Efficient acquisition and processing system for safety production monitoring data of electric power steel structure

Abstract

The application discloses a kind of efficient collection and processing system of electric power steel structure safety production monitoring data, comprising: space-time sparse sampling module analyzes electric power steel structure monitoring point space distribution density and vibration signal time domain characteristics, generates dynamic sampling strategy and executes collection operation by distributed sensing node;Data transmission scheduling module encapsulates data according to Predict Steel Core pre-research platform protocol, and directional transmission is adopted using time division multiplexing mechanism.The physical information prediction module imports the explicit time-domain physical information dynamic response prediction model, and generates state prediction data combined with the structure parameters;Threshold early warning analysis module loads online incremental learning dynamic threshold early warning algorithm to carry out real-time comparison.The platform interactive storage module stores data and provides calling interface, and the control coordination module dynamically adjusts the running parameters and data flow conversion time sequence of each module, the application guarantees the efficient operation of whole link, improves the efficiency and accuracy of electric power steel structure safety production monitoring data processing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power steel structure safety production data processing, and particularly relates to an efficient collection and processing system for power steel structure safety production monitoring data. BACKGROUND

[0002] As the core load-bearing component of power infrastructure such as power transmission towers and substation supports, power steel structures are long-term exposed to complex natural environments and alternating loads, and the structural health state thereof is directly related to the safe and stable operation of the power system. With the continuous expansion of the power grid scale and the growth of the service life, the risk of stress concentration, fatigue damage and other hidden dangers of steel structures significantly increases, and higher requirements are put forward for the real-time and comprehensiveness of safety production monitoring. Current monitoring needs to cover multi-dimensional data such as stress and strain, vibration response, and the data scale increases exponentially with the increase of monitoring point density, and needs to be combined with physical information such as structural geometric parameters and material properties to realize state judgment, which poses a severe challenge to the efficiency and collaboration of data collection and processing, and an integrated monitoring data processing system that adapts to the characteristics of power steel structures is urgently needed.

[0003] The prior art has obvious deficiencies in the processing of power steel structure monitoring data. On the one hand, the data collection lacks dynamic adaptation capability, and a fixed frequency full-quantity sampling mode is mostly used, without optimizing the sampling strategy in combination with the spatial distribution difference of monitoring points and signal time domain characteristics, resulting in a large amount of redundant invalid data, which not only increases the transmission bandwidth pressure, but also reduces the subsequent data processing efficiency. On the other hand, the early warning threshold setting lacks a dynamic updating mechanism, mostly relying on fixed experience threshold or static model calculation results, without fully integrating historical monitoring data and real-time prediction information for incremental learning and updating, and being difficult to adapt to the dynamic response changes of steel structures under different service stages and environmental conditions, prone to early warning lag or false alarm problems, and unable to provide accurate and reliable decision support for safety production. SUMMARY

[0004] In order to overcome the shortcomings and deficiencies of the prior art, the present application provides an efficient collection and processing system for power steel structure safety production monitoring data.

[0005] The technical scheme adopted by the present application is a kind of efficient acquisition and processing system of electric power steel structure safety production monitoring data, comprising: space-time sparse sampling module, based on the spatial distribution density of electric power steel structure stress-strain monitoring point and the time domain characteristics of vibration signal, calling adaptive space-time sparse sampling reconstruction model to generate dynamic sampling strategy, and executing data acquisition operation through distributed sensing node;Data transmission scheduling module, receiving the acquisition data output by the space-time sparse sampling module, encapsulating data according to the communication protocol specification of Predict Steel Core research platform, and using time division multiplexing transmission mechanism to transmit the encapsulated data to data processing unit;Physical information prediction module, obtain transmission data from data transmission scheduling module, import explicit time domain physical information dynamic response prediction model, combine electric power steel structure geometric parameters and material properties to build dynamic response equation, and solve to generate structure state prediction data;Threshold early warning analysis module, receiving the prediction data of the physical information prediction module, loading online incremental learning dynamic threshold early warning algorithm, and using the threshold parameter library trained by historical monitoring data for real-time threshold comparison analysis;Platform interactive storage module, data interaction link is established with threshold early warning analysis module, early warning analysis result and original monitoring data are classified and stored according to the data storage format of Predict Steel Core research platform, and data calling interface is provided;System control and coordination module, respectively with space-time sparse sampling module, data transmission scheduling module, physical information prediction module, threshold early warning analysis module and platform interactive storage module establish control signal connection, and dynamically adjust the operation parameters and data flow time sequence of different modules according to the monitoring scene of electric power steel structure.

[0006] Further, the adaptive space-time sparse sampling reconstruction model in the space-time sparse sampling module satisfies: wherein, is a sparse sampling data matrix, is a space-time sparse sampling weight coefficient, is a spatial basis function of electric power steel structure monitoring point, is a signal time domain sparse basis function, is a full-quantity monitoring data matrix, is the dimension of spatial basis function and time domain sparse basis function respectively, is a sampling step factor;The data transmission scheduling module determines the transmission data priority based on the output of the adaptive space-time sparse sampling reconstruction model , wherein the priority coefficient , is matrix trace operation, T is the transpose of matrix.

[0007] Further, the explicit time domain physical information dynamic response prediction model in the physical information prediction module satisfies: wherein, is the displacement response function of the electric power steel structure, is the structural damping matrix, is the structural stiffness matrix, is the external load function, is the sparse sampling data driven correction term; calculated based on the structural parameters stored in the Predict Steel Core research platform , is the structural mass matrix, is the damping coefficient, and is updated by the following formula: , , is the initial damping coefficient, are the mean and variance operations, respectively.

[0008] Further, the online incremental learning dynamic threshold warning algorithm in the threshold warning analysis module satisfies: wherein, is the dynamic warning threshold at time t, is the incremental learning rate, is the prediction data weight coefficient, is the predicted displacement output by the explicit time-domain physical information dynamic response prediction model, is the historical displacement response data stored in the Predict Steel Core research platform, is the historical time variable; the warning judgment basis is the threshold amplification coefficient, and , is the sampling data anomaly threshold, is the counting operation.

[0009] Further, the platform interaction storage module optimizes data storage based on the Predict Steel Core research platform, and the storage index is wherein, is the exclusive or operation, is the encryption hash function; data compression rate , is the compressed storage data, is the data capacity calculation function; the sparse basis function of the adaptive space-time sparse sampling reconstruction model is called in the compression process to transform the time series data, satisfying .

[0010] Further, the system control coordination module adjusts the running parameters of different modules by the following formula: wherein, is the running weight of the qth module, The data processing rate at time t for the module, The reference processing rate; The solving efficiency of the explicit time-domain physical information dynamic response prediction model is determined: , The model solving time consumption, The data volume counting function; at the same time, the analysis results of the online incremental learning dynamic threshold early warning algorithm are dynamically corrected: , The early warning proportion at time t, The early warning signal set.

[0011] Further, the data transmission scheduling module comprises: a protocol analysis unit that receives the collected data output by the space-time sparse sampling module, reads the power steel structure monitoring point number and the sampling timestamp information carried in the data header, analyzes the data type and format according to the field definition of the Predict Steel Core research platform communication protocol, extracts the valid data field and eliminates the redundant identification information; a data encapsulation unit that, according to the data type obtained by analysis, calls the encapsulation function specified by the protocol to associate and bind the valid data with the monitoring point attribute parameters, generates an encapsulated data packet including a data check code, wherein the check code is generated by an exclusive or operation of the collected data and the module running parameters; a transmission scheduling unit that obtains the priority identifier of the encapsulated data packet, combines the bandwidth occupation of the current communication link, and allocates transmission time slots according to the time division multiplexing mechanism to direct the data packet to the receiving port corresponding to the physical information prediction module; a state feedback unit that monitors the packet loss rate and delay parameters in the data transmission process in real time, converts the monitoring results into control signals and feeds them back to the system control coordination module to provide data support for running parameter adjustment.

