A method and system for monitoring temperature effects of metal roofs

Abstract

The application discloses a kind of temperature effect monitoring method and system of metal roof, specifically related to building structure health monitoring technical field, for solving the problems such as temperature monitoring data and structure mechanics response decoupling, static early warning threshold invalid in prior art;Three-dimensional temperature field data is obtained by distributed temperature sensing network, and temperature gradient eigenvalue is extracted therefrom;Piezoelectric film sensor is used to collect constraint point micro-vibration signals, and dynamic constraint counterforce spectrum eigenvalue is extracted by signal transformation and manifold learning;Temperature gradient and mechanical characteristics are coupled to establish a model to generate temperature stress evaluation value;Early warning threshold range is dynamically adjusted according to environmental parameters;When evaluation value exceeds threshold value, output structure safety early warning signal;Multi-physical field coupling analysis of temperature field distribution and structure response is realized, dynamic early warning mechanism of environment self-adaptation is constructed, and the accuracy and reliability of metal roof temperature effect monitoring are significantly improved.

Description

Technical Field

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

[0002] Metal roofs are widely used in large public buildings such as airport terminals and stadiums due to their high structural strength and ease of construction. Existing metal roof temperature monitoring technologies primarily rely on localized temperature sensors to collect data and apply fixed thresholds for safety assessments. While these solutions have proven effective in providing basic temperature information, they are limited by the single-point data collection model and struggle to fully characterize the overall temperature distribution of large-scale roofs.

[0003] The defects of existing technologies are: the temperature monitoring data is decoupled from the structural mechanical response, which is specifically manifested as: single-point temperature measurement cannot establish a three-dimensional temperature gradient field, resulting in distorted assessment of the temperature load range; there is a lack of a real-time correlation mechanism between temperature data and structural strain and displacement, and the actual impact of temperature changes on structural safety cannot be quantified; the static warning threshold is disconnected from the dynamic environmental parameters, and the risk of failure is significant under seasonal changes and extreme weather conditions. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for monitoring the temperature effect of a metal 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 method for monitoring the temperature effect of a metal roof comprises the following steps:

[0007] S1: Obtain three-dimensional temperature field data of the metal roof through a distributed temperature sensing network;

[0008] S2: Extract the temperature gradient characteristic value of the metal roof from the three-dimensional temperature field data;

[0009] S3: The micro-vibration signal sequence of the metal roof constraint points is collected through a piezoelectric film sensor array. The micro-vibration signal sequence is processed using the Hilbert-Huang transform to generate an instantaneous frequency-energy spectrum. The dynamic constraint reaction force spectrum eigenvalues ​​are extracted from the instantaneous frequency-energy spectrum using the Riemannian manifold learning method.

[0010] S4: Establish a coupling relationship model between the temperature gradient eigenvalue and the dynamic constraint reaction spectrum eigenvalue to generate the temperature stress evaluation value of the metal roof;

[0011] S5: Dynamically adjust the warning threshold range according to solar radiation intensity and ambient wind speed parameters;

[0012] S6: outputting a structure safety early warning signal of the metal roof when the temperature stress evaluation value exceeds the early warning threshold range.

[0013] Further, the three-dimensional temperature field data of the metal roof is acquired through the distributed temperature sensing network, including:

[0014] The distributed temperature sensing network includes a plurality of temperature measurement optical fiber layout subsystems at an angle to each other; the three-dimensional temperature field data is acquired through the distributed optical fiber temperature measurement technology of the temperature measurement optical fiber layout subsystems to collect discrete temperature point data of the metal roof surface, and the discrete temperature point data is subjected to spatial correlation analysis, and a three-dimensional temperature field data is reconstructed and generated through an improved interpolation algorithm.

[0015] Further, the temperature measurement optical fiber layout subsystem covers the entire metal roof area.

[0016] Further, the temperature gradient characteristic value of the metal roof is extracted from the three-dimensional temperature field data, including:

[0017] The temperature difference value of different spatial positions of the metal roof is calculated, the physical boundary constraint condition of the metal roof is identified based on the three-dimensional temperature field data, the temperature difference value is corrected in combination with the physical boundary constraint condition, and a feature vector is extracted as a temperature gradient characteristic value from the corrected temperature difference value through a principal component analysis algorithm.

[0018] Further, the micro-vibration signal sequence of the constraint point of the metal roof is acquired through the piezoelectric film sensor array, the Hilbert-Huang transform is used to process the micro-vibration signal sequence to generate an instantaneous frequency-energy spectrum, and a dynamic constraint counterforce spectrum characteristic value is extracted from the instantaneous frequency-energy spectrum through a Riemannian manifold learning method, including:

[0019] The micro-vibration signal sequence is decomposed into an intrinsic modal function component through the Hilbert-Huang transform;

[0020] The intrinsic modal function component is subjected to Hilbert spectrum analysis to generate an instantaneous frequency-energy spectrum;

[0021] The instantaneous frequency-energy spectrum is mapped to a Riemannian manifold space to construct a tangent vector field;

[0022] The geodesic line distance matrix characteristic value of the tangent vector field is calculated;

[0023] The feature vector corresponding to the maximum geodesic line distance matrix characteristic value is selected as the dynamic constraint counterforce spectrum characteristic value.

[0024] Further, the micro-vibration signal sequence is acquired through the piezoelectric film sensor array to obtain the vibration response of the constraint point of the metal roof.

[0025] Furthermore, a coupling relationship model between the temperature gradient eigenvalue and the dynamic constraint reaction spectrum eigenvalue is established to generate the temperature stress assessment value of the metal roof, including:

[0026] Construct the covariance tensor of the temperature gradient eigenvalue and the dynamic constraint reaction force spectrum eigenvalue;

[0027] Perform orthogonal decomposition on the covariance tensor to obtain the eigenmode matrix;

[0028] The mapping relationship between the characteristic mode matrix and the temperature stress value is established through the least squares regression algorithm;

[0029] The real-time eigenmode matrix is ​​input into the mapping relationship to generate the temperature stress evaluation value of the metal roof.

