An ultra-deep diaphragm wall construction dynamic supervision method and system based on an internet of things

Abstract

The application discloses a kind of based on Internet of Things's super deep diaphragm wall construction dynamic supervision method and system, comprising the following steps: construct the perception network consisting of edge computing gateway node and multiple types of sensing unit, structural response data such as displacement, inclination, strain, acceleration are collected and cached, according to the structure disturbance degree, switch data processing mode, use the Hampe l filter model with local statistical threshold adjustment mechanism to identify and eliminate outliers, generate structured filter data sequence, symbol mapping is carried out through sliding window, symbol trajectory sequence is constructed, and symbol dynamic entropy calculation is executed, form local symbol dynamic entropy sequence.Combining abnormal frequency information generates two-dimensional time sequence judgment matrix, constructs hierarchical risk coding model, and outputs digital warning coding signal with time mark and risk level to supervision platform and management terminal, realizes the dynamic identification and risk response to construction state.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of civil engineering information construction monitoring, and in particular to a method and system for dynamic supervision of super deep diaphragm wall construction based on the Internet of Things. BACKGROUND

[0002] With the development of smart construction sites and Internet of Things sensing technology, high-frequency monitoring and dynamic early warning of structural response state during super deep diaphragm wall construction have gradually become an important part of ensuring the safety control of foundation pit engineering. At present, multiple sensors are often used on construction sites to collect physical quantities such as displacement, strain and inclination in real time, and edge devices or central servers are used to perform filtering, analysis and early warning.

[0003] The existing construction monitoring technology still has obvious limitations in the face of complex working condition disturbance, multi-source data noise interference and response behavior evolution trend identification. On the one hand, conventional filtering models have difficulty in adapting to local disturbance feature changes when processing disturbed structural response signals, which can easily lead to misjudgment of abnormal values and loss of effective information, affecting the accuracy of subsequent analysis. On the other hand, current trend identification of monitoring data is mostly based on global statistics or artificial rule setting, lacking deep modeling of symbol evolution law and dynamic changes of structural complexity, making it difficult to accurately reflect the risk dynamics in the working condition change process. Therefore, how to provide a method and system for dynamic supervision of super deep diaphragm wall construction based on the Internet of Things is a problem that needs to be solved by those skilled in the art. SUMMARY

[0004] One object of the present application is to provide a method and system for dynamic supervision of super deep diaphragm wall construction based on the Internet of Things. The present application combines multi-source structural response monitoring, edge computing, Hampel filtering model and symbol dynamic entropy calculation technology, and describes in detail the whole process processing method from perception data collection, disturbance identification, abnormal filtering, symbol trajectory construction to risk coding generation, which has the advantages of fine monitoring response, accurate abnormal identification and timely risk warning.

[0005] According to the method and system for dynamic supervision of super deep diaphragm wall construction based on the Internet of Things, the following steps are included:

[0006] A multi-source Internet of Things sensing network covering the construction site of super deep diaphragm wall is constructed, a sensing unit with edge processing capability is deployed, construction structural response signals are collected, original multi-dimensional time series data are generated and edge side caching is completed;

[0007] Based on the local disturbance features of the original multi-dimensional time series data, the data processing mode is determined and the sampling strategy is switched, and the data processing input sequence corresponding to the construction working condition is output;

[0008] Applying a Hampel filtering model with a local statistical threshold adjustment mechanism to the data processing input sequence, identifying and eliminating outliers, generating a structured filtering data sequence;

[0009] Performing sliding window symbol mapping on the structured filtering data sequence to construct a symbol trajectory sequence reflecting the state evolution characteristics;

[0010] Performing symbol dynamic entropy calculation based on the symbol trajectory sequence to generate a local symbol dynamic entropy sequence;

[0011] Combining the local symbol dynamic entropy sequence with the abnormal data frequency information marked by the Hampel filtering model to generate a two-dimensional time series judgment matrix;

[0012] Constructing a hierarchical risk coding model based on the two-dimensional time series judgment matrix, and outputting a digital early warning coding signal with a time identification field and a risk level field;

[0013] Outputting the digital early warning coding signal to the super deep diaphragm wall construction monitoring platform and the remote construction management terminal for triggering a linkage response mechanism and updating construction safety supervision data.

[0014] Optionally, the generation of the original multi-dimensional time series data includes:

[0015] Constructing a multi-source Internet of Things sensing network in the super deep diaphragm wall construction area, which is composed of an edge computing gateway node, multiple sensing units and a wireless communication module. The edge computing gateway node is deployed in the center of the construction control area for unified management of data reception, caching and preliminary processing;

[0016] The multiple sensing units are configured according to the structure position and response type, specifically including: the sensing units for collecting displacement signals are arranged at the top and middle areas of the wall, the sensing units for collecting inclination signals are arranged at the edge area of the supporting structure, and the sensing units for collecting acceleration signals, strain signals, pore water pressure signals and temperature signals are arranged at the bottom of the foundation pit, the enclosure structure, the soil adjacent to the pile foundation and the heat source sensitive area of the construction area;

[0017] A stable communication link between the edge computing gateway node and each sensing unit is established through the wireless communication module. The edge computing gateway node sends sampling instructions to each sensing unit according to the set synchronous sampling period, and each sensing unit synchronously collects the structure response signals at its location;

[0018] Each sensing unit returns the collected structure response signals to the edge computing gateway node in the form of a data packet. The edge computing gateway node performs timestamp extraction and unified time correction on the received data packet to construct a structure response data set based on a unified time reference;

[0019] The edge computing gateway node integrates the structure response data set, organizes the original multi-dimensional time series data in a time sequence, and writes the original multi-dimensional time series data into a local cache area in the edge computing gateway node.

[0020] Optionally, the output of the data processing input sequence includes:

[0021] The edge computing gateway node performs sliding window analysis on the original multi-dimensional time series data, extracts the rate of change parameters of the structure response monitoring variables, including the displacement rate of change, the inclination angle rate of change, the acceleration fluctuation amplitude, the strain rate, and the pore water pressure gradient;

[0022] Based on the deviation between the rate of change parameters of the structure response monitoring variables and the corresponding stable threshold, a disturbance level label is generated, and a target data processing mode is selected in a preset sampling strategy mapping table according to the disturbance level label;

[0023] The data processing mode includes a basic data processing mode and an enhanced data processing mode, wherein the basic data processing mode defines a standard sampling frequency, a default upload period, a fixed channel range, and a regular sensor unit activation state, and the enhanced data processing mode defines an improved sampling frequency, a compressed upload period, an expanded channel range, and an activation configuration of a local supplementary sensor unit;

[0024] According to the parameter content of the target data processing mode, the corresponding sensor unit is controlled to perform structure response signal sampling, and a data processing mode label field, a sampling parameter field, and a spatial channel identification field are added to the collected structure response signal to generate a structured sampling data unit;

[0025] According to the time sequence and field consistency rules, the plurality of structured sampling data units are integrated to generate a data processing input sequence.

