Temperature field reconstruction method and system for large-span roof

Abstract

The invention discloses a temperature field reconstruction method and system for a large-span roof, particularly relates to the technical field of building structure health monitoring, and is used for solving the problems of compensation interruption and temperature field discontinuity caused by time sequence asynchronization of environmental parameters and temperature data in the prior art. Roof temperature sensors and micro-meteorological parameters are synchronously collected, and uniform timestamps are made; establishing a compensation cache region to extract uncompensated data; dynamically dividing a real-time / delay compensation area according to the change rate of the solar radiation intensity; performing dynamic compensation on the real-time area data to generate compensated data, and performing asynchronous compensation on the delay area data in a reconstruction gap; inputting the compensated data into a reconstruction algorithm to generate temperature field distribution; checking continuity based on the gradient variation of the historical temperature field and the current temperature field; zero interruption in the temperature field reconstruction process is achieved, time dimension continuity is ensured, and meanwhile the real-time requirement is met through a resource optimization strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of building structure health monitoring, and more particularly to a temperature field reconstruction method and system for a large-span roof. Background Art

[0002] Temperature field monitoring on large-span roof structures (such as stadiums and airport terminals) relies on distributed sensor networks and dynamic environmental compensation technology. Existing methods typically collect temperature data by deploying fiber optic sensors and infrared thermal imagers on the roof. These sensors, combined with micro-meteorological stations to obtain parameters such as solar radiation and wind speed, use linear or nonlinear compensation formulas to correct for the effects of environmental factors (such as changes in solar incidence angle) on the measured temperature, thereby achieving dynamic modeling of the temperature field.

[0003] In the existing technology, there is a time-series asynchrony between the real-time updates of dynamic compensation parameters and environmental data. There is an inherent delay difference between the acquisition frequency of micrometeorological parameters (such as radiation intensity and incident angle) and the temperature sensor data stream, which makes it impossible to strictly synchronize the parameter updates of the compensation formula with the temperature data acquisition time. When the environmental parameters suddenly change (such as instantaneous cloud changes causing a jump in radiation intensity), the compensation calculation needs to interrupt the current reconstruction process to wait for the latest parameters, resulting in frequent interruptions in the temperature field generation process. This not only destroys the continuity of the temperature field in the time dimension, but also causes a step-like jump in the reconstruction results, further affecting the stability of subsequent thermal stress analysis. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a temperature field reconstruction method and system for a large-span roof to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A temperature field reconstruction method for a large-span roof comprises the following steps:

[0007] S1. Synchronously collect rooftop distributed temperature sensor data and micrometeorological parameters, and add unified timestamps to the temperature sensor data stream and micrometeorological parameter stream;

[0008] S2. Establish a compensation buffer area, and when the arrival of new micro-meteorological parameters is detected, extract the uncompensated temperature sensor data within a preset time window;

[0009] S3. Divide the roof area into a real-time compensation area and a delayed compensation area based on a comparison result of the solar radiation intensity change rate in the new micrometeorological parameter and a preset threshold value;

[0010] S4. Dynamically compensate the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data. Meanwhile, mark the uncompensated temperature sensor data in the delayed compensation area as a data set to be compensated, and perform asynchronous compensation on the data set to be compensated during the temperature field reconstruction interval.

[0011] S5. Input the real-time compensated data into a preset reconstruction algorithm to generate the roof temperature field distribution at the current moment;

[0012] S6. Based on the gradient change between the previously generated roof temperature field distribution and the currently generated roof temperature field distribution, perform a time continuity check on the reconstruction result.

[0013] Furthermore, the distributed rooftop temperature sensor data and micro-meteorological parameters are collected synchronously, and the temperature sensor data stream and the micro-meteorological parameter stream are timestamped uniformly, including:

[0014] Acquire a first temperature sensor data stream through a distributed optical fiber sensor, acquire a second temperature sensor data stream through an infrared thermal imager, and acquire a solar radiation intensity data stream and a solar incidence angle change data stream through a micro-meteorological station;

[0015] The first temperature sensor data stream of the distributed optical fiber sensor, the second temperature sensor data stream of the infrared thermal imager, and the solar radiation intensity data stream and solar incidence angle change data stream of the micro-meteorological station are given a unified time stamp through the time synchronization server.

[0016] Furthermore, a compensation buffer is established, and when the arrival of new micro-meteorological parameters is detected, uncompensated temperature sensor data within a preset time window is extracted, including:

[0017] Configure a ring storage queue to cache a first temperature sensor data stream and a second temperature sensor data stream with a unified timestamp;

[0018] When a new solar radiation intensity data stream or a new solar incidence angle change data stream arrives, the preset time window is calculated forward based on the corresponding timestamp of the corresponding new micrometeorological parameter, and all uncompensated first temperature sensor data streams and uncompensated second temperature sensor data streams in the corresponding time window are extracted as the data set to be processed.

[0019] Furthermore, based on the comparison between the solar radiation intensity change rate in the new micro-meteorological parameter and the preset threshold, the roof area is divided into a real-time compensation area and a delayed compensation area, including:

[0020] The spatial coordinate distribution of the area with sudden curvature change of the preloaded roof;

[0021] Calculate the rate of change value of the new solar radiation intensity data stream;

[0022] When the rate of change exceeds the preset threshold, the entire coordinate range of the curvature mutation area is included in the real-time compensation area;

[0023] For non-curvature mutation areas, based on the comparison results of the change amplitude of the solar incident angle change data stream with the preset angle change threshold, the sub-area whose change amplitude exceeds the preset angle change threshold is classified into the real-time compensation area, and the remaining areas are classified into the delayed compensation area.

[0024] Furthermore, a dynamic compensation operation is performed on the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data, including:

[0025] Acquire an uncompensated first temperature sensor data stream and an uncompensated second temperature sensor data stream corresponding to the real-time compensation area;

[0026] Calculating a time-weighted compensation factor based on the timestamp difference between the new solar radiation intensity data stream, the new solar incidence angle change data stream, and the uncompensated temperature sensor data stream;

[0027] A dynamic compensation formula is used to perform batch compensation operations on the uncompensated first temperature sensor data stream and the uncompensated second temperature sensor data stream respectively to generate real-time compensated data, including the real-time compensated first temperature sensor data stream and the real-time compensated second temperature sensor data stream.

[0028] Furthermore, marking the uncompensated temperature sensor data in the delay compensation area as a data set to be compensated includes:

[0029] The uncompensated first temperature sensor data stream and the uncompensated second temperature sensor data stream corresponding to the delay compensation area are stored in a queue to be compensated.

[0030] Furthermore, during the temperature field reconstruction interval, an asynchronous compensation operation is performed on the data set to be compensated, including:

[0031] During the interval between reconstruction tasks, data is extracted from the queue to be compensated and the same dynamic compensation formula is used to perform compensation operations.

[0032] Furthermore, the real-time compensated data is input into a preset reconstruction algorithm to generate the roof temperature field distribution at the current moment, including:

[0033] receiving a first temperature sensor data stream after real-time compensation and a second temperature sensor data stream after real-time compensation;

[0034] Processing the first temperature sensor data stream and the second temperature sensor data stream after real-time compensation using a radial basis interpolation algorithm based on an adaptive kernel function;

[0035] During the interpolation calculation process, the kernel function parameters are dynamically adjusted according to the roof curvature change rate, and the interpolation node density is increased preferentially in areas with high curvature changes;

[0036] Interpolation operations are performed through a parallel computing architecture to generate continuous temperature field distribution data covering the entire roof as the roof temperature field distribution at the current moment.