[0030] Furthermore, the warning threshold range is dynamically adjusted according to the solar radiation intensity and ambient wind speed parameters, including:

[0031] Calling the benchmark threshold range in the historical operating condition database;

[0032] Construct a fuzzy logic rule base for solar radiation intensity and ambient wind speed parameters;

[0033] Match the correction factor in the fuzzy logic rule base according to the real-time solar radiation intensity and ambient wind speed parameters;

[0034] The dynamic warning threshold range is generated by multiplying the baseline threshold range by the correction factor.

[0035] Furthermore, when the temperature stress evaluation value exceeds the warning threshold range, a structural safety warning signal of the metal roof is output, including:

[0036] Mapping the temperature stress assessment values ​​to the three-dimensional grid coordinates of the metal roof;

[0037] Identify the grid coordinate areas that exceed the warning threshold range and generate risk area identification;

[0038] Construct risk state vector based on risk area identification;

[0039] The risk state vector is encoded into a structural safety warning signal through the orthogonal coding conversion algorithm.

[0040] In another aspect, the present invention provides a temperature effect monitoring system for metal roofs, comprising the following modules:

[0041] Temperature sensing module, used to obtain three-dimensional temperature field data of the metal roof through a distributed temperature sensing network;

[0042] Gradient extraction module, used to extract the temperature gradient characteristic value of the metal roof from the three-dimensional temperature field data;

[0043] The mechanical feature module is used to collect the micro-vibration signal sequence of the metal roof constraint points through a piezoelectric film sensor array, process the micro-vibration signal sequence using the Hilbert-Huang transform to generate an instantaneous frequency-energy spectrum, and extract the dynamic constraint reaction force spectrum eigenvalues ​​from the instantaneous frequency-energy spectrum using the Riemannian manifold learning method;

[0044] Thermomechanical coupling module is used to establish a coupling relationship model between the temperature gradient eigenvalue and the dynamic constraint reaction spectrum eigenvalue to generate the temperature stress assessment value of the metal roof;

[0045] Threshold dynamic module, used to dynamically adjust the warning threshold range according to solar radiation intensity and ambient wind speed parameters;

[0046] The safety warning module is used to output a structural safety warning signal of the metal roof when the temperature stress evaluation value exceeds the warning threshold range.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The technical solution of the present invention constructs a holographic monitoring system for the temperature effects of metal roofs through the deep collaboration of distributed temperature sensing networks and mechanical feature extraction. Compared with the traditional single-point monitoring mode, its core breakthrough lies in the real-time coupled analysis of the three-dimensional distribution of the temperature field and the structural mechanical response. Specifically, the spatially refined extraction of temperature gradient eigenvalues ​​solves the problem of distortion in the assessment of the temperature load range, while the dynamic constraint reaction force spectrum eigenvalues ​​generated based on micro-vibration signal spectrum analysis establishes a quantitative correlation mechanism between temperature changes and structural constraints in the field of engineering monitoring. This collaborative analysis mode of multi-physical field characteristics enables the temperature stress assessment process to truly reflect the thermal-mechanical coupling effects of metal roofs.

[0049] Through the adaptive early warning decision-making mechanism of environmental parameters, dynamic optimization of metal roof safety protection is achieved; based on the fuzzy rule library of solar radiation intensity and environmental wind speed parameters, the early warning threshold range is dynamically adjusted, effectively overcoming the failure risk of fixed thresholds under seasonal changes and extreme weather conditions; the real-time comparison mechanism of temperature stress assessment values ​​and dynamic thresholds, combined with orthogonal coding conversion to generate anti-interference early warning signals, not only greatly improves the accuracy of early warnings, but also forms a closed-loop control chain from data collection to risk response; this paradigm shift from "passive monitoring" to "active protection" provides reliable technical support for the full life cycle safety management of metal roofs of large buildings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flow chart of a method for monitoring the temperature effect of a metal roof according to the present invention;

[0051] Figure 2 The figure is a schematic structural diagram of a temperature effect monitoring system for a metal roof according to the present invention. DETAILED DESCRIPTION

[0052] 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.

[0053] Example 1: Figure 1 The present invention provides a method for monitoring the temperature effect of a metal roof, which comprises the following steps:

[0054] S1: Obtain three-dimensional temperature field data of the metal roof through a distributed temperature sensing network;

[0055] S2: Extract the temperature gradient characteristic value of the metal roof from the three-dimensional temperature field data;

[0056] S3: The micro-vibration signal sequence of the metal roof constraint points is collected through a piezoelectric film sensor array. The micro-vibration signal sequence is processed using the Hilbert-Huang transform to generate an instantaneous frequency-energy spectrum. The dynamic constraint reaction force spectrum eigenvalues ​​are extracted from the instantaneous frequency-energy spectrum using the Riemannian manifold learning method.

[0057] S4: Establish a coupling relationship model between the temperature gradient eigenvalue and the dynamic constraint reaction spectrum eigenvalue to generate the temperature stress evaluation value of the metal roof;

[0058] S5: Dynamically adjust the warning threshold range according to solar radiation intensity and ambient wind speed parameters;

[0059] S6: When the temperature stress evaluation value exceeds the warning threshold range, a structural safety warning signal of the metal roof is output.