[0026] Optionally, the generation of the structured filtering data sequence includes:

[0027] The edge computing gateway node divides the structure response monitoring variables in the data processing input sequence into a plurality of channel data sets according to the physical quantity type, and configures an independent sliding window for each channel data set, wherein the length of the sliding window is dynamically determined from a disturbance level-window length mapping table according to the disturbance level label of the last period;

[0028] In each sliding window, the median value of the structure response monitoring variables in the corresponding sliding window is calculated as the center reference value, and the absolute deviation of all data points in the sliding window is calculated based on the center reference value, and the median of the absolute deviation is extracted as the absolute median deviation of the corresponding sliding window;

[0029] The first discrimination threshold and the second discrimination threshold of the channel are loaded from a preset channel threshold parameter table according to the channel type, and the center reference value and the absolute median difference are used as a basis for calculation. If the absolute deviation between the structural response monitoring variable value and the center reference value exceeds the second discrimination threshold multiplied by the absolute median difference, it is identified as a strong abnormal value. If the deviation of the structural response monitoring variable value is between the first discrimination threshold and the second discrimination threshold multiplied by the absolute median difference, it is identified as a weak abnormal value. Otherwise, it is determined to be a normal value.

[0030] The structural response monitoring variable value identified as a strong abnormal value is replaced with the center reference value of the current sliding window, and the structural response monitoring variable value identified as a weak abnormal value is added with an abnormality mark field. The original data of the normal value is retained without modification.

[0031] The processed structural response monitoring variable values in all channels are time axis aligned, and a channel identification field and an abnormality mark field are added to form a structured filtering data sequence.

[0032] Optionally, the construction of the symbol trajectory sequence comprises:

[0033] The edge computing gateway node establishes a sliding symbol mapping window for each type of structural response monitoring variable in the structured filtering data sequence. The length of the sliding symbol mapping window is set according to the sampling frequency of the monitoring variable, and the sliding step is 1.

[0034] In each sliding symbol mapping window, the structural response monitoring variable values corresponding to the continuous time points in the current window are extracted. Each structural response monitoring variable value in the sliding symbol mapping window is subjected to symbol mapping processing according to a preset symbol division rule. The symbol division rule divides the value range into a plurality of non-overlapping intervals, and each value interval corresponds to a unique symbol identifier.

[0035] All symbol identifiers generated in the sliding symbol mapping window are arranged in time sequence and combined to generate a symbol segment, and a time index field and a channel identification field of the structural response monitoring variable are added in the symbol segment.

[0036] After the sliding symbol mapping window completes a sliding, the sliding symbol mapping window is continuously advanced in time sequence, and the symbol mapping and segment combination steps are repeatedly executed to gradually generate a plurality of symbol segments.

[0037] All symbol segments are integrated and reorganized in the generation time sequence to form a complete symbol trajectory sequence.

[0038] Optionally, the generation of the local symbol dynamic entropy sequence comprises:

[0039] A symbol dynamic entropy calculation sliding window is constructed for each type of structure response monitoring variable in the symbol trajectory sequence, the length of the symbol dynamic entropy calculation sliding window is parameter set according to the sampling frequency and variation characteristics of the corresponding structure response monitoring variable, and the sliding step is set as one time unit;

[0040] In each symbol dynamic entropy calculation sliding window, a time-continuous symbol identifier sequence is extracted, all symbol identifier types appearing in the current window are identified, and the ratio between the occurrence number of each symbol identifier and the total number of symbols is counted to form a symbol probability distribution corresponding to the current window;

[0041] Based on the probability values of each symbol identifier in the current symbol probability distribution, the symbol dynamic entropy value of the symbol dynamic entropy calculation sliding window is calculated, and the symbol dynamic entropy value is bound with the time label of the current sliding window and the channel identifier of the structure response monitoring variable to generate a local symbol dynamic entropy data point;

[0042] The symbol dynamic entropy calculation sliding window is slid forward along the time axis by one time unit, and the symbol extraction, probability construction, entropy value calculation and data point generation steps are repeated to gradually generate multiple local symbol dynamic entropy data points;

[0043] All local symbol dynamic entropy data points are combined in time sequence into a local symbol dynamic entropy sequence.

[0044] Optionally, the generation of the two-dimensional time sequence judgment matrix comprises:

[0045] The local symbol dynamic entropy sequence is received, the local symbol dynamic entropy data points are classified and arranged according to the channel identifier of the structure response monitoring variable, and an entropy value time distribution column is constructed based on each time unit;

[0046] The structure response monitoring variable values with the added abnormality marked field are extracted from the structured filtering data sequence, and the ratio between the number of monitoring variable values marked as abnormal in each time unit and the number of all monitoring variable values under the corresponding time unit is counted to form a corresponding abnormal data frequency value sequence;

[0047] The time labels in the local symbol dynamic entropy data points are aligned with the time labels of the abnormal data frequency value sequence, and the row index of the two-dimensional time sequence judgment matrix is established based on the time unit, and the column index is established according to the channel identifier;

[0048] In the two-dimensional time sequence judgment matrix, each cell corresponds to a combination of a time label and a structure response monitoring variable channel, and the combination expression pair of the local symbol dynamic entropy value corresponding to the time label and the corresponding abnormal data frequency value is filled in the cell to form a composite monitoring judgment unit with entropy disturbance dimension and frequency disturbance dimension;

[0049] The edge computing gateway node completes data filling of all time indexes and channel indexes, and generates a complete two-dimensional time sequence judgment matrix.

[0050] Optionally, the output of the digital early warning encoding signal includes:

[0051] A hierarchical risk encoding model is constructed based on the two-dimensional time sequence judgment matrix, and the hierarchical risk encoding model includes a risk level judgment rule library, a time label index module, a channel state evaluation module, and an encoding field generation module.

