. The initial baseline values ​​are determined based on engineering material parameters, component geometry, component connection status, theoretical stress wave propagation rate, and the deformation lag time of the overall structural mode shape. The finite element simulation analysis results of the construction phase include at least one of the following: construction phase case number, stressed component number, response propagation path, response start time, response peak time, and path stiffness parameter.

[0042] The edge computing server updates the initial baseline value by combining sensor sample data from similar historical construction projects or calibration test data from the initial construction phase of the current project site.

[0043] When there are multiple directed edges between a suspected anomaly source process node and an abnormal component node, the edge computing server generates a path-level mechanical response delay window based on the mechanical response delay window of each edge in the path. The lower limit of the path-level delay window is the cumulative value of the lower limits of the delay windows of all edges in the path, and the upper limit of the path-level delay window is the cumulative value of the upper limits of the delay windows of all edges in the path. The path-level mechanical response delay window can also be directly calibrated from the simulation analysis results of the finite element construction stage.

[0044] For topology path cost weights The edge computing server determines the weight values ​​based on the local stiffness distribution coefficients between adjacent structural components and the geometric distance between their centroids. To ensure dimensional consistency in the subsequent multi-feature comprehensive matching and scoring, the edge computing server normalizes the calculated weight values ​​so that they fall within the (0,1) interval.

[0045] When there are multiple transmission paths between a suspected abnormal source process node and an abnormal component node, the edge computing server calculates the cumulative value of the topology path cost weight in each transmission path, and takes the transmission path with the smallest cumulative topology path cost weight as the scoring path.

[0046] Generally, the higher the stiffness distribution ratio of adjacent structural components and the shorter the geometric distance between their centroids, the lower the cost corresponding to the mechanical transmission path, and the lower the topological path cost weight assigned by the edge computing server to the mechanical transmission path.

[0047] Reference Figure 4 In this embodiment, step S200 is executed collaboratively by an edge computing server, a structural monitoring terminal, and a mechanical condition acquisition terminal. This step establishes a unified time reference between the structural monitoring data and the mechanical condition data, and extracts the observed rate of change sequence from the observation data with absolute timestamps. Step S200 includes the following sub-steps: The S201 edge computing server is based on the IEEE 1588 precise time protocol and performs multi-terminal time synchronization with the structural monitoring terminal and the mechanical condition acquisition terminal.

[0048] During system initialization and operation, the edge computing server configured as the master clock sends synchronization messages to the structure monitoring terminal and the mechanical condition acquisition terminal configured as slave clocks via the on-site industrial LAN. The slave clock devices receive the synchronization messages and record the reception time, and then send a delay request message to the master clock; the master clock receives the delay request message, records the reception time, and returns a delay response message to the slave clock devices.

[0049] The structural monitoring terminal and the mechanical condition acquisition terminal calculate the network link transmission delay and local clock deviation based on the timestamps corresponding to the synchronization message, delay request message and delay response message, and calibrate the local timing unit based on the calculation results.

[0050] Each slave clock terminal uses the calculated clock deviation value to perform phase compensation calibration on its local timing unit, thereby forming a unified time reference among the distributed acquisition terminals. When the field network equipment supports hardware timestamps, the phase compensation calibration can reach the microsecond level; when the field network equipment does not support hardware timestamps, the system uses software timestamps for compensation and uses the synchronization error as a correction parameter for the tolerance threshold.

[0051] Based on a unified time reference, the structural monitoring terminal collects the original displacement or strain signals of physical components through physical state sensors, and converts them into discrete data points through analog-to-digital conversion. During the generation of discrete data points, the timestamp module of the structural monitoring terminal adds a compensated absolute timestamp to each discrete data point. The mechanical condition acquisition terminal uses the same timestamp appending method to time-stamp the mechanical condition data read from the mechanical control bus.

[0052] To ensure temporal consistency in subsequent sequence alignments, the edge computing server sorts, checks for missing values, and resamples the structural monitoring data and mechanical condition data according to absolute timestamps. When the time interval between adjacent data points exceeds a preset missing threshold, the edge computing server marks the corresponding time period as an invalid sampling segment to prevent invalid sampling segments from participating in subsequent anomaly detection.

[0053] S202, the edge computing server extracts the observation rate of change features based on the synchronized discrete observation data.

[0054] For any target component node in the construction spatiotemporal correlation network model The edge computing server extracts target component nodes according to a preset time window length. The corresponding absolutely discrete observation sequence uploaded by the structural monitoring terminal is defined as follows: .

[0055] To ensure that the discrete observation sequence reflects the dynamic response state of the component and avoids high-frequency signal aliasing, the system sets a sampling period. The Nyquist sampling theorem is satisfied, and the sampling frequency is configured to be greater than twice the highest frequency of the target analysis frequency band. The target analysis frequency band is determined comprehensively based on the natural frequency of the building structure, the excitation frequency of the construction machinery, and the sensor response bandwidth.

