Building retard-bonded prestress intelligent detection method based on Internet of Things

Abstract

The invention relates to the technical field of building detection, and discloses a building retard-bonded prestress intelligent detection method based on the Internet of Things. The method comprises the steps of collecting and preprocessing real-time monitoring data of the retard-bonded prestressed structure through an Internet of Things platform, and filtering interference signals to improve data quality; classifying the preprocessed data according to a preset stress state interval, and extracting time-space domain features of each interval to generate a stress feature set; detecting the change frequency of the stress state interval, and starting a dynamic adjustment mechanism to mark an abnormal interval if the change frequency exceeds a threshold value; according to the matching degree of the historical damage mode of the structure and the current feature, a non-abnormal interval stress feature set is decomposed to generate a damage feature subset, multi-parameter collaborative correction is carried out in combination with the incidence relation of material mechanics and environmental action parameters, and finally the position of a potential damage source is reversely traced along a damage evolution path. And determining a maintenance priority and generating a maintenance instruction.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of building detection, and in particular to a building slow-bonding prestress intelligent detection method based on Internet of Things. BACKGROUND

[0002] In the field of modern building engineering, slow-bonding prestress structures are widely used in important infrastructures such as large-span bridges, high-rise buildings and large venues due to their characteristics of considering construction convenience and structural durability. The stress state of such structures directly determines the safety performance and service life of the whole building. Once the stress distribution is abnormal or hidden damage occurs, if it is not found and handled in time, it may cause serious safety accidents such as structure cracking, load capacity reduction and even collapse, which poses a great threat to life and property safety. At present, the detection means for slow-bonding prestress structures mainly rely on traditional manual detection. Detection personnel need to use professional instruments to collect data on site, which not only consumes a lot of manpower, material resources and time cost, but also has the problems of long detection period and limited coverage, and it is difficult to realize real-time dynamic monitoring of the overall stress state of the structure. Although sensors are introduced in some projects for data collection, the existing monitoring system can only realize simple recording and transmission of a single parameter, and lacks effective preprocessing mechanism for collected data, so that the mixed environmental interference signals and equipment noise in the data cannot be effectively filtered, which directly affects the accuracy of subsequent data analysis results. The existing detection method has obvious deficiencies in stress state analysis. It cannot preset a reasonable stress state interval according to the stress characteristics of the slow-bonding prestress structure, cannot accurately classify the state of the structure under different stress levels, and thus cannot extract the spatiotemporal features reflecting the real stress state of the structure. When the stress state of the structure fluctuates frequently, the system cannot timely identify such abnormal changes and start the corresponding adjustment mechanism, which easily misses the best opportunity for early damage warning. In addition, in the process of structure damage diagnosis, the existing method mostly ignores the relevance between the historical damage pattern of the structure and the current monitoring data, and only relies on a single parameter or local data for damage judgment, which leads to incomplete damage feature extraction and does not consider the mutual influence between the mechanical parameters of the structure material and the environmental action parameters (such as temperature, humidity, load change, etc.), so that the damage feature cannot be scientifically corrected, the damage source positioning accuracy is low, the repair priority cannot be accurately determined, and finally the efficiency and effect of repair work are affected, which cannot meet the high-precision, real-time and intelligent requirements of modern building engineering for slow-bonding prestress structure safety monitoring. SUMMARY

[0003] The application aims to provide a building slow-bonding prestress intelligent detection method based on Internet of Things to solve the problems in the background art.

[0004] To achieve the above object, the application provides a building slow-bonding prestress intelligent detection method based on the Internet of Things, which comprises the following steps: Collecting real-time monitoring data of the slow-bonding prestress structure through the Internet of Things platform and preprocessing the real-time monitoring data; Classifying the preprocessed real-time monitoring data according to a preset stress state interval, extracting the space-time domain features in each stress state interval, and generating a stress feature set; Detecting the change frequency of the stress state interval, and if the change frequency exceeds a change frequency threshold, starting a dynamic adjustment mechanism to mark an abnormal stress state interval; According to the matching degree calculation result of the structure historical damage mode and the current space-time domain features, decomposing the stress feature set of the non-abnormal stress state interval to generate a structure damage feature subset; Based on the correlation between the structure material mechanics parameters and the environmental action parameters, the structure damage feature subset is subjected to multi-parameter collaborative correction; Based on the structure damage feature subset along the structure damage evolution path, the potential damage source position is traced back in reverse, the maintenance priority is determined, and the maintenance instruction is generated.

[0005] Preferably, the collecting of the real-time monitoring data of the slow-bonding prestress structure through the Internet of Things platform and the preprocessing of the real-time monitoring data comprise the following steps: Synchronously collecting strain data, displacement data and environmental temperature and humidity data of the slow-bonding prestress structure, and time-synchronously aligning the strain data, displacement data and environmental temperature and humidity data; Noise filtering is performed on the time-synchronously aligned strain data, displacement data and environmental temperature and humidity data, which includes removing dynamic interference components in the strain data by using an adaptive filtering method, and eliminating abnormal data points based on the correlation between the environmental temperature and humidity data and the displacement data; The filtered strain data, displacement data and environmental temperature and humidity data are reorganized at a preset time interval to generate a standardized monitoring data set.

[0006] Preferably, the classification of the preprocessed real-time monitoring data according to the preset stress state interval, the extraction of the space-time domain features in each stress state interval, and the generation of the stress feature set comprise the following steps: The standardized monitoring data set is divided into a plurality of stress state intervals according to the stress level, and the stress state interval includes a low stress interval, a medium stress interval and a high stress interval; The space-time domain features of the strain data, displacement data and environmental temperature and humidity data in each stress state interval are extracted respectively; The space-time domain features in the same stress state interval are aggregated according to the spatial position to generate a multi-dimensional stress feature set.

[0007] Preferably, the time-space-domain features of the strain data include strain distribution gradient, time-domain fluctuation amplitude and spatial variation coefficient, the time-space-domain features of the displacement data include displacement change rate and displacement accumulation, and the time-space-domain features of the environmental temperature and humidity data include temperature and humidity coupling coefficient and environmental action intensity.