[0060] Multiple temperature measurement fiber optic laying subsystems are deployed on the metal roof at angles to each other, with the angles between the temperature measurement fiber optic laying subsystems set between 45 and 135 degrees to ensure that the laying paths of all temperature measurement fiber optic laying subsystems cross-cover the entire metal roof area and that there are no monitoring blind spots. During specific implementation, the temperature measurement fiber optic laying subsystem lays the first temperature measurement fiber optic path along the ridge of the metal roof, the second temperature measurement fiber optic path along the eaves of the metal roof edge, and the third temperature measurement fiber optic path along the diagonal direction of the metal roof. The triangular grid layout formed by the three groups of temperature measurement fiber optic paths ensures that any roof area is covered by at least two temperature measurement fiber optic paths. The density of the fiber optic sensing nodes of the temperature measurement fiber optic laying subsystem is controlled to be no less than 4 measurement points per square meter. For example, in a large terminal project, a total of 48,600 measurement points are set up in an area with a roof area of ​​12,000 square meters.

[0061] Distributed fiber optic temperature measurement technology collects discrete temperature data from the metal roof surface. This technology involves: a laser pulse generator transmits a laser pulse signal with a central wavelength of 1550 nanometers to the temperature measurement fiber deployment subsystem, with a single pulse width controlled to the order of 100 nanoseconds. The receiver collects the backscattered light signal and performs time-domain analysis. The absolute temperature value at each sensor node along the fiber path is calculated based on the intensity ratio of the Stokes and anti-Stokes light from the Raman scattering effect. Each temperature value corresponds to a specific coordinate position in the three-dimensional spatial coordinate system of the metal roof surface, with a coordinate positioning error of less than 2 mm. This ultimately generates a discrete temperature point dataset containing spatial coordinates and temperature values. For example, during implementation, the discrete temperature point data recorded by a single sensor node is formatted as (X coordinate value, Y coordinate value, Z coordinate value, temperature value).

[0062] A spatial correlation analysis of discrete temperature point data consists of two core processing stages: the first stage establishes a three-dimensional spatial coordinate grid matrix for the metal roof, with a grid cell size of 0.25 meters by 0.25 meters. Each discrete temperature point data point is mapped to the center coordinate of the corresponding grid cell. The second stage calculates the temperature autocorrelation function of adjacent grid cells and calculates the correlation coefficient using the Moran index formula in spatial statistics. The Moran index calculation requires constructing a spatial weight matrix. The weight coefficient is determined by the Euclidean distance between the grid cells. For example, the weight coefficient for a grid cell with a 1-meter spacing is set to 0.85, while the weight coefficient for a grid cell with a 2-meter spacing is reduced to 0.45. The output of the spatial correlation analysis identifies the coordinates of areas with abnormal temperature fluctuations.

[0063] Three-dimensional temperature field data is reconstructed using an improved interpolation algorithm. This algorithm implements three specific optimizations within the standard kriging interpolation framework: First, a heat conduction boundary constraint is introduced, defining fixed connections on the metal roof as interpolation boundaries that cannot be crossed by isotherms. For example, the ridge weld is defined as an isothermal boundary. Second, a gradient direction correction factor is added. When the angle between the interpolation direction and the measured temperature gradient is greater than 45 degrees, the interpolation weight in that direction is automatically multiplied by a correction factor of 0.6. Third, the iteration termination condition is set when the overall temperature change between two consecutive interpolation results is less than 0.5 degrees Celsius. The reconstruction process detects and removes outlier temperature points in real time. The outlier criterion is that a data point exceeds the average temperature of the surrounding eight neighborhoods by more than 10 degrees Celsius. The resulting three-dimensional temperature field data achieves a spatial resolution of 0.0625 square meters. For example, the reconstruction for the stadium project contains 192,000 temperature data points.

[0064] The temperature measurement fiber optic deployment subsystem uses armored distributed temperature measurement optical cables with an IP68 protection rating. Spatial correlation analysis uses an ellipsoid model to calculate the spatial weight coefficient, with the long axis of the ellipsoid aligning with the roof slope. An improved interpolation algorithm dynamically adjusts the interpolation step size during data processing, automatically reducing it to 60% of the original value when a temperature gradient greater than 15 degrees Celsius per meter is detected. The relative error of the three-dimensional temperature field data obtained through this implementation process does not exceed 1.2%. For example, the deviation between the measured temperature value and the reconstructed value is less than 0.46 degrees Celsius in a high-temperature environment of 38 degrees Celsius. All processing steps are based on the geometric topology of the metal roof and the thermal conductivity characteristics of the material, ensuring that the three-dimensional temperature field data truly reflects the thermodynamic state of the roof.

[0065] The three-dimensional temperature field data is collected in real time by a distributed temperature sensing network. This data contains temperature measurements at each spatial coordinate point on the metal roof surface. When calculating temperature differences at different spatial locations on the metal roof, the Euclidean distance between adjacent nodes in the three-dimensional grid coordinate system is used as the basis for calculation. The specific implementation process involves first determining the distance threshold between spatial locations within a range of 1 to 3 meters. For example, grid nodes with a distance of 2 meters on the roof plane are selected as calculation units. Then, within each calculation unit, the temperature values ​​of multiple spatial locations are compared. The temperature difference is calculated using an absolute temperature difference algorithm, which calculates the absolute value of the difference between the current spatial location and the adjacent spatial location. For example, at the ridge of a metal roof, if the temperature is 52 degrees Celsius and the adjacent eaves point is 38 degrees Celsius, the temperature difference in this area is 14 degrees Celsius. During implementation, a temperature difference matrix is ​​established to store the calculation results for all spatial locations. The distance threshold is set to ensure that the grid units cover the temperature-sensitive area and meet the required 3D reconstruction accuracy.