[0056] Considering that absolute physical quantity values ​​are easily affected by changes in ambient temperature, sensor zero-point drift, and random noise, edge computing servers do not directly use absolute physical quantity values ​​for state feature comparison. Before performing first-order difference calculations, the edge computing server processes the observation sequence... Amplification filtering, median filtering, or low-pass filtering are performed to suppress the amplification effect of random high-frequency noise on the differential characteristics.

[0057] As one implementation method, edge computing servers process observation sequences. Perform first-order difference calculations to generate the observed rate of change series. The formula for calculating the first-order difference eigenvalue is as follows: ; in, For the first The measured rate of change characteristic values ​​for each sampling interval and These are the absolute physical quantity values ​​recorded at the current sampling point and the previous adjacent sampling point, respectively. The sampling period is set uniformly for the entire system. This is to extract the total length of the sequence within the time window.

[0058] First-order difference processing is used to mitigate the impact of zero-point drift in the constant term and slow temperature disturbances on the observation sequence. Thermal expansion and contraction deformation of components at construction sites due to natural temperature differences typically exhibits low-frequency, slow changes, while zero-point drift generated by long-term sensor operation usually manifests as a relatively stable offset. In contrast, the mechanical response of components induced by construction operations typically changes within a shorter timeframe. Through first-order difference processing, the edge computing server can extract dynamic response change characteristics related to construction operations from the observation sequence.

[0059] Reference Figure 5 In this embodiment, step S300 is executed by the edge computing server. It is used to correct the static time axis corresponding to the theoretical response benchmark data based on the operation rate parameters uploaded by the machinery condition acquisition terminal, and to generate a dynamic benchmark response sequence that matches the on-site construction rhythm. Step S300 includes the following sub-steps: S301, the edge computing server determines the currently executing process node, reads the discrete sequence of theoretical response change rate and theoretical static time axis, and generates a continuous actual operation rate function based on the machine condition data.

[0060] For the current execution process node The edge computing server determines the current execution process node based on at least one of the following: the BIM four-dimensional schedule plan, the status of the construction task sheet, the on-site process confirmation signal, and the operation timestamp uploaded by the mechanical condition acquisition terminal.

[0061] The edge computing server reads from the finite element simulation analysis results of the construction phase stored on the BIM server, or generates them by calling the finite element solution module associated with the BIM server. Construction operations then affect the associated components. The discrete sequence of theoretical response rate of change when the action occurs and the discrete sequence corresponding to the theoretical response rate of change Theoretical static time axis Discrete sequence of theoretical response rate of change With the theoretical static time axis Discrete sequences of theoretical response rates of change with the same data index Each data point in the data corresponds to a theoretical static time axis. One of the theoretical time points.

[0062] Theoretical Static Timeline This represents the duration of the work process at a standard construction speed. To obtain the actual construction rhythm on site, the edge computing server retrieves the currently executing process node. The mechanical operating data of the corresponding construction machinery within the same time period are used to extract the discrete sequence of operating rates with absolute timestamps.

[0063] The discrete sequence of operating rates is determined according to the type of operation. In the hoisting operation, the operating rate parameters include at least one of the following: the winch wire rope winding speed, the hook lifting speed, or the slewing angular velocity; in the support dismantling operation, the operating rate parameters include at least one of the following: the hydraulic cylinder displacement speed, the oil pressure change rate, or the cutting equipment feed speed.

[0064] The edge computing server sorts the discrete sequence of job rates according to absolute timestamps and removes or interpolates missing or abruptly changed data points. After data processing, the edge computing server uses linear interpolation, Lagrange interpolation, or polynomial fitting to convert the discrete sequence of job rates into a continuous actual job rate function. .

[0065] S302, the edge computing server establishes an equivalent integral relationship of job duration based on job rate and solves for the actual mapping time.

[0066] During actual construction, the operating speed of construction machinery can fluctuate due to changes in the site environment, load weight, support status, or machinery control strategy, causing the stress response of components to occur earlier or later than the theoretical response. To establish a time mapping relationship between the actual construction progress and the theoretical construction progress, the edge computing server uses the nameplate calibration value of the construction machinery or the simulation setting value of the finite element construction stage as the theoretical standard operating speed. And establish a normalized cumulative equation for the actual work rate.

[0067] Edge computing servers based on a theoretical static timeline Using any theoretical time point as the objective variable, solve for the actual mapping time. This makes the actual mapping time The following equivalent integral relationship for operation duration must be satisfied: ; in, This is the time integration variable.

