A system and method for monitoring and managing displacement of civil structures of a thermal power plant

Abstract

The application relates to the technical field of health monitoring and management of civil structures, and discloses a kind of thermal power plant civil structure displacement monitoring management system and method, the method includes through multi-source heterogeneous sensor microsecond level synchronous acquisition, space-time unified data processing, structure type classification modeling, displacement component decoupling estimation, voxelized three-dimensional deformation field reconstruction, pattern matching identification and three-level dynamic early warning mechanism.The application can realize high-precision, real-time and intelligent structure displacement monitoring and risk prediction, and form a closed-loop emergency response through linkage with a DCS system.

Description

Technical Field

[0001] This invention relates to the field of civil engineering structure health monitoring and management technology, and more specifically, to a displacement monitoring and management system and method for civil engineering structures in thermal power plants. Background Technology

[0002] The civil engineering structure of a thermal power plant, centered on key structures such as the main plant building, boiler steel frame, turbine foundation, chimney, and cooling tower, is susceptible to slow displacements or sudden deformations on the order of micrometers to millimeters during long-term service due to the combined effects of multiple factors, including temperature gradients, unit vibration, foundation settlement, accumulated coal ash loads, and environmental corrosion. This directly impacts equipment alignment accuracy, pipeline stress distribution, and overall operational safety. Therefore, effective monitoring and management of the displacement status of the civil engineering structure in a thermal power plant is crucial for ensuring the safe and stable operation of the unit.

[0003] Existing methods for monitoring displacement of civil structures in thermal power plants mainly employ discrete point measurements, relying on manual intervention or periodic fieldwork. The acquired data is discontinuous in time and mostly single-dimensional in space, resulting in low system integration and failing to fully meet the demands for high-precision, real-time, and intelligent structural displacement management. Some solutions focus on process collaboration and data visualization on information platforms, but lack in-depth research on the synchronous access, spatiotemporal alignment, and data fusion mechanisms of multi-source heterogeneous sensors (such as GNSS, hydrostatic levels, inclinometers, and fiber optic gratings) in the complex electromagnetic and high-temperature environment of thermal power plants. Furthermore, there is a lack of dedicated analytical models for the displacement evolution characteristics of typical structures such as turbine foundations and chimneys, leaving the diagnosis of structural health and risk prediction without refined support.

[0004] Therefore, how to integrate multi-source heterogeneous sensor data, achieve unified spatiotemporal modeling, adaptively analyze according to different structural types, and support dynamic early warning and emergency response for the degree of displacement of civil engineering structures, so as to improve the intelligence level of structural health monitoring and the ability to ensure operational safety, has become an urgent technical problem to be solved. Summary of the Invention

[0005] This invention provides a management system and method for monitoring and managing the displacement of civil structures in thermal power plants. It solves the technical problems in the prior art of integrating multi-source heterogeneous sensor data, achieving unified spatiotemporal modeling, adaptive analysis according to different structural types, and supporting dynamic early warning and emergency response for the degree of displacement of civil structures, so as to improve the level of intelligence of structural health monitoring and the ability to ensure operational safety.

[0006] This invention provides a management system and method for monitoring and managing the displacement of civil structures in thermal power plants, including: Firstly, a method for monitoring and managing displacement of civil structures in thermal power plants includes: Original monitoring data of key civil engineering structures in thermal power plants are collected from multiple dimensions to generate original monitoring datasets. The original monitoring data is obtained by multi-source heterogeneous sensors deployed on key civil engineering structures. The multi-source heterogeneous sensors include a global navigation satellite system receiver, a hydrostatic level, a dual-axis inclinometer, and a fiber optic strain sensor. The original monitoring dataset is subjected to spatiotemporal synchronization processing to obtain a spatiotemporally unified monitoring dataset; Among them, multiple sensors are triggered by the same pulse signal to perform synchronous sampling, and a timestamp sequence is generated based on the built-in high-stability crystal oscillators of multiple sensors. Data from different sampling frequencies are uniformly interpolated to a preset reference sampling frequency, and the physical quantities output by each sensor are converted to a unified spatial coordinate system based on the control network of the thermal power plant area, thus obtaining a spatiotemporally unified monitoring dataset. Based on the structural types and service environment characteristics of different structures in thermal power plants, the monitoring dataset is classified and divided to generate multiple structural monitoring categories. The structural monitoring categories include at least the turbine foundation monitoring category and the chimney monitoring category. A classification analysis model was established based on multiple structural monitoring categories to obtain displacement component estimation results for different structural categories; Among them, for the turbine foundation monitoring category, a joint analysis model of foundation settlement-vibration response based on the coupling of unit operation status is constructed. The vertical displacement sequence measured by the hydrostatic level is the main input, and the horizontal tilt angle change measured by the inclinometer and the local strain gradient measured by the fiber optic grating sensor are the auxiliary constraints. Turbine speed, load rate and bearing vibration spectrum are introduced as external excitation parameters, and the foundation settlement rate and vibration-induced additional displacement components are dynamically estimated by Kalman filter. For chimney monitoring categories, a hot air gravity multi-field coupled deformation analytical model is constructed. Based on the three-dimensional position time series data provided by the global navigation satellite system receiver, it is combined with chimney wall temperature distribution data, real-time wind speed and direction data and flue gas emission flow data. By using the finite element shape function mapping relationship to separate the axial elongation component caused by thermal expansion, the lateral sway component caused by wind-induced vortex-induced vibration, and the creep settlement component under long-term load, the displacement component estimation results for different structural categories are obtained. Based on the displacement component estimation results, the displacement evolution rate, cumulative offset and deformation stability index of different structural categories are obtained, and a displacement evaluation index set is generated. The displacement assessment index set is compared with the preset displacement safety threshold range. When any index exceeds the corresponding threshold range, a graded early warning message is generated and pushed to the operation and maintenance management terminal.

[0007] Furthermore, spatiotemporal synchronization processing includes: Differential calculation is performed on the dual-frequency carrier phase signal output by the global navigation satellite system receiver to obtain a high-precision three-dimensional coordinate sequence; The analog signal of the liquid level height from the hydrostatic level instrument is converted from analog to digital and the ambient temperature drift is corrected to obtain the corrected liquid level height data. The wavelength shift signal of the fiber optic strain sensor is demodulated and thermal strain compensation is performed in combination with distributed temperature sensing data to generate a compensated strain signal. The Kalman filter algorithm is applied to the raw angle data output by the MEMS inclinometer to suppress high-frequency noise and obtain filtered angle data. Subtract the non-structural displacement component caused by thermal expansion from the output signal of the temperature-compensated displacement gauge, and extract the net structural displacement component. The high-precision three-dimensional coordinate sequence, the corrected liquid level height data, the compensated strain signal, the filtered angle data, and the net structural displacement components are summarized, assigned a unified timestamp, and converted to a unified three-dimensional geographic coordinate system of the plant area to generate the spatiotemporally unified monitoring dataset.