[0066] The physical boundary constraints of metal roofs are identified based on three-dimensional temperature field data. Physical boundary constraints are defined as the structural characteristics of fixed joints in metal roofs. The identification process consists of three stages: The first stage locates potential boundary coordinates by detecting sudden temperature gradients. The boundary identification mechanism is triggered when the temperature change rate of adjacent nodes exceeds 10 degrees Celsius per meter. The second stage verifies the validity of the identified physical boundaries by comparing them with a database of structural joint locations in the building design drawings. The third stage establishes a constraint attribute classification system, categorizing constraint types into rigid connection constraints, sliding connection constraints, and free boundary constraints. For example, weld seams in metal roof panels are classified as rigid connection constraints, while roof expansion joints are classified as sliding connection constraints. The final output includes a feature description table containing the spatial coordinates of the boundary, the constraint radius, and the direction of the constraint force. The constraint radius can be set from 30 cm to 150 cm, depending on the size of the connector.

[0067] The temperature difference value is corrected in conjunction with the physical boundary constraints, and the correction process implements a differentiated processing mechanism based on the constraint type. For rigid connection constraint areas, the temperature difference value is multiplied by the heat transfer suppression coefficient. The heat transfer suppression coefficient range is set between 0.5 and 0.9 and is determined by the thermal conductivity characteristics of the fixed connection material. For example, the coefficient for stainless steel connection areas is 0.7. For sliding connection constraint areas, a displacement compensation factor is added to the temperature difference value. The compensation factor value is equal to the theoretical displacement prediction value divided by the material linear expansion coefficient. For free boundary areas, the temperature difference value is directly used as the original calculation result without correction. In the implementation case of a large terminal building, after the metal roof expansion joint area is identified as a sliding connection constraint type, a displacement compensation factor of 0.28 is added to all temperature difference values ​​within an 80 cm radius of the area. When the correction mechanism is executed, the parameters in the physical boundary constraint database are automatically matched to establish a corresponding relationship matrix between temperature difference values ​​and constraint characteristics.

[0068] The principal component analysis algorithm is used to extract eigenvectors from the corrected temperature difference values ​​as temperature gradient eigenvalues. The implementation process of the principal component analysis algorithm includes four core steps. First, the covariance matrix of the corrected temperature difference values ​​is constructed. The matrix dimension is consistent with the number of roof space grid cells. Second, the covariance matrix is ​​subjected to eigenvalue decomposition to solve the eigenvector set. Then, the dominant eigenvectors are selected based on the eigenvalue size. The selection criteria are set to the principal component direction with a cumulative contribution rate of more than 80%. Finally, the selected eigenvectors are normalized. The extracted temperature gradient eigenvalues ​​include two key parameters: the directional eigenvector component and the modulus eigenvalue component. The directional eigenvector component represents the projection angle of the dominant direction of the heat flow in the three-dimensional space coordinate system, and the modulus eigenvalue component represents the intensity of the temperature gradient. During implementation, the eigenvector dimensions are kept consistent with the three-dimensional coordinate system. For example, the temperature gradient eigenvalue extracted from a gymnasium roof has a directional eigenvalue component of (0.89, 0.44, 0.09) and a modulus eigenvalue component of 28.3 degrees Celsius per meter, indicating that the heat flow is at a 26-degree angle to the roof plane and the gradient intensity is 28.3 degrees Celsius per meter. The iteration termination criteria for the principal component analysis algorithm are set to be a change in the eigenvector direction of less than 0.15 degrees or a change in the modulus of less than 0.4 degrees Celsius per meter.

[0069] The implementation process is configured with a special scenario processing mechanism: when rain or snow is detected by the weather station, the surface temperature difference correction mode is automatically activated. This mode excludes temperature changes caused by surface water evaporation when performing temperature difference value calculations; when the temperature gradient direction deviates from the roof slope direction by more than 30 degrees, the system triggers the direction recalibration process, which optimizes the result by weighted averaging the three sets of characteristic vectors in the adjacent area. The physical properties of metal roof materials are stored in the system characteristic parameter database. For example, the linear expansion coefficient parameter of steel is 12×10-6 / ℃, the linear expansion coefficient parameter of aluminum alloy material is 23×10 -6 / °C, providing a precise calculation basis for temperature difference correction. All extracted temperature gradient eigenvalues ​​are displayed in real time within the building model via a 3D visualization system. Different color depths are used to identify modulus eigenvalue components, and arrow vectors are used to identify directional eigenvector components, forming a complete roof heat flow distribution map. The final output is automatically linked to the temperature effect assessment process, providing thermodynamic characteristic input for subsequent temperature stress analysis.

[0070] Through the temperature gradient eigenvalue extraction process, the thermal conductivity characteristics of the metal roof are quantified into mathematical features that can be calculated and analyzed. For example, during the period of changing solar radiation intensity in winter, the maximum modulus eigenvalue component recorded in the south area of ​​the roof reached 42.7 degrees Celsius per meter. This quantification result provides accurate input parameters for temperature stress assessment. During the implementation process, a data quality monitoring mechanism was established. When it was detected that the modulus value of the temperature gradient eigenvalue extracted five times in the same area changed by more than 15%, the recalibration instruction of the distributed temperature sensing network was automatically triggered to ensure the continuity and reliability of the temperature effect monitoring data. The implementation case of the feature extraction algorithm showed that in the terminal building project with a roof area of ​​30,000 square meters, the complete temperature gradient eigenvalue extraction cycle was completed within 25 seconds, meeting the real-time requirements of structural health monitoring.