[0068] As a specific computational method, edge computing servers employ Simpson's integral rule or the trapezoidal rule to discretize and accumulate the equivalent integral relationship of the job duration. When the cumulative normalized job load reaches the theoretical time point... When dealing with the corresponding workload, the maximum integral value is taken as the actual mapping time. .

[0069] If the actual operating rate of the machinery on site is consistently lower than the theoretical standard operating rate The normalized workload accumulates at a slower rate, resulting in a shorter actual mapping time. Greater than the corresponding theoretical time point Edge computing servers for theoretical static timelines Each discrete time point in the sequence is solved one by one to obtain the actual mapped time series with a non-uniform distribution.

[0070] To avoid prolonged downtime of construction machinery leading to insufficient mapping time, the edge computing server presets a maximum allowable duration threshold for each process. During the discretization solution process, if the current iteration time exceeds the maximum allowable duration threshold for the process, and the accumulated normalized workload has not yet reached the theoretical time point, [further action will be taken]. The edge computing server determines the current execution process node based on the corresponding workload. If the process is in an abnormal interruption state, stop the currently executing process node. The integral mapping operation is performed, and a process stagnation marker is generated.

[0071] By using the equivalent integral relationship of the operation duration, the theoretical static time axis can be nonlinearly stretched or compressed according to the changes in the on-site mechanical operation rate, thereby obtaining the actual mapped time series corresponding to the on-site construction rhythm.

[0072] S303, the edge computing server resamples the actual mapped time series to generate a dynamic baseline response sequence.

[0073] The actual mapped time series obtained after step S302 exhibits a non-uniform distribution on the time axis, making direct point-by-point comparison with the observed rate of change series collected under a fixed sampling period impossible. To unify the time scale, the edge computing server uses the actual mapped time series as the x-axis and the theoretical response rate of change as the discrete sequence. Using the data points as the ordinate, a discrete scatter plot is constructed.

[0074] The edge computing server uses a cubic spline interpolation algorithm to fit a discrete set of points and generate a baseline rate of change interpolation curve. The baseline rate of change interpolation curve is used to represent the relationship between the theoretical response rate of change and the actual mapping time.

[0075] Subsequently, the edge computing server uses a uniform sampling period. Using a step size, perform equidistant resampling on the baseline rate of change interpolation curve, and keep the resampling time window consistent with the truncation window of the observed rate of change sequence.

[0076] Capture the window at the currently executing process node The actual start timestamp, the time when the machine operation rate first exceeds the start threshold, or the time corresponding to the on-site process confirmation signal are used as the starting point, with the preset sliding window length or the currently executing process node as the starting point. The end time is taken as the endpoint.

[0077] The edge computing server ultimately outputs a dynamic baseline response sequence after time-scale synchronization. ,in This represents the total length of the sequence generated after resampling. Dynamic benchmark response sequence. The theoretical response change trend obtained from the finite element construction stage simulation is retained, and the correlation between the theoretical response and the actual mechanical working condition data on site is maintained on the time scale.

[0078] Reference Figure 6 In this embodiment, step S400 is executed by the edge computing server to compare the observed rate of change sequences. With dynamic reference response sequence The cumulative distance of morphological differences between them is calculated, and an anomaly judgment result for the component node status is generated based on the tolerance threshold. Step S400 includes the following sub-steps: S401, the edge computing server constructs a local distance matrix between the observed rate of change sequence and the dynamic benchmark response sequence.

[0079] The edge computing server obtains the length as Observed rate of change series and length is Dynamic reference response sequence .

[0080] Edge computing servers based on observed rate of change sequences and dynamic reference response sequence The construction dimension is Two-dimensional local distance matrix Local distance matrix elements in Used to represent the observed rate of change sequence The Middle Eigenvalues ​​and Dynamic Benchmark Response Sequence The Middle The local difference between the eigenvalues ​​is calculated using the squared Euclidean distance, as shown in the following formula: ; in, , Local distance matrix Used to provide basic cost data for subsequent dynamic time warping calculations.

[0081] S402, the edge computing server performs dynamic time warping calculations based on the local distance matrix to obtain the cumulative distance of morphological differences.

[0082] To calculate the observed rate of change series With dynamic reference response sequence The overall morphological differences between them, and the local distance matrix of edge computing servers Find a regular path from the starting point to the ending point of the matrix, and minimize the sum of the local differences along the regular path.

[0083] Edge computing server defines cumulative cost function Cumulative cost function This represents the distance from the matrix starting point (1,1) to the matrix coordinates. The minimum cumulative distance. Before the recursive calculation begins, the edge computing server sets boundary initialization conditions to avoid index out-of-bounds errors. Specifically, the edge computing server sets the starting point of the matrix... and auxiliary boundaries and Initialize to positive infinity.