[0037] Furthermore, based on the gradient change between the previously generated roof temperature field distribution and the currently generated roof temperature field distribution, the reconstruction result is checked for temporal continuity, including:

[0038] Extract the historical temperature gradient distribution data from the previously generated roof temperature field distribution;

[0039] Calculate the temperature gradient change at the same spatial coordinate point in the currently generated roof temperature field distribution;

[0040] comparing the temperature gradient change with a preset gradient threshold;

[0041] When the temperature gradient change exceeds the preset gradient threshold, the coordinates of the corresponding roof curvature mutation area are located;

[0042] A local recompensation instruction is generated according to the positioning result, and the local recompensation instruction includes the spatial coordinate range that needs to be recompensed.

[0043] In another aspect, the present invention provides a temperature field reconstruction system for a large-span roof, comprising the following modules:

[0044] Synchronous acquisition module, used to synchronously collect rooftop distributed temperature sensor data and micro-meteorological parameters, and stamp the temperature sensor data stream and micro-meteorological parameter stream with a unified time stamp;

[0045] A cache extraction module is used to establish a compensation buffer area and extract uncompensated temperature sensor data within a preset time window when the arrival of new micrometeorological parameters is detected;

[0046] A zoning decision module is used to divide the roof area into a real-time compensation area and a delayed compensation area based on the comparison result of the solar radiation intensity change rate in the new micro-meteorological parameters and the preset threshold;

[0047] The dynamic compensation module is used to dynamically compensate the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data. At the same time, the uncompensated temperature sensor data in the delayed compensation area is marked as a data set to be compensated, and asynchronous compensation operations are performed on the data set to be compensated during the temperature field reconstruction interval.

[0048] The field reconstruction module is used to input the real-time compensated data into the preset reconstruction algorithm to generate the roof temperature field distribution at the current moment;

[0049] The verification and positioning module is used to verify the time continuity of the reconstruction result based on the gradient change of the roof temperature field distribution generated previously and the roof temperature field distribution generated currently.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. Synchronous acquisition and a unified timestamp mechanism are used to eliminate timing deviations in multi-source data. Combined with a dynamic compensation buffer and real-time / delayed partitioning strategies, only local compensation calculations need to be interrupted when environmental parameters suddenly change, ensuring the continuous operation of the temperature field reconstruction process. A continuity verification mechanism based on the previous and current temperature field gradient changes accurately identifies jump areas and triggers local re-compensation, fundamentally avoiding the temperature field step distortion caused by global waiting in traditional methods. This solution ensures high continuity of the temperature field of large-span roofs in the time dimension, providing a stable and reliable data foundation for subsequent thermal stress analysis.

[0052] 2. A zoning decision-making mechanism driven by the rate of change of radiation intensity is adopted to prioritize high-sensitivity areas (areas with sudden changes in curvature) and implement delayed compensation for low-impact areas, effectively reducing the real-time computing load. Asynchronous compensation operations are performed during the temperature field reconstruction interval, and resource conflicts between the compensation calculation and reconstruction processes are avoided by reusing resources during idle periods. Combined with the temperature field reconstruction algorithm accelerated by GPU parallelism, the entire process is compressed to minutes while ensuring spatial resolution, meeting the stringent real-time requirements of large-span roof safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a temperature field reconstruction method for a large-span roof according to the present invention;

[0054] Figure 2 The figure is a schematic structural diagram of a temperature field reconstruction system for a large-span roof according to the present invention. DETAILED DESCRIPTION

[0055] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0056] Example 1: Figure 1 The present invention provides a temperature field reconstruction method for a large-span roof, which includes the following steps:

[0057] S1. Synchronously collect rooftop distributed temperature sensor data and micrometeorological parameters, and add unified timestamps to the temperature sensor data stream and micrometeorological parameter stream;

[0058] S2. Establish a compensation buffer area, and when the arrival of new micro-meteorological parameters is detected, extract the uncompensated temperature sensor data within a preset time window;

[0059] S3. Divide the roof area into a real-time compensation area and a delayed compensation area based on a comparison result of the solar radiation intensity change rate in the new micrometeorological parameter and a preset threshold value;

[0060] S4. Dynamically compensate the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data. Meanwhile, mark the uncompensated temperature sensor data in the delayed compensation area as a data set to be compensated, and perform asynchronous compensation on the data set to be compensated during the temperature field reconstruction interval.

[0061] S5. Input the real-time compensated data into a preset reconstruction algorithm to generate the roof temperature field distribution at the current moment;

[0062] S6. Based on the gradient change between the previously generated roof temperature field distribution and the currently generated roof temperature field distribution, perform a time continuity check on the reconstruction result.

[0063] The first temperature sensor data stream is obtained through a distributed fiber optic sensor. The distributed fiber optic sensor is arranged at a specific position on the surface of the roof steel structure grid, specifically at the intersection of the main beam and the secondary beam of the roof. The sensor spacing is controlled within a range of 2 meters (for example, 2 meters, 3 meters, etc. can be adjusted according to the specific roof span). The sensor works based on the Raman scattering effect to measure temperature. The specific process is as follows: the pulsed laser transmitted inside the optical fiber interacts with the thermal vibration of the material molecules to produce backscattered light. The intensity of the scattered light signal is proportional to the temperature change. The scattered light is converted into a voltage signal by a photoelectric converter, and then sampled by an analog-to-digital converter to form a discrete temperature value. The sampling frequency is set to a specific value (for example, once per second or once every 2 seconds). Each sampling point corresponds to a pre-set three-dimensional spatial coordinate position on the roof, thereby forming a continuous spatially distributed roof temperature data sequence, namely the first temperature sensor data stream.

[0064] The infrared thermal imager is installed at a fixed position on a high-rise building around the roof or on a mobile drone platform. The imaging spectrum selects the long-wave infrared range (for example, the wavelength range of 8-14 microns), the single-pixel spatial coverage size is set to a specific resolution (for example, 0.5 meters × 0.5 meters or 1 meter × 1 meter), and the acquisition frequency is set to a specific interval (for example, one frame of thermal image is obtained every 2 seconds or every 5 seconds). When processing the thermal image, the infrared radiation intensity value of each pixel area is segmented, and the blackbody radiation correction curve is used to convert the radiation value into the corresponding steel structure surface temperature value. The temperature value of each pixel is associated with the longitude and latitude coordinates to generate spatial array data, namely the second temperature sensor data stream.

[0065] The micro-meteorological station is installed in an unobstructed area at the highest point of the roof. The radiation sensor uses a thermopile pyranometer. The sensing element contains a constantan thermocouple stack structure. It generates a voltage signal by measuring the temperature differential electromotive force caused by solar radiation. The voltage signal is collected at a specific sampling interval (for example, 0.5 seconds or 1 second). The voltage value is converted into a unit area radiation power value (in watts / square meter) using a calibration formula, thereby forming a solar radiation intensity data stream; the change in solar incidence angle is measured using a two-axis electronic inclinometer. The inclinometer is fixed on a vertical mounting base. The direction of the gravity vector is solved in real time through the built-in three-axis accelerometer and three-axis gyroscope. The angle change between the gravity vector and the roof normal vector is calculated, and the angle change value is recorded at a fixed period (for example, 0.5 seconds or 1 second). The change is processed by a specific normalization function (for example, the inverse tangent function) to form a solar incidence angle change data stream.

[0066] The first temperature sensor data stream of the distributed optical fiber sensor, the second temperature sensor data stream of the infrared thermal imager, and the solar radiation intensity data stream and the solar incidence angle change data stream of the micro-meteorological station are uniformly timestamped. During the specific implementation, a high-precision time synchronization server is deployed. The server uses the global positioning system satellite signal as the reference time source. The receiver continuously receives the coordinated universal time signals sent by multiple positioning satellites (for example, 4 or more), corrects the local clock offset through a time calibration algorithm (for example, a least squares estimation algorithm), and the internally configured precision time base device (for example, a rubidium atomic clock) controls the clock error within a specific accuracy range (for example, less than 1 microsecond).