[0071] An array of piezoelectric film sensors is installed at the metal roof's constraint points, including fixed support nodes, weld joints, and expansion joint connectors. The effective measurement range of each piezoelectric film sensor covers a 12 mm diameter area, and the sensor spacing is set to a maximum distance of no more than 1.8 meters between adjacent constraint points. For example, the sensor array is arranged in a 0.9-meter equidistant grid in the key area of ​​the roof ridge. The sensor array is directly attached to the metal surface and converts mechanical vibrations into electrical signal waveforms through a charge conversion circuit. The sampling frequency is set to 20,000 Hz, and the signal resolution reaches 0.001 meters per second squared. The collected micro-vibration signal sequence contains time series and spatial coordinate information. For example, the coordinates of a weld position are recorded as three-dimensional values, corresponding to a continuous sequence of vibration data points collected over a period of 5 seconds.

[0072] The piezoelectric film sensor array acquires the vibration responses of the metal roof's restraint points in real time, generating a micro-vibration signal sequence. This sequence is then transmitted to the signal processing unit via a data acquisition system. Upon receiving the raw signal, the signal processing unit first performs preprocessing: it automatically detects abnormal pulse signals and eliminates interference points using a median filter with a filter window width set to 51 sampling points. The signal is then normalized to map all amplitude values ​​to a standard range of -1.0 to 1.0. The processed micro-vibration signal sequence is stored in a time-space matrix format, with rows representing time series indices and columns representing spatial sensor node numbers. For example, 128 sensor nodes form a 128-column data matrix.

[0073] The micro-vibration signal sequence is decomposed into intrinsic mode function components using the Hilbert-Huang transform, which implements the empirical mode decomposition algorithm. The decomposition process begins with an iterative screening operation: first, the local maximum and minimum points of the input signal are identified. These extreme points are then connected using cubic spline interpolation to form upper and lower envelopes. The mean curve of the upper and lower envelopes is then calculated, and finally, this mean curve is subtracted from the original signal to obtain the preliminary modal components. After each iteration, the components are checked to see if they meet the intrinsic mode function conditions: the difference between the number of zero crossings and the number of extreme points does not exceed one; and the local mean at all locations approaches zero. The iterative termination condition is set to a standard deviation of less than 0.3 over three consecutive screenings. Finally, multiple eigenvimotor components that meet the conditions are output. For example, a weld vibration signal is decomposed into five groups of intrinsic mode function components with a frequency distribution between 180 Hz and 1900 Hz.

[0074] Hilbert spectrum analysis is performed on the intrinsic mode function components to generate an instantaneous frequency-energy spectrum. The analysis process uses orthogonal demodulation technology. First, a Hilbert transform is performed on each intrinsic mode function component to generate a corresponding analytical signal. Second, the instantaneous phase angle of the analytical signal is calculated, and the instantaneous frequency value is solved by phase angle differentiation. The frequency calculation time step is set to 0.05 milliseconds. Finally, a frequency-energy relationship matrix is ​​constructed based on the instantaneous frequency value and the envelope amplitude value at the corresponding moment. The instantaneous frequency-energy spectrum ultimately forms a two-dimensional distribution diagram, with the horizontal axis representing the frequency value and the vertical axis representing the energy density value. The spectrum resolution is set to 0.5 Hz. For example, a peak energy density of 1.8 volts per hertz was recorded at 125 Hz at a certain expansion joint location.

[0075] The instantaneous frequency-energy spectrum is mapped to a Riemannian manifold space to construct a tangent vector field. This mapping process is based on differential geometry theory. First, an orthogonal tangent space coordinate system is defined on the frequency-energy plane, with the origin of the coordinate system set at the lowest energy point in the spectrum. Next, each data point in the instantaneous frequency-energy spectrum is converted into a tangent vector. The length of the tangent vector is determined by the natural logarithm of the energy density value, and the directional angle is determined by the phase difference between the frequency value and the reference frequency value. Once the tangent vector field is constructed, a set of spatial vectors is formed, each associated with a specific spatial coordinate position. For example, a tangent vector with a modulus of 0.43 and a directional angle of 27.5 degrees is generated for a frequency of 325 Hz. This vector set is automatically mapped to the roof's three-dimensional coordinate system.

[0076] The eigenvalues ​​of the geodesic distance matrix of the tangent vector field are calculated using an iterative optimization algorithm. First, a connection diagram of the tangent vector positions is constructed, with the connection conditions set to a frequency difference of less than 5 Hz and a spatial distance of less than 0.7 meters. Next, the geodesic distance between each pair of connection points is calculated, representing the shortest path length on the manifold surface. The shortest path is iteratively calculated using the Floyd algorithm, with the iterative convergence threshold set to a maximum difference of no more than 0.03 between the two adjacent distances. Finally, the calculated results are stored as a symmetric distance matrix with a dimension equal to the total number of tangent vectors. In this case study, the distance matrix formed by 256 tangent vectors in a roof region is subjected to eigenvalue decomposition using the Jacobi method.

[0077] The eigenvector corresponding to the maximum geodesic distance matrix eigenvalue is selected as the eigenvalue of the dynamic constraint reaction force spectrum, and the process implements a standardized processing flow. First, the entire eigenvalue set is obtained through the eigenvalue decomposition algorithm, and the eigenvalue decomposition calculation accuracy is controlled at 0.001; secondly, all eigenvalues ​​are sorted in descending order, and the maximum eigenvalue is selected as the main eigencomponent; then the eigenvector corresponding to the main eigenvalue is extracted and vector normalization is performed; finally, the normalized eigenvector is associated with the corresponding physical space coordinate and stored. The final expression of the eigenvalue of the dynamic constraint reaction force spectrum includes two components: the modulus and the phase angle. For example, the modulus characteristic of the eigenvector extracted from a fixed support node is 0.94, and the phase angle characteristic is 31.6 degrees, indicating the strength and direction characteristics of the constraint reaction force at that position. The calculation time of the entire feature extraction algorithm is controlled to be completed within 350 milliseconds.