[0084] After completing the boundary initialization, the edge computing server calculates the cumulative cost function corresponding to each element in the matrix according to the following dynamic programming recursive formula: ; Considering the temporal continuity of the mechanical response of a building structure during continuous construction, edge computing servers can configure window constraints during recursive calculations to limit the deviation of the regularized path from the matrix diagonal. As one implementation method, the edge computing server introduces a window width parameter. When the coordinate index satisfies At that time, the edge computing server will use the corresponding cumulative cost function. The value is assigned to positive infinity to avoid forced matching of sequence points with large time span differences.

[0085] After the edge computing server completes the traversal calculation, it will return the value corresponding to the endpoint of the matrix. As observed rate of change sequence With dynamic reference response sequence The cumulative distance of morphological differences between the nodes. The cumulative distance of morphological differences is used to characterize the degree to which the measured response change trend of the target component node deviates from the dynamic reference response trend.

[0086] Window width parameter Based on the duration of the construction phase and the unified sampling cycle The fluctuation range of mechanical operation rate or the time deviation range in historical normal construction samples are determined and stored in the parameter library of BIM server or edge computing server.

[0087] S403, the edge computing server determines the status of component nodes based on the tolerance threshold and generates an initial anomaly marker.

[0088] Edge computing servers acquire cumulative distance of morphological differences Afterwards, With respect to the preset tolerance threshold Perform numerical comparison.

[0089] Tolerance threshold This data is derived from finite element simulation calibration data. An edge computing server analyzes historical simulation sample sequences within the range of no structural damage and normal construction deviations, extracts the statistical distribution of dynamic time-normalized distances between samples, and uses the upper limit of a specific confidence interval in the statistical distribution as a tolerance threshold. .

[0090] Tolerance threshold It can also be corrected by combining the on-site calibration samples from the initial construction phase of the current project; the on-site calibration samples include the observed rate of change sequence under normal construction conditions, the corresponding dynamic benchmark response sequence, and the dynamic time warping distance between the two types of sequences.

[0091] When the system uses software timestamp compensation, or when the on-site network synchronization error exceeds the preset error benchmark, the edge computing server adjusts the tolerance threshold based on the synchronization error. Make corrections.

[0092] When the following conditions are met: At that time, the edge computing server determines the target component node. The mechanical response characteristics are within the allowable construction fluctuation range, no initial anomaly marker is generated, and the process ends for the target component node. The abnormal judgment process.

[0093] When the following conditions are met: At that time, the edge computing server determines the target component node. The mechanical response characteristics exceeded the allowable construction fluctuation range, and the target component node was determined. An anomaly exists. The edge computing server is a target component node in the system database. Apply an initial exception flag.

[0094] The initial anomaly marker serves as the trigger condition for reverse tracing of subsequent anomaly source processes, driving the edge computing server to enter the reverse traversal search process for suspected anomaly source process nodes.

[0095] Reference Figure 7 In this embodiment, step S500 is executed by the edge computing server. Starting from the component node with an abnormal state, the edge computing server retrieves the process nodes in the working state from the construction spatiotemporal correlation network model, and then determines the abnormal source process through discrete cross-correlation calculation, mechanical response delay constraints, and multi-feature comprehensive matching scoring. Step S500 includes the following sub-steps: S501, the edge computing server performs a reverse breadth-first search in the construction spatiotemporal correlation network model to generate a set of suspected anomaly source process nodes.

[0096] Step S400 is the target component node After generating the initial anomaly marker, the edge computing server initiates the anomaly tracing process. The edge computing server uses the anomaly component node... Starting from the search point, in the construction spatiotemporal correlation network model Search layer by layer along the opposite direction of the directed edge.

[0097] The reverse retrieval process includes: along the mechanical transfer edge of the component. The reverse force transmission path is retrieved from the preceding component nodes, and the associated edges of the components along the process are also retrieved. The reverse action path retrieval retrieves the process node associated with the preceding component node.

[0098] To control the search range and computational load, the edge computing server is configured with a maximum search depth. The maximum search depth is determined based on the force transmission hierarchy of the building structure, the number of component connections, the main response paths obtained from finite element simulation, and the effective tracing range in historical construction condition samples. As one implementation, the maximum search depth is set to 2 to 3 physical topology levels extending outward from the anomalous component nodes.

[0099] During the retrieval process, the edge computing server determines whether a process node is in operation based on the absolute timestamp corresponding to the process node, combined with at least one of the following: BIM four-dimensional progress plan, construction task sheet status, operation rate parameters uploaded by the machinery condition acquisition terminal, and on-site process confirmation signal.

[0100] The edge computing server deletes process nodes that are down, not scheduled, not yet started, or have already ended, and aggregates process nodes that meet the maximum search depth and are in a working state, generating a set of process nodes suspected of being sources of anomalies. .