[0067] Each data acquisition terminal device is installed with a network time protocol communication module, the distributed fiber optic sensors are connected via fiber optic Ethernet, the infrared thermal imager and micro-meteorological station are connected via wireless communication modules, and all terminal devices send clock synchronization request packets to the time synchronization server at specific time intervals (for example, every minute).

[0068] The time synchronization server uses a two-way time transfer mechanism to respond. The specific process records the arrival time of the request packet and the sending time of the response packet. The client device records the time when the response packet is received. The four time values ​​of request sending time, request arrival time, response sending time and response receiving time are used to calculate the communication path transmission delay, and the local clock deviation is corrected through the time compensation algorithm.

[0069] After completing time synchronization, each acquisition terminal adds a time tag in a specific format (for example, a 64-bit UTC time code) each time it generates data. The time tag accuracy reaches a specific level (for example, microsecond level). The time tag of the first temperature sensor data stream is embedded in a specific field of the message header of the fiber optic demodulator output data. The time tag of the second temperature sensor data stream is written into the metadata area of ​​the thermal image file. The time tag of the solar radiation intensity data stream is placed in the status word area of ​​the serial communication protocol data packet. The time tag of the solar incidence angle change data stream is stored in the index mark area of ​​the circular buffer.

[0070] After all data streams converge to the central data processing unit, the time tag information of each data stream is extracted, and the difference in time tags between different data streams is compared. When it is detected that the time deviation exceeds a specific threshold (for example, 50 milliseconds), the time series interpolation algorithm is enabled to generate transition data points to achieve timeline alignment. Finally, the first temperature sensor data stream, the second temperature sensor data stream, the solar radiation intensity data stream, and the solar incidence angle change data stream with strictly matched timestamps are output.

[0071] The time synchronization quality verification adopts a spatial sampling method, randomly selecting some roof measurement points (for example, sampling points that account for 5% of the roof area), and comparing the difference between the recording time of the optical fiber sensor at that location and the exposure time of the infrared thermal imager. When the time difference exceeds a specific alarm threshold (for example, 5 milliseconds), the exception handling process is triggered.

[0072] Fiber-optic timing signal relays are deployed in areas with weak signal coverage, such as roof edges. These relays receive optical synchronization signals transmitted from the backbone fiber, regenerate clock signals locally, and distribute them to surrounding terminal devices via optical switching equipment. The resulting unified timestamp data set is stored in a time series database system, which is indexed by millisecond timelines to support fast retrieval and provide a foundational data set with strict time alignment for subsequent data processing.

[0073] A circular storage queue is configured to cache the first and second temperature sensor data streams with unified timestamps. The specific implementation process is as follows: a continuous address area is allocated in the memory space of the central processing unit as the carrier of the circular storage queue. This storage area manages data according to the first-in-first-out rule. Its capacity is determined by the temperature sensor sampling frequency and the maximum allowable delay (for example, when the sampling interval is 2 seconds, the capacity is designed to store 30 minutes of data). The data writing mechanism is as follows: when the first and second temperature sensor data streams with unified timestamps are received, the parser extracts the time tag and spatial coordinate information from the data packets and stores the data packets in chronological order into continuous memory blocks of the circular storage queue. The memory block size matches the number of bytes in the temperature data packet (for example, a single temperature data packet contains coordinate values ​​and temperature values, totaling 64 bytes). The circular storage queue is set with dual identifiers: a write pointer and a read pointer. The write pointer always points to the latest data storage location. When the queue is full, it automatically overwrites the oldest historical data. The overwriting rule is set according to the preset data validity period (for example, the default validity period is 15 minutes). The first temperature sensor data stream and the second temperature sensor data stream are partitioned and stored in the queue. The first temperature sensor data stream is stored at the front address of the queue (for example, the address range starts at 0000 and ends at 7FFF), and the second temperature sensor data stream is stored at the back address (for example, the address range starts at 8000 and ends at FFFF). When storing, an additional status mark bit is added to identify the data compensation status (the uncompensated mark is marked with status code 00, and the compensated mark is marked with status code 01).

[0074] The implementation of detecting the arrival of a new solar radiation intensity data stream or a new solar incidence angle variation data stream includes: deploying an interrupt trigger mechanism at the micro-meteorological data receiving port, which continuously monitors the data port input status. When the start frame signal of the new solar radiation intensity data stream (such as the Ethernet frame synchronization header AA) or the start identifier of the new solar incidence angle variation data stream (such as the serial port data header 55) is detected, the interrupt service program is immediately activated. The interrupt service program first parses the time tag field in the new data packet (such as the 64-bit UTC time code defined in the previous step), and at the same time verifies the checksum of the data packet to ensure integrity. For the new solar radiation intensity data stream, its time tag value and the newly arrived radiation value sequence are extracted; for the new solar incidence angle variation data stream, the time tag value and the newly arrived angle change sequence are extracted. After detecting the arrival of new data, an event log is generated, which records the event type (radiation update or angle update) and the arrival time accurate to microseconds.

[0075] The operational process for forward-calculating the preset time window based on the timestamp of the new micrometeorological parameter is as follows: the time tag of the newly arrived micrometeorological parameter is taken as the reference time point, and the preset time window length is set to the window duration (the window duration value is determined by the heat conduction response speed. For example, if the thermal response time constant of steel is 30 seconds, the window duration is set to 15 to 60 seconds and can be adjusted). The starting time point of the time window is calculated as the reference time point minus the window duration. The storage area is scanned in the circular storage queue, and index matching is performed on the first temperature sensor data stream and the second temperature sensor data stream respectively: the time tag field of each data block in the queue is traversed, and the data blocks with time tags greater than or equal to the time window starting time point and less than or equal to the reference time point are filtered. At the same time, the status flag bit is verified to be in the uncompensated state (status code 00). The extracted data blocks form a temporary data set. If there are multiple time point data at the same spatial location (such as the overlapping area of ​​infrared and optical fiber), the latest valid data is retained.

[0076] Extracting uncompensated temperature sensor data within the corresponding time window as the dataset to be processed involves spatially integrating the selected first and second temperature sensor data streams. Using a unified coordinate system conversion (geographic coordinate system X-axis, Y-axis, and Z-axis), the first temperature sensor data stream from the fiber optic sensor is mapped to a three-dimensional grid based on the installation path coordinates. The second temperature sensor data stream from the infrared thermal imager is projected onto the same grid based on pixel latitude and longitude. The grid size is set based on the roof accuracy requirements (e.g., a 0.5 meter by 0.5 meter grid). After coordinate conversion, data fusion is performed: if both the first and second temperature sensor data streams exist within the same grid cell, the average value is taken as the cell temperature value (or weighted processing is performed based on sensor accuracy). For cells with a single data source, the original value is used directly. This ultimately generates a spatiotemporally aligned dataset to be processed. This dataset's data record format includes the spatial grid index number, the original temperature value, the compensation status flag, and the source sensor type identifier.

[0077] The preset time window length is derived from a thermal model: a transient heat conduction equation is constructed based on the specific heat capacity of the roof panel material (e.g., 502 joules per kilogram Kelvin for steel), thermal conductivity (50 watts per meter Kelvin), and ambient wind speed. The temperature response lag time constant is determined through numerical simulation. The preset window length must satisfy the relationship that half the lag time constant is less than the window length, and the window length is less than twice the lag time constant (e.g., if the lag time constant for a steel roof is approximately 30 seconds, the window length should be between 15 and 60 seconds). In engineering applications, material properties are input through the parameter calibration interface, and simulations are performed to generate a recommended window length. A linear buffer expansion strategy is used to process the time window boundaries: when the reference time point approaches the oldest data in the queue (e.g., the queue storage start time), if the time window start time point is less than the storage start time, the window range is adaptively adjusted from the storage start time to the reference time point to avoid out-of-range indexing. A system alarm is triggered to indicate the risk of data overwrite. The alarm threshold is set to a specific number of window duration violations per hour (e.g., 5). The dataset to be processed is ultimately stored in a dedicated buffer with a time window identifier attached, awaiting the invocation of the compensation operation. The buffer structure is stored as a two-dimensional matrix, with rows corresponding to spatial grid indices and columns arranging temperature values ​​in time series, enabling fast access in both spatial and temporal dimensions.