[0052] The risk level judgment rule library defines a plurality of risk level threshold intervals, each risk level threshold interval including a combined range of a local symbolic dynamic entropy value interval and an abnormal data frequency value interval.

[0053] The time label index module extracts all composite monitoring judgment units of each time unit corresponding to all channels in the two-dimensional time sequence judgment matrix according to the row index, and inputs them to the channel state evaluation module.

[0054] The channel state evaluation module performs level mapping on the composite monitoring judgment units under each channel according to the risk level judgment rule library, obtains the corresponding channel risk level label, and counts the maximum risk level of all channels in the time unit.

[0055] The encoding field generation module constructs a digital early warning encoding signal including a time field and a risk level field according to the time label of the time unit and the corresponding maximum risk level.

[0056] Optionally, the following modules are included:

[0057] A multi-source Internet of Things sensing system is used to deploy a sensing network in a construction area, the sensing network including a plurality of sensing units with edge processing capability, an edge computing gateway node, and a wireless communication module, each sensing unit collecting structural response monitoring variables and generating original multi-dimensional time series data;

[0058] An edge data processing module is used to extract local disturbance features of the original multi-dimensional time series data, switch data processing modes and sampling strategies according to disturbance levels, and output data processing input sequences.

[0059] An anomaly identification and filtering module includes a Hampel filtering model module, which is used to perform sliding window processing on the data processing input sequences by channel, identify and correct anomalies based on a central reference value and an absolute median difference, and generate a structured filtering data sequence.

[0060] A symbolic modeling module is used to map the structured filtering data sequence to a symbolic trajectory sequence, construct a sliding symbolic mapping window, and generate a symbolic fragment with time index and channel identification.

[0061] a symbol dynamic entropy calculation module, configured to construct a sliding window based on the symbol trajectory sequence, calculate a local symbol probability distribution, and generate a local symbol dynamic entropy sequence;

[0062] a two-dimensional time sequence judgment matrix generation module, configured to align and combine the local symbol dynamic entropy sequence with the abnormal data frequency marked in the structured filtering data sequence, and construct a two-dimensional time sequence judgment matrix;

[0063] a hierarchical risk coding model module, configured to generate a digital early warning coding signal with a time field and a risk level field based on the two-dimensional time sequence judgment matrix;

[0064] a linkage output and supervision interaction module, configured to output the digital early warning coding signal to a construction monitoring platform and a remote construction management terminal.

[0065] The present application has the following beneficial effects:

[0066] (1) The present application realizes the on-site real-time collection, classification caching and dynamic processing of structural response monitoring data by constructing a multi-source Internet of Things sensing network and combining an edge computing architecture, effectively supporting the dynamic sensing and data closed-loop management of the construction process under complex working conditions.

[0067] (2) The present application realizes fine identification and hierarchical marking processing of strong and weak abnormal values in multi-channel response signals by introducing an improved Hampel filtering model with a local statistical threshold adjustment mechanism, while ensuring effective preservation of original information, improving the pertinence and robustness of abnormal value elimination.

[0068] (3) The present application constructs a local symbol dynamic entropy sequence that can quantitatively reflect the complexity of structural response by designing a sliding window symbol mapping and symbol dynamic entropy calculation mechanism, and generates a two-dimensional time sequence judgment matrix by combining the abnormal frequency value, making the risk identification have both disturbance frequency and disturbance structure, and improving the traceability of risk evolution trend.

[0069] (4) The present application outputs a digital early warning coding signal with a time field and a risk level field by constructing a hierarchical risk coding model, and links the construction monitoring platform and the remote terminal for response, effectively enhancing the timeliness and operability of safety warning, and constructing a construction dynamic supervision closed-loop system integrating sensing, judgment and warning. BRIEF DESCRIPTION OF DRAWINGS

[0070] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application and are used to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0071] Figure 1A flow chart of a method and system for dynamic supervision of super deep diaphragm wall construction based on Internet of Things is provided. DETAILED DESCRIPTION

[0072] The application will now be described in further detail with reference to the drawings. These drawings are simplified schematic diagrams which show the basic structure of the application in a schematic manner only, and thus only show the components relevant to the application.

[0073] Reference Figure 1 A method and system for dynamic supervision of super deep diaphragm wall construction based on Internet of Things includes the following steps:

[0074] A multi-source Internet of Things sensing network covering the construction site of the super deep diaphragm wall is constructed, a sensing unit with edge processing capability is deployed, construction structure response signals are collected, original multi-dimensional time series data are generated, and edge side caching is completed;

[0075] Based on the local disturbance features of the original multi-dimensional time series data, the data processing mode is determined and the sampling strategy is switched, and the data processing input sequence corresponding to the construction condition is output;

[0076] The Hampel filtering model with local statistical threshold adjustment mechanism is applied to the data processing input sequence, and the abnormal values are identified and removed, and the structured filtering data sequence is generated;

[0077] The structured filtering data sequence is subjected to sliding window symbol mapping, and the symbol trajectory sequence reflecting the state evolution features is constructed;

[0078] Based on the symbol trajectory sequence, symbol dynamic entropy calculation is performed, and the local symbol dynamic entropy sequence is generated;

[0079] The local symbol dynamic entropy sequence and the abnormal data frequency information marked by the Hampel filtering model are combined to generate a two-dimensional time sequence judgment matrix;

[0080] Based on the two-dimensional time sequence judgment matrix, a hierarchical risk coding model is constructed, and a digital early warning coding signal with a time identification field and a risk level field is output;

[0081] The digital early warning coding signal is output to the super deep diaphragm wall construction monitoring platform and the remote construction management terminal, which is used to trigger the linkage response mechanism and update the construction safety supervision data.

[0082] In this embodiment, the generation of the original multi-dimensional time series data includes:

[0083] A multi-source Internet of Things sensing network is constructed in the construction area of the super-deep diaphragm wall, and the multi-source Internet of Things sensing network is composed of an edge computing gateway node, a plurality of sensing units and a wireless communication module. The edge computing gateway node is deployed in the center of the construction control area and is used for unified management of data reception, caching and preliminary processing.

[0084] The plurality of sensing units are configured according to the structural position and the response type. Specifically, the sensing units for collecting displacement signals are arranged at the top and middle regions of the wall, the sensing units for collecting inclination signals are arranged at the edge region of the supporting structure, and the sensing units for collecting acceleration signals, strain signals, pore water pressure signals and temperature signals are arranged at the bottom of the foundation pit, the enclosure structure, the soil body adjacent to the pile foundation and the heat source sensitive region of the construction area, respectively.