[0101] When the reverse retrieval fails to find any process nodes in the working state, the edge computing server outputs a component anomaly alarm for the process with an undetermined anomaly source and stops targeting the abnormal component nodes. The subsequent cross-correlation calculation.

[0102] S502, the edge computing server performs discrete cross-correlation calculations on the mechanical condition sequence and component observation change rate sequence corresponding to the suspected abnormal source process node, and extracts the physical delay time.

[0103] For the set of process nodes suspected of being abnormal sources Any suspected abnormal source process node in The edge computing server retrieves the mechanical condition data uploaded by the corresponding mechanical condition acquisition terminal and generates a mechanical condition change rate sequence using the same first-order difference processing method as the component observation data. .

[0104] Edge computing servers use absolute timestamps to analyze the mechanical operating condition change rate sequence. Extracting and observing the rate of change sequence Same time window, and according to a uniform sampling period Resampling is performed. When the effective time ranges of the two sequences are not completely consistent, the edge computing server truncates the sequences according to the overlapping time interval; when there are a few missing sampling points within the overlapping time interval, the edge computing server uses linear interpolation to fill in the missing sampling points.

[0105] To unify the direction of change in mechanical operating conditions with the direction of change in component response, the edge computing server determines the expected correlation direction between the two sequences based on finite element simulation results from the construction phase or on-site calibration samples. When the expected correlation direction is negative, the edge computing server adjusts the mechanical operating condition rate of change sequence... Perform sign transformation so that subsequent cross-correlation peaks are compared in the direction of positive correlation.

[0106] Edge computing servers calculate observed rate of change sequences With the series of mechanical operating conditions change rate Discrete cross-correlation function between This is used to represent the degree of waveform correlation between two sequences at different time offsets. The formula for calculating the discrete cross-correlation function is: ; in, Let be the delay step size parameter for sliding alignment of two sequences, and ; and These are the observed rate of change sequences. and the series of mechanical operating conditions change rates The arithmetic mean within the calculation window.

[0107] The edge computing server iterates through the cross-correlation function within a preset delay step range. And determine the cross-correlation function. The delay step size corresponding to the maximum positive correlation peak value The preset delay step range is based on the path-level mechanical response delay window and uniform sampling period generated in step S100. Sure.

[0108] Based on the unified time base established in step S200, the edge computing server follows... Delay step size Convert to physical delay time .

[0109] S503, the edge computing server, calculates the physical latency based on the path-level mechanical response delay window. Perform a consistency check on execution latency.

[0110] The edge computing server reads suspected anomaly source process nodes from the construction spatiotemporal correlation network model established in step S100. To abnormal component node Corresponding path-level mechanical response delay window And determine the physical delay time. Whether it is within the path-level mechanical response delay window.

[0111] when or Physical delay time The mechanical response delay constraints corresponding to the current structural materials, component connection state, and geometric transfer path are not met. The edge computing server identifies the corresponding cross-correlation peak as a result that does not conform to the force transmission timing relationship and removes the corresponding suspected abnormal source process node from the candidate nodes.

[0112] when Physical delay time To satisfy the path-level mechanical response delay constraint, the edge computing server retains the corresponding suspected abnormal source process node and transmits the corresponding cross-correlation peak, physical delay time and path information to step S504.

[0113] When all suspected anomaly source process nodes fail the latency consistency check, the edge computing server outputs a component anomaly alarm for which the anomaly source process has not been determined, and no longer performs multi-feature comprehensive matching scoring.

[0114] S504, the edge computing server performs multi-feature comprehensive matching and scoring on candidate process nodes that have passed the latency consistency verification, and outputs anomaly tracing alarms based on the scoring results.

[0115] For candidate process nodes that pass the latency consistency check, the edge computing server calculates a comprehensive matching score. Used to determine nodes of abnormal components The process node with the highest degree of matching in response change relationship. Overall matching score. The calculation formula includes three dimensionless features: normalized cross-correlation peak value, time delay matching degree, and topological path cost. ; in, This represents the normalized cross-correlation peak value. The edge computing server will use the original cross-correlation peak value... Divided by the observed rate of change series With the series of mechanical operating conditions change rate The geometric mean of the zero-delay autocorrelation peaks of each sensor is used to obtain the normalized cross-correlation peak, and the normalized cross-correlation peak is limited to a preset value range to reduce the impact of differences in sensor range and signal amplitude on the scoring results.

[0116] This represents the median of the path-level mechanical response delay window. The delay matching term in the overall matching score is used to represent the physical delay time. The degree of closeness to the path-level mechanical response delay window value. When the path-level mechanical response delay window value is zero, the edge computing server uses a preset positive number to replace the denominator to avoid division by zero in the scoring calculation.