[0078] In practice, the preset time window length is determined based on the thermophysical properties of the roof panel material (such as steel density, specific heat capacity, and thermal conductivity) and the ambient wind speed. The transient heat conduction equation is a universal physical equation used in building thermal engineering to calculate the time-dependent temperature change of a material. It describes the relationship between the rate of heat transfer within a material and its properties (density, specific heat capacity, and thermal conductivity) and the ambient heat input (solar radiation). The temperature response lag time constant is determined using a standard laboratory method: a standard specimen made of the same material as the roof is subjected to a simulated step increase in solar radiation intensity (e.g., from 200 watts per square meter to 800 watts per square meter) in a controlled environmental chamber. A high-precision temperature sensor is used to record the time required for the specimen surface temperature to reach approximately 63.2% of the new equilibrium state. For example, for a 0.6 mm thick Q345B steel plate, this time is measured to be approximately 28 seconds in no wind conditions. The preset time window length is between 1 and 1.5 times this measured time value (e.g., 28 to 42 seconds). In engineering applications, the time constant can be directly entered into the system configuration interface, and the system automatically sets the recommended window length.

[0079] In specific implementation, the transient heat conduction process is realized through standard thermal tests:

[0080] Material thermal response test: select a standard specimen made of the same material as the roof (such as 0.6mm thick Q345B steel plate) and set the radiation intensity to 800W / m in the environmental chamber. 2, continue heating until the temperature stabilizes, then suddenly reduce the radiation to 200W / m 2 , and record the temperature drop curve using an infrared thermal imager (sampling rate 1 Hz).

[0081] Determination of hysteresis time: When the temperature drops to 63.2% of the initial temperature difference (such as from 45℃→37.6℃), record the required time as the hysteresis time constant (typical value for steel is 28 seconds).

[0082] Relationship explanation: The heat transfer rate is expressed as follows: the larger the density / specific heat capacity value, the slower the temperature change (for example, steel is 3 times slower than concrete); the higher the thermal conductivity coefficient, the faster the environmental radiation influence is transferred (aluminum is 2 times faster than steel).

[0083] System configuration: After entering the material parameters in the software interface (density 7850kg / m 3 , specific heat capacity 502 J / kg·K), the system automatically calculates the recommended time window (28-42 seconds).

[0084] The implementation process for preloading the spatial coordinate distribution of roof curvature mutation areas involves extracting the roof's 3D geometric model data from the building information model (BIM). The model format is an industry-standard file and contains a non-uniform rational B-spline mathematical representation of all roof surface curves. Gaussian curvature is calculated for each surface patch using the following algorithm: The principal curvature radius of any point on the surface is taken, with the Gaussian curvature being equal to the product of the two principal curvature values. When the absolute value of the Gaussian curvature at a point exceeds a set threshold (e.g., 0.05 per meter), the point is marked as a curvature mutation point. All curvature mutation points are projected onto a 2D coordinate system (using the projected coordinate system of the building plane) to generate a density distribution cloud map. Cluster analysis is used to identify clusters of mutation points. The cluster radius is set to a specific value (e.g., 2 meters). The boundaries of each cluster are calculated using a convex hull algorithm, ultimately outputting a set of polygonal vertex coordinates. For example, in a stadium project (e.g., a cable-membrane structure with a span of 250 meters), 15 regions of sudden curvature changes, including ridges and valleys, were identified. Each region was stored as an independent set of coordinate ranges (e.g., a sequence of polygonal vertices) with a specific accuracy (e.g., 0.01 meters). The coordinates of these sudden curvature changes were stored in a spatial database and indexed and mapped to the roof grid.

[0085] The process for calculating the rate of change of a new solar radiation intensity data stream is as follows: Take the two most recently arriving solar radiation intensity data streams: the radiation value R1 corresponding to time tag T1 and the radiation value R2 corresponding to time tag T2 (where T2 > T1). Calculate the time difference ΔT = T2 - T1 (in seconds) and the change in radiation value ΔR = R2 - R1 (in watts per square meter). The rate of change is defined as the time derivative of the radiation change: ΔR divided by ΔT (in watts per square meter per second). When a new data stream arrives for the first time (i.e., no historical data exists), the rate of change is set to a specific initial value (e.g., 0). In engineering implementation, the rate of change calculation results are stored in a dynamic array, with the array index corresponding to the micrometeorological data time tag. To prevent noise interference, a minimum time difference threshold (e.g., 0.5 seconds) is set. When ΔT falls below this threshold, the time window is automatically expanded to include the first three consecutive data points, and the linear regression slope is used as the rate of change value. Output a data sequence of time tags and corresponding rate of change values ​​for subsequent partitioning decisions.

[0086] When the rate of change value exceeds the preset threshold, the entire coordinate range of the curvature mutation area is included in the real-time compensation area in the following manner: the preset threshold is determined by material thermal sensitivity analysis. The specific method is to establish a transient heat conduction differential equation for the roof panel material (such as aluminum-magnesium-manganese alloy) and solve the boundary condition where the critical radiation change rate causes the surface temperature rise rate to exceed a specific value (such as 2 degrees Celsius per minute). Under typical working conditions, the preset threshold is set to a specific value (such as 100 watts per square meter second). The decision-making process is: compare the current rate of change value with the preset threshold in real time. When the rate of change value is detected to be greater than the preset threshold, retrieve the coordinate range of all curvature mutation areas from the spatial database. Mark the roof grid cells corresponding to these coordinate ranges as real-time compensation status (status code 01), and generate a set of area identifiers. In large-scale airport roof applications (for example, an area of ​​30,000 square meters), when clouds move rapidly at noon in summer, the measured radiation change rate reaches a specific value (for example, 120 watts per square meter second), triggering the system to complete the status marking of 12 curvature mutation areas (with a total area of ​​approximately 5,000 square meters) within a specific time (for example, within 200 milliseconds).

[0087] Method for determining the critical radiation change rate threshold:

[0088] Engineering measurement:

[0089] A radiometer (Apogee SP-510) and an infrared thermal imager (FLIR A655sc) were installed in a typical area of ​​the roof. When the radiation intensity changes ΔR (unit: W / m 2 ), record the temperature rise rate ΔT / Δt (℃ / min); statistically analyze 100 sets of data and find that: when ΔR / Δt≥100W / m 2·s, ΔT / Δt>2℃ / min in 90% of cases.

[0090] Software Verification:

[0091] Use ANSYS Thermal module to build the roof panel model (material parameters are the same as in Section 0035); set up transient analysis: apply solar radiation heat flux to the surface and insulate the bottom surface; parametrically sweep the radiation change rate (50-150W / m 2 ·s), output the temperature rise rate curve; when the slope of the curve is greater than 2℃ / min, the corresponding radiation change rate is marked as the critical value.

[0092] Equation Description: The differential equation is encapsulated in the ANSYS solver. Its physical essence is "solar radiation absorbed by a material = internal heat accumulation + surface heat loss." The system only requires calling the software interface (command: / SOLVE), eliminating the need for manual equation processing.

[0093] The partitioning operation for the non-curvature mutation area includes: first excluding the coordinate range of the curvature mutation area from the global roof grid, and the remaining area is defined as the non-curvature mutation area. Obtain the current solar incidence angle change data stream, which contains the angle change value of the time series (in degrees). The sliding window standard deviation algorithm is used to calculate the change amplitude: take the incidence angle change data set within a specific time window (for example, 5 minutes), and calculate the statistical standard deviation of the data set as the change amplitude indicator. The preset angular change threshold is determined based on the solar trajectory simulation. The specific method is: establish a time-varying model of the solar altitude angle / azimuth angle at the geographical location, calculate the theoretical maximum angular velocity value, and take a specific ratio (for example, 50%) as the threshold benchmark. In engineering applications, the preset angular change threshold is set to a specific value (for example, 0.3 degrees per second).