[0085] A stable communication link is established between the edge computing gateway node and each sensing unit through the wireless communication module. The edge computing gateway node sends a sampling instruction to each sensing unit according to a set synchronous sampling period, and each sensing unit synchronously collects the structural response signals at its location.

[0086] Each sensing unit returns the collected structural response signals to the edge computing gateway node in the form of a data packet. The edge computing gateway node performs timestamp extraction and unified time correction on the received data packet to construct a structural response data set based on a unified time reference.

[0087] The edge computing gateway node integrates and processes the structural response data set, organizes the original multi-dimensional time series data in the order of sampling time, and writes the original multi-dimensional time series data into the local cache area in the edge computing gateway node.

[0088] In this embodiment, the output of the data processing input sequence includes:

[0089] The edge computing gateway node performs sliding window analysis on the original multi-dimensional time series data to extract the change rate parameters of the structural response monitoring variables, including the displacement change rate, the inclination angle change rate, the acceleration fluctuation amplitude, the strain rate and the pore water pressure gradient.

[0090] Based on the deviation between the change rate parameters of the structural response monitoring variables and the corresponding stable threshold, a disturbance level label is generated, and a target data processing mode is selected in a preset sampling strategy mapping table according to the disturbance level label.

[0091] The data processing mode includes a basic data processing mode and an enhanced data processing mode, wherein the basic data processing mode defines a standard sampling frequency, a default upload period, a fixed channel range and a regular sensor unit activation state, and the enhanced data processing mode defines an improved sampling frequency, a compressed upload period, an extended channel range and an activation configuration of a local supplementary sensor unit;

[0092] According to the parameter content of the target data processing mode, the structural response signal sampling of the corresponding sensor unit is controlled, and a data processing mode label field, a sampling parameter field and a spatial channel identification field are added to the collected structural response signal to generate a structured sampling data unit;

[0093] According to the time sequence and field consistency rule, the plurality of structured sampling data units are integrated to generate a data processing input sequence;

[0094] The deviation degree between the change rate parameter of the structural response monitoring variable and the corresponding stable threshold value is that the actual change rate value of each structural response monitoring variable obtained in the sliding time window is compared with the stable reference threshold value set for each type of structural response monitoring variable to form a deviation degree evaluation result reflecting whether the current structural response is in a stable state. The stable reference threshold value is an interval value set in advance based on historical monitoring data, design specifications or field working condition experience, and is used to represent the change rate range of each structural response monitoring variable in a normal state. The deviation degree is judged based on whether the actual change rate value exceeds the stable reference threshold interval or the relative distance from the reference threshold.

[0095] In the embodiment, the generation of the structured filtering data sequence includes:

[0096] The edge computing gateway node divides the structural response monitoring variables in the data processing input sequence into a plurality of channel data sets according to the physical quantity type, and configures an independent sliding window for each channel data set, wherein the length of the sliding window is dynamically determined from the disturbance level-window length mapping table according to the disturbance level label of the last period;

[0097] In each sliding window, the median value of the structural response monitoring variables in the corresponding sliding window is calculated as the center reference value, and the absolute deviation of all data points in the sliding window is calculated based on the center reference value, and the median of the absolute deviation is extracted as the absolute median deviation of the corresponding sliding window;

[0098] The first discrimination threshold and the second discrimination threshold of the channel are loaded from a preset channel threshold parameter table according to the channel type, and the center reference value and the absolute median difference are used as the basis for calculation. If the absolute deviation between the structural response monitoring variable value and the center reference value exceeds the second discrimination threshold multiplied by the absolute median difference, it is identified as a strong abnormal value. If the deviation of the structural response monitoring variable value is between the first discrimination threshold and the second discrimination threshold multiplied by the absolute median difference, it is identified as a weak abnormal value. Otherwise, it is determined to be a normal value.

[0099] The structural response monitoring variable value identified as a strong abnormal value is replaced by the center reference value of the current sliding window, and the structural response monitoring variable value identified as a weak abnormal value is marked with an abnormality mark field. The original data of the normal value is retained without modification.

[0100] The processed structural response monitoring variable values in all channels are time axis aligned, and channel identification fields and abnormality mark fields are added to form a structured filtering data sequence.

[0101] The preset channel threshold parameter table is a structured configuration data set stored in the edge computing gateway node, which is used to set corresponding abnormality identification parameters for different types of structural response monitoring variables. The channel threshold parameter table is indexed and organized by monitoring variable type, and each record includes monitoring variable name, first discrimination threshold, second discrimination threshold, and corresponding channel identification. The first discrimination threshold is used to identify the upper limit of the deviation range of weak abnormal values, and the second discrimination threshold is used to identify the lower limit of the deviation range of strong abnormal values. Both are set with the absolute median difference as the scale reference. This parameter table can be generated by engineering experience data, historical monitoring data statistical analysis or expert system preset, and can be dynamically issued to the edge computing gateway node by the remote supervision system. In actual operation, the channel type is called in real time for abnormal value identification and judgment;

[0102] The replacement processing refers to the replacement of the original value of the data point with the center reference value in the sliding window to which the data point belongs according to the set processing rule for the structural response monitoring variable data determined as a strong abnormal value. The center reference value is the median of all monitoring variable values in the sliding window, which is used to reflect the local data trend. The replacement processing is automatically executed by the edge computing gateway node after the abnormality discrimination is completed. The execution process includes extracting the median of the current sliding window, positioning the index position of the target abnormal data point in the data sequence, and covering the original value for writing while retaining the abnormal state mark information of the original data point.

[0103] In this embodiment, the construction of the symbol trajectory sequence includes:

[0104] The edge computing gateway node establishes a sliding symbol mapping window for each type of structure in the structured filtered data sequence, the length of the sliding symbol mapping window is set according to the sampling frequency of the monitoring variable, and the sliding step is 1;

[0105] In each sliding symbol mapping window, the values of the structure response monitoring variables corresponding to the continuous time points in the current window are extracted, and each value of the structure response monitoring variable in the sliding symbol mapping window is subjected to symbol mapping processing according to a preset symbol division rule, the symbol division rule divides the value range into a plurality of non-overlapping intervals, and each value interval corresponds to a unique symbol identifier;

[0106] All symbol identifiers generated in the sliding symbol mapping window are arranged in time sequence and combined to generate a symbol segment, and a time index field and a channel identifier field of the structure response monitoring variable are added in the symbol segment;

[0107] After the sliding symbol mapping window completes a sliding, the sliding symbol mapping window is continuously advanced in time sequence, and the symbol mapping and segment combination steps are repeatedly executed to gradually generate a plurality of symbol segments;

[0108] All symbol segments are integrated and reorganized in the generation time sequence to form a complete symbol trajectory sequence.