[0117] From candidate process node to abnormal component node The cumulative topology path cost weight is the transmission path with the lowest cumulative topology path cost weight. The higher the cumulative topology path cost weight, the better the chances of finding a candidate process node and an abnormal component node. The more indirect the transmission path, the greater the path deduction value in the overall matching score.

[0118] , and These are the weighting coefficients. , and All are non-negative numbers and have been normalized to make them... The edge computing server is determined based on finite element simulation samples, historical construction condition samples, or the initial calibration samples of the current project. , and The value of .

[0119] Edge computing servers compare the comprehensive matching scores of each candidate process node. The highest comprehensive matching score is compared with the preset source matching threshold.

[0120] When the highest comprehensive matching score is lower than the source matching threshold, the edge computing server does not determine the unique abnormal source process and outputs a component abnormality alarm for which the abnormal source process is not determined.

[0121] When the highest comprehensive matching score is not lower than the source matching threshold, the edge computing server will determine the candidate process node with the highest comprehensive matching score as the node causing the abnormal component. The process node with the highest probability of an abnormal status will be identified, and an abnormality tracing alarm will be output.

[0122] When the difference in the comprehensive matching scores among multiple candidate process nodes is less than a preset difference threshold, the edge computing server outputs multiple suspected anomaly source process nodes and their corresponding rankings according to the comprehensive matching scores from high to low, and marks the abnormal component nodes in the control visualization interface and BIM space model. And the corresponding process nodes suspected of being sources of abnormality.

[0123] The parameters of source matching threshold, preset difference threshold, mechanical operation start threshold, maximum allowable duration threshold of process, maximum search depth and window width are determined based on finite element simulation samples, historical construction condition samples or current project initial calibration samples, and stored in the parameter library of BIM server or edge computing server.

[0124] To further illustrate the data processing process of the BIM and IoT-based construction process monitoring method in the construction site, the following example uses a scenario of dual-machine lifting and assembly of a large-span steel main beam.

[0125] Reference Figure 8 In the construction scenario of this embodiment, the building project performs a dual-crane lifting and assembly operation of a large-span steel main beam. The edge computing server reads the four-dimensional progress model and finite element construction stage simulation analysis results from the BIM server, and establishes the lifting process nodes and steel main beam component nodes corresponding to the two crawler cranes in the initial node set.

[0126] The edge computing server establishes process component association edges from the lifting process node to the steel main beam component node, and presets the path-level mechanical response delay window and topology path cost weight in the database for the transmission of the lifting point effect to the mid-span monitoring point of the steel main beam.

[0127] During the lifting operation, the edge computing server in the on-site industrial local area network sends synchronization messages to the structural monitoring terminal and the mechanical condition acquisition terminal through the Precision Time Protocol, enabling the local timing unit of each terminal to complete time calibration.

[0128] Strain sensors deployed at the mid-span of the main steel beam continuously collect strain signals from the component. The structural monitoring terminal adds an absolute timestamp based on a unified time base to the collected strain data and uploads the timestamped strain data to the edge computing server.

[0129] Edge computing servers perform first-order difference calculations on discrete strain data to generate observed rate-of-change sequences that mitigate the effects of static zero-point drift and slow temperature disturbances, such as... Figure 8 The solid line waveform of part (a) is shown in the figure.

[0130] Simultaneously, the mechanical condition acquisition terminals within the control buses of the two crawler cranes acquire the real-time operating rate of the winch and append an absolute timestamp based on the same time reference to the real-time operating rate of the winch. The mechanical condition acquisition terminals then upload the mechanical condition data with the absolute timestamp to the edge computing server.

[0131] The edge computing server, based on the actual operating rates of the two crawler cranes, uses the equivalent integral relationship of the operation duration to solve point-by-point the theoretical static time axis corresponding to the theoretical response benchmark data, obtaining the actual mapped time series. Subsequently, the edge computing server generates a dynamic benchmark response sequence that varies with the on-site lifting rhythm through cubic spline interpolation, such as... Figure 8 The dashed waveform in part (a) is shown.

[0132] During a certain period of the lifting operation, the hydraulic system of the winch of one of the crawler cranes experienced a transient jam, causing an unexpected drop in the actual lifting rate. This transient jam in the winch's hydraulic system triggered changes in the stress response at the mid-span of the steel main beam, and the observed rate of change sequence showed abnormal fluctuations within the corresponding time period.

[0133] Edge computing servers utilize a dynamic time warping algorithm to calculate the local distance matrix between the observed rate of change sequence and the dynamic baseline response sequence, and recursively search for a warped path under a set window width constraint, such as... Figure 8 As shown in section (b), when the cumulative distance of morphological differences calculated by the edge computing server exceeds the tolerance threshold, it is determined that there is a state anomaly in the steel main beam component node, and an initial anomaly mark is added to the steel main beam component node.