[0094] During specific implementation, the preset angular change threshold (such as 0.3 degrees per second) is determined based on the longitude and latitude of the project site and the date. The solar altitude angle / azimuth time-varying model refers to an internationally accepted astronomical calculation method (such as the ISO 19115 standard or the NRELSOLPOS algorithm). The theoretical maximum angular velocity value is obtained through the following executable steps: using standard astronomical calculation software, input the precise longitude and latitude of the project site and the target date (such as the summer solstice), and calculate the numerical sequence of the solar azimuth angle at each second from sunrise to sunset; calculate the absolute value of the angle change between adjacent second data, and the maximum value in the sequence is the theoretical maximum angular velocity value. For example, the calculated value of the maximum change rate of the azimuth angle at noon on the summer solstice in Guangzhou is approximately 0.25 degrees per second. The preset angular change threshold is 1.2 times this value (i.e. 0.3 degrees per second) as a safety margin.

[0095] The partition decision logic is as follows: traverse each grid cell in the non-curvature mutation area, extract the incident angle change data corresponding to the cell (matched according to the cell's geographical location), and calculate the change amplitude value. When the change amplitude value is greater than the preset angle change threshold, the grid cell is classified into the real-time compensation area; otherwise, it is classified into the delayed compensation area. In the gently curved area (such as the central roof), when the wind speed suddenly changes and the amplitude of the solar incident angle change reaches a specific value (for example, 0.5 degrees per second), the system automatically marks the area as a real-time compensation area. The partition result is stored as a compensation state matrix, where the matrix rows correspond to the grid row numbers and the columns correspond to the grid column numbers. The element value 0 indicates a delayed compensation area, and 1 indicates a real-time compensation area. A bidirectional index is established between the state matrix and the roof space coordinates to support fast area queries. Finally, the real-time compensation area coordinate set and the delayed compensation area coordinate set are output for subsequent compensation operations.

[0096] The specific implementation process for obtaining the uncompensated first and second temperature sensor data streams corresponding to the real-time compensation zone is as follows: Based on the real-time compensation zone coordinate range set generated in the previous step, data records with matching spatial locations are retrieved from the circular storage queue. Spatial matching utilizes a grid index query mechanism, based on the three-dimensional grid coordinate system of the roof division, to extract all data entries whose grid index numbers fall within the real-time compensation zone. For each matching entry, its data type identifier (first temperature sensor data stream identifier or second temperature sensor data stream identifier) ​​is identified, and the uncompensated first and second temperature sensor data streams are separated. Uncompensated status verification is achieved by checking the status flag (status code 00). If a data block has been marked as compensated (status code 01), it is skipped. Integrity checks are performed during data extraction. If data at a specific grid point is missing (e.g., due to packet loss), a temporary data block is generated using spatial neighbor interpolation. The temporary data block is marked as pending verification and a special flag is added.

[0097] Calculating a time-weighted compensation factor based on the timestamp differences between the new solar radiation intensity data stream, the new solar incidence angle change data stream, and the uncompensated temperature sensor data stream involves the following operations: Assume the time tag of the uncompensated temperature data is Ttemp, the time tag of the newly arrived solar radiation intensity data stream is Trad, and the time tag of the solar incidence angle change data stream is Tang. The radiation time difference is calculated as |Ttemp - Trad|, and the incidence angle time difference is calculated as |Ttemp - Tang|. The time-weighted compensation factor is calculated as follows: the radiation compensation weight is equal to the quotient of 1 minus the radiation time difference divided by the preset maximum valid time difference, and the incidence angle compensation weight is equal to the quotient of 1 minus the incidence angle time difference divided by the preset maximum valid time difference. The maximum valid time difference is determined through thermodynamic analysis based on the material's thermal relaxation time constant (e.g., 30 seconds for steel). When the time difference exceeds the maximum valid time difference, the corresponding weight is reset to zero. The weight calculation results are stored in vector form. The vector dimension is consistent with the number of temperature data points. Each data point is associated with two weight values ​​(radiation weight and incident angle weight).

[0098] The dynamic compensation formula is used to perform batch compensation operations on the uncompensated first temperature sensor data stream and the uncompensated second temperature sensor data stream as follows: The core structure of the dynamic compensation formula includes a temperature reference term and an environmental disturbance term. The temperature reference term corresponds to the original uncompensated temperature value, and the environmental disturbance term is composed of the linear superposition of the radiation compensation component and the incident angle compensation component. The specific calculation process is: the compensated temperature value is equal to the original temperature value plus an adjustment amount, which is equal to the radiation compensation weight multiplied by the new solar radiation intensity value multiplied by the solar radiation coefficient plus the incident angle compensation weight multiplied by the new solar incident angle change value multiplied by the incident angle coefficient. The solar radiation coefficient is calibrated through material light and heat absorption experiments (for example, the value of aluminum plate is 0.65 watts per square meter per degree Celsius), and the incident angle coefficient is obtained by mapping from an angle-heat flux comparison table. During batch processing, three calculation steps are performed in parallel for each uncompensated temperature data point: the first step is to calculate the radiation compensation component value, which is equal to the radiation compensation weight of the point multiplied by the new solar radiation intensity value multiplied by the solar radiation coefficient; the second step is to calculate the incident angle compensation component value, which is equal to the incident angle compensation weight of the point multiplied by the new solar incident angle change value multiplied by the incident angle coefficient; the third step is to calculate the final compensated temperature value, which is equal to the original temperature value plus the radiation compensation component value plus the incident angle compensation component value. The compensation calculation result is stored as a new data block, which inherits the original spatial coordinate information and updates the status flag to the compensated state (status code 01). An anomaly detection mechanism is set in batch processing. When the compensation value exceeds the allowable temperature range of the material (for example, -40 to 200 degrees Celsius for steel), the boundary truncation process is automatically triggered and the abnormal event log is recorded.

[0099] In practice, the calculation of radiation compensation weights and incident angle compensation weights relies solely on the time difference data: the weight is equal to 1 minus the "difference between the temperature data timestamp and the corresponding meteorological parameter timestamp" divided by the "temperature response lag time constant" (see paragraph 0034 for definition). If the time difference exceeds this time constant, the weight is set to 0. For example, if the temperature data timestamp is 10:00:00 and the solar radiation data timestamp is 10:00:05, with a time constant of 28 seconds, the radiation compensation weight is 1-(5 / 28) ≈ 0.82.

[0100] Storing the uncompensated first and second temperature sensor data streams corresponding to the delay compensation zone into the pending compensation queue involves filtering relevant entries in the circular storage queue based on the delay compensation zone coordinate range set. Data extraction rules are the same as for the real-time compensation zone, but with the addition of a priority control strategy. Priority is arranged in reverse chronological order (with the most recent time stamp receiving the highest priority). The queue storage structure utilizes a two-level index design, with the primary index being the time segment and the secondary index being the spatial grid number. Data compression is performed during storage to remove redundant spatial coordinate information (retaining only the starting coordinate for identical grid points).