[0109] In the embodiment, the generation of the local symbol dynamic entropy sequence includes:

[0110] A symbol dynamic entropy calculation sliding window is constructed for each type of structure response monitoring variable in the symbol trajectory sequence, the length of the symbol dynamic entropy calculation sliding window is parameter set according to the sampling frequency and the variation characteristic of the corresponding structure response monitoring variable, and the sliding step is set as one time unit;

[0111] In each symbol dynamic entropy calculation sliding window, a time-continuous symbol identifier sequence is extracted, all symbol identifier types appearing in the current window are identified, and the ratio between the occurrence number of each symbol identifier and the total number of symbol identifiers is calculated to form a symbol probability distribution corresponding to the current window;

[0112] Based on the probability values of the symbol identifiers in the current symbol probability distribution, a symbol dynamic entropy value of the symbol dynamic entropy calculation sliding window is calculated, and the symbol dynamic entropy value is bound with the time label of the current sliding window and the channel identifier of the structure response monitoring variable to generate a local symbol dynamic entropy data point;

[0113] The symbol dynamic entropy calculation sliding window is slid forward along the time axis by one time unit, and the symbol extraction, probability construction, entropy value calculation and data point generation steps are repeatedly executed to gradually generate a plurality of local symbol dynamic entropy data points;

[0114] combining all local symbolic dynamic entropy data points in time sequence into a local symbolic dynamic entropy sequence;

[0115] The sampling frequency of the structure response monitoring variable refers to the number of times of signal collection of a specific physical response by the sensing unit in a unit time controlled by the edge computing gateway node, and specifically includes the sampling frequencies of displacement, inclination, acceleration, strain, pore water pressure and other physical quantities; different types of structure response monitoring variables are set with different initial sampling frequencies according to their dynamic change speed and disturbance sensitivity in the construction process of the super deep diaphragm wall, for example, the acceleration signal is usually configured with a higher sampling frequency because of its dramatic change, and the temperature or strain signal is configured with a lower sampling frequency because of its relatively gentle change; the variation characteristic refers to the degree of fluctuation of the value of the monitoring variable in a period of time, which is usually measured by calculating the difference amplitude, change rate or standard deviation between the continuous sampling values; the edge computing gateway node evaluates the dynamic state of the structure response monitoring variable based on the current sampling frequency and the real-time calculated variation characteristic, and adjusts the length of the sliding window for calculating the symbolic dynamic entropy, so as to adapt to the capture needs of the time sequence characteristics and evolution trend of the monitoring variable.

[0116] In the embodiment, the generation of the two-dimensional time sequence judgment matrix includes:

[0117] The local symbolic dynamic entropy sequence is received, the local symbolic dynamic entropy data points are classified and arranged according to the channel identifier of the structure response monitoring variable, and an entropy value time distribution column is constructed based on each time unit;

[0118] The structure response monitoring variable values with the added abnormality marked field are extracted from the structured filtered data sequence, and the ratio between the number of monitoring variable values marked as abnormal in each time unit and the number of all monitoring variable values in the corresponding time unit is calculated to form a corresponding abnormal data frequency value sequence, wherein the abnormal data frequency value is used to represent the short-time disturbance intensity;

[0119] The time labels in the local symbolic dynamic entropy data points are aligned with the time labels of the abnormal data frequency value sequence, and the row index of the two-dimensional time sequence judgment matrix is established based on the time unit, and the column index is established according to the channel identifier;

[0120] In the two-dimensional time sequence judgment matrix, each cell corresponds to a combination of a time label and a structure response monitoring variable channel, and the combination expression pair of the local symbolic dynamic entropy value corresponding to the time label and the corresponding abnormal data frequency value is filled in the cell to form a composite monitoring judgment cell with the entropy disturbance dimension and the frequency disturbance dimension;

[0121] The edge computing gateway node completes the data filling of all time indexes and channel indexes to generate a complete two-dimensional time sequence judgment matrix.

[0122] In this embodiment, the output of the digital early warning encoding signal includes:

[0123] A hierarchical risk encoding model is constructed based on the two-dimensional time sequence judgment matrix, and the hierarchical risk encoding model includes a risk level judgment rule library, a time label index module, a channel state evaluation module, and an encoding field generation module.

[0124] The risk level judgment rule library defines a plurality of risk level threshold intervals, each risk level threshold interval contains a combination range of a local symbolic dynamic entropy value interval and an abnormal data frequency value interval, and is used to distinguish the risk levels corresponding to disturbances of different intensities.

[0125] The time label index module extracts all composite monitoring judgment units of all channels corresponding to each time unit according to the row index in the two-dimensional time sequence judgment matrix, and inputs them into the channel state evaluation module.

[0126] The channel state evaluation module performs level mapping on the composite monitoring judgment units under each channel according to the risk level judgment rule library, obtains the corresponding channel risk level label, and counts the maximum risk level of all channels in the time unit.

[0127] The encoding field generation module constructs a digital early warning encoding signal containing a time field and a risk level field according to the time label of the time unit and the corresponding maximum risk level.

[0128] In this embodiment, the following modules are included:

[0129] A multi-source Internet of Things sensing system is used to deploy a sensing network in a construction area, and the sensing network includes a plurality of sensing units with edge processing capability, edge computing gateway nodes, and wireless communication modules. Each sensing unit collects structural response monitoring variables and generates original multi-dimensional time series data.

[0130] An edge data processing module is used to extract local disturbance features of the original multi-dimensional time series data, switch data processing modes and sampling strategies according to disturbance levels, and output data processing input sequences.

[0131] An anomaly identification and filtering module includes a Hampel filtering model module, which is used to perform sliding window processing on the data processing input sequences by channel, identify and correct anomalies based on the center reference value and the absolute median difference, and generate a structured filtering data sequence.

[0132] A symbolic modeling module is used to map the structured filtering data sequence to a symbolic trajectory sequence, construct a sliding symbolic mapping window, and generate a symbolic fragment with time index and channel identification.