[0008] Furthermore, differentiated spatial interpolation is performed on different structure areas to generate interpolation results for each area. The spatial interpolation includes: For the turbine foundation area, spatial densification interpolation is performed using radial basis functions, and zero displacement or fixed support constraints are applied at the foundation boundary to obtain the interpolation results for the turbine foundation area. For the chimney structure, a spiral path interpolation model is constructed in cylindrical coordinates, and the rate of change of torsion angle along the height direction is introduced as an interpolation constraint parameter to obtain the interpolation results of the chimney structure. For the member connection node region of the boiler steel frame, based on the finite element mesh generation results, constrained spatial interpolation is performed using the interpolation basis function consistent with the element shape function to obtain the interpolation results of the boiler steel frame region; The three types of interpolation results are respectively filled into the voxel mesh of the corresponding structures to form a three-dimensional displacement vector field with physical consistency; The smoothness test and energy conservation verification are performed on the three-dimensional displacement vector field to identify and eliminate abnormal voxel elements that do not meet the structural mechanics continuity condition, and generate a verified voxelized three-dimensional deformation field.

[0009] Furthermore, in the process of calculating the displacement evaluation index set based on the displacement component estimation results, pattern matching and recognition are performed, including: Extract feature vectors from four dimensions of the current three-dimensional deformation field: displacement amplitude sequence, displacement rate time history, displacement acceleration peak value, and main peak frequency of the spectrum, and generate the current feature vector set. Call the typical displacement mode library in the standard model database to obtain the mode template set. The typical displacement mode library includes four types of templates: turbine foundation settlement and vibration coupling mode, chimney wind-induced vortex oscillation mode, cooling tower foundation differential settlement mode, and main plant thermal expansion accumulation mode. For each type of template, a dynamic time warping algorithm is executed with the current feature vector set to calculate the minimum cumulative distance under path alignment, thereby obtaining the minimum cumulative distance value corresponding to each template. Each minimum cumulative distance value is normalized to a similarity score between 0 and 1, generating a similarity score set for each template; A threshold judgment is performed on the similarity score set. If the similarity score of any template is greater than the preset matching threshold, it is determined that the current displacement behavior is successfully matched with the template, and the matching type and score value are recorded to generate a pattern matching result.

[0010] Furthermore, the generation of tiered early warning information includes dynamic tiered early warning decision-making, including: Real-time monitoring of the displacement rate of each structure; when the displacement rate exceeds the preset rate threshold and the duration is longer than the first time window, a first-level warning is triggered. The cumulative displacement value is calculated based on the displacement evaluation index set and compared with the proportion of the design allowable displacement limit. When the proportion exceeds the second threshold and the current displacement pattern is successfully matched, and the similarity score is higher than the third threshold, a level 2 warning is triggered. Based on the pattern matching results and the residual energy of the deformation field, the existence of a displacement surge event is detected; a displacement surge is defined as the displacement increment per unit time exceeding the fourth threshold, or the current displacement pattern matching fails and the residual energy of the deformation field exceeds the fifth threshold; a level three warning is triggered if either condition is met.

[0011] Furthermore, upon triggering a Level 3 warning, an emergency response mechanism is established, which includes: Automatically execute operation commands to cut off unnecessary power supplies and generate a power cut-off command set. The unnecessary power supplies refer to power supply circuits other than safety instrument systems, fire protection systems and emergency lighting systems. Push alarm information including location, type, and severity level to the operation and maintenance terminal; Initiate a cloud-based in-depth analysis process, specifically: retrieve the building information model, material property parameters, and historical maintenance records corresponding to the alarmed structure from the standard model database, and generate a basic information set for the alarmed structure. Based on the building information model and the reconstructed voxelized three-dimensional deformation field, a spatial superposition comparison is performed to identify the component units that have experienced abnormal displacement and their adjacent areas, and to generate abnormal area location results. Based on the handling plans for similar displacement events in historical maintenance records, multi-factor influencing parameters including the current cumulative displacement value, displacement evolution rate, ambient temperature change rate and current load conditions are extracted to construct a structural safety status assessment matrix under the influence of multiple factors. Using the finite element inversion algorithm, with the current voxelized three-dimensional deformation field as the boundary condition, the local stress-strain field of the abnormal region is reconstructed, and the safety margin of the key nodes is calculated to obtain the safety margin calculation results. Based on the safety margin calculation results and the structural safety status assessment matrix, a structural health status assessment report is generated, and a set of disposal instructions containing risk level, recommended disposal measures and priority order is pushed to the operation and maintenance management platform.

[0012] Furthermore, during the generation of tiered early warning information, a real-time response mechanism is executed, including: After completing the voxelized three-dimensional deformation field reconstruction, the edge computing host executes the real-time discrimination logic of the first, second and third level early warning conditions in parallel to obtain the early warning discrimination results; When the warning judgment result meets any level of warning triggering condition, the local emergency control interface is immediately activated, and an instruction to cut off unnecessary power is sent to the power plant's distributed control system to complete the emergency power control. Simultaneously, the alarm event is encapsulated into a structured message, which includes the structure identifier, warning level, coordinates of the location of the sudden displacement increase, matching mode type, and deformation field residual energy value, to generate a structured alarm message; The structured alarm messages are pushed to designated operation and maintenance terminals and central monitoring screens through an encrypted channel via an industrial firewall. After the alarm information is transmitted, a local caching mechanism is activated to continuously record the original sensor data stream within a preset time window after the alarm is triggered, generating a data source record for subsequent in-depth cloud analysis.

[0013] Furthermore, the smoothness test and energy conservation verification include: For each voxel element in the voxelized three-dimensional deformation field, calculate its displacement gradient with neighboring voxel elements in the X, Y, and Z coordinate directions to generate a voxel element displacement gradient set. The smoothness of the displacement gradient set of the voxel element is checked. If the absolute value of the displacement gradient in any coordinate direction exceeds the critical gradient value corresponding to the allowable strain limit of the structure material, the voxel element is marked as an abnormal element, and an abnormal voxel element mark set is generated. The strain energy density distribution of the entire structure is calculated based on the displacement vector field, and the total strain energy is obtained by integrating over the entire structure volume. The total strain energy is compared with the theoretical energy input by external loads, including the vibration power of the steam turbine generator set, wind load, and thermal expansion force, to obtain the energy deviation value. If the relative deviation between the total strain energy and the theoretical energy exceeds a preset energy conservation tolerance threshold, it is determined that there is a physical inconsistency in the deformation field, and all abnormal voxel elements in the abnormal voxel element marker set are removed. The preset energy conservation tolerance threshold is determined based on the condition number of the structural stiffness matrix and the sensor measurement uncertainty, and is usually selected as 5%-15% for concrete structures. The interpolation algorithm is triggered to recalculate the local region after removing abnormal voxel units, generating a corrected 3D deformation field.