[0012] Further, the physical information prediction module comprises: a parameter import unit that obtains encapsulated data through the receiving port of the data transmission scheduling module, analyzes and extracts the geometric size, material elastic modulus and Poisson's ratio basic parameters of the power steel structure, and standardizes the conversion according to the parameter format requirements of the explicit time-domain physical information dynamic response prediction model; an equation construction unit that calculates the structure mass matrix and stiffness matrix based on the converted basic parameters, combines the external load monitoring data and sparse sampling data correction items to construct a dynamic response differential equation including time derivative items, and determines the initial boundary conditions and solving domain of the equation; a model solving unit that uses an explicit integration algorithm to numerically solve the dynamic response differential equation, sets the solving step and convergence criterion, and iteratively calculates the structure displacement and stress prediction data at different times, and stores them in a temporary data buffer; a result output unit that reads the prediction data from the temporary data buffer, organizes and encapsulates them according to the preset format, transmits them to the threshold early warning analysis module after adding data validity identifiers, and feeds back the solving process parameters to the platform interactive storage module.

[0013] Further, the threshold early warning analysis module comprises: a data receiving unit, establishing a communication link with the physical information prediction module, receiving an encapsulated package including structure state prediction data, parsing and extracting prediction values and corresponding timestamp information, synchronously calling the historical database interface of the Predict Steel Core research platform to obtain historical monitoring data under similar working conditions; a threshold updating unit, loading an online incremental learning dynamic threshold early warning algorithm, inputting historical monitoring data and current prediction data into the algorithm model, iteratively updating the weight coefficient and learning rate of threshold calculation, and generating real-time dynamic early warning threshold; a comparison and analysis unit, comparing the current structure state prediction data with the dynamic early warning threshold point by point, calculating the deviation value and comparing it with the judgment threshold, marking the data points exceeding the judgment range and recording the corresponding monitoring position and time information; an early warning generation unit, statistically analyzing the marked abnormal data points, generating early warning signals according to the degree of abnormality, adding signal generation time and associated monitoring point information, and transmitting to the platform interaction storage module.

[0014] Further, the platform interaction storage module comprises: an interface adaptation unit, establishing a data interaction link with the threshold early warning analysis module, developing a data calling interface conforming to the Predict Steel Core research platform specification, receiving and format converting early warning data, prediction data and original sampling data; a classified storage unit, dividing storage areas according to data types, indexing and storing original sampling data according to monitoring points and timestamps, storing prediction data and early warning results in association, and compressing historical data space using a sparse matrix storage method; a data indexing unit, generating unique index identifiers for various types of data based on an encryption hash function, establishing a mapping relationship table between indexes and storage addresses, optimizing data retrieval paths, and improving the data calling efficiency of the Predict Steel Core research platform; a permission management unit, receiving permission control signals from the system control and coordination module, managing access permissions to data calling interfaces in stages, recording data read and write operation logs, and ensuring the security and traceability of stored data.

[0015] Beneficial effects: The present application proposes an efficient acquisition and processing system for power steel structure safety production monitoring data, the space-time sparse sampling module relies on the accurate capture of the spatial distribution density of the power steel structure monitoring point and the time domain characteristics of the vibration signal, generates a dynamic sampling strategy, replaces the traditional fixed full sampling mode, and reduces invalid data from the source; The data transmission scheduling module follows, encapsulates data according to the pre-research platform communication protocol and uses a time division multiplexing transmission mechanism to further reduce the bandwidth occupancy pressure, and the linkage of the two makes the data flow efficiency get qualitative improvement. The physical information prediction module fuses the core physical information such as the geometric parameters and material properties of the power steel structure to construct a prediction model and generate high-precision structure state prediction data; The threshold early warning analysis module continuously absorbs historical monitoring data to optimize the threshold parameter library through an online incremental learning algorithm, realizes real-time dynamic updating of the early warning threshold, fully adapts to the response changes of the steel structure in different service stages and different environments, and effectively reduces the early warning lag and false alarm. The classification storage of the platform interactive storage module and the dynamic parameter regulation of the system control coordination module form support, fully guarantee the efficiency and accuracy of monitoring data processing, and build a technical defense line for the safety production of power steel structures. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The system module composition diagram of the present application;

[0017] Figure 2 The system running step flow chart of the present application. DETAILED DESCRIPTION

[0018] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict, and the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0019] As Figure 1 shown, an efficient acquisition and processing system for power steel structure safety production monitoring data, comprising:

[0020] A space-time sparse sampling module based on the spatial distribution density of the power steel structure stress-strain monitoring point and the time domain characteristics of the vibration signal, calls an adaptive space-time sparse sampling reconstruction model to generate a dynamic sampling strategy, and executes data acquisition operation through a distributed sensing node;

[0021] Specifically, the space-time sparse sampling module is the core starting point of system data collection. Through the dynamic adaptive sampling strategy, the effectiveness of data and the cost of collection are balanced, and the waste of resources caused by full sampling is avoided. In the implementation process, the monitoring point layout data of the power steel structure is obtained, and the spatial distribution density parameters of the monitoring points are determined. The monitoring point spacing of the power transmission tower type steel structure is usually set to 5 meters, and the key nodes are encrypted to 2 meters. At the same time, the time domain characteristic parameters of the vibration signal are collected, including the signal frequency range 0.1-50Hz and the sampling initial period 0.02 seconds. Based on these parameters, the adaptive space-time sparse sampling reconstruction model is called. The model first analyzes the stress-strain correlation degree of different monitoring points in the spatial dimension. The sampling density is reduced in the area with a correlation degree higher than 0.8, and the density is increased in the area with a correlation degree lower than 0.3. Then, combined with the signal fluctuation amplitude in the time domain, the sampling period is shortened to 0.01 seconds when the fluctuation amplitude exceeds the reference value by 30%, and the sampling period is lengthened to 0.05 seconds when the fluctuation amplitude is less than 5%, and a dynamic sampling strategy is generated. The strategy is issued to the distributed sensing nodes. The sampling accuracy of the sensing nodes is set to 0.001MPa, and the working voltage is maintained at 3.3V. According to the strategy, the stress-strain and vibration signal data collection operations are executed at regular intervals. The collected data is temporarily stored in the 8GB local storage of the node, and waits for the transmission instruction.

[0022] The data transmission scheduling module receives the collected data output by the space-time sparse sampling module, encapsulates the data according to the communication protocol specification of the Predict SteelCore research platform, and uses the time division multiplexing transmission mechanism to transmit the encapsulated data to the data processing unit.

[0023] Specifically, the data transmission scheduling module undertakes the functions of transit processing and directional transmission of collected data, ensuring the integrity and efficiency of data in the transmission process, and connecting the acquisition end and the processing end. In the implementation process, the acquisition data uploaded by the distributed sensor nodes in the space-time sparse sampling module is received through the RS485 communication interface, the interface communication rate is set to 115200bps, the data packet size of each received data is limited to 1024 bytes, and the receiving interval and the sampling period are kept synchronous. After receiving, the data is encapsulated according to the communication protocol specification of the Predict Steel Core research platform. In the encapsulation process, the core fields such as the monitoring point number, the sampling timestamp and the data value in the data are extracted first, arranged in the order of the fields specified by the protocol, 16 bytes of frame identifier are added to the head of the data packet, 4 bytes of check code are added to the tail, and the check code is generated by using the CRC32 algorithm. After encapsulation, the module monitors the bandwidth occupation of the current communication link in real time, the monitoring frequency is 0.1 seconds / time, when the bandwidth occupation rate is less than 60%, the single transmission mode is adopted; when it is higher than 60%, the time division multiplexing transmission mechanism is started, the encapsulated data is divided into 5 transmission channels according to the monitoring point area, each channel is allocated 20% of the bandwidth resource, the transmission time interval is set to 0.02 seconds, the encapsulated data is transmitted to the data receiving port of the physical information prediction module through the Ethernet interface, and the retransmission check is performed once every 10 data packets in the transmission process to ensure that there is no data loss.