[0008] Preferably, the change frequency of the stress state interval is detected, and if the change frequency exceeds a change frequency threshold, a dynamic adjustment mechanism is started to mark an abnormal stress state interval, including: The number of changes of the stress state interval is counted based on a preset monitoring period, the change frequency per unit time is calculated, and it is determined whether the change frequency exceeds a change frequency threshold; If the change frequency exceeds the change frequency threshold, the time-space-domain features of adjacent stress state intervals are aligned through a space-time registration method, and a cross-interval consistency index is extracted; If the cross-interval consistency index is lower than a consistency threshold, the stress state interval corresponding to the current monitoring period is marked as an abnormal stress state interval, and re-extraction of the time-space-domain features in the abnormal stress state interval is triggered; otherwise, it is marked as a non-abnormal stress state interval.

[0009] Preferably, the extraction of the cross-interval consistency index includes: aligning the time-space-domain features of adjacent stress state intervals, calculating similarity measures of the aligned strain distribution gradient, displacement change rate and temperature and humidity coupling coefficient respectively, and taking the weighted average of the similarity measures as the cross-interval consistency index.

[0010] Preferably, the stress feature set of the non-abnormal stress state interval is decomposed according to the matching degree calculation result of the structure historical damage mode and the current time-space-domain feature, to generate a structure damage feature subset, including: For the current time-space-domain feature in the stress feature set of the non-abnormal stress state interval, it is determined according to the matching degree calculation result of the structure historical damage mode label and the current time-space-domain feature, whether the current time-space-domain feature matches the structure historical damage mode; If the matching degree is greater than or equal to a preset matching threshold, a spatial frequency band decomposition mode is used to extract a structure damage feature subset from the stress feature set of the non-abnormal stress state interval; If the matching degree is less than the preset matching threshold, a spatial frequency band that responds significantly is selected for focused decomposition based on the correlation between the material nonlinear response of the strain distribution and the resonance characteristics of the structure dynamic characteristics in the stress feature set of the non-abnormal stress state interval, to generate a structure damage feature subset.

[0011] Preferably, the structure damage feature subset is subjected to multi-parameter collaborative correction based on the correlation between the structure material mechanics parameters and the environmental action parameters, including: According to the constitutive relationship of the elastic modulus and the strain response in the structural material mechanics parameters, stiffness characteristics are corrected for the strain distribution in the structural damage feature subset; Based on the aging relationship of the humidity diffusion coefficient and the temperature and humidity coupling in the environmental action parameters, environmental effect compensation correction is performed on the temperature and humidity coupling coefficient in the structural damage feature subset; The corrected strain distribution and temperature and humidity coupling coefficient are fused with the displacement cumulative quantity feature to generate a multi-parameter cooperatively corrected structural damage feature subset.

[0012] Preferably, the potential damage source position is traced along the structural damage evolution path based on the corrected structural damage feature subset, the maintenance priority is determined, and the maintenance instruction is generated, including: The structural damage evolution path is obtained, and the structural damage evolution path is constructed based on the structural connection relationship and the internal force transmission path, and contains the damage propagation direction and the influence weight between the structure positions; Abnormal features are extracted from the multi-parameter cooperatively corrected structural damage feature subset, and the corresponding positions in the structural damage evolution path are matched according to the abnormal feature types; The potential damage source position of the abnormal feature is located by traversing in the damage propagation direction of the structural damage evolution path in reverse; The maintenance priority is determined according to the number and distribution depth of the abnormal features of the potential damage source position; Based on the maintenance priority and the remaining safe life of the structure, a maintenance instruction containing the damage source position identifier and the maintenance time is generated.

[0013] Preferably, after the maintenance priority is determined according to the number and distribution depth of the abnormal features of the potential damage source position, the method further includes: The time-varying characteristics of the abnormal features of each potential damage source position are analyzed, and the feature degradation rate is extracted; The maintenance priority is adjusted according to the feature degradation rate and the damage tolerance value in the structural material mechanics parameters; The maintenance instruction is updated based on the adjusted maintenance priority.