[0014] Furthermore, the closed-loop emergency response of the voxelized three-dimensional deformation field and the distributed control system works in concert in the following ways: In the reconstructed voxelized three-dimensional deformation field, the spatial coordinate clusters of abnormal voxel units whose displacement exceeds a preset sudden increase threshold are mapped in real time to the building information model of the power plant, and the corresponding key equipment identifiers and power supply circuit numbers are extracted to generate a list of threatened equipment and a list of associated power sources. Based on the pattern matching results and displacement evolution rate, the remaining safe time window from the triggering of the warning to the expected structural intervention is dynamically calculated; based on the remaining safe time window and the warning level, a set of emergency operation instructions with graded delays or immediate execution is generated. The emergency operation instruction set is directly sent to the control logic layer of the power plant's distributed control system through a secure encryption protocol, driving its execution module to complete at least one of the following operations in a preset priority order: cutting off the power supply to unnecessary loads in the associated power supply list, adjusting the load rate of the affected units, and starting the auxiliary support system or linking the fire protection and ventilation system to isolate the area. While performing emergency operations, the sensor data streams of the abnormal voxel unit and its surrounding area are continuously collected, the voxelized three-dimensional deformation field is updated in real time, and the displacement response attenuation rate and deformation field stability index after control intervention are calculated. The displacement response attenuation rate, deformation field stability index and the execution status of the emergency operation instruction set are compared and fed back in real time. If the deformation does not stabilize as expected or continues to deteriorate, the upgraded handling process is automatically triggered or human intervention is notified, forming a closed-loop emergency response link of perception, decision-making, control and evaluation.

[0015] Furthermore, it also includes a model update mechanism: The cloud analytics center periodically aggregates de-identified displacement feature vector sequences from various edge computing nodes. These feature vectors include displacement amplitude, displacement rate, displacement acceleration, and the main peak frequency of the spectrum, generating an aggregated displacement feature vector set. Based on the aggregated displacement feature vector set, a multivariate time series training set is constructed, and it is divided into four subsets according to the structure type: turbine foundation, chimney, cooling tower and main plant, to obtain the classification training subset; For each of the aforementioned classification training subsets, a dedicated long short-term memory neural network prediction model is trained independently. A sliding window mechanism is used to input the feature vectors of the past N time steps and output the displacement evolution trend of the next M time steps to obtain a preliminary prediction model. When training the preliminary prediction model, a physical constraint loss function is introduced to force the prediction results to satisfy the structural continuity condition and the energy conservation principle, thus obtaining the physical constraint prediction model. Online evaluation is performed based on the physical constraint prediction model. When newly collected data causes the model prediction error to exceed the preset tolerance threshold, the online fine-tuning process of the model is triggered to generate updated model parameters. The updated model parameters and corresponding early warning threshold parameters are packaged and distributed to relevant edge computing nodes to complete the model parameter distribution and deployment.

[0016] Secondly, a displacement monitoring and management system for civil structures in a thermal power plant includes: Data acquisition module: used to collect raw monitoring data of key civil engineering structures in thermal power plants from multiple dimensions and generate monitoring datasets; Data processing module: used to perform spatiotemporal synchronization processing on the monitoring dataset to obtain a spatiotemporally unified monitoring dataset; Data segmentation module: used to classify and segment the spatiotemporally unified monitoring dataset according to the structural type and service environment characteristics of different structures in the thermal power plant, and generate multiple structural monitoring categories; The index generation module is used to establish a classification analysis model based on multiple structural monitoring categories, obtain displacement component estimation results for different structural categories, and, based on the displacement component estimation results, obtain displacement evolution rate, cumulative offset and deformation stability index for different structural categories, and generate a displacement evaluation index set. Early warning module: It is used to compare the displacement evaluation index set with the preset displacement safety threshold range. When any index exceeds the corresponding threshold range, it generates a graded early warning information and pushes it to the operation and maintenance management terminal.

[0017] The beneficial effects of this invention are as follows: This invention achieves microsecond-level synchronous sampling of multi-source heterogeneous sensors through hardware pulse triggering and high-stability crystal oscillator collaboration, solving the problem of data spatiotemporal asynchrony in the complex electromagnetic environment of thermal power plants; it constructs dedicated analysis models driven by physical mechanisms for typical structures such as turbine foundations and chimneys, achieving accurate decoupling and estimation of displacement components; it adopts a spatial interpolation method combining voxelized three-dimensional deformation fields and structural mechanical constraints, improving the physical consistency of displacement field reconstruction; it introduces pattern matching and multi-level dynamic early warning mechanisms, upgrading risk identification from threshold exceedance judgment to behavioral pattern recognition, significantly improving the accuracy and timeliness of early warnings; the emergency response mechanism is deeply integrated with the power plant's DCS system, realizing closed-loop linkage from perception, diagnosis to control. By deeply coupling the spatial positioning accuracy of the voxelized three-dimensional deformation field with the real-time control capability of the DCS system, it achieves millisecond-level closed-loop linkage from precise structural anomaly positioning to automatic protection measure execution, greatly shortening emergency response time and improving the intelligent level of active safety protection in power plants. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of a method for monitoring and managing the displacement of civil structures in a thermal power plant, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a displacement monitoring and management system module for civil structures in a thermal power plant, provided in an embodiment of the present invention. Detailed Implementation

[0019] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0020] At least one embodiment of the present invention discloses a displacement monitoring and management system and method for civil structures in thermal power plants, comprising: like Figure 1 As shown, a method for monitoring and managing the displacement of civil structures in a thermal power plant includes the following steps: Step 1: Collect raw monitoring data of key civil engineering structures in thermal power plants from multiple dimensions and generate a monitoring dataset; Step 2: Perform spatiotemporal synchronization processing on the monitoring dataset to obtain a spatiotemporally unified monitoring dataset; Step 3: Based on the structural types and service environment characteristics of different structures within the thermal power plant, classify and divide the spatiotemporally unified monitoring dataset to generate multiple structural monitoring categories; The monitoring dataset is divided into monitoring categories based on the structure attributes: turbine foundation, chimney, boiler steel frame, and cooling tower. Pre-set physical mechanism-driven analysis models are called for each category, with the turbine foundation model incorporating unit operating parameters as external excitation. Step 4: Establish a classification analysis model based on multiple structural monitoring categories to obtain displacement component estimation results for different structural categories; and based on the displacement component estimation results, obtain the displacement evolution rate, cumulative offset, and deformation stability index for different structural categories to generate a displacement evaluation index set. Step 5: Compare the displacement evaluation index set with the preset displacement safety threshold range. When any index exceeds the corresponding threshold range, generate a graded early warning message and push it to the operation and maintenance management terminal.