[0024] The physical information prediction module obtains the transmission data from the data transmission scheduling module, imports the explicit time-domain physical information dynamic response prediction model, constructs the dynamic response equation combined with the geometric parameters and material properties of the power steel structure, and solves to generate the structure state prediction data.

[0025] Specifically, the physical information prediction module is the key to realize the structure state prediction, which combines data and physical laws to improve the accuracy of structure state evaluation. In the implementation process, the module receives the encapsulated data transmitted by the data transmission scheduling module through the Ethernet receiving port, the buffer size of the receiving port is set to 64KB, the receiving timeout time is set to 5 seconds, after receiving, the data packet is parsed, and the stress, strain, vibration signal data and corresponding monitoring point information are extracted. Import the explicit time-domain physical information dynamic response prediction model, load the basic parameters of the power steel structure when the model starts, including the rod length (error control within ±1mm) and section size (precision 0.1mm) in the structure geometric parameters, the elastic modulus (value 206GPa), Poisson's ratio (value 0.3) and density (value 7850kg / m³) in the material properties, these parameters are stored in the local database of the module in advance, and the calling response time is not more than 0.01 seconds. Based on these parameters, the dynamic response equation is constructed, and the boundary conditions need to be determined in the equation construction process combined with the spatial coordinates of the monitoring points, the fixed end constraint is set to displacement 0, and the hinged end constraint is set to rotation angle 0, at the same time, the collected data obtained by analysis is taken as the external excitation item into the equation. The model uses explicit central difference method for solution, the solution step is set to 0.001 seconds, the iteration number is limited to 1000 times, and the convergence precision is controlled within 1e-6, the residual change is monitored in real time during the solution process, the iteration is stopped after the residual is stable, and the structure displacement, stress and other state prediction data at different times are generated, the time resolution of the prediction data is consistent with the sampling period, and the prediction data is stored in the 256GB cache.

[0026] The threshold warning analysis module receives the prediction data of the physical information prediction module, loads the online incremental learning dynamic threshold warning algorithm, and performs real-time threshold comparison analysis by using the threshold parameter library trained by the historical monitoring data;

[0027] Specifically, the threshold early warning analysis module is responsible for real-time research and judgment of the structural safety state, and timely identification of potential safety hazards to provide early warning support for safety production. In the implementation process, the state prediction data stored in the cache by the physical information prediction module is received through the internal data bus, the transmission rate of the data bus is set to 1 Gbps, the data reading rate is not less than 10 MB / s, and the corresponding monitoring point number and prediction timestamp are associated during reading. After receiving, load the online incremental learning dynamic threshold early warning algorithm, after the algorithm is started, call the historical database interface of the Predict Steel Core research platform to obtain the monitoring data of the power steel structure in the past 3 years, the data volume is controlled within 100 GB, and the abnormal values (data exceeding 5 times the standard deviation of the mean value) are removed through data screening to obtain effective historical data. Use the effective historical data to train the threshold parameter library, the learning rate is set to 0.01 during training, and the training iteration number is 500 times. The parameter library after training includes the basic threshold and update coefficient under different monitoring points and different working conditions. Compare and analyze the prediction data received in real time with the corresponding threshold in the parameter library, compare one by one according to the monitoring point, calculate the deviation value of the prediction data and the threshold, and the deviation value is calculated by using the absolute error method. The comparison frequency is synchronized with the prediction data generation frequency, and after completing the comparison once, the comparison result (normal, abnormal) is marked and temporarily stored, waiting for subsequent processing.

[0028] The platform interaction storage module establishes a data interaction link with the threshold early warning analysis module, classifies and stores the early warning analysis results and original monitoring data according to the data storage format of the Predict Steel Core research platform, and provides a data calling interface at the same time;

[0029] Specifically, the platform interaction storage module realizes the standardized storage and convenient calling of data, guarantees the traceability and reusability of data, and supports subsequent analysis and decision-making. In the implementation process, the module establishes a data interaction link with the threshold early warning analysis module through the TCP / IP protocol, sets the link connection timeout time to 10 seconds, and after successful connection, it receives the early warning analysis results and the corresponding original monitoring data and prediction data in real time. The received data formats include text format and binary format, and the module automatically identifies and converts them into a unified JSON format. According to the data storage format of the Predict Steel Core research platform, the original monitoring data is stored according to the directory structure of "date-monitoring point", and the prediction data and early warning analysis results are stored according to the directory structure of "date-analysis batch". The storage medium uses a 1TB solid state hard disk with a read / write speed of not less than 500MB / s. During storage, a data sharding storage strategy is used, with each 500MB divided into a data block, and a unique identification code added to each data block. At the same time, the module develops a data calling interface that meets the RESTful specification, which supports GET and POST request methods, with a request response time of not more than 1 second. The interface permission is divided into administrator, analyst, and visitor levels, with different levels corresponding to different data access ranges. In addition, the module regularly checks the integrity of the stored data, with a verification period of 24 hours. The verification uses the MD5 hash value comparison method, and when data damage is found, the data is automatically retrieved from the backup node (the backup node has the same storage capacity as the main node) for repair.

[0030] The system control coordination module is connected with the space-time sparse sampling module, the data transmission scheduling module, the physical information prediction module, the threshold early warning analysis module, and the platform interaction storage module through control signals, and dynamically adjusts the operating parameters and data flow time sequence of different modules according to the power steel structure monitoring scene.

[0031] Specifically, the system control coordination module is the central command unit of the system, which coordinates the operation of each module to ensure that the system adapts to different monitoring scenarios. During implementation, the system control coordination module establishes a control signal connection with the space-time sparse sampling module, the data transmission scheduling module, the physical information prediction module, the threshold early warning analysis module, and the platform interaction storage module through the CAN bus. The bus communication rate is set to 500 kbps, and the control signal transmission period is 0.05 seconds. The module receives the operating state parameters uploaded by each module in real time, including the sampling frequency of the space-time sparse sampling module, the working voltage of the sensor node, the bandwidth occupancy rate and packet loss rate of the data transmission scheduling module, the model solution time and convergence accuracy of the physical information prediction module, the comparison frequency and abnormal data proportion of the threshold early warning analysis module, the storage usage rate and interface response time of the platform interaction storage module, and the parameter acquisition accuracy is controlled within ±1%. Combined with the real-time changes of the power steel structure monitoring scene, such as when the environmental wind speed exceeds 15 m / s, the system control coordination module sends a control signal to the space-time sparse sampling module to increase the sampling frequency to 1.5 times the original frequency; when the data transmission packet loss rate exceeds 5%, the system control coordination module sends a signal to the data transmission scheduling module to increase the number of transmission channels to 8; when the model solution time exceeds 2 seconds, the system control coordination module sends a signal to the physical information prediction module to appropriately reduce the convergence accuracy to 5e-6. At the same time, the module dynamically adjusts the data flow time sequence of each module to ensure seamless connection of sampling, transmission, prediction, analysis, and storage. The operating parameters and time sequence adjustment instructions are sent in real time through the control signal, and each module updates the parameters within 1 second after receiving the signal.