[0078] During the implementation process, special operating condition processing rules are configured: when the ambient vibration intensity exceeds a set threshold, noise reduction mode is automatically activated; when the temperature exceeds 80 degrees Celsius, a temperature compensation algorithm is activated to calibrate sensor sensitivity. The mechanical parameters of metal roofing materials are stored in the system's physical property database. For example, the elastic modulus of steel is 206 GPa and the Poisson's ratio is 0.28, providing a basis for feature extraction. All generated dynamic constraint reaction force spectrum eigenvalues ​​are displayed in real time within the 3D model via a color cloud map. The color depth of the cloud map corresponds to the eigenvector modulus value, and the vector direction indicates the direction of the reaction force.

[0079] The feature extraction system incorporates a data quality monitoring mechanism: when the convergence time of three consecutive eigenvalue calculations exceeds 450 milliseconds, the algorithm parameter configuration is automatically optimized; when the eigenvector angular fluctuation exceeds 15 degrees, a directional recalibration process is triggered. In a field test on a stadium roof project, piezoelectric film sensors covered 32 key constraint points, and the eigenvalue convergence iterations for the geodesic distance matrix calculation averaged 17 times, with the entire process completing within 1.9 seconds. The correlation coefficient between the eigenvalues ​​of the dynamic constraint reaction spectrum and the temperature load experimental data reached 0.93, verifying the technical reliability of monitoring the temperature effects of metal roofs.

[0080] The temperature gradient eigenvalue is derived from the thermodynamic state data of the metal roof and consists of two elements: a directional eigenvector component and a modulus eigenvalue component. The directional eigenvector component identifies the angular projection of the main direction of heat flow in the three-dimensional coordinate system, while the modulus eigenvalue component reflects the intensity of the temperature change rate (measured in degrees Celsius per meter). The dynamic constraint reaction spectrum eigenvalue characterizes the structural mechanical response and consists of two parameters: the modulus eigenvalue represents the constraint force intensity (dimensionless), and the phase angle eigenvalue represents the reaction force direction angle (measured in degrees). These two eigenvalues ​​are synchronously input into the coupled modeling system through a data interface, with input timestamp alignment accuracy controlled within 20 milliseconds to ensure data consistency across the temporal dimension.

[0081] To construct the covariance tensor for the temperature gradient eigenvalues ​​and the dynamic constraint reaction spectrum eigenvalues, a three-dimensional tensor structure is first defined: the first dimension corresponds to the temperature gradient eigenvalue modulus eigenvalue component, the second dimension corresponds to the dynamic constraint reaction spectrum eigenvalue modulus eigenvalue quantity, and the third dimension corresponds to the directional correlation angle difference between the two eigenvalue types. The covariance formula for the tensor elements is used: the covariance value within each spatial grid cell is the product of the temperature gradient eigenvalue modulus eigenvalue component and the dynamic constraint reaction spectrum eigenvalue modulus eigenvalue quantity within that cell, minus the product of their respective means. Data normalization is performed before calculation: the temperature gradient eigenvalue modulus eigenvalue component is dimensionalized by dividing it by its historical maximum value, and the dynamic constraint reaction spectrum eigenvalue modulus eigenvalue quantity is dimensionless by dividing it by its historical mean. The resulting covariance tensor has the same dimensions as the number of roof grid partitions. For example, a 32-partition system generates a 32×32×32 three-dimensional tensor matrix, with each tensor element representing the thermal-mechanical coupling strength of a specific spatial region.

[0082] The covariance tensor is orthogonally decomposed to obtain the eigenmode matrix, and a high-order singular value decomposition algorithm is implemented. The decomposition process consists of four stages: the first stage expands the three-dimensional tensor into a matrix sequence, with each slice matrix corresponding to a specific directional correlation angle difference interval; the second stage performs singular value decomposition on each slice matrix, retaining the principal components with singular value contributions exceeding 85%; the third stage reconstructs the core tensor and calculates the eigenvectors; and the fourth stage sorts the eigenvectors by variance contribution to form the eigenmode matrix. The decomposition iterations are terminated when the difference between two consecutive eigenvectors is less than 0.03, and the maximum number of iterations is limited to 100. Each column vector of the eigenmode matrix corresponds to an independent eigenmode, and the matrix row numbers are associated with specific spatial grid coordinates. For example, the eigenvectors corresponding to a roof panel region (X3, Y7) are [0.48, 0.32, 0.21, 0.07], representing the weight coefficient distribution of the four main modes in that region.

[0083] A least-squares regression algorithm was used to establish a mapping relationship between the eigenmode matrix and the temperature stress values. This process employed a step-by-step training mechanism. First, a historical operating condition dataset was collected: the eigenmode matrix records and the corresponding measured temperature stress values ​​(in megapascals) were retrieved from the metal roof structural health monitoring system. The dataset consisted of at least 500 valid samples. Next, a multivariate linear regression equation was constructed: the temperature stress values ​​were set as the dependent variable and the weight coefficient vector of the eigenmode matrix as the independent variable. The regression coefficients were then calculated using a residual sum-of-squares minimization algorithm, with a residual threshold set at 0.25 MPa. Finally, the model accuracy was verified using a ten-fold cross-validation method to examine the correlation coefficients between the predicted and measured values. The final representation of the mapping relationship was a weight coefficient vector. For example, the coefficient vector [0.87, 0.35, -0.12, 0.09] in the regression equation represents the contribution weights of the four eigenmodes to the temperature stress. During model training, outlier points were automatically excluded; the outlier criterion was that a data point exceeded three times the standard deviation.