[0021] Based on the above steps, the present invention provides a preferred embodiment, specifically: In a typical 600MW coal-fired power plant, a multi-source heterogeneous sensor network is first deployed at key civil engineering structural locations. Specifically, five Global Navigation Satellite System (GNSS) receivers, model Trimble BD982, supporting L1 and L2 dual-frequency carrier phase differential modes, are deployed at the four corners and center of the No. 1 turbine generator foundation; four sets of hydrostatic levels are installed at the junction of the foundation slab and the concrete cushion layer, each set containing a temperature-compensated liquid level sensor and a 24-bit analog-to-digital converter, with the liquid level sensor integrating a Pt100 platinum resistance thermometer for real-time liquid temperature measurement; three sets of dual-axis inclinometers are installed near the turbine bearing housing, each inclinometer having a built-in MEMS inertial measurement unit (IMU) with a sampling frequency of 100Hz; eight fiber optic strain sensors are attached along the principal stress direction of the turbine foundation, each sensor being equipped with a distributed Raman temperature sensing channel for thermal strain compensation. All the aforementioned sensors are connected to the same edge computing host via industrial Ethernet or RS485 bus. This host is equipped with an FPGA chip to generate a unified hardware pulse signal, which is connected to the external trigger pins of each sensor via dedicated trigger cables. Each sensor is equipped with a high-stability temperature-compensated crystal oscillator with a frequency stability better than ±0.1ppm. Upon receiving the hardware pulse, it generates a timestamp sequence with microsecond-level precision based on the local crystal oscillator, ensuring that all raw monitoring data are strictly aligned in the time dimension. The raw monitoring dataset contains five types of physical quantities: three-dimensional coordinates (X, Y, Z) output from the GNSS receiver, liquid level height h output from the hydrostatic level, pitch angle θ and roll angle φ output from the dual-axis inclinometer, center wavelength offset output from the fiber optic grating sensor, and ambient temperature T output from the built-in temperature sensor of each sensor.

[0022] After receiving the raw monitoring data, the edge computing host initiates a spatiotemporal synchronization processing flow. First, it performs non-differential, non-combined precise single-point positioning calculations on the L1 / L2 dual-frequency carrier phase observations output from the GNSS receiver. Using precise ephemeris and clock bias products provided by IGS, it estimates the receiver position, clock bias, and tropospheric delay parameters through Kalman filtering, ultimately obtaining a three-dimensional coordinate time series with millimeter-level accuracy. Second, it digitizes the analog voltage signal output from the hydrostatic level via a 24-bit analog-to-digital converter and reads the synchronously acquired Pt100 temperature measurement values. Based on the empirical relationship between liquid density and temperature, it performs temperature drift correction, which considers the influence of liquid density and volume expansion coefficient at the reference temperature on the density at the current temperature, thus obtaining the corrected liquid level height data. Third, a matched-filter demodulation algorithm is used to extract the center wavelength shift from the reflection spectrum signal of the fiber optic grating strain sensor. Simultaneously, distributed Raman temperature sensing data along the same path is read to obtain the temperature distribution along the fiber length. Thermo-optic coefficient and thermal expansion coefficient are used to calculate the thermally induced wavelength drift, which is subtracted from the total shift. This is then combined with the strain sensitivity coefficient and grating period parameters to generate the net strain signal. Fourth, the angular velocity and acceleration signals from the original output of the dual-axis inclinometer are first filtered by a fifth-order Butterworth low-pass filter (cutoff frequency 5Hz) to remove high-frequency vibration noise, and then input into an extended Kalman filter for attitude fusion to suppress instantaneous angle jumps caused by mechanical vibration, outputting smoothed inclinometer data. After the above processing, all data streams are aligned with a unified timestamp, and the observations in the local coordinate system are transformed to the geocentric geographic coordinate system defined by the thermal power plant control network using a seven-parameter Bursa coordinate transformation model, forming a spatiotemporally unified monitoring dataset.

[0023] Subsequently, the system automatically categorizes the monitoring points based on the component attribute tags in the Building Information Model (BIM). Each component in the BIM model has tags for structural type, service environment, and functional attributes. The system reads the BIM component ID corresponding to each sensor and classifies them into four major structural monitoring categories: turbine foundation monitoring, chimney monitoring, boiler steel frame monitoring, or cooling tower monitoring. For example, the sensor located in the No. 1 turbine foundation area is classified into the turbine foundation monitoring category and associated with the unit number #1, the bearing position (front / rear bearing), and the foundation support type (fixed / sliding support); the sensor located in the chimney elevation range of 50m–60m is classified into the chimney monitoring category and associated with the chimney height range of 50–60m, the flue gas temperature zone (high temperature zone > 150℃), and the windward side exposed to wind load. Each monitoring category maintains its own sensor topology graph. The nodes in the graph are unique sensor IDs (such as GNSS_01, IL_03, etc.). The edge weights are determined by multiplying the physical distance between the sensors by the connection stiffness of the structural components in which they are located. The connection stiffness is calculated from the material properties and cross-sectional geometric parameters in the BIM model.

[0024] For different structural monitoring categories, the system calls the corresponding classification analysis model to estimate displacement components. For the turbine foundation monitoring category, a joint analysis model of foundation settlement and vibration response is constructed. This model uses the vertical displacement sequence output by the hydrostatic level as the state variable, and the state vector includes three components: settlement, settlement rate, and vibration-induced displacement. The observation equation includes the horizontal rotation angle output by the inclinometer and the strain gradient output by the fiber optic grating sensor, which together constitute the observation constraints. Simultaneously, the system obtains the speed, load rate, and bearing vibration acceleration spectrum of Unit 1 from the power plant's DCS system in real time via the OPC UA protocol, and uses this as the external excitation input to the state transition equation. An unscented Kalman filter is used to iteratively update the state vector, outputting the posterior estimates of settlement rate and vibration-induced displacement every 10 seconds. For the chimney monitoring category, a thermal-wind-gravity multi-field coupled deformation analytical model is constructed. The model uses the three-dimensional position and time sequence output from a GNSS receiver as the main input, integrating the temperature field of the chimney wall acquired by an infrared thermal imager, the wind speed and direction output from the plant's meteorological station, and the flue gas flow rate output from the desulfurization system's DCS. Based on the pre-established finite element model of the chimney, its shape function matrix is ​​extracted, and the total displacement is decomposed into three parts: the axial elongation component caused by thermal expansion is obtained by integrating the temperature difference between the chimney wall and the reference temperature along the height direction and multiplying it by the linear expansion coefficient; the lateral oscillation component caused by wind-induced vortex-induced displacement is calculated based on the drag coefficient, air density, the square of the wind speed, the cube of the chimney height, and the bending stiffness; and the settlement component caused by concrete creep. The three components are decoupled and separated by solving the overdetermined equations using the least squares method. The equations show that the measured total displacement is equal to the sum of the products of each component and its corresponding shape function submatrix.