[0101] During the interval between reconstruction tasks, asynchronous compensation is performed on the dataset to be compensated as follows: the system continuously monitors the status of the temperature field reconstruction process. When a signal indicating the completion of the reconstruction algorithm is detected (e.g., a result code returned by the GPU kernel), an idle time slot timer is started. The minimum available idle time slot duration is determined by the reconstruction cycle (e.g., if the reconstruction takes 20 seconds, the minimum idle time slot is set to 5 milliseconds). The asynchronous compensation scheduler checks the length of the queue to be compensated. If the queue is not empty, it extracts data blocks in order of priority (the number of blocks extracted at a time is determined by the idle time slot length). After extraction, the same dynamic compensation formula as the real-time compensation is immediately applied. The compensation results are stored in a separate storage area (different from the real-time compensation results) and marked as the source of delay compensation (status code 10). The compensation process implements a resource throttling strategy: dynamically allocated computing resources do not exceed a specific proportion of the total system resources (e.g., 30%) to prevent impacting real-time reconstruction performance.

[0102] The specific operations for receiving the first and second temperature sensor data streams after real-time compensation include: reading data blocks marked as completed from the compensation result storage area. These data blocks contain the first and second temperature sensor data streams after real-time compensation generated in the previous step. A double-buffer mechanism is implemented on the receiving end: while the reconstruction algorithm processes the current data, the next batch of data is simultaneously written to the backup buffer, implementing pipeline operation. The data transmission interface utilizes a direct memory access engine that automatically identifies the data block header identifier (the first temperature sensor data stream identifier or the second temperature sensor data stream identifier) ​​and stores it in different memory partitions according to data type. The memory partitions are configured as high-speed cache memory areas with access speeds reaching a specific level (e.g., 6 GB / s bandwidth). Data integrity checks are performed during reception, calculating the cyclic redundancy check value for each data block and comparing it with the header checksum. If an error is detected, the source end is immediately requested to retransmit the missing data block. For rooftops with extremely long spans (e.g., exceeding 300 meters), distributed receiving nodes are deployed, with transmission status synchronized between nodes via a fiber optic data bus.

[0103] The implementation process of the adaptive kernel-based radial basis interpolation algorithm involves the following: The algorithm core utilizes an improved radial basis function model, which consists of two components: a basis function and an adaptive adjustment mechanism. The basis function is a Gaussian kernel function, whose characteristic parameters include two adjustable variables: the support radius and the shape parameter. The interpolation calculation principle is to treat each sensor temperature data point as the center point of a spatial basis function, and construct a continuous temperature field by superimposing the basis functions. The specific implementation steps are divided into three phases: initialization, matrix construction, and equation solution. The initialization phase loads the roof spatial grid structure definition (for meshing, see step S2) and establishes a mapping between grid points and sensor data points. The matrix construction phase creates the interpolation system matrix. The matrix element calculation rule is: the distance between the grid point and the sensor point is input into the kernel function to generate the influence coefficient. The equation solution phase uses the preconditioned conjugate gradient method to solve large sparse linear equations. The iteration termination criterion is set when the residual norm is less than a specific threshold (e.g., 10 to the power of -6). Algorithm optimization strategies include excluding low-temperature differential areas during initialization (using the nearest neighbor algorithm for areas with temperature differences less than a set value, such as 1°C) and using a spatial tree index to accelerate distance retrieval. In an airport rooftop application (with 500,000 grid points), the algorithm's initialization phase was kept within a specific timeframe (e.g., 700 milliseconds).

[0104] The implementation method of dynamically adjusting the kernel function parameters according to the roof curvature change rate during the interpolation calculation process is: retrieve the curvature data of the current processing area from the pre-stored curvature distribution database (see the S3 step embodiment). The kernel function parameter adjustment strategy is: the larger the curvature change rate value, the smaller the support radius value is set, and the larger the shape parameter value is set. The specific conversion rule is determined by the experimental calibration curve: a mapping table of the curvature change rate and the optimal parameter group is established. When the curvature change rate increases by a specific value (for example, 0.1 per meter), the support radius is reduced by a specific proportion (for example, 5%), and the shape parameter is increased by a specific proportion (for example, 8%). The parameter dynamic loading mechanism adopts the regional grouping method to divide the roof into multiple processing blocks (block size is about 100×100 grids) according to the curvature characteristics, and each block uses a unified parameter group. A transition buffer is set when the parameters are switched to prevent step errors caused by sudden changes in the boundaries.

[0105] During the system deployment phase, the curvature distribution database is constructed according to the following process: 3D roof surface data is extracted from the project's BIM model. Standard curvature calculation tools in computer-aided design software (such as Rhino's "CurvatureAnalysis" command) are used to calculate Gaussian curvature values ​​at densely sampled points on the roof (e.g., one point every 0.5 meters). The coordinates and curvature values ​​of points whose absolute curvature exceeds a set threshold (e.g., 0.05) are stored in the spatial database. For example, the roof ridge of a stadium is automatically identified as a high-curvature area, and its coordinate boundaries are stored.

[0106] Curvature distribution database construction process:

[0107] Data input: Export the roof BIM model (.rvt format) from Revit and convert it into a .3dm file readable by Rhino;

[0108] Automated Processing: Executing Scripts in Rhino:

[0109]

[0110]

[0111] Regional aggregation: Use the "Cluster Points" component of Grasshopper (cluster radius 2m) to generate 15 mutation zones;

[0112] Boundary generation: Execute the "Convex Hull" command on each point set and output polygon vertices (e.g., ridge area: [(12.3,45.6),(12.5,45.8)...]);

[0113] Database storage: Vertex coordinates are stored in the PostGIS database, and an R-tree spatial index is established.

[0114] The operation process of giving priority to increasing the density of interpolation nodes in areas with high curvature changes is as follows: define high curvature change areas as areas where the curvature change rate exceeds a set threshold (for example, 0.15 per meter). The node encryption strategy adopts a quadtree subdivision algorithm: first detect the boundary of the high curvature area, and then perform a recursive subdivision operation inside the boundary. The specific steps are: calculate the temperature gradient value in the original grid unit. If the gradient value exceeds the set temperature gradient threshold (for example, 3 degrees Celsius per meter) and is in a high curvature area, the unit is subdivided into four sub-units. The temperature value of the newly added node is calculated by the cubic spline interpolation algorithm, and the spline control points are selected from the 9 surrounding original nodes. The encryption depth is linearly controlled by the curvature change rate value. For every specific increase in the curvature change rate (for example, 0.05 per meter), the recursive depth increases by one level (the maximum depth is limited to a specific value, such as level 3).

[0115] The implementation of interpolation calculations using a parallel computing architecture includes: The parallel architecture is designed as a CPU-GPU heterogeneous computing model, with the CPU responsible for logic control and data preprocessing, and the GPU performing the core computing tasks. The computational task decomposition strategy utilizes spatial domain decomposition: the roof is divided into multiple computational unit blocks (e.g., 64 blocks), each of which is assigned to an independent GPU compute core. Two levels of parallelism are implemented within the GPU compute core: block-level parallelism (coarse-grained) and grid-point-level parallelism (fine-grained). Memory management utilizes zero-copy technology, transferring data directly from host fixed memory to GPU video memory. The core compute kernel is configured with 2048 concurrent threads, each responsible for interpolating a specific number of grid points (e.g., 32). Page-locked memory barriers are used as a synchronization mechanism to ensure data consistency across block boundaries. On typical hardware configurations (e.g., NVIDIA V100 GPUs), the computation time for 100,000 interpolation points is reduced to a specific time (e.g., 80 milliseconds), achieving a specific speedup (e.g., 11x).

[0116] Generating continuous temperature field distribution data covering the entire roof includes: the final output data format is designed as a four-dimensional array structure, the first three dimensions represent spatial coordinate indexes (X, Y, Z axis grid indexes), and the fourth dimension stores temperature values. Spatial continuity is achieved by using cubic spline surface fitting, and a continuous surface is generated with the interpolation calculation results as control points. Data verification performs the following checks: randomly select a specific proportion (e.g. 5%) of point positions on the roof surface, compare the interpolated temperature with the real-time infrared scanning temperature, and if the deviation exceeds a specific threshold (e.g. ±1.5 degrees Celsius), automatically mark the abnormal area and trigger local recalculation. The result data is output to the temperature field database, which supports spatial range queries and time series backtracking. The final generated roof temperature field distribution data format complies with the ISO geospatial data exchange standard, achieving seamless integration with the building information system.