[0084] The real-time eigenmode matrix input is mapped to generate a temperature stress assessment value for the metal roof, and the generation process is dynamically verified. First, the current eigenmode matrix data is read and the weight coefficient vector trained historically is matched by the grid coordinate index. Next, the temperature stress assessment value is calculated: the temperature stress value for each grid cell is equal to the dot product of the regression coefficient vector of the row vector of the eigenmode matrix for that cell. Finally, a physical range check is performed: if the calculated result exceeds the yield strength range of the metal material, a secondary calculation is automatically triggered. The assessment value is output as a spatially distributed matrix, with the matrix elements representing the temperature stress value (in megapascals) at each grid point. For example, the assessment value for the ridge area (X5, Y9) is 152.7 MPa. The system establishes an output quality monitoring mechanism: when a sudden change in stress value exceeding 25 MPa is detected in adjacent grids, the input eigenvalue quality is automatically verified. When the material temperature exceeds 120 degrees Celsius, a high-temperature stress correction algorithm is activated to adjust the output value.

[0085] During the implementation process, an adaptive material property mechanism is configured: Based on the actual material type of the metal roof panels, the corresponding material parameter database is automatically matched. For example, aluminum alloy panels are matched with an elastic modulus of 70,000 MPa and a thermal expansion coefficient of 0.000023 degrees Celsius, while galvanized steel panels are matched with an elastic modulus of 206,000 MPa and a thermal expansion coefficient of 0.000012 degrees Celsius. Special working condition handling rules include: activating dynamic load compensation mode during typhoon warnings, adjusting stress assessment values ​​by adding a wind speed correction factor; and activating a freeze-thaw effect correction algorithm when ice and snow loads are detected.

[0086] The system evaluates the system setting parameter self-optimization function: when the cumulative monitoring data exceeds 10,000 groups, the least squares regression coefficient is automatically updated; when the characteristic modal change rate exceeds 15% for 10 consecutive times, the orthogonal decomposition recalibration is triggered. In the terminal building application case, the roof is divided into 256 evaluation grids, the characteristic modal matrix contains 8 main modal components, and the spatial resolution of the temperature stress evaluation value output by the regression calculation reaches 0.5 meters. The evaluation results are displayed in real time through a three-dimensional cloud chart, and a gradient color from blue to red is used to identify the stress range from 0 to 250 megapascals, and the warm-toned area identifies the high-risk part. The entire temperature stress evaluation process takes an average of 1.3 seconds on a standard hardware platform, meeting the real-time monitoring needs of large building roofs.

[0087] The system maintenance unit configures the data traceability mechanism: stores the characteristic modal matrix snapshots of the last 30 days, which can be analyzed back when detecting evaluation value abnormalities; establishes a material aging correction model to automatically adjust the elastic modulus parameter according to the service life of the metal roof. The final output temperature stress evaluation value matrix is transmitted to the structure health monitoring platform through the safety warning interface, providing quantitative basis for metal roof maintenance decisions.

[0088] When calling the reference threshold range in the historical working condition database, the historical working condition database system stores the complete temperature stress record data of the metal roof structure in the past five monitoring years. The reference threshold range is defined as the critical interval of the temperature stress that the metal roof can safely bear, which is determined by statistical analysis of historical monitoring data: first, extract the statistical distribution characteristics of all temperature stress monitoring values, calculate the percentile value distribution, and take the 95th percentile value as the upper limit reference threshold initial value and the 5th percentile value as the lower limit reference threshold initial value; then, exclude extreme working condition data, excluding abnormal records exceeding 3 times the standard deviation range; finally, form a reference threshold range spatial distribution matrix, with the matrix rows corresponding to the roof space grid partition numbers and the list indicating the classification results of different seasonal types. The reference threshold range spatial distribution matrix performs data update operations in March every year, and when updating, the newly added complete annual monitoring data is included to recalculate the percentile value distribution. For example, the ridge area of a terminal building project has a summer reference threshold range of 112 megapascals to 168 megapascals, which is obtained by analyzing 12,500 groups of effective data from five summer monitoring periods from 2019 to 2023.

[0089] In constructing a fuzzy logic rule base for solar radiation intensity and ambient wind speed parameters, the fuzzy sets of input and output variables were clearly defined: solar radiation intensity parameters were divided into three fuzzy subsets: weak, medium, and strong radiation intensity levels; and ambient wind speed parameters were divided into three fuzzy subsets: breeze, moderate, and strong wind levels. The fuzzy sets defining the output correction factors were divided into three categories: decrease, maintain, and increase. The rule base construction involved two phases: rule generation and parameter setting. The rule generation phase determined the input-output correlation through historical data regression analysis and established nine core fuzzy rules. The parameter setting phase determined the membership function parameters for each fuzzy set. For example, the radiation parameter value boundaries for the weak radiation intensity level were set between 0 watts per square meter and 400 watts per square meter, while those for the strong radiation intensity level were set between 800 watts per square meter and 1200 watts per square meter. All fuzzy rules were expressed using standard conditional statements. For example, Rule 6 states that when the solar radiation intensity is at the medium radiation intensity level and the ambient wind speed is at the moderate wind level, the correction factor is 1.0, the median value of the maintain level. After the rule base was established, it was verified and tested with 300 sets of historical data, and the fitting error was controlled within 5.3%.

[0090] Matching correction factors in a fuzzy logic rule base based on real-time solar radiation intensity and ambient wind speed parameters involves three consecutive processing stages. The fuzzification stage converts the real-time measured solar radiation intensity (in watts per square meter) and ambient wind speed (in meters per second) into membership values ​​for each fuzzy subset using triangular membership functions. For example, a radiation value of 650 watts per square meter has a membership of 0.82 in the medium radiation intensity category. The rule triggering stage calculates the applicability of all nine rules, which is the algebraic product of the membership degrees of the input variables. The defuzzification stage uses a centroid calculation method to weighted average the outputs of all rules to calculate the final correction factor. The real-time matching process is performed every 10 minutes, automatically shortened to once a minute in extreme weather conditions. For example, when the real-time solar radiation intensity is 880 watts per square meter (a membership of 0.92 in the strong radiation intensity category) and the ambient wind speed is 3.8 meters per second (a membership of 0.67 in the moderate wind category), the correction factor output is calculated to be 1.15.