[0025] After obtaining the displacement component estimation results for each structural category, the system performs differentiated spatial interpolation for different structural regions. In the turbine foundation region, spatial densification interpolation is performed using multiple quadratic radial basis functions. The interpolation function considers the distance between the interpolation point and the sensor location, as well as shape parameters. The shape parameter is set to 0.8 times the average sensor spacing. The center point of the interpolation kernel function is strictly set at the physical coordinates of the deployed sensors. At the junction of the foundation and the concrete pad, a zero-displacement boundary condition is applied, meaning that the displacement in all three directions is zero. In the chimney structure, a cylindrical coordinate system is established, with an interpolation loop set every 5 meters along the height direction, and 12 interpolation points are evenly distributed on each loop. The rate of change of the torsional angle is introduced as a derivative constraint for the radial displacement. The rate of change of the radial displacement along the height direction is equal to the product of the radius at that location and the rate of change of the torsional angle, ensuring that the interpolation results conform to the physical laws of torsional deformation of the chimney. In the boiler steel frame member node region, the system reads the tetrahedral mesh model generated by ANSYS, which contains node coordinates and element connection relationships. The displacement field inside the member was reconstructed using a 10-node Lagrange interpolation basis function consistent with the Tet10 element shape function. At the hinged ends of the member, the moment degree of freedom was released, i.e., a boundary condition with zero displacement normal derivative was applied. After interpolation, the three types of results were filled into the corresponding 10cm×10cm×10cm voxel mesh of the structure to form the initial three-dimensional displacement vector field. Subsequently, a smoothness check was performed on the voxelized deformation field: the L2 norm of the displacement gradient between adjacent voxels was calculated, and if it exceeded 1.5 times the material yield strain, it was marked as anomaly; at the same time, energy conservation verification was performed: the total strain energy was calculated by integrating stress and strain over volume, and if the deviation from the external force work exceeded 10%, the anomalous voxel elements that did not meet the structural mechanical continuity condition were removed, and finally, the verified voxelized three-dimensional deformation field was generated.

[0026] Based on the deformation field, the system calculates a set of displacement evaluation indices and performs pattern matching and recognition. First, it extracts four-dimensional feature vectors from the current three-dimensional deformation field: displacement amplitude sequence, displacement rate time history, peak displacement acceleration, and main peak frequency of the spectrum, generating the current feature vector set. The system calls the typical displacement pattern library in the standard model database, which contains four types of templates: turbine foundation settlement-vibration coupling mode, chimney wind-induced vortex-induced oscillation mode, cooling tower foundation differential settlement mode, and main plant thermal expansion accumulation mode. For each type of template, a dynamic time warping algorithm is executed with the current feature vector set to calculate the minimum cumulative distance under nonlinear alignment of the time axis. Then, the minimum cumulative distance is normalized into a similarity score, which is equal to 1 minus the ratio of the minimum cumulative distance to the maximum possible distance, generating a similarity score set. A matching threshold of 0.8 is set. If the similarity score of any template is greater than 0.8, the current displacement behavior is considered to have successfully matched that template, and the matching type and score value are recorded, generating the pattern matching result.

[0027] The system executes dynamic hierarchical early warning decisions based on the displacement assessment index set and pattern matching results. The first-level early warning trigger condition is: the displacement rate of any structure exceeds 0.05 mm / h and the duration is greater than 2 hours. The second-level early warning trigger condition is: the ratio of the cumulative displacement value to the design allowable displacement limit exceeds 70%, the current displacement pattern matches successfully, and the similarity score is higher than 0.85. The third-level early warning trigger condition is: a sudden increase in displacement is detected, defined as a displacement increment exceeding 0.3 mm within a unit time (10 minutes), or the current displacement pattern matching fails (all template similarity scores do not exceed 0.8), and the residual energy of the deformation field exceeds 15% of the total strain energy. The residual energy is calculated using the square norm of the difference between the measured displacement and the model-predicted displacement. When a Level 3 warning is triggered, the system immediately activates the emergency response mechanism: the edge computing host sends digital output commands to the power plant's distributed control system via the Modbus TCP protocol to cut off non-essential power supply circuits, retaining power only for the safety instrument system, fire protection system, and emergency lighting system; at the same time, it pushes alarm information to the operation and maintenance terminal, including the structure's BIM code (e.g., CHIMNEY-01), warning level (Level 3), abnormal voxel coordinates (x=125.3m, y=87.6m, z=45.2m), and pattern matching type (e.g., matching failed). The system synchronously initiates a cloud-based deep analysis process: retrieving the BIM model of the corresponding structure, concrete strength grade (e.g., C40), steel reinforcement ratio (e.g., 1.8%), and historical maintenance work orders from the standard model database to generate a basic information set of the alarming structure; performing Boolean intersection operations based on the BIM geometric model and the current voxelized 3D deformation field to identify the component units experiencing abnormal displacement and their adjacent areas; combining the handling plans for similar events in historical maintenance records to extract the current cumulative displacement value, displacement evolution rate, ambient temperature change rate, and current load conditions to construct a structural safety status assessment matrix; using the inverse finite element module of ABAQUS software, with the current voxelized 3D deformation field as the displacement boundary condition, reconstructing the local stress-strain field of the abnormal area, calculating the safety margin of key nodes, which is determined by the ratio of material yield strength to actual stress; finally generating a structural health status assessment report and pushing a set of handling instructions to the operation and maintenance management platform, including risk level (high, medium, low), recommended handling measures (e.g., immediate shutdown to check foundation cracks), and priority ranking.

[0028] When the system triggers a Level 3 warning, the reconstructed voxelized 3D deformation field becomes the direct basis for driving the DCS to perform emergency operations. The specific process is as follows: The system extracts anomalous voxel clusters with sudden displacement increases from the voxelized 3D deformation field and quickly maps them to the power plant's building information model (BIM) using spatial indexing based on their 3D coordinates. The model automatically associates critical equipment directly affected or threatened by the anomalous displacement (such as turbine bearings, main steam pipe supports, and generator wiring devices) with their power supply circuit numbers. For example, if an anomalous voxel cluster is located under a certain pier of the turbine foundation, the system will automatically lock the power supply circuits of auxiliary equipment such as the lubrication jacking pump and turning gear motor corresponding to that foundation.

[0029] The system uses the coupling mode matching results (such as rapid foundation settlement) and the current displacement evolution rate to estimate in real time the remaining safe time window from the triggering of the warning to the possible structural instability or equipment interference (e.g., based on rate extrapolation, it is estimated that the displacement will exceed the safety limit after 15 minutes). Based on the length of this time window and the warning level, the system dynamically generates tiered emergency instructions: for emergencies with extremely short time windows (e.g., <5 minutes), immediate execution instructions are generated; for situations with longer time windows, delayed execution or pre-execution instructions can be generated, providing a buffer period for operators to intervene.

[0030] The generated emergency operation command set (such as disconnecting unnecessary loads in list A, reducing the load of Unit #1 to 50%) is directly sent to the control logic layer of the DCS system using a secure protocol with OPC UA over TSN or Modbus TCP with AES encryption. The DCS drives its execution modules (such as PLCs and intelligent circuit breakers) to automatically complete the operation according to a preset priority order. The entire process requires no manual confirmation, ensuring that critical power disconnection or load adjustment is completed within tens of milliseconds to several seconds.

[0031] Immediately after the emergency operation is executed, the control effect evaluation phase begins, utilizing real-time feedback data to form a closed loop: Sensor data streams from the abnormal voxel cluster and its surrounding area are continuously collected, and the voxelized 3D deformation field of this area is locally reconstructed at a higher frequency (e.g., once per second). By comparing the deformation field before and after the control action, key indicators are obtained. Key indicators include displacement response attenuation rate and deformation field stability index; Displacement response decay rate: the percentage decrease in displacement rate after control.

[0032] Stability indices of the deformation field: the degree of convergence of the displacement vector direction and the change of the displacement gradient.