[0014] Compared with the prior art, the beneficial effects of the present application are: Compared with traditional manual detection and single sensor monitoring, the overall collection of real-time monitoring data of the slow-bonding prestressed structure is realized through the Internet of Things platform, which greatly expands the coverage range and real-time performance of data collection, and effectively reduces the labor and time cost without the need for detection personnel to frequently operate on site. At the same time, the stress changes of the structure under different working conditions can be captured in real time, providing comprehensive and continuous data support for subsequent analysis. In the data processing link, the real-time monitoring data is preprocessed, which can effectively filter out the useless information such as environmental interference and equipment noise in the data, improve the data quality, avoid the deviation of the subsequent analysis results caused by the poor quality of the original data, and ensure that the stress features extracted based on the data can truly reflect the actual stress state of the structure. According to the preset stress state interval, the preprocessed data is classified, and the spatiotemporal features in each interval are extracted to generate a stress feature set, which can accurately analyze the characteristics of different stress levels of the structure, and clearly show the state difference of the structure in different stress stages, providing more targeted feature basis for subsequent damage diagnosis. Compared with the data analysis method without classification mechanism in the existing method, the effectiveness and accuracy of feature extraction are significantly improved. By detecting the change frequency of the stress state interval, when the change frequency exceeds the threshold, the dynamic adjustment mechanism is started to mark the abnormal stress state interval, which can timely identify the abnormal fluctuation of the stress state of the structure, quickly lock the area that may have safety hazards, avoid delay in damage warning due to failure to find abnormalities in time, and reduce the risk of safety accidents. According to the matching degree calculation result of the historical damage mode of the structure and the current spatiotemporal features, the stress feature set of the non-abnormal stress state interval is decomposed to generate a structure damage feature subset, which fully utilizes the past damage data and experience of the structure, makes the damage feature extraction more in line with the actual damage evolution law of the structure, avoids the one-sidedness of damage features caused by relying only on current data, and improves the comprehensiveness and reliability of damage features. Based on the correlation between the mechanical parameters of the structure material and the environmental action parameters, the structure damage feature subset is modified in multiple parameters, which considers the joint influence of the material properties and external environmental factors on the structure damage state, for example, temperature changes may change the mechanical properties of the material, and then affect the stress distribution and damage development of the structure. Through collaborative correction, the interference caused by these factors can be eliminated, and the damage features are closer to the actual damage of the structure, providing accurate basis for subsequent damage source positioning. Based on the modified structure damage feature subset, the potential damage source position is traced along the structure damage evolution path, which can accurately locate the specific position of the damage occurrence, avoid the problem of fuzzy damage positioning in traditional detection, and determine the repair priority and generate repair instructions according to the position and damage degree of the damage source. It can guide the repair personnel to carry out work in an orderly manner, avoid the low efficiency caused by blind repair, ensure that the repair resources can be preferentially invested in key damage positions, improve the pertinence and effectiveness of repair work, and ensure the safety performance and service life of the slow-bonding prestressed structure. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The working principle diagram of the building slow-bonding prestressed intelligent detection method based on the Internet of Things is described. Figure 2Flow chart for real-time monitoring data collection and preprocessing; Figure 3 Flow chart for multi-parameter collaborative correction. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0017] Please refer to Figure 1 The present application provides a building retard-bond prestress intelligent detection method based on the Internet of Things, which comprises the following steps: through the Internet of Things sensing nodes deployed at the key nodes of the building retard-bond prestress structure, real-time collection of strain data, displacement data and environmental temperature and humidity data of the structure, and transmission of these data to an Internet of Things platform for centralized processing; the Internet of Things platform performs preprocessing on the received real-time monitoring data, including time synchronization alignment, noise filtering and data reorganization, to generate a standardized monitoring data set; then, the standardized monitoring data set is classified according to a preset stress state interval, divided into a low stress interval, a medium stress interval and a high stress interval, and the spatiotemporal domain features in each interval are extracted to form a multi-dimensional stress feature set; the system continuously monitors the change frequency of the stress state interval, and when the change frequency exceeds a preset threshold, a dynamic adjustment mechanism is started, the features of adjacent intervals are aligned through a space-time registration method, and a cross-interval consistency index is calculated, and if the index is lower than a threshold, it is marked as an abnormal stress state interval; for a non-abnormal stress state interval, the system calculates the matching degree of the current spatiotemporal domain features and the historical damage pattern of the structure, generates a structure damage feature subset according to the matching result using a spatial frequency band decomposition mode or a focused decomposition mode; based on the correlation between the mechanical parameters of the structure material and the environmental action parameters, the structure damage feature subset is subjected to multi-parameter collaborative correction, including stiffness characteristic correction and environmental effect compensation correction; finally, based on the corrected feature subset, the location of a potential damage source is traced back along a pre-constructed structure damage evolution path, the repair priority is determined in combination with the number, distribution depth and feature degradation rate of abnormal features, and an instruction containing the location identification and repair time is generated.

[0018] Embodiment 1: Please refer to Figure 2The Internet of Things platform synchronously collects monitoring data of the slow-bonding prestressed structure through distributed deployment of strain sensors, displacement sensors and temperature and humidity sensors. High-precision clock synchronization protocol is used in the collection process to ensure consistency of the data timestamps of all sensor nodes, and time synchronization alignment of strain data, displacement data and environmental temperature and humidity data is achieved. The synchronization protocol is realized based on the IEEE 1588 precise time protocol, and through deployment of transparent clock devices at the network switch level, the link delay fluctuation in the data transmission process is compensated, so that the sampling time deviation of all sensor nodes is controlled within milliseconds. The sensor nodes are designed in a modular manner, each node including a signal conditioning circuit, an analog-to-digital converter and a network interface. The strain sensor is composed of a resistance strain gauge and a Wheatstone bridge, the displacement sensor uses the laser ranging principle, and the temperature and humidity sensor uses a digital integrated sensing element. All sensor nodes are connected to the data collection gateway through industrial Ethernet, and the gateway has a built-in FPGA chip to realize parallel data reception and timestamp marking functions, ensuring that the massive monitoring data has strict time sequence characteristics at the collection end.

[0019] The aligned data is subjected to adaptive filtering processing, and the filtering method dynamically adjusts the filtering parameters according to the frequency domain characteristics of the signal to effectively remove dynamic interference components in the strain data. The adaptive filtering algorithm uses the LMS least mean square algorithm as the core, and updates the filter coefficients by real-time calculation of the correlation between the reference signal and the main signal, wherein the reference signal comes from the environmental vibration sensor data unrelated to the structure. The filtering process is divided into two stages, the first stage designs a band-stop filter for high-frequency noise, and the stop-band frequency range is set according to 1.5 times the natural frequency of the structure, effectively eliminating harmonic interference caused by the operation of electromechanical equipment; the second stage uses a sliding average filter to process low-frequency drift, and the sliding window length is dynamically adjusted according to the signal sampling rate, and the number of data points in the window is kept in the range of 200-500 sampling points. For correlation analysis of environmental temperature and humidity data and displacement data, a temperature and humidity-displacement coupling model is established, which is based on historical data to obtain the transfer function of temperature and humidity change and displacement response. When the temperature and humidity in real-time monitoring data suddenly changes and the displacement response deviates from the predicted value of the transfer function by more than three standard deviations, the data point is determined as an abnormal point and is removed.

[0020] The data after filtering out the noise is reorganized according to the preset 5-minute time interval, the continuous data stream is divided into equal-length data segments, and the data in each data segment is normalized. The data reorganization process uses a ring buffer technology, and the buffer capacity can accommodate 5 minutes of sampling data. When the buffer is full, the data processing thread is triggered. The normalization process uses the Z-score standardization method, which processes the data of each sensor channel separately. After subtracting the mean value of the channel in the buffer and dividing by the standard deviation, all sensor data is converted to dimensionless values. The processed data segment adds a timestamp index and a sensor position code to form a structured data packet. The data packet is packaged in JSON format, including packet header identification, data body length verification and data content. The standardized monitoring data set is finally stored in a time series database. The database uses a hierarchical storage architecture, with hot data stored in an in-memory database for real-time analysis, and cold data stored in a distributed file system for long-term storage.