[0032] Preferably, the adaptive space-time sparse sampling reconstruction model in the space-time sparse sampling module satisfies: wherein, is a sparse sampling data matrix, is a space-time sparse sampling weight coefficient, is a power steel structure monitoring point spatial basis function, is a signal time-domain sparse basis function, is a full-quantity monitoring data matrix, are the dimensions of the spatial basis function and the time-domain sparse basis function, respectively, is a sampling step factor; the data transmission scheduling module determines the transmission data priority based on the output of the adaptive space-time sparse sampling reconstruction model , wherein the priority coefficient is , is a matrix trace operation, T is a matrix transpose.

[0033] Specifically, the spatio-temporal sparse sampling and transmission scheduling are cooperatively optimized, and the pertinence of data acquisition and the orderliness of transmission are further improved through quantization model and priority mechanism. In the implementation process, when the spatio-temporal sparse sampling module runs, the dimension of the spatial basis function is determined based on the spatial distribution characteristics of the power steel structure monitoring points, and is set to 20*25 according to the monitoring area division. The dimension of the time domain sparse basis function is set to 30 according to the signal frequency range, and the sampling step factor is set to 5 according to the initial sampling period. The module generates a sparse sampling data matrix through model operation, and the iteration update step of the weight coefficient in the operation process is controlled to be 0.005, so that the sampling data can reflect the structure state and avoid redundancy. After the data transmission scheduling module receives the matrix, the matrix trace operation program is started, the operation precision is retained to 6 decimal places, the priority coefficient is calculated through the ratio of the front and rear matrix traces, and the coefficient value range is controlled to be between 0.3 and 0.8. When the coefficient is higher than 0.6, the corresponding data is marked as high priority and is allocated with 80% of the transmission bandwidth; when the coefficient is lower than 0.4, the corresponding data is marked as low priority and is allocated with 10% of the transmission bandwidth; and the rest is marked as medium priority and is allocated with 10% of the transmission bandwidth. The priority coefficient is recalculated every 0.2 seconds during the transmission process, the bandwidth allocation ratio is dynamically adjusted, the error rate of the interface transmission is controlled to be below 1e-6, and the high-value data is guaranteed to be transmitted preferentially.

[0034] Preferably, the explicit time-domain physical information dynamic response prediction model in the physical information prediction module satisfies: , wherein is a power steel structure displacement response function, is a structure damping matrix, is a structure stiffness matrix, is an external load function, is a sparse sampling data driven correction term; the structure parameters are calculated based on the Predict Steel Core pre-research platform , is a structure mass matrix, is a damping coefficient, and is updated through the following formula: , , is an initial damping coefficient, are mean and variance operations, respectively.

[0035] Specifically, the parameter adaptability of the physical information prediction model is optimized, and the fitting degree of the structural dynamic response prediction is improved by dynamically correcting the damping parameter. In the implementation process, after the physical information prediction module loads the explicit time domain model, the structural mass matrix stored in the Predict Steel Core research platform is called, the matrix dimension is set to 150x150 according to the number of steel structure members, and the calculation accuracy is controlled within ±0.001. In the initial damping coefficient, the mass proportional damping coefficient is set to 0.02, and the stiffness proportional damping coefficient is set to 0.005. When the model runs, every 10 groups of sparse sampling data are received, the damping coefficient updating program is started, the mean and variance of the sampling data are calculated, the mean is calculated by using the sliding window method, the window size is set to 20, and the variance calculation precision is retained to 4 decimal places. The mass proportional damping coefficient is corrected by 1% of the mean in the positive direction, and the stiffness proportional damping coefficient is corrected by 0.5% of the variance in the reverse direction, and the corrected coefficient value is limited to 0.8-1.2 times of the initial value. The updated damping coefficient is substituted into the dynamic response equation, the time step of the external load function is consistent with the sampling period when solving the equation, the weight of the correction term is adjusted according to the data reliability, the weight is set to 0.7 when the reliability is higher than 0.9, and the weight is set to 0.3 when the reliability is lower than 0.7, so that the deviation between the prediction result and the actual structure response is controlled within 5%.

[0036] Preferably, the online incremental learning dynamic threshold early warning algorithm in the threshold early warning analysis module satisfies: , wherein, is the dynamic early warning threshold at time t, is the incremental learning rate, is the prediction data weight coefficient, is the predicted displacement output by the explicit time domain physical information dynamic response prediction model, is the historical displacement response data stored in the Predict Steel Core research platform, is the historical time variable; early warning judgment basis , is the threshold amplification coefficient, and , is the sampling data anomaly threshold, is the counting operation.

[0037] Specifically, the dynamic updating mechanism of the threshold early warning integrates historical and real-time data through incremental learning to improve the accuracy of early warning determination. During implementation, the threshold early warning analysis module loads the algorithm, sets the incremental learning rate to 0.015, which is determined according to the fluctuation degree of historical data. The larger the fluctuation, the higher the value but not more than 0.03. The prediction data weight coefficient is initially set to 0.6 and dynamically adjusted with the prediction accuracy. When the prediction deviation is less than 3%, it is increased to 0.7, and when it is higher than 8%, it is reduced to 0.4. The algorithm calculates the dynamic early warning threshold every 0.5 seconds. When calculating, the maximum value in the historical displacement response data is extracted first. The time span of historical data is set to the last 1 hour, and the data quantity is controlled within 500 groups. The predicted displacement and the maximum value of historical data are weighted and summed according to the weight coefficient, and then combined with the learning rate and the threshold value at the last moment to calculate the current threshold value. The threshold value fluctuation amplitude is controlled within ±10%. When early warning is determined, the threshold amplification coefficient is initially set to 1.2. Every time 1 sampling data exceeds the abnormal threshold value, the coefficient increases by 0.02, and at most increases to 1.5. If there is no abnormal data for 5 seconds in succession, the initial value is restored. The response time of the determination process is controlled within 0.1 seconds to ensure that the early warning signal is generated in time.

[0038] Preferably, the platform interaction storage module optimizes data storage based on the Predict Steel Core research platform, and the storage index , is an exclusive or operation, is an encryption hash function; data compression rate , is the stored data after compression, is a data capacity calculation function; during compression, the sparse basis function of the adaptive space-time sparse sampling reconstruction model is called The time series data is transformed to satisfy .

[0039] Specifically, the efficiency and security of platform interaction storage are realized through hash index and sparse compression to achieve efficient data storage and fast retrieval. During implementation, the platform interaction storage module sets the output length of the encryption hash function to 256 bits after establishing the data interaction link, and the response time of the hash operation is controlled within 0.001 seconds. The sparse sampling data, prediction data and early warning threshold are subjected to XOR operation, and the operation sequence is arranged according to the data generation time. The operation result is used as the input of the hash function to generate a unique storage index. When compressing data, the time domain sparse basis function is called, and the truncation error of the basis function is controlled below 1e-5. The time series data is converted into a sparse matrix form, and the compression ratio is adjusted according to the data redundancy. When the redundancy is higher than 60%, the compression ratio is set to 1:5, and when the redundancy is lower than 30%, the compression ratio is set to 1:2. The storage index and compressed data are stored in the form of "index-data" key-value pairs, the write delay of the storage medium is controlled within 0.05 seconds, and the read rate is not less than 80MB / s. When calculating the data capacity, the ratio of the capacity difference between the full data and the compressed data to the total capacity is the compression rate, which is stable between 60%-80%. At the same time, the hash index is checked once an hour to ensure the correspondence between the index and the data.