[0033] The calculated displacement response attenuation rate and deformation field stability index are compared in real time with the expected control effect model (based on historical successful treatment cases or simulation).

[0034] If the effect meets or exceeds expectations (e.g., attenuation rate > 70%, stability index enters the "safe" range within 10 minutes), the system determines that the emergency measures are effective, continues to monitor and awaits further manual intervention.

[0035] If the effect does not meet expectations or the deformation continues to worsen (e.g., attenuation rate <30%, stability indicators continue to deteriorate), the system will automatically trigger an upgrade process: for example, perform a wider power cut-off, start redundant support systems, or immediately push the highest priority alarm to the operation and maintenance terminal, requiring emergency manual intervention.

[0036] Each complete "early warning, control, and assessment" event records its full-link data (including the deformation field characteristics at the time of triggering, the executed instructions, and the subsequent deformation response curve) and uses it to update the control strategy model. By learning the actual effects of different control actions in historical events, the system continuously optimizes its decision-making logic (such as more accurately predicting the safe time window and selecting the optimal control instruction for similar pattern matching results in the future), thereby achieving continuous self-improvement of the closed-loop system.

[0037] To enable those skilled in the art to fully understand and implement this invention, the following supplementary explanation of the implementation principles of each key technical aspect of this invention is provided in conjunction with the actual operation scenario of the aforementioned 600MW coal-fired power plant.

[0038] During the synchronous sampling phase of multi-source heterogeneous sensors, the FPGA chip inside the edge computing host generates a hardware pulse signal with a period of 10ms, which is simultaneously transmitted to the external trigger pins of each type of sensor via a dedicated shielded trigger cable. Since each sensor is equipped with a temperature-compensated crystal oscillator with a frequency stability better than ±0.1ppm, upon receiving the same rising edge trigger signal, each sensor generates a microsecond-level timestamp based on its local high-stability clock. This mechanism avoids the millisecond-level jitter caused by protocol stack delays, switch queuing, and electromagnetic interference in industrial Ethernet, as is common in traditional software time synchronization. This ensures that the GNSS receiver's position output, the hydrostatic level's liquid level reading, the inclinometer's attitude angle, and the fiber optic grating's wavelength offset are recorded at the same physical moment, providing a rigorous time reference for subsequent spatiotemporal fusion.

[0039] During the temperature drift correction process of the hydrostatic level, the system uses a Pt100 platinum resistance thermometer to measure the liquid temperature in real time. The liquid density at the current operating temperature is then calculated using an empirical formula, which is based on the liquid density at the reference temperature, the coefficient of volumetric expansion, and the temperature difference. Since the hydrostatic level's measurement principle relies on the height difference of the liquid within the communicating vessels to reflect the settlement, and changes in liquid volume with temperature can lead to false displacement signals, the true liquid column height must be deduced from the density. Specifically, the original voltage signal is converted into a digital value by a 24-bit analog-to-digital converter, corresponding to the hydrostatic pressure generated by the liquid column. This hydrostatic pressure is determined by the product of the liquid density at the current temperature, gravitational acceleration, and the original liquid column height. Using the known gravitational acceleration and the calibrated sensor sensitivity coefficient, the system first calculates the original liquid column height, then corrects it based on the ratio of the density at the current temperature to the reference density to obtain the corrected height. This eliminates systematic deviations caused by temperature and achieves sub-millimeter-level vertical displacement monitoring accuracy.

[0040] In the fiber grating thermal strain compensation stage, distributed Raman temperature sensing channels synchronously acquire temperature distribution along the same fiber path. The system then calculates the thermally induced wavelength drift at each point, which is determined by the sum of the material's thermal expansion coefficient and thermo-optic coefficient, the grating period, and the temperature rise relative to the reference temperature. Since the fiber grating center wavelength shift includes both mechanical strain and temperature effects, failure to separate them will lead to strain misjudgment. By subtracting the thermally induced wavelength drift from the total shift, retaining only the net wavelength change caused by structural stress, and combining this with the calibrated strain sensitivity coefficient and grating period parameters, the true strain can be accurately reconstructed, effectively solving the cross-sensitivity problem of strain sensing in high-temperature environments of thermal power plants.

[0041] In the decoupling process of turbine foundation displacement components, the state vector of the unscented Kalman filter contains three components: settlement amount, representing long-term slow foundation settlement; settlement rate, representing the settlement trend; and vibration-induced displacement, representing high-frequency vibration caused by unit rotational imbalance. The observation equation integrates the tilt angle output by the inclinometer (reflecting foundation tilt) and the strain gradient measured by the fiber optic grating (reflecting local bending deformation), both of which jointly constrain the state estimation. Simultaneously, the rotational speed provided by the DCS system, as a periodic excitation source, is modeled as a known input term in the state transition equation, enabling the filter to distinguish between transient displacement caused by mechanical vibration and irreversible foundation settlement. This joint model outputs a posterior estimate of the settlement rate and vibration-induced displacement every 10 seconds, thereby achieving early detection of settlement trends and dynamic separation of vibration responses.

[0042] In the decoupling of chimney deformation components, the system pre-establishes a three-dimensional finite element model based on the design drawings and extracts its shape function matrix. When the measured displacement from GNSS is input, the system represents the total displacement as a linear combination of three physical mechanisms: the thermal expansion component is obtained by integrating the temperature field measured by the infrared thermal imager on the chimney wall; the wind-induced vortex-induced component calculates the theoretical sway based on wind speed, wind direction, and aerodynamic parameters of the chimney; and the creep settlement component is estimated from the concrete age and load history. Since the corresponding shape function submatrices are linearly independent, they constitute an overdetermined system of equations. The measured displacement is equal to the sum of the products of each component and its corresponding shape function submatrix. Solving this system using the least squares method uniquely determines the contribution of each component, achieving quantitative separation of multi-physics coupled deformation.

[0043] During the spatial interpolation stage of the boiler steel frame, the system imports a tetrahedral mesh model generated by ANSYS, which accurately describes the geometry of the members and the connection relationships between nodes. The interpolation uses a 10-node Lagrange basis function consistent with the Tet10 element, whose shape function exhibits quadratic continuity within the element, accurately reflecting the bending and axial deformation of the members. At the hinged joints, the mechanical boundary conditions require zero bending moment, i.e., zero normal derivative of displacement. The system enforces this derivative constraint during interpolation to avoid non-physical displacement abrupt changes at the hinges. This method ensures that the reconstructed displacement field conforms to both sensor-measured data and the fundamental laws of structural mechanics.

[0044] In the pattern matching and recognition stage, the Dynamic Time Warping (VTW) algorithm allows the template to be non-linearly aligned with the current feature sequence on the time axis. For example, the settlement-vibration coupling mode of a steam turbine foundation may appear as low-frequency settlement superimposed with 30Hz vibration, but in actual events, the vibration frequency may drift to 28Hz or 32Hz due to load changes. The VTW algorithm constructs a cumulative distance matrix and finds the optimal curvature path to calculate the minimum cumulative distance, thus tolerating small shifts in frequency and time. The normalized score is obtained by subtracting the ratio of the minimum cumulative distance to the maximum possible distance from 1, mapping the distance to the interval between 0 and 1, making the similarity between different patterns comparable and avoiding misjudgments due to differences in feature dimensions.