[0021] The hardware platform of the data acquisition system uses an industrial-grade embedded device, and the main controller uses a multi-core ARM processor. It is equipped with dual-gigabit network interfaces to realize parallel data transmission. The sensor node power supply uses POE power supply mode, which transmits data and power through network cable at the same time, reducing the complexity of wiring. The system software platform is developed based on the Linux real-time kernel, and the data acquisition program uses a multi-threaded architecture. Each sensor node is assigned an independent data reception thread, and the thread priority is set according to the real-time requirements of the data. The data preprocessing algorithm is written in C++ and compiled into a dynamic link library for the main program to call and execute. The intermediate variables and state information during algorithm execution are saved in shared memory to ensure data processing continuity.

[0022] The network architecture of the monitoring system is divided into three layers. The sensor layer is composed of various sensor nodes connected into a local area network through a switch. The transmission layer uses industrial routers to forward local area network data to the cloud platform. The platform layer is deployed on the cloud server and runs data storage and analysis services. Network communication uses the MQTT protocol publish / subscribe mode. Sensor nodes act as publishers to send data to the message broker server, and data processing programs act as subscribers to obtain data from the server. Communication data packets are added with digital signatures and encryption protection to prevent data tampering and leakage.

[0023] The system implements continuous self-diagnosis function during operation, regularly checks the working state of the sensor node, monitors the online state of the node through the heartbeat packet mechanism, alarms the offline node and tries to reconnect. The data quality evaluation module calculates the signal-to-noise ratio and data integrity rate in real time, and triggers the data re-sampling or compensation mechanism when the index is lower than the threshold. All operation logs and system state information are recorded in the system log, which supports query by time range and event type, and provides basis for system maintenance. Before the system is deployed and implemented, on-site survey is needed to determine the sensor installation position and wiring path, and the installation position should be selected to avoid obstacles and strong electromagnetic interference sources. The sensor is fixed by a special fixture to ensure that the sensor is in close contact with the measured structure, and the calibration test is performed after installation to verify the measurement accuracy and reliability. Joint testing is performed during system debugging to verify the data acquisition synchronization and preprocessing effect, and the algorithm parameters are adjusted to the best state before formal operation.

[0024] In embodiment 2, the standardized monitoring data set is divided into a low stress interval, a medium stress interval and a high stress interval according to a preset stress threshold. The stress threshold is set according to the allowable stress range in the structure design specification, and the specific division standard refers to the stress limit value of the prestressed concrete structure in the Concrete Structure Design Specification GB50010. The low stress interval corresponds to the stress level of the structure under normal working condition, which is usually set to 0-40% of the design allowable stress; the medium stress interval reflects the stress state of the structure under large load, which is 40%-70% of the design allowable stress; the high stress interval indicates that the structure is close to or reaches the design limit state, which is 70%-100% of the design allowable stress. The boundary value of each interval is adjusted according to the structure characteristics of the specific project, considering the influence of material aging, environmental factors, etc., and a proper buffer zone is set to avoid frequent jumping at the interval boundary.

[0025] In each stress state interval, three types of spatio-temporal features are extracted from the strain data: strain distribution gradient reflects the strain change rate between adjacent measuring points, which is obtained by calculating the difference value of spatially adjacent sensor measuring point data at the same time point, and the difference direction is along the arrangement direction of the prestressed tendon; the time domain fluctuation amplitude represents the fluctuation characteristics of strain data in time dimension, which is calculated by using the standard deviation of each sensor measuring point in the current time window, and the time window length is consistent with the data reorganization interval; the spatial variation coefficient reflects the dispersion degree of strain data in spatial distribution, which is obtained by calculating the variation coefficient of all sensor measuring point data at the same time point, that is, the ratio of standard deviation to mean value. These strain features describe the stress state of the structure from different dimensions, the strain distribution gradient reveals the stress concentration phenomenon, the time domain fluctuation amplitude reflects the dynamic load action characteristics, and the spatial variation coefficient reflects the uniformity of the overall stress of the structure.

[0026] Two types of features are extracted from displacement data: displacement change rate is obtained by calculating the first-order time derivative of displacement data of each measuring point, and the instantaneous change rate is calculated by using the central difference method, which effectively reflects the deformation speed of the structure; displacement accumulation is calculated by time integration of displacement data of each measuring point, and the integration time starts from the beginning of the monitoring period, and the cumulative value reflects the permanent deformation development trend of the structure. Displacement change rate can be used to identify sudden deformation, and displacement accumulation can be used to evaluate long-term deformation accumulation. The feature extraction of environmental temperature and humidity data includes temperature and humidity coupling coefficient and environmental action intensity: the temperature and humidity coupling coefficient is calculated by using the Pearson correlation coefficient calculation method, and the correlation coefficient of temperature and humidity data in the same time series is calculated, and the coefficient value ranges from -1 to 1, and a positive value indicates that the temperature and humidity change in the same direction, and a negative value indicates that the temperature and humidity change in the opposite direction; the environmental action intensity is obtained by calculating the weighted Euclidean norm of temperature and humidity data, and temperature and humidity are respectively given different weight coefficients, and the weight value is determined according to the material temperature and humidity expansion coefficient. The temperature and humidity coupling coefficient reflects the cooperative change characteristics of environmental parameters, and the environmental action intensity quantifies the comprehensive influence degree of environmental factors on the structure.

[0027] All extracted spatiotemporal features are aggregated according to the spatial position of the sensor to form a multi-dimensional stress feature set with spatial coordinates as the core. The aggregation process is based on the topological structure of the sensor arrangement, and the feature data of each spatial position point contains all the feature quantities of the point, and the feature data of adjacent points is recorded to facilitate spatial correlation analysis. The feature set is stored in the form of a matrix, the row dimension corresponds to the spatial position point, the column dimension corresponds to different feature types, and the time dimension is represented by multiple time slice matrices. The spatial position code uses a global coordinate system, and the coordinates of each sensor point are accurately recorded to facilitate the visualization analysis of spatial features. The feature extraction algorithm adopts modular design, and each feature type corresponds to an independent calculation module, and the modules communicate with each other through a standard data interface. The calculation process fully utilizes the advantages of parallel computing, and the feature calculation tasks of different sensor points are allocated to multiple computing cores for simultaneous execution. The feature data storage uses a column-oriented database, which facilitates fast retrieval and analysis by feature type. The feature extraction process implements quality monitoring, range checking and consistency verification are performed on each feature value, and abnormal feature values are marked and trigger the review mechanism.