[0040] Preferably, the system control coordination module adjusts the operating parameters of different modules through the following formula: wherein, is the operating weight of the qth module, is the data processing rate of the module at time t, is the reference processing rate; The solving efficiency of the explicit time domain physical information dynamic response prediction model is determined: is the model solving time, is the data volume counting function; and the analysis result of the online incremental learning dynamic threshold early warning algorithm is dynamically corrected: is the early warning proportion at time t, is the early warning signal set.

[0041] Specifically, the dynamic regulation of each module of the system allocates resources through weight adjustment to ensure the operation efficiency of the whole link. During the implementation, the system control and coordination module sets the initial operation weight of each module to 0.2 when initialized, and sets the reference processing rate according to the hardware performance of the module, with the time-space sampling module set to 10 MB / s, the transmission scheduling module set to 15 MB / s, the prediction module set to 8 MB / s, the early warning module set to 12 MB / s, and the storage module set to 16 MB / s. The actual processing rate of each module is collected every 0.1 seconds, and the instantaneous sampling method is used for rate calculation with a sampling interval of 0.01 seconds. The average value of 5 samplings is taken as the result. The operation weight is adjusted according to the ratio of the actual rate to the reference rate. When the ratio is higher than 1.2, the weight is increased by 0.05, and when the ratio is lower than 0.8, the weight is decreased by 0.03. The weight value range is limited to 0.1-0.3. For the prediction module, the solving efficiency is calculated every 0.3 seconds. When the solving time exceeds 2 seconds, the weight is decreased by 0.04. For the early warning module, the early warning proportion is calculated every 0.2 seconds. When the proportion exceeds 10%, the weight is increased by 0.06. The weight adjustment instruction is issued through the control bus, and each module updates the parameters within 1 second after receiving the instruction to ensure that the whole system processing time delay is controlled within 3 seconds.

[0042] Preferably, the data transmission scheduling module comprises: a protocol analysis unit that receives the collected data output by the time-space sparse sampling module, reads the power steel structure monitoring point number and sampling timestamp information carried in the data header, analyzes the data type and format according to the field definition of the PredictSteel Core research platform communication protocol, extracts the valid data field and eliminates the redundant identification information; a data encapsulation unit that associates and binds the valid data with the monitoring point attribute parameters according to the data type obtained by analysis, generates an encapsulated data packet including a data check code by calling the encapsulation function specified by the protocol, wherein the check code is generated by the exclusive OR operation of the collected data and the module operation parameters; a transmission scheduling unit that obtains the priority identifier of the encapsulated data packet, combines the bandwidth occupation of the current communication link, and allocates transmission time slots according to the time division multiplexing mechanism to direct the data packet to the receiving port corresponding to the physical information prediction module; a state feedback unit that monitors the packet loss rate and delay parameters in the data transmission process in real time, converts the monitoring results into control signals and feeds them back to the system control and coordination module to provide data support for operation parameter adjustment.

[0043] Specifically, the data transmission scheduling module guarantees the integrity and orderliness of the collected data from receiving to transmission through the collaborative processing of different units. During implementation, the protocol analysis unit accesses the output link of the space-time sparse sampling module through a dedicated interface, with the interface response time set to 0.01 seconds. It reads the 16-bit monitoring point number and 32-bit sampling timestamp information included in the data header, extracts the effective data fields such as stress and strain, vibration signals, etc. according to the 28-field definition of the Predict Steel Core research platform communication protocol, and uses byte stream analysis to extract the effective data fields. The analysis error is controlled within ±1 byte, and invalid information such as frame header redundancy identifiers is synchronously removed. The data packaging unit calls the built-in packaging function of the protocol to associate and bind the effective data with 8 attribute parameters such as monitoring point material and installation position, generates a packaging data packet including a 32-bit CRC check code, and the check code generation time does not exceed 0.005 seconds. The transmission scheduling unit monitors the communication link bandwidth every 0.02 seconds. When the bandwidth occupancy rate is less than 50%, it allocates 3 transmission time slots, and when it is higher than 50%, it expands to 6, and then sends the data packet to the receiving port of the physical information prediction module. The port reception success rate needs to reach more than 99.9%. The state feedback unit monitors the packet loss rate and delay parameters with a period of 0.1 seconds. When the packet loss rate exceeds 1% or the delay exceeds 0.5 seconds, a 4-byte control signal is immediately generated and fed back to the system control coordination module.

[0044] Preferably, the physical information prediction module comprises: a parameter import unit that obtains the packaged data through the receiving port of the data transmission scheduling module, analyzes and extracts the geometric size, material elastic modulus, and Poisson's ratio basic parameters of the power steel structure, and standardizes and converts them according to the parameter format requirements of the explicit time-domain physical information dynamic response prediction model; an equation construction unit that calculates the structure mass matrix and stiffness matrix based on the converted basic parameters, combines the external load monitoring data and sparse sampling data correction items, constructs the dynamic response differential equation including the time derivative item, and determines the initial boundary conditions and solution domain of the equation; a model solving unit that uses an explicit integration algorithm to numerically solve the dynamic response differential equation, sets the solution step and convergence criterion, and iteratively calculates the structure displacement and stress prediction data at different times, and stores them in the temporary data buffer area; a result output unit that reads the prediction data from the temporary data buffer area, organizes and packages them according to the preset format, adds data validity identifiers, and transmits them to the threshold early warning analysis module, while feeding back the solving process parameters to the platform interactive storage module.

[0045] Specifically, the physical information prediction module improves the accuracy and stability of the structure state prediction through different unit division and refined model operation links. During implementation, the parameter import unit obtains encapsulated data through a gigabit Ethernet interface, sets the interface buffer to 32 KB, sets the receive timeout to 3 seconds, extracts basic parameters such as steel structure geometric dimensions (accuracy ± 0.5 mm), elastic modulus (206 ± 2 GPa), and Poisson's ratio (0.30 ± 0.01), converts them to floating-point data according to the parameter format of the explicit time-domain model, and controls the conversion time to within 0.03 seconds. The equation construction unit calculates a 120x120 mass matrix and a stiffness matrix based on the converted parameters, with a calculation accuracy of up to 4 decimal places, fuses external load monitoring data and sparse sampling correction terms, constructs dynamic response differential equations including second-order time derivative terms, determines the boundary conditions of fixed end displacement and hinge end rotation angle as 0, and automatically checks the coefficient consistency after equation construction. The model solving unit uses an explicit integration algorithm, sets a solving step size of 0.002 seconds and a convergence criterion of 1e-5, iteratively calculates the structure displacement and stress prediction data every 0.01 seconds, temporarily stores them in a 128 GB cache, and the cache response time does not exceed 0.001 seconds. The result output unit arranges and encapsulates the prediction data according to the preset 16-byte data header format, adds a 2-bit validity identifier, and transmits it to the threshold warning analysis module, while synchronously feeding back 6 process parameters such as solving step size and iteration number to the platform interaction storage module.

[0046] Preferably, the threshold warning analysis module includes: a data receiving unit that establishes a communication link with the physical information prediction module, receives encapsulated packages including structure state prediction data, extracts prediction values and corresponding timestamp information, and synchronously calls the historical database interface of the Predict Steel Core research platform to obtain historical monitoring data under similar working conditions; a threshold updating unit that loads an online incremental learning dynamic threshold warning algorithm, inputs the historical monitoring data and current prediction data into the algorithm model, iteratively updates the weight coefficient and learning rate of threshold calculation, and generates real-time dynamic warning thresholds; a comparison and analysis unit that compares the current structure state prediction data with the dynamic warning thresholds point by point, calculates the deviation value and compares it with the judgment threshold, marks the data points that exceed the judgment range and records the corresponding monitoring location and time information; a warning generation unit that statistically analyzes the marked abnormal data points, generates warning signals according to the abnormality degree, adds signal generation time and associated monitoring point information, and transmits them to the platform interaction storage module.