[0134] Reference Figure 9 After the initial anomaly markers are generated, the edge computing server performs a reverse breadth-first search in the construction spatiotemporal correlation network model, starting from the steel main beam component nodes. Within the maximum search depth range, the edge computing server extracts the lifting process nodes corresponding to the two crawler cranes currently in operation and generates a set of suspected anomaly source process nodes.

[0135] For the set of suspected abnormal source process nodes, the edge computing server calculates the discrete cross-correlation function between the mechanical condition change rate sequence of the two crawler cranes and the observed change rate sequence of the steel main beam, finds the delay step corresponding to the maximum positive correlation peak, and converts the delay step into physical delay time.

[0136] like Figure 9 As shown in part (a), the cross-correlation function corresponding to the crawler crane that experiences hydraulic jamming exhibits a positive correlation peak at a specific delay step, and the calculated physical delay time is within the preset path-level mechanical response delay window, thus passing the delay consistency check.

[0137] Another normally operating crawler crane was removed from the screening process by the edge computing server because the cross-correlation function did not show the required positive correlation peak due to the small change in working conditions, or the calculated physical delay time exceeded the path-level mechanical response delay window.

[0138] To identify the source of the anomaly, the edge computing server performs a multi-feature comprehensive matching score on candidate process nodes that have passed the latency consistency check. The edge computing server substitutes the normalized cross-correlation peak value, latency matching degree, and topology path cost weight into the comprehensive matching score formula.

[0139] like Figure 9 As shown in part (b), the crawler crane lifting process node that experiences jamming has a high normalized cross-correlation peak value and time delay matching degree, and its comprehensive matching score is higher than the preset traceability matching threshold.

[0140] The edge computing server identifies the tracked crane lifting process that is stuck as the most likely source of the abnormality that could cause the steel main beam to be in an abnormal state, and outputs an abnormality tracing alarm on the control interface.

[0141] The technical features of the above embodiments can be combined in any way (as long as there is no contradiction in the combination of these technical features). For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written should also be considered to be within the scope of this specification.

Claims

1. A method for monitoring the construction process of building engineering based on BIM and the Internet of Things, characterized in that, Includes the following steps: A construction spatiotemporal relationship network model containing process nodes and component nodes is constructed based on building information model data, and a mechanical response delay window is preset. Based on the time synchronization protocol, the time base of the structural monitoring terminal and the mechanical condition acquisition terminal is unified, and the observation data corresponding to the component nodes is obtained. The observation rate of change sequence is extracted by first-order difference calculation. Obtain the operating rate parameter from the mechanical operating condition data, establish the equivalent integral relationship of the operating duration by combining it with the theoretical response benchmark data, solve the actual mapping time and resample to generate a dynamic benchmark response sequence; Calculate the cumulative distance of morphological differences between the observed rate of change sequence and the dynamic baseline response sequence. When the cumulative distance of morphological differences is greater than the tolerance threshold, it is determined that the corresponding component node has an abnormal state. When a corresponding component node is determined to have an abnormal state, a set of suspected abnormal source process nodes is retrieved in reverse from the construction spatiotemporal correlation network model; the physical delay time is obtained by cross-correlation calculation of the mechanical condition change rate sequence and the observed change rate sequence corresponding to the suspected abnormal source process node set; based on the mechanical response delay window between the suspected abnormal source process node and the component node with an abnormal state, the physical delay time is checked for consistency; a comprehensive matching score is calculated for the candidate process nodes that pass the consistency check, and an abnormal source tracing alarm is output.

2. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, A construction spatiotemporal relationship network model containing process nodes and component nodes is constructed based on building information model data, including: Extract the association information between construction procedures and physical components, and establish an initial node set containing procedure nodes and component nodes; Establish process sequence edges, mechanical transmission edges, and process component association edges in the initial node set; By fusing the process sequence subgraph connected by the process sequence edge with the component mechanics transfer subgraph connected by the mechanics transfer edge, a construction spatiotemporal relationship network model is generated through the process component association edge.

3. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, Based on a time synchronization protocol, the time base of the structural monitoring terminal and the mechanical condition acquisition terminal is unified, and the observation data corresponding to the component nodes is acquired, including: The master clock device of the time synchronization protocol sends synchronization messages to the structural monitoring terminal and the mechanical condition acquisition terminal; The structural monitoring terminal and the mechanical condition acquisition terminal send delay request messages to the master clock device; After receiving the delay request message, the master clock device returns a delay response message to the structural monitoring terminal and the mechanical condition acquisition terminal. The structural monitoring terminal and the mechanical condition acquisition terminal calculate the network link transmission delay and local clock deviation based on the timestamps corresponding to the synchronization message, delay request message and delay response message; The structural monitoring terminal and the mechanical condition acquisition terminal use local clock deviation to perform phase compensation calibration on the local timing unit, generate a unified time reference, and add an absolute timestamp when acquiring observation data and mechanical condition data.

4. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, By combining theoretical response baseline data, an equivalent integral relationship for the operation duration is established, and the actual mapping time is solved, including: Obtain the theoretical static time axis from the theoretical response baseline data; Convert the operation rate parameter into a continuous actual operation rate function; Based on the preset theoretical standard operating rate, and with the theoretical time point on the theoretical static time axis as the target quantity, a normalized cumulative operating quantity equation for the continuous actual operating rate function is constructed. The normalized workload cumulative equation is discretized and summed for solution. When the cumulative normalized workload equals the workload corresponding to the theoretical time point, the upper limit of the integral is taken as the corresponding actual mapping time.

5. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, Solving for the actual mapping time and resampling to generate a dynamic baseline response sequence includes: Obtain the discrete sequence of the rate of change of the theoretical response from the theoretical response benchmark data; A discrete scatter set is constructed using the actual mapped time series as the x-axis and the data points of the discrete sequence of the theoretical response rate of change as the y-axis. A baseline rate of change interpolation curve is generated by fitting a discrete scatter point set using a cubic spline interpolation algorithm. Resampling is performed on the benchmark rate of change interpolation curve with a uniform sampling period as the step size to generate a dynamic benchmark response sequence.

6. The construction process monitoring method for building engineering based on BIM and IoT according to claim 1, characterized in that, Calculate the cumulative morphological distance between the observed rate of change series and the dynamic baseline response series, including: Calculate the squared Euclidean distance between each eigenvalue in the observed rate of change sequence and each eigenvalue in the dynamic baseline response sequence, and construct a two-dimensional local distance matrix; Find regular paths in the two-dimensional local distance matrix and recursively calculate the cumulative cost function corresponding to each element in the two-dimensional local distance matrix; During the recursive calculation process, a window width parameter is introduced to limit the deviation of the regularized path from the diagonal of the two-dimensional local distance matrix. The cumulative cost function value corresponding to the endpoint of the two-dimensional local distance matrix is ​​used as the cumulative distance of morphological difference.

7. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, In the construction spatiotemporal correlation network model, a set of suspected anomaly source process nodes is retrieved in reverse, including: Starting from the component node that is determined to have an abnormal state, the preceding component node and the process node associated with the preceding component node are retrieved layer by layer in the opposite direction of the directed edge in the construction spatiotemporal association network model. Determine whether the retrieved process node is in a working state; Within the maximum search depth, aggregate the process nodes that are in operation to generate a set of process nodes suspected of being sources of anomalies.

8. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, The physical delay time is obtained by cross-correlation calculation between the mechanical condition change rate sequence and the observed change rate sequence corresponding to the set of suspected anomaly source process nodes, including: Calculate the discrete cross-correlation function between the mechanical operating condition rate of change sequence and the observed rate of change sequence at different time offsets; Traverse the discrete cross-correlation function within the preset delay step range to determine the delay step corresponding to the maximum positive correlation peak value of the discrete cross-correlation function. The physical delay time is obtained by multiplying the delay step size by the uniform sampling period.

9. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, Based on the mechanical response delay window between the suspected anomaly source process node and the component node with an abnormal state, a delay consistency check is performed on the physical delay time, including: The path-level mechanical response delay window between the suspected abnormal source process node and the component node with an abnormal state is used as the mechanical response delay window; Compare the physical delay time with the mechanical response delay window values; When the physical delay time is within the mechanical response delay window, the corresponding suspected anomaly source process node is retained as a candidate process node. When the physical delay time is not within the mechanical response delay window, the corresponding suspected abnormal source process node is deleted from the suspected abnormal source process node set.

10. The construction process monitoring method for building engineering based on BIM and the Internet of Things according to claim 1, characterized in that, Calculate a comprehensive matching score for candidate process nodes that pass the latency consistency check, including: Obtain the normalized cross-correlation peak value, latency matching degree, and topology path cost of candidate process nodes; The normalized cross-correlation peak value is calculated by dividing the original cross-correlation peak value by the geometric mean of the zero-delay autocorrelation peak values ​​of the mechanical condition rate of change sequence and the observed rate of change sequence. The delay matching degree is determined based on the absolute value of the difference between the physical delay time and the midpoint of the mechanical response delay window; The construction spatiotemporal network model has a preset topology path cost weight; the topology path cost is determined based on the transmission path with the minimum cumulative topology path cost weight between the candidate process node and the component node with an abnormal state. The normalized cross-correlation peak value, time delay matching degree, and topology path cost are weighted according to the preset weight coefficients to obtain the comprehensive matching score, in which the topology path cost is deducted.