[0028] The aggregation of spatio-temporal features considers the temporal consistency of data, ensuring the time synchronization of feature data within the same time slice. Spatial interpolation techniques are used to handle individual sensor missing cases, and inverse distance weighting method is used to interpolate and complete the feature data based on adjacent sensors. Standardization of feature data eliminates the influence of dimension, making different features comparable. The aggregated feature set establishes version management, records the parameter settings and calculation timestamps of feature extraction, and ensures the traceability of feature data. The application of feature set supports various analysis needs, real-time monitoring stage uses sliding window method to update the latest feature data, and historical data analysis stage supports query feature data by time range. Visualization of feature data uses multi-dimensional data rendering technology, supports feature distribution cloud map display based on spatial location, time domain feature change curve display, and feature correlation matrix visualization. Feature data export supports standard data format, which is convenient for third-party analysis tools for further processing. The system provides feature management interface, allows users to customize feature extraction parameters, adjusts feature calculation method and aggregation method. Feature calculation performance is continuously optimized, using memory computing technology to improve the processing speed of large-scale feature data, and using data compression technology to reduce feature storage space occupation. Feature data security protection uses access control mechanism, different users set different feature data access permissions. Feature data analysis results are seamlessly integrated with the structural health state evaluation module.

[0029] In the embodiment 3, the system counts the number of changes in the stress state interval in a monitoring period of 1 hour, calculates the change frequency in unit time, and the frequency value reflects the fluctuation characteristics of the structure stress state. The monitoring period is set according to the structure characteristics and engineering requirements, and can be fixed period or adaptive adjustment mode. The fixed period mode is simple and easy to implement, and the adaptive mode dynamically adjusts the period length according to the data fluctuation degree. The change number statistics is based on the transition sequence of the stress state interval, and each interval jump records a change event, including various transition types such as from low to medium, from medium to high, from high to medium, from medium to low. The change frequency calculation uses the average value of the number of changes in unit time, and the frequency value exceeding the set threshold value indicates that the structure may be in an abnormal working state. When the change frequency exceeds the set threshold value, the dynamic adjustment mechanism is started, which includes two main steps of feature alignment and consistency check. The trigger condition of dynamic adjustment considers the duration and amplitude of the frequency value, and only when the frequency value exceeds the limit continuously and exceeds a certain amplitude, the adjustment is started to avoid accidental fluctuation caused by false triggering. In the adjustment process, the space-time domain features of adjacent stress state intervals are aligned through space-time registration method. The space-time registration uses a feature point-based matching algorithm, which identifies feature points with significant characteristics in two intervals as reference points. The feature point selection considers the significance and stability of the feature, and preferentially selects feature points with large gradient value and obvious fluctuation as the registration reference. The registration process adjusts the time delay and spatial offset, the time delay compensation adjusts the time sequence by interpolation method, and the spatial offset correction unifies the spatial reference system by coordinate transformation method.

[0030] After alignment, the cross-interval consistency index is extracted, which evaluates the rationality of state transition by calculating the similarity of adjacent interval features. The calculation formula of the cross-interval consistency index is:

[0031] Wherein: represents the cross-interval consistency index, is the number of feature types, is the weight coefficient of the th feature type, represents the similarity calculation function, and represent the values of the th feature in the adjacent time interval. The weight coefficient is assigned according to the importance of the feature, and the feature with high importance is given a larger weight, and the sum of all weight coefficients is 1. The similarity calculation function uses different calculation methods for different types of features. The cosine similarity is used for continuous numerical features, the Jaccard similarity coefficient is used for classification features, and the Hausdorff distance is used for spatial distribution features.

[0032] The similarity calculation of strain distribution gradient adopts an improved cosine similarity method, which considers the spatial distribution characteristics and compares not only the gradient size but also the gradient direction. The similarity calculation of displacement change rate is based on rate change pattern matching and uses dynamic time warping algorithm to handle possible time shift phenomena. The similarity calculation of temperature and humidity coupling coefficient directly uses correlation coefficient comparison to evaluate the consistency of environmental action patterns. The similarity calculation results of each feature are normalized to the range of 0-1 to facilitate weighted average calculation.

[0033] If the cross-interval consistency index is lower than the preset consistency threshold, the stress state interval corresponding to the current monitoring period is marked as an abnormal stress state interval, and the consistency threshold is set according to historical data and engineering experience. The abnormal marking process records the occurrence time, duration and abnormal feature type, providing a basis for subsequent analysis. After marking as abnormal, the re-extraction process of the spatio-temporal features in the interval is triggered, and the re-extraction uses more refined parameter settings to improve the accuracy and reliability of feature extraction. The re-extracted feature data is used to verify the abnormality and distinguish between real abnormalities and measurement errors. For non-abnormal stress state intervals, the system maintains the normal monitoring process, but strengthens the monitoring frequency and data quality check of these intervals. The data processing of abnormal and non-abnormal intervals adopts differentiated strategies, with abnormal interval data being processed first and reported immediately, and non-abnormal interval data being processed according to the regular process. The processing results of all intervals are recorded in the monitoring log, including interval type, feature data, consistency index and processing status, etc. The system provides a consistency threshold adjustment function, allowing users to adjust the threshold size according to actual conditions. The threshold adjustment considers the influence of seasonal factors, load changes, etc., and different threshold standards can be used in winter and summer. When the load is larger, the threshold tolerance is appropriately increased, and when the load is smaller, the threshold requirement is reduced. The threshold adjustment process records the adjustment reason, adjustment time and adjustment personnel information, ensuring the traceability of threshold setting.