[0047] Specifically, the early warning mechanism of the threshold early warning analysis module realizes dynamic updating and accurate judgment of the threshold through the cooperation of different units, and improves the timeliness of safety early warning. In the implementation process, the data receiving unit establishes a full-duplex communication link with the physical information prediction module through the RS485 interface, the link transmission rate is set to 115200bps, receives the encapsulated package including the prediction data and the 32-bit timestamp, parses and extracts the prediction value, and then calls the Predict Steel Core research platform historical database through the HTTP interface to obtain 1000 groups of historical monitoring data of the same working condition within 72 hours. The data calling response time is not more than 0.2 seconds. The threshold updating unit loads the online incremental learning algorithm, inputs the historical data and the current prediction data into the model in a 7:3 ratio, sets the learning rate to 0.02 to update the weight coefficient, generates a real-time dynamic early warning threshold every 0.1 seconds, and controls the threshold fluctuation range within ±5%. The comparison and analysis unit calculates the absolute deviation of the prediction data and the dynamic threshold point by point, and marks the abnormal data point when the deviation exceeds 15% of the threshold. 8-bit monitoring position code and 32-bit time information are recorded synchronously, and the comparison processing rate is not less than 1000 points / second. The early warning generation unit classifies the abnormal data points into three levels according to the severity, adds 16-bit signal generation time and monitoring point information, generates an early warning signal, and transmits it to the platform interaction storage module. The time delay from signal generation to transmission is not more than 0.05 seconds.

[0048] Preferably, the platform interaction storage module comprises: an interface adaptation unit, which establishes a data interaction link with the threshold early warning analysis module, develops a data calling interface conforming to the Predict Steel Core research platform specification, and receives and formats converts early warning data, prediction data and original sampling data; a classified storage unit, which divides the storage area according to the data type, indexes and stores the original sampling data according to the monitoring point and the timestamp, and stores the prediction data and the early warning result in association, and compresses the historical data storage space by using a sparse matrix storage method; a data indexing unit, which generates unique index identifiers for various types of data based on an encryption hash function, establishes a mapping relationship table of indexes and storage addresses, optimizes the data retrieval path, and improves the data calling efficiency of the Predict Steel Core research platform; and a permission management unit, which receives the permission control signal of the system control and coordination module, manages the access permission of the data calling interface in stages, records the data read and write operation log, and ensures the security and traceability of the stored data.

[0049] Specifically, the efficient operation and maintenance of the platform interaction storage module realizes data security storage and convenient calling through different unit function subdivision, supporting system data management capability. In the implementation process, the interface adaptation unit adopts TCP / IP protocol to establish a data link with the threshold early warning analysis module, the link connection timeout time is set to 8 seconds, the RESTful calling interface conforming to the pre-research platform specification is developed, the interface supports GET and POST two request modes, the early warning data, prediction data and the like are converted into JSON format, and the format conversion success rate needs to reach 100%. The classified storage unit divides 3 independent storage areas, the original sampling data is stored according to the directory structure of “monitoring point-date”, and the storage space occupied by a single data is controlled within 64 bytes; the prediction data and the early warning result adopt an associated storage mode, and the storage density is set to 512 data per sector. The data index unit adopts SHA-256 encryption hash function to generate a 256-bit unique index, establishes a mapping relationship table of the index and the storage address, the table entry update frequency is 0.05 seconds / time, and when the Predict Steel Core pre-research platform calls data, the index matching time-consuming does not exceed 0.01 seconds. The permission management unit receives the 8-bit permission control signal of the system control and coordination module, divides the access permission into 3 levels of administrator, operator and visitor, the administrator can read and write all data, the operator can only read the prediction and early warning data, the visitor can only view the early warning result, 128-bit operation logs are synchronously recorded, and the log storage time is not less than 90 days.

[0050] The adaptive space-time sparse sampling reconstruction model is the core algorithm model for efficient acquisition of monitoring data in the application. The essence of the model is a dynamic sampling strategy generation tool that combines spatial distribution characteristics and time domain signal rules. The model can reduce redundant acquisition while ensuring data quality by accurately selecting effective monitoring data. The implementation process of the model can be divided into three key steps: parameter initialization, model loading, and spatial layout data acquisition of the power steel structure monitoring points. The monitoring point spacing reference value is determined according to different structure types such as transmission towers and substation supports (usually 5 meters, with key nodes encrypted to 2 meters). At the same time, the initial time domain characteristics of the vibration signal are collected, including the frequency range (0.1-50Hz), the fluctuation amplitude reference value, and other basic parameters. The second step is to perform space-time characteristic analysis. In the spatial dimension, the stress-strain correlation of different monitoring points is calculated. The area with a correlation degree higher than 0.8 is determined as a data redundancy area, and the area with a correlation degree lower than 0.3 is determined as a key monitoring area. In the time domain dimension, the signal fluctuation amplitude is tracked in real time. When the fluctuation exceeds the reference value by 30%, it is marked as a high dynamic period, and when it is lower than 5%, it is marked as a low dynamic period. The last step is to generate a dynamic sampling strategy. For the spatial redundancy area, the sampling density is reduced (the interval is extended to twice the original interval), the key monitoring area is improved (the interval is shortened to half of the original interval), the high dynamic period is shortened (the sampling period is 0.01 seconds), and the low dynamic period is extended (the period is 0.05 seconds). The strategy is sent to the distributed sensing nodes for execution. The core function of the model is to replace the traditional fixed full sampling mode, control the data volume from the source, and significantly reduce the bandwidth pressure of subsequent data transmission and the computational load of processing links. At the same time, it avoids resource waste caused by excessive sampling, lays a data foundation for the efficient operation of the monitoring system, and provides a reference for the development of similar models.

[0051] The explicit time-domain physical information dynamic response prediction model is the core tool for realizing the state prediction of the electric power steel structure. The model is a structure response deduction model that combines data-driven and physical laws. By combining monitoring data and structure physical parameters, the model realizes accurate prediction of the future state of the structure. The implementation process of the model is carried out around the three steps of "parameter loading-equation construction-numerical solution". In the parameter loading stage, after the model is started, the basic physical parameters of the electric power steel structure are called from the system storage module, including geometric parameters (rod length error ±1mm, cross-section size accuracy 0.1mm), material properties (elastic modulus 206GPa, Poisson's ratio 0.3, density 7850kg / m³), and sparse sampling data transmitted by the data transmission module are received as the input source of the external excitation term. In the equation construction stage, based on the above parameters, the structure mass matrix and stiffness matrix are calculated, and the dynamic response differential equation including the second-order time derivative term is constructed based on the explicit time-domain theory. The damping matrix adopts the Rayleigh damping model, which is dynamically corrected through the mass proportional damping coefficient (initially 0.02) and the stiffness proportional damping coefficient (initially 0.005). The correction basis is the mean and variance of the sparse sampling data (the mass proportional coefficient is adjusted positively by 1% for every change in the mean, and the stiffness proportional coefficient is adjusted inversely by 0.5% for every change in the variance). In the numerical solution stage, the explicit central difference method is used to iteratively solve the equation, with a solution step size of 0.001 seconds, a convergence precision of 1e-6, and an iteration number limit of 1000 times. After the residual error is stable, the predicted data of the structure displacement, stress and other parameters at different times are output. The function of the model is to convert discrete monitoring data into continuous structure state evolution law, breaking through the limitation of pure data-driven model lacking physical constraints, and improving the accuracy and reliability of structure state evaluation, providing scientific prediction basis for safety warning.