[0045] In the emergency response following the triggering of a Level 3 warning, the edge computing host sends a DO command to the DCS system via the Modbus TCP protocol. This command directly affects the PLC's digital output module, cutting off the power supply circuit for non-critical loads. This process does not rely on upper-level monitoring software but achieves millisecond-level power cutoff through hard-wired logic, ensuring rapid reduction of secondary risks in the event of sudden structural instability. Simultaneously, the abnormal voxel coordinates (x=125.3m, y=87.6m, z=45.2m) can be precisely located near the circumferential weld of the 9th section of the chimney using BIM model lookup, providing centimeter-level positioning guidance for on-site maintenance.

[0046] In cloud-based inverse finite element analysis, ABAQUS's inverse analysis module uses the measured voxelized displacement field as the displacement boundary condition to invert the internal stress field. Since forward finite element analysis typically derives displacement from loads, while the inverse problem derives stress from displacement, which is ill-conditioned, the system introduces a Tikhonov regularization term to suppress noise amplification and considers material nonlinearity by incorporating a concrete constitutive model (such as Concrete Damaged Plasticity). The safety margin is determined by the ratio of the material's yield strength to the actual stress, where the yield strength is taken as the design value of the compressive strength of C40 concrete (19.1 MPa), and the actual stress is the maximum principal stress obtained from the inverse analysis. When the safety margin is less than 1.2, it is considered high-risk and requires immediate shutdown.

[0047] The above implementation principles demonstrate that this invention, through hardware synchronization, physical model-driven, structural constraint interpolation, and closed-loop emergency linkage, achieves a leap from discrete point alarm to full-field state perception, mechanism diagnosis, and precise response in the displacement monitoring of civil engineering structures in thermal power plants.

[0048] like Figure 2 As shown, a displacement monitoring and management system for civil structures in a thermal power plant includes: Data acquisition module: used to collect raw monitoring data of key civil engineering structures in thermal power plants from multiple dimensions and generate monitoring datasets; Data processing module: used to perform spatiotemporal synchronization processing on the monitoring dataset to obtain a spatiotemporally unified monitoring dataset; Data segmentation module: used to classify and segment the spatiotemporally unified monitoring dataset according to the structural type and service environment characteristics of different structures in the thermal power plant, and generate multiple structural monitoring categories; The index generation module is used to establish a classification analysis model based on multiple structural monitoring categories, obtain displacement component estimation results for different structural categories, and, based on the displacement component estimation results, obtain displacement evolution rate, cumulative offset and deformation stability index for different structural categories, and generate a displacement evaluation index set. Early warning module: It is used to compare the displacement evaluation index set with the preset displacement safety threshold range. When any index exceeds the corresponding threshold range, it generates a graded early warning information and pushes it to the operation and maintenance management terminal.

[0049] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for monitoring and managing displacement of civil structures in a thermal power plant, characterized in that, include: Original monitoring data of key civil engineering structures in thermal power plants are collected from multiple dimensions to generate a monitoring dataset; The monitoring dataset is subjected to spatiotemporal synchronization processing to obtain a spatiotemporally unified monitoring dataset; Based on the structural types and service environment characteristics of different structures within a thermal power plant, the spatiotemporally unified monitoring dataset is classified and divided to generate multiple structural monitoring categories. A classification analysis model was established based on multiple structural monitoring categories to obtain displacement component estimation results for different structural categories; Based on the displacement component estimation results, the displacement evolution rate, cumulative offset and deformation stability index of different structural categories are obtained, and a displacement evaluation index set is generated. The displacement assessment index set is compared with the preset displacement safety threshold range. When any index exceeds the corresponding threshold range, a graded early warning message is generated and pushed to the operation and maintenance management terminal.

2. The method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 1, characterized in that, Spatiotemporal synchronization processing includes: Differential calculation is performed on the dual-frequency carrier phase signal output by the global navigation satellite system receiver to obtain a high-precision three-dimensional coordinate sequence; The analog signal of the liquid level height from the hydrostatic level instrument is converted from analog to digital and the ambient temperature drift is corrected to obtain the corrected liquid level height data. The wavelength shift signal of the fiber optic strain sensor is demodulated and thermal strain compensation is performed in combination with distributed temperature sensing data to generate a compensated strain signal. The Kalman filter algorithm is applied to the raw angle data output by the MEMS inclinometer to suppress high-frequency noise and obtain filtered angle data. Subtract the non-structural displacement component caused by thermal expansion from the output signal of the temperature-compensated displacement gauge, and extract the net structural displacement component. The high-precision three-dimensional coordinate sequence, the corrected liquid level height data, the compensated strain signal, the filtered angle data, and the net structural displacement components are summarized, assigned a unified timestamp, and converted to a unified three-dimensional geographic coordinate system of the plant area to generate the spatiotemporally unified monitoring dataset.

3. The method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 1, characterized in that, Differential spatial interpolation is performed on different structure areas to generate interpolation results for each area. The spatial interpolation includes: For the turbine foundation area, spatial densification interpolation is performed using radial basis functions, and zero displacement or fixed support constraints are applied at the foundation boundary to obtain the interpolation results for the turbine foundation area. For the chimney structure, a spiral path interpolation model is constructed in cylindrical coordinates, and the rate of change of torsion angle along the height direction is introduced as an interpolation constraint parameter to obtain the interpolation results of the chimney structure. For the member connection node region of the boiler steel frame, based on the finite element mesh generation results, constrained spatial interpolation is performed using interpolation basis functions consistent with the element shape functions to obtain the interpolation results for the boiler steel frame region; The three types of interpolation results are respectively filled into the voxel mesh of the corresponding structures to form a three-dimensional displacement vector field with physical consistency; The smoothness test and energy conservation verification are performed on the three-dimensional displacement vector field to identify and eliminate abnormal voxel elements that do not meet the structural mechanics continuity condition, and generate a verified voxelized three-dimensional deformation field.

4. The method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 3, characterized in that, In the process of calculating the displacement evaluation index set based on the displacement component estimation results, pattern matching and recognition are performed, including: Extract feature vectors from four dimensions of the current three-dimensional deformation field: displacement amplitude sequence, displacement rate time history, displacement acceleration peak value, and main peak frequency of the spectrum, and generate the current feature vector set. Call the typical displacement mode library in the standard model database to obtain the mode template set. The typical displacement mode library includes four types of templates: turbine foundation settlement and vibration coupling mode, chimney wind-induced vortex oscillation mode, cooling tower foundation differential settlement mode, and main plant thermal expansion accumulation mode. For each type of template, a dynamic time warping algorithm is executed with the current feature vector set to calculate the minimum cumulative distance under path alignment, thereby obtaining the minimum cumulative distance value corresponding to each template. Each minimum cumulative distance value is normalized to a similarity score between 0 and 1, generating a similarity score set for each template; A threshold judgment is performed on the similarity score set. If the similarity score of any template is greater than the preset matching threshold, it is determined that the current displacement behavior is successfully matched with the template, and the matching type and score value are recorded to generate a pattern matching result.