[0034] The processing result of the abnormal stress state interval generates an abnormal report, which contains the time, location, type, severity, and processing suggestions of the abnormal occurrence. The report is prepared in a standard format and can be exported in multiple file formats. The abnormal report is sent to relevant personnel in a timely manner through message pushing to ensure that the abnormal situation is handled in a timely manner. The report history data is stored in a database, which supports query and statistical analysis based on time, location, type, and other conditions. During the implementation of the system, an abnormal processing flow specification is established, and the processing time limit and responsible personnel for each type of abnormality are clearly defined. The abnormality is confirmed by a multi-person review mechanism, and important abnormalities require confirmation by the technical responsible person. The abnormal processing result is fed back to the system to form a closed-loop management. The system generates a statistical analysis report of abnormalities on a regular basis to summarize the occurrence rules and trends of abnormalities and provide decision support for structure maintenance. The data of the abnormal stress state interval is stored separately and marked with an abnormality label to facilitate subsequent key analysis. Abnormal data can be used for machine learning model training to improve the system's ability to identify abnormal patterns. As data accumulates, the system can establish an abnormal pattern library to achieve automatic classification and identification of abnormal types. The results of abnormal data analysis are fed back to the monitoring parameter optimization to continuously improve the performance and reliability of the monitoring system.

[0035] In embodiment 4, refer to Figure 3 For the stress feature set of the non-abnormal stress state interval, the system loads the structure historical damage pattern database, which stores the feature patterns under typical damage conditions, including the spatial and temporal feature labels of common damage types such as concrete cracking, prestressed steel slip, and anchor loosening. The database organizes data in a time sequence chart structure, and each damage pattern contains feature vectors, occurrence location, development process, and other metadata information. The matching degree calculation uses an improved Euclidean distance algorithm that considers the dimension difference and importance weight of the feature quantities to measure the similarity between the current spatio-temporal features and the historical pattern labels. The matching degree threshold is set according to the importance of the project and the tolerance of false positives, usually floating between 0.7-0.9.

[0036] When the matching degree is greater than or equal to the preset threshold, the system starts the spatial domain frequency band decomposition mode to extract the structural damage feature subset. This mode is based on the principle of wavelet packet transform to decompose the spatial domain features into different frequency band components. The decomposition process uses Daubechies wavelet basis function, and the decomposition layer is automatically determined according to the feature complexity, usually 3-5 layers. The energy value of each frequency band is calculated by the sum of squares method, and the frequency band with energy significantly higher than the background noise is selected as the damage sensitive frequency band. The determination of sensitive frequency band refers to the frequency domain feature distribution of historical damage cases, and preferentially selects the frequency band range with higher probability of occurrence in previous damage. If the matching degree is less than the preset threshold, the system analyzes the correlation between the material nonlinear response of strain distribution in the stress feature set and the resonance characteristics of structural dynamic characteristics, and identifies abnormal frequency band by establishing strain-frequency response function. The response function calculation uses the transfer function estimation method, the input is the environmental excitation data, and the output is the strain response data, and the resonance peak with gain significantly higher than the average level is found in the frequency domain. Focus decomposition refines the analysis of the frequency band corresponding to these resonance peaks, extracts the feature components in the specific frequency band by band-pass filtering, and generates the structural damage feature subset. Referring to Table 1, the feature matching analysis results of a prestressed beam during the monitoring period are shown.

[0037] Table 1: Prestressed beam monitoring feature matching analysis record

[0038] Based on the constitutive relation between elastic modulus and strain response in structural material mechanics parameters, the stiffness characteristics of the strain distribution in the damage feature subset are corrected. The correction process considers the time-varying characteristics of the concrete elastic modulus, and adjusts the modulus value according to the material age and curing conditions. The stiffness correction uses an iterative algorithm, which updates the modulus value and recalculates the strain distribution at each iteration until the error between the calculated and measured strain values converges within the allowed range. The corrected strain distribution more accurately reflects the actual stress state of the structure. Based on the time-dependent relationship between the humidity diffusion coefficient and the temperature and humidity coupling in environmental action parameters, the temperature and humidity coupling coefficient in the damage feature subset is compensated and corrected for environmental effects. The humidity diffusion coefficient is determined according to the concrete mix proportion and curing records, and the Arrhenius formula is used for temperature compensation considering the effect of temperature on diffusion speed. The correction process establishes a humidity field finite element model, calculates the actual diffusion coefficient through back analysis, and adjusts the temperature and humidity coupling relationship according to the diffusion coefficient. The compensated and corrected temperature and humidity coupling coefficient eliminates the interference of environmental factors and more accurately reflects the characteristics of the structure itself. The corrected strain distribution and temperature and humidity coupling coefficient are fused with the displacement cumulative quantity feature, and the principal component analysis method is used for multi-parameter collaborative correction. The principal component analysis calculates the eigenvalues and eigenvectors, and selects the principal components with cumulative contribution rate of more than 85% as the new features after fusion. The fusion process considers the physical meaning and dimension of each feature, and standardizes different types of features to ensure the reasonableness of the fusion results. The fused feature subset contains most of the information of the original features, while reducing the data dimension, which facilitates subsequent analysis and processing. The structure damage feature subset after multi-parameter collaborative correction is stored in a special data structure, containing feature values, correction parameters, time stamps, and spatial positions. The storage format supports fast retrieval and batch processing, facilitating historical data comparison and analysis. The visualization of the feature subset uses multi-dimensional scatter plots and parallel coordinate plots to intuitively display the relationship and trend between different parameters. The quality control of the feature subset includes integrity check, range verification, and consistency test to ensure data reliability.

[0039] The system provides a feature correction parameter adjustment interface, allowing experienced engineers to adjust the correction coefficients according to actual conditions. The parameter adjustment record saves adjustment history, including adjustment time, adjustment personnel, adjustment reason and adjustment effect, etc. The feature correction model is updated regularly, and the model parameters are retrained according to the newly accumulated monitoring data to maintain the accuracy of the model. The correction effect evaluation is carried out by comparing the degree of agreement between the feature values before and after correction and the measured values, and the evaluation results guide the optimization of subsequent correction strategies. The analysis results of the damage feature subset are connected with the structure safety evaluation module to provide data support for damage identification and state evaluation. The time series analysis of the feature subset reveals the damage development trend, and the spatial distribution analysis locates the possible position of the damage. The long-term accumulation of the feature subset forms a structure health file, which provides the basis for life prediction and maintenance decision. All processing processes are automatically run while retaining manual intervention interfaces to ensure the balance between system intelligence and reliability. The feature data processing flow establishes a quality traceability mechanism, and each feature data can be traced back to the original monitoring data and processing parameters. The data version management records the operation records and parameter settings of each processing, supporting the reproduction and verification of processing results. The system running state is monitored in real time, and the backup scheme is automatically started and an alarm is sent when an exception occurs. Regularly generate feature processing reports to summarize processing effects and improvement suggestions, and continuously improve system performance.