[0052] The online incremental learning dynamic threshold early warning algorithm is the core algorithm for realizing real-time research and judgment of the safety state in the application, which is an intelligent analysis tool for continuously optimizing the early warning standard based on historical data and real-time prediction results, and realizes accurate identification of structural abnormalities through dynamic updating of the threshold. The implementation process of the algorithm includes four key links: first, historical data preprocessing, when the algorithm starts, the historical database of the Predict Steel Core research platform is called to obtain the monitoring data of the past 3 years (the data volume is controlled within 100 GB), and the rule of mean ± 5 times standard deviation is used to eliminate outliers to obtain the effective historical data set; second, initialize the threshold parameter, calculate the basic threshold of different monitoring points and different working conditions based on the effective historical data, and set the incremental learning rate (initially 0.015) and the prediction data weight coefficient (initially 0.6), the learning rate is adjusted according to the fluctuation degree of the historical data, and the weight coefficient is dynamically optimized with the prediction accuracy (the prediction deviation is lower than 3% to increase to 0.7, higher than 8% to decrease to 0.4); then, the threshold dynamic updating is executed, the output data of the physical information prediction model is received once every 0.5 seconds, the current threshold is iteratively updated according to the formula "last time threshold × (1- learning rate) + learning rate × (predicted data × weight coefficient + historical maximum value × (1- weight coefficient))", and the fluctuation range of the threshold is controlled within ± 10%; finally, the early warning judgment is performed, the absolute deviation of the real-time prediction data and the dynamic threshold is calculated, when the deviation exceeds 15% of the threshold, the abnormality is marked, and the three-level early warning signal is generated according to the abnormality degree. The algorithm compares the structure state with the safety standard in real time, identifies potential hidden dangers in time, solves the problem that the traditional fixed threshold is difficult to adapt to the structure service stage and the change of environmental conditions, greatly reduces the early warning lag and the false alarm rate, and provides accurate risk prompt for the safety production of the electric power steel structure.

[0053] The Predict Steel Core research platform is a data management and interaction hub supporting the operation of the whole system in the present application, and is an integrated software platform specially designed for the monitoring scene of electric power steel structures, including communication protocol specification, data storage management, interface adaptation and other core functions, providing basic support for the collaborative operation of various modules. The implementation of the platform is carried out around the four dimensions of "protocol definition-data storage-interface development-right management": in terms of protocol definition, the platform formulates unified communication protocol specifications, clearly defines the frame structure of data transmission (16-byte frame identifier in the header, core field in the middle, 4-byte CRC32 check code in the tail), field order (monitoring point number, timestamp, data value, etc.) and interface parameters (RS485 interface rate 115200bps, Ethernet interface rate 1Gbps), and ensures the standardization of data interaction between the acquisition end and the processing end; in terms of data storage management, the platform adopts a hierarchical storage architecture, divides the original sampling data area, the prediction data area and the historical database area, stores the original data according to the "monitoring point-date" directory, stores the prediction data and the early warning result in association, generates a 256-bit unique index using the SHA-256 hash function, establishes the mapping relationship between the index and the storage address, and improves the data retrieval efficiency (index matching time ≤0.01 seconds); in terms of interface development, the platform provides multiple types of interfaces, including data receiving interface (buffer size 64KB, timeout time 5 seconds), data calling interface (RESTful specification, supporting GET / POST requests), and algorithm calling interface (for parameter calling of prediction model and early warning algorithm); in terms of right management, the platform sets three levels of rights of administrator, operator and visitor, the administrator can read and write all data, the operator can only read prediction and early warning data, the visitor can only view early warning results, and 128-bit operation logs (storage ≥90 days) are recorded synchronously. The role of the platform is to realize the standardized transmission, standardized storage and convenient calling of monitoring data, break the data barriers between modules, ensure the collaborative and efficient operation of the whole link of sampling, transmission, prediction and early warning, and provide key platform support for the landing of the system technical scheme.

[0054] As Figure 2As shown, an efficient acquisition and processing system for power steel structure safety production monitoring data, the system operation includes: step S1, calling the distributed sensing node of the space-time sparse sampling module, according to the power steel structure monitoring point layout and vibration signal characteristics, generating dynamic sampling strategy through adaptive space-time sparse sampling reconstruction model, executing multi-measuring point synchronous data acquisition operation; step S2, receiving the collected data through the data transmission scheduling module, packaging and processing according to the Predict Steel Core pre-research platform communication protocol, combining the link bandwidth state to adopt time division multiplexing mechanism to transmit to the physical information prediction module; step S3, in the physical information prediction module, analyze and extract the structure basic parameters, combine the sparse sampling data correction term to build the differential equation of the explicit time domain physical information dynamic response prediction model, and solve the structure state prediction data by using the explicit integral algorithm; step S4, import the prediction data into the threshold early warning analysis module, load the online incremental learning dynamic threshold early warning algorithm, calculate the real-time dynamic threshold by using the parameter library obtained by training the historical data, and compare and analyze the prediction data and the threshold; step S5, receiving the early warning analysis result and various monitoring data through the platform interactive storage module, storing according to the Predict Steel Core pre-research platform storage specification, generating encrypted index and establishing data calling interface; step S6, real-time acquisition of different module operation parameters and data processing state through the system control coordination module, combining the early warning analysis result and the transmission delay parameter, dynamically adjusting the sampling frequency, the solving step length and the transmission time slot allocation strategy of different modules.

[0055] The system has the primary advantage of solving the problems of redundancy and poor adaptability of traditional data collection. In view of the problem of invalid data accumulation caused by fixed full-sampling of the prior art, the space-time sparse sampling module generates a dynamic sampling strategy by analyzing the spatial distribution of the power steel structure monitoring points and the time domain characteristics of the vibration signals, and filters effective data from the collection source. The data transmission scheduling module then encapsulates the data according to the pre-research platform protocol and transmits them using the time division multiplexing mechanism. The combination of the two greatly reduces redundant data, reduces bandwidth occupation, and significantly improves data flow efficiency. Secondly, the system accurately solves the problem of rigid early warning threshold. In the face of the defects that the traditional fixed threshold cannot adapt to the dynamic changes of the structure, the physical information prediction module fuses the physical parameters such as the geometry and material of the steel structure to construct a model and generate high-precision state prediction data. The threshold warning analysis module continuously absorbs historical data to optimize the threshold library through online incremental learning algorithm, realizes real-time updating of the threshold, fully adapts to the structure response under different service stages and environments, and effectively reduces the early warning lag and false alarm. In addition, the multi-module collaborative support forms a comprehensive advantage. The classification storage mechanism of the platform interactive storage module improves the data access efficiency, and the system control coordination module dynamically adjusts the running parameters of each module to ensure efficient connection of the whole link of sampling, transmission, prediction and early warning. This collaborative design not only overcomes the defects of the prior art, but also realizes the efficiency and accuracy of monitoring data processing, providing strong technical support for the safe production of power steel structures.

[0056] In the description of the present application, it should be pointed out that, unless otherwise specified and limited, the terms "arrangement", "installation", "connection", "connection", "fixing" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be direct connection, or indirect connection through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0057] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various equivalent changes, modifications, replacements and variations of the embodiments can be made without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalent scope.