[0133] a symbol dynamic entropy calculation module, configured to construct a sliding window based on the symbol trajectory sequence, calculate a local symbol probability distribution, and generate a local symbol dynamic entropy sequence;

[0134] a two-dimensional time sequence judgment matrix generation module, configured to align and combine the local symbol dynamic entropy sequence with the abnormal data frequency marked in the structured filtering data sequence, and construct a two-dimensional time sequence judgment matrix;

[0135] a hierarchical risk coding model module, configured to generate a digital early warning coding signal with a time field and a risk level field based on the two-dimensional time sequence judgment matrix;

[0136] a linkage output and regulatory interaction module, configured to output the digital early warning coding signal to a construction monitoring platform and a remote construction management terminal.

[0137] Embodiment 1:

[0138] In order to verify the feasibility of the application in implementation, the application is applied to a certain plot of super-deep diaphragm wall foundation pit engineering, the foundation pit excavation depth reaches 38.5 meters, the surrounding environment is complex, the underground structure is dense, and the construction interference risk is relatively high. In order to improve the construction dynamic supervision capability, the project adopts the super-deep diaphragm wall construction dynamic supervision system based on the Internet of Things.

[0139] A multi-source Internet of Things sensing network is deployed inside the construction control area of the project, and the network structure is composed of an edge computing gateway node, 48 sensing units and LoRa wireless communication modules. The edge computing gateway node is installed in the north side control shed, and has edge processing functions such as high-frequency cache, data analysis, model execution and field reconstruction. The 48 sensing units are arranged as follows: displacement sensing units are arranged at the top and middle of the east, west and south walls; inclination sensing units are arranged at the southwest corner and northeast corner support structures; acceleration sensing units are arranged at the center and sinking area of the foundation pit bottom plate; strain sensing units are arranged at the main reinforcement of the enclosure pile and the outer edge of the waterproof curtain; pore water pressure sensing units are arranged in the backfill soil outside the south wall; temperature sensing units are arranged at the pile body and heat source collection and distribution points in the construction area.

[0140] All sensing units establish a communication link with the edge computing gateway node through the LoRa module, and the system sets the sampling synchronization period to 60 seconds. The edge computing gateway node performs unified timestamp calibration on all collected structure response signals according to the description of claim 2, organizes them into original multidimensional time series data in the order of sampling, and writes them into the edge side cache.

[0141] The system further applies a data processing mode control mechanism based on disturbance level label switching, dynamically analyzes the change rate of the monitoring variables, and switches to an enhanced sampling strategy in real time. For example, during the support dismantling operation, if the change rate of the inclination angle exceeds the upper limit of the stable threshold range, the system will switch the basic sampling mode to the enhanced mode, automatically increase the sampling frequency to 1 time / 30 seconds, activate the local backup channel, and compress the upload period to 1 minute, achieving high-frequency disturbance capture.

[0142] The raw data is input into the structured Hampel filter model for anomaly identification. This model calculates the median and absolute median difference in the channel-level sliding window, identifies weak and strong abnormal data through a pre-set double-threshold parameter table, labels the fields or replaces them with a central reference value, and generates a structured filtered data sequence.

[0143] The system sets a sliding symbol mapping window for the structured filtered data sequence by channel, maps continuous numerical data into discrete symbol identifiers according to the partition interval, generates symbol fragments in the sliding window, and assembles them into a symbol trajectory sequence. The symbol sequence is input to drive the symbol dynamic entropy algorithm, which calculates the probability distribution of each type of symbol in the local sliding window, calculates the corresponding entropy value, and generates a local symbol dynamic entropy sequence with time index and channel field.

[0144] The abnormal frequency information marked in the Hampel filter is also counted by time unit, synchronized with the local symbol dynamic entropy, and a two-dimensional time sequence judgment matrix is constructed. Each cell of the matrix represents the combination of entropy disturbance value and abnormal frequency value of a channel at a certain time, which is read by the subsequent risk coding judgment module.

[0145] The system takes the two-dimensional matrix as input and calls the hierarchical risk coding model, which maps the entropy value and frequency value into a risk level label according to the pre-set rule library, and generates a digital warning code signal combined with the time index, such as "20250606-CH04-L2" representing moderate disturbance in channel 4. The coding signal is pushed to the monitoring platform by the edge node, and triggers the sound and light alarm and remote terminal display refresh simultaneously. The system operation data statistics are as follows:

[0146] Table 1 Structure response disturbance and digital warning statistics

[0147]

[0148] The table data shows that the maximum symbol dynamic entropy of channel CH03 reaches 0.94, and the abnormal frequency value is 0.48, which belongs to the case of higher fluctuation and abnormal concentration distribution, and is determined as L3 level high risk disturbance, and the system completes the early warning response within 5 seconds. The simultaneous occurrence of high entropy value and high frequency usually reflects the complex structure response behavior caused by construction disturbance, which has strong early warning significance. The maximum symbol dynamic entropy of CH05 is 0.78, and the abnormal frequency value is 0.33, which is determined as L2 level medium risk, indicating that local disturbance exists but has not reached the strong disturbance threshold, and the system also completes feedback within 5 seconds, proving that the model still has accurate identification ability under non-extreme conditions. The dynamic entropy value of channel CH08 decreases to 0.61, and the abnormal frequency value is only 0.25, corresponding to L1 level low risk disturbance, reflecting that the structure response of the channel is stable as a whole, and the disturbance is controllable. CH04 again presents high dynamic entropy (0.92) and frequency value (0.46), and the system successfully determines L3 level and outputs the early warning code, indicating that the model has high sensitivity and repeated identification ability. Finally, CH06 channel shows dynamic entropy of 0.84 and frequency value of 0.39, and is determined as L2 level, and the system completes processing within 5 seconds, showing stable early warning response delay. In summary, the system maintains rapid judgment and accurate coding under multiple channels and multiple disturbance intensities, has the ability of hierarchical perception and timely response to the trend of dynamic disturbance, and the data distribution and risk level output have high consistency, verifying the reliability and practicality of the proposed technical path.

[0149] Therefore, the embodiment comprehensively demonstrates the complete technical path from perception network construction, data analysis, sliding window judgment, filtering and removal, symbol mapping, entropy sequence construction to coding output, and provides a highly feasible technical scheme for real-time dynamic supervision of super deep foundation disturbance behavior.