[0040] In embodiment 5, the system pre-stores a structure damage evolution path based on the structural connection relationship and internal force transmission path of the structure, which is represented by a directed graph data structure. The nodes represent key structural components such as prestressed anchorage ends, bending sections, shear connection points, etc., and the edges represent damage propagation direction and influence weight. Taking a large prestressed concrete bridge as an example, its damage evolution path contains 156 nodes and 283 edges. The node weight is determined according to the component importance coefficient, and the edge weight is calculated according to the internal force transmission efficiency and damage diffusion probability. This path model considers the structure's geometric topology, material properties and load transmission mechanism, and its accuracy is verified by finite element analysis and historical damage data. From the structure damage feature subset after multi-parameter collaborative correction, abnormal features are extracted, including strain gradient distribution exceeding the threshold, abnormal displacement change rate or deviation of temperature and humidity coupling coefficient from the normal range. In the actual monitoring of a bridge structure, the system detects a strain gradient of 0.58 at the position of the box girder web, which exceeds the preset threshold of 0.35, and at the same time, the displacement change rate is found to be abnormally increased to 0.18 mm / h at the corresponding position. These abnormal features are matched with the nodes in the damage evolution path, and the abnormal features are mapped to specific structural positions through feature pattern recognition algorithms, establishing a correspondence between abnormal features and spatial positions.

[0041] The potential damage source location of abnormal features is located by using the improved depth-first search algorithm, which traverses the damage propagation direction of the structural damage evolution path in reverse. The traversal process starts from the feature node where the anomaly is detected, and searches along the reverse path of the directed edge. The damage probability of each possible path is calculated considering the edge weight and node importance. In the above bridge case, the system starts from the box girder web node (number G5-22) where the abnormal strain is detected, and locates the prestressed anchorage zone (number A2-07) as the most likely damage source location after traversing 3 propagation nodes. The damage propagation probability of this path reaches 0.87. The repair priority is determined according to the number and distribution depth of abnormal features at the potential damage source location. The weighted scoring method is used for priority calculation. The weight of the number of abnormal features is 0.6, and the weight of the distribution depth is 0.4, where the distribution depth refers to the shortest path length from the damage source to the detection point. The repair priority evaluation of a certain sea-crossing bridge shows that the number of abnormal features at the main tower foundation location is 8, the distribution depth is 1, and the priority score is 9.2; while the number of abnormal features of the approach bridge slab is 5, the distribution depth is 3, and the priority score is 6.8. The system automatically lists the main tower foundation as the priority repair object. The time-varying characteristics of the abnormal features at each potential damage source location are analyzed, and the feature degradation rate is extracted. The analysis uses the time series regression method to calculate the trend of feature values in the last 72 hours. In the monitoring of a certain stadium roof structure, the system found that the strain feature of the cable-strut node was growing at a rate of 0.05% per hour, and the acceleration continued to increase. According to the damage tolerance value in the structural material mechanics parameters, the system compares the current degradation rate with the allowable rate threshold value, and automatically raises the repair priority when it exceeds 150% of the tolerance value.

[0042] Based on the adjusted repair priority and the remaining safe life of the structure, the repair instruction containing the damage source location identification and the repair time is generated. The instruction generation module adopts a templated design, including information such as damage location coordinates, damage type, severity level, and recommended repair time window. For a certain subway tunnel monitoring project, the system generates an instruction requiring the repair of the ring joint (location K23+156) within 14 days. The location damage priority is urgent, and the predicted remaining safe life is 21 days. The repair instruction is automatically assigned to the maintenance department through the work order management system, and is sent to the relevant technical personnel at the same time. The repair instruction execution process establishes a closed-loop management mechanism, and each instruction has corresponding state tracking and result feedback. The maintenance personnel upload the repair results on site through the mobile terminal, including damage confirmation, repair measures, and treatment effect. The system updates the damage evolution path model based on the feedback information, and corrects the edge weight and propagation probability parameters. For unconfirmed damage warnings, the system automatically adjusts the detection threshold and algorithm parameters to reduce the false positive rate. All repair history records form a knowledge base, providing data support for subsequent damage prediction.

[0043] The system generates a maintenance priority analysis report periodically, which counts the damage occurrence frequency, evolution law and maintenance effect of each structure area. The report uses visual charts to show the priority distribution, damage type statistics and maintenance resource allocation. Based on long-term monitoring data, the system establishes a damage prediction model, which can predict the possible damage types and development trends in advance, providing decision basis for preventive maintenance. The maintenance resource optimization module automatically generates the optimal maintenance plan and resource allocation scheme according to the priority distribution and resource constraints. The maintenance instruction management system is integrated with related platforms such as asset management system and financial system, realizing the informatization management and data sharing of the whole maintenance process. The instruction state is updated in real time, and the management personnel can check the execution progress and processing results of each instruction at any time. The system sets up a multi-level early warning mechanism, and for high-priority damage warning, it automatically starts the emergency response process, notifies the relevant responsible person and upgrades the processing authority. All operations leave traces and can be traced back, ensuring the standardization and safety of the maintenance process.

[0044] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0045] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives and variations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart detection method for loosely bonded prestressed building structures based on the Internet of Things, characterized in that, include: Real-time monitoring data of the loosely bonded prestressed structure is collected through an IoT platform, and the real-time monitoring data is preprocessed. The preprocessed real-time monitoring data is classified according to the preset stress state intervals, and the spatiotemporal features of each stress state interval are extracted to generate a stress feature set. The frequency of change in the stress state range is detected. If the frequency of change exceeds the frequency threshold, a dynamic adjustment mechanism is activated to mark the abnormal stress state range. Based on the matching results of the historical damage patterns of the structure and the current spatiotemporal characteristics, the stress feature set of the non-abnormal stress state interval is decomposed to generate a subset of structural damage features. Based on the correlation between the mechanical parameters of structural materials and the parameters of environmental effects, a multi-parameter collaborative correction is performed on a subset of structural damage characteristics. Based on the modified subset of structural damage features, the location of potential damage sources is traced back along the structural damage evolution path to determine the maintenance priority and generate maintenance instructions.