Claims

1. An efficient acquisition and processing system for monitoring data of safe production of an electric power steel structure, characterized in that, The application relates to a power steel structure monitoring system based on Predict Steel Core research platform. The system comprises a space-time sparse sampling module, a data transmission scheduling module, a physical information prediction module, a threshold early warning analysis module and a platform interactive storage module. The space-time sparse sampling module generates a dynamic sampling strategy based on the spatial distribution density of power steel structure stress-strain monitoring points and time domain characteristics of vibration signals, and executes a data collection operation through distributed sensing nodes. The data transmission scheduling module receives collected data output by the space-time sparse sampling module, encapsulates the data according to a communication protocol specification of the Predict Steel Core research platform, and adopts a time division multiplexing transmission mechanism to transmit the encapsulated data to a data processing unit. The physical information prediction module obtains transmission data from the data transmission scheduling module, imports an explicit time domain physical information dynamic response prediction model, constructs a dynamic response equation in combination with geometric parameters and material properties of the power steel structure, and solves the equation to generate structure state prediction data. The adaptive space-time sparse sampling reconstruction model in the space-time sparse sampling module satisfies: Wherein, is a sparse sampling data matrix, is a space-time sparse sampling weight coefficient, is a power steel structure monitoring point space base function, is a signal time domain sparse base function, is a full-quantity monitoring data matrix, is the dimension of the space base function and the time domain sparse base function respectively, is a sampling step factor; the data transmission scheduling module determines the transmission data priority based on the adaptive space-time sparse sampling reconstruction model output , wherein the priority coefficient , is a matrix trace operation, T is a matrix transpose; The explicit time-domain physical information dynamic response prediction model in the physical information prediction module satisfies: wherein, is a power steel structure displacement response function, is a structure damping matrix, is a structure stiffness matrix, is an external load function, is a sparse sampling data-driven correction term; the structure parameters are calculated based on the Predict Steel Core pre-research platform , is a structure mass matrix, is a damping coefficient, and is updated by the following formula: , , is an initial damping coefficient, is the mean and variance operation, respectively; The online incremental learning dynamic threshold early warning algorithm in the threshold early warning analysis module meets: Wherein, is a dynamic early warning threshold at time t, is an incremental learning rate, is a prediction data weight coefficient, is a predicted displacement output by an explicit time domain physical information dynamic response prediction model, is historical displacement response data stored by a Predict Steel Core pre-research platform, is a historical time variable; early warning judgment basis , is a threshold amplification coefficient, and , is a sampling data anomaly threshold, is a counting operation; The platform interaction storage module optimizes data storage based on the Predict Steel Core pre-research platform, and stores indexes wherein, is an exclusive or operation, is an encryption hash function; data compression rate , is the compressed storage data, is a data capacity calculation function; during compression, the sparse basis function of the adaptive space-time sparse sampling reconstruction model is called Transform the time series data to meet .

2. The system according to claim 1, wherein, The system control coordination module adjusts the running parameters of different modules by the following formula: Wherein, is the running weight of the qth module, is the data processing rate of the module at time t, is the reference processing rate; The solving efficiency of the explicit time domain physical information dynamic response prediction model is determined: is the model solving time consumption, is the data volume counting function; and the analysis result is dynamically corrected in combination with the online incremental learning dynamic threshold early warning algorithm: is the early warning proportion at time t, and alert(t) is the early warning signal set.

3. The system of claim 1, wherein, The threshold early warning analysis module receives prediction data from the physical information prediction module, loads an online incremental learning dynamic threshold early warning algorithm, and performs real-time threshold comparison and analysis by using a threshold parameter library obtained by training historical monitoring data. The platform interactive storage module establishes a data interaction link with the threshold early warning analysis module, stores early warning analysis results and original monitoring data in a classified manner according to a data storage format of the Predict Steel Core research platform, and provides a data calling interface. The system control coordination module establishes a control signal connection with the space-time sparse sampling module, the data transmission scheduling module, the physical information prediction module, the threshold early warning analysis module and the platform interactive storage module, and dynamically adjusts running parameters and data flow time sequences of different modules according to a power steel structure monitoring scene. The data transmission scheduling module comprises a protocol analysis unit, a data encapsulation unit and a transmission scheduling unit. The protocol analysis unit receives collected data output by the space-time sparse sampling module, reads power steel structure monitoring point numbers and sampling time stamp information carried in data headers, analyzes data types and formats according to field definitions of a communication protocol of the Predict Steel Core research platform, extracts valid data fields and eliminates redundant identification information. The data encapsulation unit binds valid data and monitoring point attribute parameters by calling an encapsulation function defined by the protocol according to the analyzed data types, generates an encapsulated data packet comprising a data check code, and generates the check code through exclusive or operation of the collected data and module running parameters. The transmission scheduling unit obtains a priority identifier of the encapsulated data packet, allocates a transmission time slot according to a time division multiplexing mechanism in combination with bandwidth occupation of a current communication link, and directly sends the data packet to a receiving port corresponding to the physical information prediction module. The state feedback unit monitors a packet loss rate and a delay parameter in a data transmission process in real time, converts monitoring results into control signals, and feeds back the control signals to the system control coordination module to provide data support for running parameter adjustment.

4. The system according to claim 1, wherein, The physical information prediction module comprises: a parameter import unit that obtains encapsulated data through a receiving port of a data transmission scheduling module, parses and extracts basic parameters of a power steel structure, such as geometric dimensions, material elastic modulus and Poisson's ratio, and performs standardized conversion according to parameter format requirements of an explicit time-domain physical information dynamic response prediction model; an equation construction unit that calculates a structure mass matrix and a stiffness matrix based on the converted basic parameters, combines external load monitoring data and sparse sampling data correction items, constructs a dynamic response differential equation including a time derivative item, and determines initial boundary conditions and a solution domain of the equation; a model solving unit that performs numerical solution of the dynamic response differential equation by using an explicit integration algorithm, sets a solution step and a convergence criterion, and iteratively calculates to obtain structure displacement and stress prediction data at different times, which are stored in a temporary data buffer; and a result output unit that reads the prediction data from the temporary data buffer, organizes and encapsulates them according to a preset format, adds a data validity identifier, and then transmits them to the threshold early warning analysis module, while feeding back solving process parameters to the platform interaction storage module.

5. The system of claim 1, wherein, The threshold early warning analysis module comprises: a data receiving unit that establishes a communication link with the physical information prediction module, receives an encapsulated package including structure state prediction data, parses and extracts prediction values and corresponding timestamp information, and synchronously calls a historical database interface of the Predict Steel Core research platform to obtain historical monitoring data under similar working conditions; a threshold updating unit that loads an online incremental learning dynamic threshold early warning algorithm, inputs the historical monitoring data and the current prediction data into the algorithm model, iteratively updates weight coefficients and learning rates of threshold calculation, and generates real-time dynamic early warning thresholds; a comparison and analysis unit that compares the current structure state prediction data with the dynamic early warning thresholds point by point, calculates deviation values and compares them with a judgment threshold, marks data points that exceed the judgment range and records corresponding monitoring positions and time information; and an early warning generation unit that statistically analyzes the marked abnormal data points, generates early warning signals according to the degree of abnormality, adds signal generation time and associated monitoring point information, and transmits them to the platform interaction storage module.

6. The system of claim 1, wherein, The platform interaction storage module comprises an interface adaptation unit, a threshold early warning analysis module, a data interaction link, a data calling interface conforming to the Predict Steel Core pre-research platform specification, early warning data, prediction data and original sampling data receiving and format conversion; a classification storage unit, a storage area is divided according to the data type, the original sampling data is stored according to the monitoring point and the time stamp, the prediction data and the early warning result are stored in association, the history data occupies the space is compressed by adopting the sparse matrix storage mode; a data index unit, a unique index identification of various data is generated based on an encryption hash function, an index and a storage address mapping relationship table is established, a data retrieval path is optimized, and the data calling efficiency of the Predict Steel Core pre-research platform is improved; a permission management unit, a permission control signal of a system control coordination module, an access permission of a data calling interface is managed in stages, a data read-write operation log is recorded, and the safety and traceability of the stored data are ensured.

Images