[0150] The above describes only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can make equivalent replacement or change according to the technical scheme and inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for dynamically monitoring the construction of a diaphragm wall based on the Internet of Things, characterized in that, The application relates to a method for constructing a multi-source Internet of Things (IoT) sensing network covering a super-deep diaphragm wall construction site, deploying a sensing unit with edge processing capability, collecting construction structure response signals, generating original multi-dimensional time series data and completing edge side caching. Based on the local disturbance features of the original multi-dimensional time series data, a data processing mode is determined and a sampling strategy is switched, and a data processing input sequence corresponding to the construction condition is output. A Hampel filtering model with a local statistical threshold adjustment mechanism is applied to the data processing input sequence to identify and eliminate abnormal values, and a structured filtering data sequence is generated. The structured filtering data sequence is subjected to sliding window symbol mapping to construct a symbol trajectory sequence reflecting state evolution features. Symbol dynamic entropy calculation is performed based on the symbol trajectory sequence to generate a local symbol dynamic entropy sequence. The local symbol dynamic entropy sequence and abnormal data frequency information marked by the Hampel filtering model are combined to generate a two-dimensional time sequence judgment matrix. A hierarchical risk coding model is constructed based on the two-dimensional time sequence judgment matrix, and a digital early warning coding signal with a time identification field and a risk level field is output. The digital early warning coding signal is output to a super-deep diaphragm wall construction monitoring platform and a remote construction management terminal to trigger a linkage response mechanism and update construction safety supervision data. The generation of the two-dimensional time sequence judgment matrix comprises the following steps: Receiving the local symbol dynamic entropy sequence, sorting the local symbol dynamic entropy data points according to the channel identification of the structure response monitoring variables, and constructing an entropy value time distribution column based on each time unit. Extracting the structure response monitoring variable values with an abnormal mark field from the structured filtering data sequence, and calculating the ratio between the number of monitoring variable values marked as abnormal in each time unit and the total number of monitoring variable values in the corresponding time unit to form a corresponding abnormal data frequency value sequence. Aligning the time labels in the local symbol dynamic entropy data points with the time labels in the abnormal data frequency value sequence, and establishing the row index of the two-dimensional time sequence judgment matrix based on the time unit and the column index based on the channel identification. In the two-dimensional time sequence judgment matrix, each cell corresponds to a combination of a time label and a structure response monitoring variable channel, and the combination expression pair of the local symbol dynamic entropy value corresponding to the time label and the corresponding abnormal data frequency value is filled in the cell to form a complex monitoring judgment cell with an entropy disturbance dimension and a frequency disturbance dimension. The edge computing gateway node completes data filling of all time indexes and channel indexes to generate a complete two-dimensional time sequence judgment matrix. The generation of the original multi-dimensional time series data comprises the following steps:

2. The method according to claim 1, wherein, A multi-source IoT sensing network is constructed in the super-deep diaphragm wall construction area, and the multi-source IoT sensing network is composed of an edge computing gateway node, a plurality of sensing units and a wireless communication module, and the edge computing gateway node is deployed in the center of the construction control area. ​ The plurality of sensing units are configured according to structural positions and response types, specifically including: the sensing units for collecting displacement signals are arranged at the top and middle regions of the wall body, the sensing units for collecting inclination signals are arranged at the edge regions of the supporting structure, and the sensing units for collecting acceleration signals, strain signals, pore water pressure signals and temperature signals are arranged at the bottom of the foundation pit, the enclosure structure, the soil body adjacent to the pile foundation and the heat source sensitive region of the construction area, respectively; A stable communication link between the edge computing gateway node and each sensing unit is established through a wireless communication module, the edge computing gateway node sends a sampling instruction to each sensing unit according to a set synchronous sampling period, and each sensing unit synchronously collects the structural response signals at the position thereof; Each sensing unit returns the collected structural response signals to the edge computing gateway node in the form of a data packet, the edge computing gateway node performs timestamp extraction and unified time correction on the received data packet, and constructs a structural response data set based on a unified time reference; The edge computing gateway node integrates and processes the structural response data set, organizes the original multidimensional time series data in the order of sampling time, and writes the original multidimensional time series data into a local cache area in the edge computing gateway node.

3. The method according to claim 1, characterized in that, The output of the data processing input sequence includes: The edge computing gateway node performs sliding window analysis on the original multidimensional time series data, and extracts the rate of change parameter of the structural response monitoring variable; Based on the deviation between the rate of change parameter of the structural response monitoring variable and the corresponding stable threshold, a disturbance level label is generated, and a target data processing mode is selected in a preset sampling strategy mapping table according to the disturbance level label; The data processing mode includes a basic data processing mode and an enhanced data processing mode, wherein the basic data processing mode defines a standard sampling frequency, a default upload period, a fixed channel range and a conventional sensing unit activation state, and the enhanced data processing mode defines an improved sampling frequency, a compressed upload period, an extended channel range and an activation configuration of a local supplementary sensing unit; According to the parameter content of the target data processing mode, the corresponding sensing unit is controlled to perform structural response signal sampling, and the structural response signal collected is added with a data processing mode label field, a sampling parameter field and a spatial channel identification field to generate a structured sampling data unit; According to the time sequence and field consistency rule, the plurality of structured sampling data units are integrated to generate a data processing input sequence.

4. The method according to claim 1, characterized in that, The generation of the structured filtering data sequence includes: The edge computing gateway node divides the structural response monitoring variables in the data processing input sequence into a plurality of channel data sets according to the physical quantity type, and configures an independent sliding window for each channel data set, the length of the sliding window being dynamically determined from a disturbance level-window length mapping table according to the disturbance level label of the last period; In each sliding window, the median value of the structural response monitoring variables in the corresponding sliding window is calculated as a center reference value, and based on the center reference value, the absolute deviation of all data points in the sliding window is calculated, and the median of the absolute deviation is extracted as the absolute median deviation of the corresponding sliding window; The first discrimination threshold and the second discrimination threshold of the channel are loaded from a preset channel threshold parameter table according to the channel type, and a center reference value and an absolute median difference are used as a basis for calculation, if the absolute deviation between the structural response monitoring variable value and the center reference value exceeds the second discrimination threshold multiplied by the absolute median difference, it is identified as a strong abnormal value; if the deviation of the structural response monitoring variable value is between the first discrimination threshold and the second discrimination threshold multiplied by the absolute median difference, it is identified as a weak abnormal value; otherwise, it is determined to be a normal value; The structural response monitoring variable value identified as a strong abnormal value is replaced by the center reference value of the current sliding window, and the structural response monitoring variable value identified as a weak abnormal value is added with an abnormal mark field, and the original data of the normal value is kept unchanged; The processed structural response monitoring variable values in all channels are time axis aligned, and channel identification fields and abnormal mark fields are added to form a structured filtering data sequence.