2. The intelligent detection method for prestressed concrete in buildings based on the Internet of Things according to claim 1, characterized in that, The process of collecting real-time monitoring data of the loosely bonded prestressed structure through an IoT platform and preprocessing the real-time monitoring data includes: Synchronously collect strain data, displacement data, and ambient temperature and humidity data of the loosely bonded prestressed structure, and synchronize the strain data, displacement data, and ambient temperature and humidity data in time. Noise filtering is performed on the time-synchronized strain data, displacement data, and ambient temperature and humidity data. Noise filtering includes using an adaptive filtering method to remove dynamic interference components in the strain data and removing abnormal data points based on the correlation between ambient temperature and humidity data and displacement data. The filtered strain data, displacement data, and ambient temperature and humidity data are reorganized at preset time intervals to generate a standardized monitoring dataset.

3. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 1, characterized in that, The preprocessed real-time monitoring data is classified according to a preset stress state interval, and the spatiotemporal features within each stress state interval are extracted to generate a stress feature set, including: The standardized monitoring dataset is divided into multiple stress state intervals according to stress level, including low stress interval, medium stress interval and high stress interval. Spatiotemporal features were extracted from strain data, displacement data, and ambient temperature and humidity data within each stress state interval. The spatiotemporal features within the same stress state range are aggregated according to their spatial location to generate a multidimensional stress feature set.

4. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 3, characterized in that, The spatiotemporal characteristics of strain data include strain distribution gradient, temporal fluctuation amplitude, and spatial variation coefficient; the spatiotemporal characteristics of displacement data include displacement change rate and displacement accumulation; and the spatiotemporal characteristics of environmental temperature and humidity data include temperature and humidity coupling coefficient and environmental influence intensity.

5. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 1, characterized in that, If the frequency of change in the detected stress state range exceeds a frequency threshold, a dynamic adjustment mechanism is activated to mark the abnormal stress state range, including: Based on the number of changes in the stress state interval according to the preset monitoring cycle, the change frequency per unit time is calculated, and it is determined whether the change frequency exceeds the change frequency threshold. If the frequency of change exceeds the frequency of change threshold, the spatiotemporal features of adjacent stress state intervals are aligned using a spatiotemporal registration method, and cross-interval consistency index is extracted. If the cross-interval consistency index is lower than the consistency threshold, the stress state interval corresponding to the current monitoring period is marked as an abnormal stress state interval, and the spatiotemporal features within the abnormal stress state interval are re-extracted; otherwise, it is marked as a non-abnormal stress state interval.

6. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 5, characterized in that, The extraction of cross-interval consistency index includes: aligning the spatiotemporal features of adjacent stress state intervals, calculating similarity measures for the aligned strain distribution gradient, displacement change rate, and temperature and humidity coupling coefficient, and using the weighted average of the similarity measures as the cross-interval consistency index.

7. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 1, characterized in that, The set of stress features in the non-abnormal stress state interval is decomposed based on the matching degree calculation results between the historical damage pattern of the structure and the current spatiotemporal characteristics to generate a subset of structural damage features, including: For the current spatiotemporal domain features in the stress feature set of non-abnormal stress state intervals, the degree of matching between the current spatiotemporal domain features and the current spatiotemporal domain features is determined based on the matching degree calculation results between the structural historical damage mode label and the current spatiotemporal domain features. If the matching degree is greater than or equal to the preset matching threshold, the structural damage feature subset is extracted from the stress feature set of the non-abnormal stress state interval using the spatial frequency band decomposition mode. If the matching degree is less than the preset matching threshold, based on the correlation between the material nonlinear response of strain distribution and the resonance characteristics of structural dynamic characteristics in the stress feature set of non-abnormal stress state interval, the spatial frequency band with significant response is selected for focused decomposition to generate a subset of structural damage features.

8. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 1, characterized in that, The multi-parameter collaborative correction of the structural damage feature subset based on the correlation between structural material mechanical parameters and environmental action parameters includes: Based on the constitutive relationship between the elastic modulus and strain response in the mechanical parameters of structural materials, the strain distribution in the structural damage feature subset is corrected for stiffness characteristics. Based on the time-dependent relationship between the humidity diffusion coefficient and the temperature-humidity coupling in the environmental action parameters, environmental effect compensation correction is applied to the temperature-humidity coupling coefficient in the structural damage feature subset. The modified strain distribution and temperature-humidity coupling coefficient are fused with the displacement accumulation characteristics to generate a subset of structural damage characteristics after multi-parameter collaborative correction.

9. The intelligent detection method for loosely bonded prestressed building structure based on the Internet of Things according to claim 1, characterized in that, The process of tracing the location of potential damage sources along the structural damage evolution path based on the modified subset of structural damage features, determining maintenance priorities, and generating maintenance instructions includes: Obtain the structural damage evolution path, which is constructed based on the structural connection relationship and internal force transmission path, including the damage propagation direction and influence weight between structural locations; Abnormal features are extracted from the structural damage feature subset after multi-parameter collaborative correction, and the corresponding positions in the structural damage evolution path are matched according to the type of abnormal features. Traverse backwards along the damage propagation direction of the structural damage evolution path to locate the potential damage source of abnormal features. The maintenance priority is determined based on the number and depth of abnormal features at the location of potential damage sources. Based on maintenance priority and the remaining safe life of the structure, maintenance instructions are generated, which include the location of the damage source and the maintenance time.

10. The intelligent detection method for building prestressed concrete based on the Internet of Things according to claim 9, characterized in that, After determining the maintenance priority based on the number and distribution depth of abnormal features at the location of potential damage sources, the process also includes: Time-varying characteristic analysis was performed on the abnormal features of each potential damage source location to extract the feature degradation rate; Adjust maintenance priorities based on characteristic degradation rates and damage tolerance values ​​in the mechanical parameters of structural materials; Update maintenance instructions based on the adjusted maintenance priority.

Images