[0117] Extracting historical temperature gradient data from a previously generated roof temperature field distribution is as follows: Retrieve a roof temperature field distribution dataset with a time tag Tprev from the temperature field history database. This dataset contains the three-dimensional coordinates and temperature values ​​of all roof grid points. The temperature gradient is calculated using the central difference method: for any grid point P(x,y), the X-direction temperature gradient is the temperature of the right neighbor minus the temperature of the left neighbor divided by twice the X-direction grid spacing. The Y-direction temperature gradient is the temperature of the upper neighbor minus the temperature of the lower neighbor divided by twice the Y-direction grid spacing. The combined gradient is the Euclidean norm of the X-direction and Y-direction gradients. The calculated gradient for each grid point is stored in a two-dimensional gradient matrix, with the matrix row and column indices corresponding to the roof grid coordinate system. Data integrity verification is performed during the data extraction process: When a boundary point is detected with missing adjacent points, the virtual mirror point algorithm is automatically used to process the boundary. The temperature value of the virtual point is linearly extrapolated from the three actual adjacent points. The historical temperature gradient distribution data storage structure contains three key fields: spatial coordinate index, gradient value, and calculation time tag.

[0118] The process for calculating the temperature gradient change at the same spatial coordinate point in the currently generated roof temperature field distribution is as follows: First, align the time series, taking the previous generation time Tprev and the current generation time Tcurrent, ensuring that the time interval meets the preset requirement (e.g., 1 to 5 minutes). Spatial coordinate alignment is achieved through grid index matching: the current temperature field distribution data is mapped to the same roof grid coordinate system. The temperature gradient change is defined as the current temperature gradient value minus the previous temperature gradient value. The specific calculation formula is: the change is equal to Gcurrent minus Gprev, where Gcurrent is the value in the current gradient matrix and Gprev is the value at the same grid index in the historical gradient matrix. The calculated change results in a new matrix with the same spatial dimensions as the roof grid, where each element represents the gradient change at the corresponding grid point. In areas of sudden gradient changes (such as roof ridges), a secondary gradient change is additionally calculated: the rate of change of the gradient change. This value is stored in the additional data layer as an auxiliary judgment factor.

[0119] Comparing the temperature gradient change with a preset gradient threshold involves deriving the preset gradient threshold through a material thermal stress model. Specifically, the preset gradient threshold is derived by establishing a mechanical relationship between the roofing steel's thermal expansion coefficient (e.g., 11×10^-6 per degree Celsius), elastic modulus (210 GPa), and temperature gradient. The critical temperature gradient corresponding to the critical thermal stress is determined through numerical simulation (e.g., the gradient threshold corresponding to a finite element model under a stress of 10 MPa). The final threshold is set to a specific proportion of the critical value (e.g., 80%). The comparison process utilizes matrix parallel computing technology: an element-wise comparison operation is performed between the gradient change matrix and a constant value threshold matrix, and the comparison results are generated into a Boolean flag matrix. A true value in the flag matrix indicates exceeding the threshold, while a false value indicates normal. In engineering applications, the preset gradient threshold is dynamically adjusted based on the ambient temperature range: when the ambient temperature falls below a specific value (e.g., 10 degrees Celsius), the threshold is reduced by a specific proportion (e.g., 20%) to address the risk of increased steel brittleness.

[0120] In practice, the preset gradient threshold (e.g., 2.15 Kelvin per meter) is determined based on the mechanical properties of the material. This mechanical relationship refers to a standard method for calculating thermal stress in structural engineering: thermal stress equals the elastic modulus multiplied by the thermal expansion coefficient multiplied by the temperature difference. The critical temperature gradient value is determined by the following steps: first, setting the material's allowable safe thermal stress value (e.g., 10 MPa for steel); then, using the material parameters (elastic modulus 210 GPa, thermal expansion coefficient 0.000011 per Kelvin), inferring the maximum allowable local temperature difference; and finally, dividing this by the temperature sensor spacing (e.g., 2 meters) to obtain the critical gradient value. For example, if the maximum allowable temperature difference = 10 MPa / (210 GPa × 0.000011) ≈ 4.3 Kelvin, the critical gradient = 4.3 Kelvin / 2 meters = 2.15 Kelvin per meter. The preset gradient threshold is 80% of this critical value (i.e., 1.72 Kelvin per meter).

[0121] When the temperature gradient exceeds a preset gradient threshold, locating the corresponding roof curvature mutation area coordinates includes: first, filtering all true-value elements in the Boolean label matrix to obtain a set of super-threshold grid coordinates. The coordinate space association uses a surface nearest neighbor algorithm: with each super-threshold point as the center, searching for curvature mutation areas within a specific radius (e.g., 2 meters). The curvature mutation area coordinate data is retrieved from a preloaded spatial database (see step S3), which stores a sequence of vertex coordinates of the roof curvature mutation area polygons. The positioning method is: calculating the minimum distance from each super-threshold point to each curvature mutation area polygon. If the minimum distance is less than a preset relevant distance threshold (e.g., 0.5 meters), the point is determined to belong to the curvature mutation area. The relevant distance threshold is set according to the thermal conductivity of the roof material: the larger the material thermal conductivity, the larger the relevant distance threshold. The positioning result is output as a list of area codes, each code corresponding to the center coordinates and boundary range of a curvature mutation area.

[0122] Generating local recompensation instructions based on positioning results involves designing the instruction data structure as a machine-readable instruction sequence, consisting of an instruction header, a payload, and a checksum. The payload contains the definition of the spatial coordinate range to be recompensed: For each located area with a sudden change in curvature, the polygon vertex coordinate sequence is obtained and the bounding rectangle is calculated. The bounding rectangle is defined as: minimum X coordinate, maximum X coordinate, minimum Y coordinate, and maximum Y coordinate. This coordinate range is encapsulated as a spatial range object and attached to the instruction payload. Once generated, the instruction is immediately sent to a distributed message queue, where the queue routing mechanism automatically distributes it to the processing node in the corresponding area based on the spatial coordinates.

[0123] The instruction priority setting strategy is: the area with greater gradient changes has higher priority, and the priority is divided into levels 0-3 (level 0 is the highest). The priority value is attached to the instruction header as instruction metadata. During continuous operation, the system maintains an instruction execution status tracking table, triggers a timeout retransmission mechanism for instructions that are not responded to in time, and the timeout threshold is set to a specific time (for example, 30 seconds), and the maximum number of retries is limited to 3 times. The local re-compensation instruction is eventually delivered for compensation execution, providing accurate spatial positioning guidance for the re-acquisition and processing of local area temperature data. The effectiveness of the instruction is verified by infrared photography, and the test results show that the instruction positioning accuracy reaches a certain level (for example, the spatial deviation is less than 0.3 meters).

[0124] Example 2: Figure 2 The present invention provides a schematic structural diagram of a temperature field reconstruction system for a large-span roof, which includes the following modules:

[0125] Synchronous acquisition module, used to synchronously collect rooftop distributed temperature sensor data and micro-meteorological parameters, and stamp the temperature sensor data stream and micro-meteorological parameter stream with a unified time stamp;

[0126] A cache extraction module is used to establish a compensation buffer area and extract uncompensated temperature sensor data within a preset time window when the arrival of new micrometeorological parameters is detected;

[0127] A zoning decision module is used to divide the roof area into a real-time compensation area and a delayed compensation area based on the comparison result of the solar radiation intensity change rate in the new micro-meteorological parameters and the preset threshold;

[0128] The dynamic compensation module is used to dynamically compensate the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data. At the same time, the uncompensated temperature sensor data in the delayed compensation area is marked as a data set to be compensated, and asynchronous compensation operations are performed on the data set to be compensated during the temperature field reconstruction interval.