5. A method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 4, characterized in that, The generation of tiered early warning information includes dynamic tiered early warning decisions, including: Real-time monitoring of the displacement rate of each structure; when the displacement rate exceeds the preset rate threshold and the duration is longer than the first time window, a first-level warning is triggered. The cumulative displacement value is calculated based on the displacement evaluation index set and compared with the proportion of the design allowable displacement limit. When the proportion exceeds the second threshold and the current displacement pattern is successfully matched, and the similarity score is higher than the third threshold, a level 2 warning is triggered. Based on the pattern matching results and the residual energy of the deformation field, the existence of a displacement surge event is detected; a displacement surge is defined as the displacement increment per unit time exceeding the fourth threshold, or the current displacement pattern matching fails and the residual energy of the deformation field exceeds the fifth threshold; a level three warning is triggered if either condition is met.

6. The method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 5, characterized in that, An emergency response mechanism is established upon triggering a Level 3 warning. This emergency response mechanism includes: Automatically execute operation commands to cut off unnecessary power supplies and generate a power cut-off command set. The unnecessary power supplies refer to power supply circuits other than safety instrument systems, fire protection systems and emergency lighting systems. Push alarm information including location, type, and severity level to the operation and maintenance terminal; Initiate a cloud-based in-depth analysis process, specifically: retrieve the building information model, material property parameters, and historical maintenance records corresponding to the alarmed structure from the standard model database, and generate a basic information set for the alarmed structure. Based on the building information model and the reconstructed voxelized three-dimensional deformation field, a spatial superposition comparison is performed to identify the component units that have experienced abnormal displacement and their adjacent areas, and to generate abnormal area location results. Based on the handling plans for similar displacement events in historical maintenance records, multi-factor influencing parameters including the current cumulative displacement value, displacement evolution rate, ambient temperature change rate and current load conditions are extracted to construct a structural safety status assessment matrix under the influence of multiple factors. Using the finite element inversion algorithm, with the current voxelized three-dimensional deformation field as the boundary condition, the local stress-strain field of the abnormal region is reconstructed, and the safety margin of the key nodes is calculated to obtain the safety margin calculation results. Based on the safety margin calculation results and the structural safety status assessment matrix, a structural health status assessment report is generated, and a set of disposal instructions containing risk level, recommended disposal measures and priority order is pushed to the operation and maintenance management platform.

7. The method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 1, characterized in that, During the generation of tiered early warning information, a real-time response mechanism is executed, including: After completing the voxelized three-dimensional deformation field reconstruction, the edge computing host executes the real-time discrimination logic of the first, second and third level early warning conditions in parallel to obtain the early warning discrimination results; When the warning judgment result meets any level of warning triggering condition, the local emergency control interface is immediately activated, and an instruction to cut off unnecessary power is sent to the power plant's distributed control system to complete the emergency power control. Simultaneously, the alarm event is encapsulated into a structured message, which includes the structure identifier, warning level, coordinates of the location of the sudden displacement increase, matching mode type, and deformation field residual energy value, to generate a structured alarm message; The structured alarm messages are pushed to designated operation and maintenance terminals and central monitoring screens through an encrypted channel via an industrial firewall. After the alarm information is transmitted, a local caching mechanism is activated to continuously record the original sensor data stream within a preset time window after the alarm is triggered, generating a data traceability record for subsequent in-depth cloud analysis.

8. A method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 3, characterized in that, The smoothness test and energy conservation verification include: For each voxel element in the voxelized three-dimensional deformation field, calculate its displacement gradient with neighboring voxel elements in the X, Y, and Z coordinate directions to generate a voxel element displacement gradient set. The smoothness of the displacement gradient set of the voxel element is checked. If the absolute value of the displacement gradient in any coordinate direction exceeds the critical gradient value corresponding to the allowable strain limit of the structure material, the voxel element is marked as an abnormal element, and an abnormal voxel element mark set is generated. The strain energy density distribution of the entire structure is calculated based on the displacement vector field, and the total strain energy is obtained by integrating over the entire structure volume. The total strain energy is compared with the theoretical energy input by external loads, including the vibration power of the steam turbine generator set, wind load, and thermal expansion force, to obtain the energy deviation value. If the relative deviation between the total strain energy and the theoretical energy exceeds the preset energy conservation tolerance threshold, it is determined that there is a physical inconsistency in the deformation field, and all abnormal voxel units in the abnormal voxel unit marker set are removed. The interpolation algorithm is triggered to recalculate the local region after removing abnormal voxel units, generating a corrected 3D deformation field.

9. A method for monitoring and managing displacement of civil structures in a thermal power plant according to claim 1, characterized in that, It also includes a model update mechanism: The cloud analytics center periodically aggregates de-identified displacement feature vector sequences from various edge computing nodes. These feature vectors include displacement amplitude, displacement rate, displacement acceleration, and the main peak frequency of the spectrum, generating an aggregated displacement feature vector set. Based on the aggregated displacement feature vector set, a multivariate time series training set is constructed, and it is divided into four subsets according to the structure type: turbine foundation, chimney, cooling tower and main plant, to obtain the classification training subset; For each of the aforementioned classification training subsets, a dedicated long short-term memory neural network prediction model is trained independently. A sliding window mechanism is used to input the feature vectors of the past N time steps and output the displacement evolution trend of the next M time steps to obtain a preliminary prediction model. When training the preliminary prediction model, a physical constraint loss function is introduced to force the prediction results to satisfy the structural continuity condition and the energy conservation principle, thus obtaining the physical constraint prediction model. Based on the physical constraint prediction model, an online evaluation is performed. When newly collected data causes the model prediction error to exceed the preset tolerance threshold, the online fine-tuning process of the model is triggered to generate updated model parameters. The updated model parameters and corresponding early warning threshold parameters are packaged and distributed to relevant edge computing nodes to complete the model parameter distribution and deployment.

10. A displacement monitoring and management system for civil structures of a thermal power plant, used to execute the displacement monitoring and management method for civil structures of a thermal power plant as described in any one of claims 1-9, characterized in that, include: Data acquisition module: used to collect raw monitoring data of key civil engineering structures in thermal power plants from multiple dimensions and generate monitoring datasets; Data processing module: used to perform spatiotemporal synchronization processing on the monitoring dataset to obtain a spatiotemporally unified monitoring dataset; Data segmentation module: used to classify and segment the spatiotemporally unified monitoring dataset according to the structural type and service environment characteristics of different structures in the thermal power plant, and generate multiple structural monitoring categories; Index generation module: used to establish a classification analysis model based on multiple structural monitoring categories to obtain displacement component estimation results for different structural categories; Based on the displacement component estimation results, the displacement evolution rate, cumulative offset and deformation stability index of different structural categories are obtained, and a displacement evaluation index set is generated. Early warning module: It is used to compare the displacement evaluation index set with the preset displacement safety threshold range. When any index exceeds the corresponding threshold range, it generates a graded early warning information and pushes it to the operation and maintenance management terminal.

Images