5. The method according to claim 1, wherein, The construction of the symbol trajectory sequence includes: The edge computing gateway node establishes a sliding symbol mapping window for each type of structural response monitoring variable in the structured filtering data sequence, the length of the sliding symbol mapping window is set according to the sampling frequency of the monitoring variable, and the sliding step is 1; In each sliding symbol mapping window, the structural response monitoring variable values corresponding to the continuous time points in the current window are extracted, and each structural response monitoring variable value in the sliding symbol mapping window is subjected to symbol mapping processing according to a preset symbol division rule, the symbol division rule divides the value range into a plurality of non-overlapping intervals, and each value interval corresponds to a unique symbol identifier; All symbol identifiers generated in the sliding symbol mapping window are arranged in time sequence and combined to generate a symbol segment, and a time index field and a channel identification field of the structural response monitoring variable are added in the symbol segment; After the sliding symbol mapping window completes a sliding, the sliding symbol mapping window is continuously advanced in time sequence, and the symbol mapping and segment combination steps are repeatedly executed to gradually generate a plurality of symbol segments; All symbol segments are integrated and reorganized in the generation time sequence to form a complete symbol trajectory sequence.

6. The method according to claim 1, wherein, The generation of the local symbol dynamic entropy sequence includes: A symbol dynamic entropy calculation sliding window is constructed for each type of structural response monitoring variable in the symbol trajectory sequence, the length of the symbol dynamic entropy calculation sliding window is set according to the sampling frequency and variation characteristics of the corresponding structural response monitoring variable, and the sliding step is set to one time unit; In each symbol dynamic entropy calculation sliding window, a time-continuous symbol identifier sequence is extracted, all symbol identifier types appearing in the current window are identified, and the ratio between the occurrence number of each symbol identifier and the total number of symbol identifiers is calculated to form a symbol probability distribution corresponding to the current window; Based on the probability values of the symbol identifiers in the current symbol probability distribution, the symbol dynamic entropy value of the symbol dynamic entropy calculation sliding window is calculated, and the symbol dynamic entropy value is bound with the time label of the current sliding window and the channel identification of the structural response monitoring variable to generate a local symbol dynamic entropy data point; The symbol dynamic entropy calculation sliding window is slid forward along the time axis by one time unit, and the symbol extraction, probability construction, entropy value calculation and data point generation steps are repeatedly performed to gradually generate multiple local symbol dynamic entropy data points; All local symbol dynamic entropy data points are combined in time sequence to form a local symbol dynamic entropy sequence.

7. The method according to claim 1, wherein the method is characterized by, The output of the digital early warning coding signal includes: A hierarchical risk coding model is constructed based on a two-dimensional time sequence judgment matrix, which includes a risk level judgment rule base, a time tag index module, a channel state evaluation module, and a coding field generation module; The risk level judgment rule base defines multiple risk level threshold intervals, each risk level threshold interval containing a combination range of local symbol dynamic entropy value intervals and abnormal data frequency value intervals; The time tag index module extracts all composite monitoring judgment units of all channels corresponding to each time unit according to the row index in the two-dimensional time sequence judgment matrix, and inputs them into the channel state evaluation module; The channel state evaluation module maps the composite monitoring judgment units under each channel according to the risk level judgment rule base to obtain the corresponding channel risk level label, and counts the maximum risk level of all channels in the time unit; The coding field generation module constructs a digital early warning coding signal containing a time field and a risk level field according to the time tag and the corresponding maximum risk level of the time unit.

8. An Internet of Things-based diaphragm wall construction dynamic supervision system, characterized in that, It includes the following modules: A multi-source Internet of Things sensing system is used to deploy a sensing network in a construction area, which includes multiple sensing units with edge processing capability, edge computing gateway nodes and wireless communication modules, each sensing unit collects structural response monitoring variables and generates original multi-dimensional time series data; An edge data processing module is used to extract local disturbance features of the original multi-dimensional time series data, switch data processing modes and sampling strategies according to disturbance levels, and output data processing input sequences; An anomaly identification and filtering module includes a Hampel filtering model module, which is used to perform sliding window processing on the data processing input sequences by channel, identify and correct anomalies based on the central reference value and absolute median deviation, and generate a structured filtering data sequence; A symbol modeling module is used to map the structured filtering data sequence to a symbol trajectory sequence, construct a sliding symbol mapping window, and generate a symbol segment with time index and channel identification; A symbol dynamic entropy calculation module is used to construct a sliding window based on the symbol trajectory sequence, calculate local symbol probability distribution, and generate a local symbol dynamic entropy sequence; A two-dimensional time sequence judgment matrix generation module is used to align and combine the local symbol dynamic entropy sequence and the abnormal data frequency marked in the structured filtering data sequence to construct a two-dimensional time sequence judgment matrix, wherein the generation of the two-dimensional time sequence judgment matrix includes: Receiving the local symbol dynamic entropy sequence, sorting the local symbol dynamic entropy data points by channel identification of the structural response monitoring variable, and constructing an entropy value time distribution column based on each time unit; Extract the structural response monitoring variable values with added abnormality flag field from the structured filtering data sequence, and count the ratio between the number of monitoring variable values marked as abnormal and the total number of monitoring variable values in each time unit to form a corresponding abnormal data frequency value sequence; Align the time labels in the local symbolic dynamic entropy data points with the time labels of the abnormal data frequency value sequence, and establish the row index of the two-dimensional time sequence judgment matrix based on the time unit, and establish the column index according to the channel identifier; In the two-dimensional time sequence judgment matrix, each cell corresponds to a combination of a time label and a structural response monitoring variable channel, and the cell is filled with the combination expression pair of the local symbolic dynamic entropy value corresponding to the time label and the corresponding abnormal data frequency value, forming a composite monitoring judgment cell with entropy disturbance dimension and frequency disturbance dimension; The edge computing gateway node completes the data filling of all time indexes and channel indexes to generate a complete two-dimensional time sequence judgment matrix; The hierarchical risk coding model module is used to generate a digital early warning coding signal with a time field and a risk level field based on the two-dimensional time sequence judgment matrix; The linkage output and supervision interaction module is used to output the digital early warning coding signal to the construction monitoring platform and the remote construction management terminal.

Images