[0129] The field reconstruction module is used to input the real-time compensated data into the preset reconstruction algorithm to generate the roof temperature field distribution at the current moment;

[0130] The verification and positioning module is used to verify the time continuity of the reconstruction result based on the gradient change of the roof temperature field distribution generated previously and the roof temperature field distribution generated currently.

[0131] The calculations involved in the embodiments are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to actual conditions.

[0132] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0133] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0134] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0135] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0136] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0137] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A temperature field reconstruction method for a large-span roof, characterized in that: The steps include: S1. Synchronously collect rooftop distributed temperature sensor data and micrometeorological parameters, and add unified timestamps to the temperature sensor data stream and micrometeorological parameter stream; S2. Establish a compensation buffer area, and when the arrival of new micro-meteorological parameters is detected, extract the uncompensated temperature sensor data within a preset time window; S3. Divide the roof area into a real-time compensation area and a delayed compensation area based on a comparison result of the solar radiation intensity change rate in the new micrometeorological parameter and a preset threshold value; S4. Dynamically compensate the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data. Meanwhile, mark the uncompensated temperature sensor data in the delayed compensation area as a data set to be compensated, and perform asynchronous compensation on the data set to be compensated during the temperature field reconstruction interval. S5. Input the real-time compensated data into a preset reconstruction algorithm to generate the roof temperature field distribution at the current moment; S6. Based on the gradient change between the previously generated roof temperature field distribution and the currently generated roof temperature field distribution, perform a time continuity check on the reconstruction result.

2. The temperature field reconstruction method for a large-span roof according to claim 1 is characterized in that: Synchronously collect rooftop distributed temperature sensor data and micro-meteorological parameters, and assign unified timestamps to the temperature sensor data stream and micro-meteorological parameter stream, including: Acquire a first temperature sensor data stream through a distributed optical fiber sensor, acquire a second temperature sensor data stream through an infrared thermal imager, and acquire a solar radiation intensity data stream and a solar incidence angle change data stream through a micro-meteorological station; The first temperature sensor data stream of the distributed optical fiber sensor, the second temperature sensor data stream of the infrared thermal imager, and the solar radiation intensity data stream and solar incidence angle change data stream of the micro-meteorological station are given a unified time stamp through the time synchronization server.

3. The temperature field reconstruction method for a large-span roof according to claim 2 is characterized in that: A compensation buffer is established. When new micro-meteorological parameters are detected, the uncompensated temperature sensor data within the preset time window is extracted, including: Configure a ring storage queue to cache a first temperature sensor data stream and a second temperature sensor data stream with a unified timestamp; When a new solar radiation intensity data stream or a new solar incidence angle change data stream arrives, the preset time window is calculated forward based on the corresponding timestamp of the corresponding new micrometeorological parameter, and all uncompensated first temperature sensor data streams and uncompensated second temperature sensor data streams in the corresponding time window are extracted as the data set to be processed.

4. The temperature field reconstruction method for a large-span roof according to claim 3 is characterized in that: Based on the comparison between the solar radiation intensity change rate in the new micro-meteorological parameters and the preset threshold, the roof area is divided into a real-time compensation area and a delayed compensation area, including: The spatial coordinate distribution of the area with sudden curvature change of the preloaded roof; Calculate the rate of change value of the new solar radiation intensity data stream; When the rate of change exceeds the preset threshold, the entire coordinate range of the curvature mutation area is included in the real-time compensation area; For non-curvature mutation areas, based on the comparison results of the change amplitude of the solar incident angle change data stream with the preset angle change threshold, the sub-area whose change amplitude exceeds the preset angle change threshold is classified into the real-time compensation area, and the remaining areas are classified into the delayed compensation area.

5. The temperature field reconstruction method for a large-span roof according to claim 4 is characterized in that: Perform dynamic compensation on the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data, including: Acquire an uncompensated first temperature sensor data stream and an uncompensated second temperature sensor data stream corresponding to the real-time compensation area; Calculating a time-weighted compensation factor based on the timestamp difference between the new solar radiation intensity data stream, the new solar incidence angle change data stream, and the uncompensated temperature sensor data stream; A dynamic compensation formula is used to perform batch compensation operations on the uncompensated first temperature sensor data stream and the uncompensated second temperature sensor data stream respectively to generate real-time compensated data, including the real-time compensated first temperature sensor data stream and the real-time compensated second temperature sensor data stream.

6. The temperature field reconstruction method for a large-span roof according to claim 5 is characterized in that: The uncompensated temperature sensor data in the delay compensation area is marked as a data set to be compensated, including: The uncompensated first temperature sensor data stream and the uncompensated second temperature sensor data stream corresponding to the delay compensation area are stored in a queue to be compensated.

7. The temperature field reconstruction method for a large-span roof according to claim 6, characterized in that: During the temperature field reconstruction interval, asynchronous compensation operations are performed on the data set to be compensated, including: During the interval between reconstruction tasks, data is extracted from the queue to be compensated and the same dynamic compensation formula is used to perform compensation operations.

8. The temperature field reconstruction method for a large-span roof according to claim 7 is characterized in that: The real-time compensated data is input into the preset reconstruction algorithm to generate the roof temperature field distribution at the current moment, including: receiving a first temperature sensor data stream after real-time compensation and a second temperature sensor data stream after real-time compensation; Processing the first temperature sensor data stream and the second temperature sensor data stream after real-time compensation using a radial basis interpolation algorithm based on an adaptive kernel function; During the interpolation calculation process, the kernel function parameters are dynamically adjusted according to the roof curvature change rate, and the interpolation node density is increased preferentially in areas with high curvature changes; Interpolation operations are performed through a parallel computing architecture to generate continuous temperature field distribution data covering the entire roof as the roof temperature field distribution at the current moment.

9. The temperature field reconstruction method for a large-span roof according to claim 8, characterized in that: Based on the gradient change between the previously generated roof temperature field distribution and the currently generated roof temperature field distribution, the reconstruction result is checked for temporal continuity, including: Extract the historical temperature gradient distribution data from the previously generated roof temperature field distribution; Calculate the temperature gradient change at the same spatial coordinate point in the currently generated roof temperature field distribution; comparing the temperature gradient change with a preset gradient threshold; When the temperature gradient change exceeds the preset gradient threshold, the coordinates of the corresponding roof curvature mutation area are located; A local recompensation instruction is generated according to the positioning result, and the local recompensation instruction includes the spatial coordinate range that needs to be recompensed.

10. A temperature field reconstruction system for a large-span roof, used to implement a temperature field reconstruction method for a large-span roof according to any one of claims 1 to 9, characterized in that: Includes the following modules: Synchronous acquisition module, used to synchronously collect rooftop distributed temperature sensor data and micro-meteorological parameters, and stamp the temperature sensor data stream and micro-meteorological parameter stream with a unified time stamp; A cache extraction module is used to establish a compensation buffer area and extract uncompensated temperature sensor data within a preset time window when the arrival of new micrometeorological parameters is detected; A zoning decision module is used to divide the roof area into a real-time compensation area and a delayed compensation area based on the comparison result of the solar radiation intensity change rate in the new micro-meteorological parameters and the preset threshold; The dynamic compensation module is used to dynamically compensate the uncompensated temperature sensor data in the real-time compensation area to generate real-time compensated data. At the same time, the uncompensated temperature sensor data in the delayed compensation area is marked as a data set to be compensated, and asynchronous compensation operations are performed on the data set to be compensated during the temperature field reconstruction interval. The field reconstruction module is used to input the real-time compensated data into the preset reconstruction algorithm to generate the roof temperature field distribution at the current moment; The verification and positioning module is used to verify the time continuity of the reconstruction result based on the gradient change of the roof temperature field distribution generated previously and the roof temperature field distribution generated currently.

Images