[0091] When multiplying the baseline threshold range by the correction factor to generate the dynamic warning threshold range, the base value for the current spatial grid cell is first extracted from the baseline threshold range's spatial distribution matrix. A matrix multiplication operation is then performed: the baseline threshold lower limit is multiplied by the correction factor, and the baseline threshold upper limit is multiplied by the correction factor. Physical boundary verification is then performed: if the calculated result exceeds the allowable strength of the metal material, it is automatically truncated to 90% of the allowable strength. The dynamic warning threshold range is output as a spatial matrix, with the matrix elements containing pairs of upper and lower limit values. For example, for a metal roof panel area, the original baseline threshold range was 128 MPa to 184 MPa. When the correction factor is 1.15, the new dynamic warning threshold range is 147 MPa to 211 MPa. The system has established a threshold plausibility monitoring mechanism: when the dynamic threshold lower limit exceeds 85% of the upper limit, a manual review process is automatically initiated. A data source verification alarm is triggered if a spatial cell fails to update after three consecutive calculations.

[0092] During implementation, automatic processing rules for special scenarios were configured: When the rainfall sensor is activated, a rain correction mode is activated, forcing the solar radiation intensity parameter back to zero; when the ambient wind speed sensor fails, the average wind speed over the past three hours is used as a substitute parameter; and when the metal roof temperature exceeds 150°C, a high-temperature material softening correction factor is activated. Metal roof material characteristic parameters are retrieved in real time from the material database. For example, the yield strength of 304 stainless steel is set at 205 MPa and the ultimate strength is set at 515 MPa, providing a basis for threshold boundary verification. Under standard operating conditions, the dynamic warning threshold range is updated every 10 minutes, and the update process completes within 0.7 seconds.

[0093] The system implements adaptive optimization: It automatically updates the baseline threshold range every quarter, incorporating data from newly added monitoring periods into statistical calculations. If the matching error within the fuzzy rule base exceeds 10% for 15 consecutive days, it automatically initiates a rule parameter relearning process. In a field test project involving industrial plants, the dynamic warning threshold range reduced the system's false alarm rate from 28.7% using traditional methods to 7.3%. The resulting dynamic warning threshold range data is transmitted to the early warning decision system via a secure interface, providing a dynamic benchmark for assessing the safety status of metal roofs.

[0094] The monitoring platform establishes a full-cycle traceability function: it stores the last 45 days of dynamic threshold data sequences and supports playback and analysis of historical threshold trajectory. If a spatial partition exceeds the dynamic threshold three times in a row, it is automatically marked as a key monitoring area and the monitoring frequency is increased. The entire dynamic adjustment process takes an average of 0.8 seconds on edge computing devices, meeting the real-time response requirements for monitoring the temperature effects of metal roofs.

[0095] The temperature stress evaluation value is derived from the output data of the temperature effect evaluation system, which is stored in the form of a spatial matrix, and the row and column numbers of the matrix correspond to the three-dimensional grid coordinate system of the metal roof. When mapping the temperature stress evaluation value to the three-dimensional grid coordinates of the metal roof, a spatial coordinate index mapping table is established: the index table contains the three-dimensional coordinate values of the center point of each grid element, and the coordinate values have a precision of 0.01 meters; fast positioning is achieved through a hash mapping algorithm, and the hash key value is generated by combining the grid row and column numbers, and the hash value stores the corresponding coordinate (X, Y, Z) numerical pair. The mapping process ensures that each temperature stress evaluation value is accurately associated with the physical location of the roof, for example, the temperature stress evaluation value 148 MPa is recorded at the coordinate position (12.5 meters, 24.8 meters, 8.7 meters) of the roof partition number grid number 5.

[0096] When identifying the grid coordinate area that exceeds the warning threshold range to generate the risk area identifier, the warning threshold range uses dynamically updated spatial matrix data. The identification process implements three key operations: first, perform point-to-point threshold comparison operations, compare the temperature stress evaluation value of each grid element with the upper and lower limit values of the dynamic warning threshold range at the same position in real time; second, mark the risk grid element, mark it as a risk state when the temperature stress evaluation value is greater than the upper limit value or less than the lower limit value of the dynamic warning threshold range; finally, perform regional clustering analysis, identify the spatially continuous risk grid set using the breadth-first search algorithm, and form a closed polygon risk area. The output of the risk area identifier includes the polygon vertex coordinate sequence and a unique area identifier code, for example, the vertex sequence of area number RISK-2023-109 includes coordinate points (X7, Y8), (X7, Y9), (X8, Y9), etc. 8 position points.

[0097] When constructing a risk state vector based on the risk area identifier, the risk state vector is defined as a multi-dimensional feature array. Vector construction includes four dimension parameters: the first dimension parameter calculates the projected area value of the risk area (in square meters), which is obtained by summing the areas of the triangular grid through the polygon triangulation method, for example, the projected area of a certain area is recorded as 3.8 square meters; the second dimension parameter extracts the maximum stress exceedance value (in MPa), which takes the maximum absolute value of the part exceeding the threshold value in all grid elements in the area; the third dimension parameter calculates the risk intensity index, which is equal to the product of the projected area value and the maximum stress exceedance value; the fourth dimension parameter records the barycentric coordinate value of the area, which is calculated by the average value of the polygon vertex coordinates. The vector construction implements normalization processing: the projected area value is divided by the total area normalization coefficient of the roof, and the maximum stress exceedance value is divided by the material yield strength normalization coefficient.

[0098] When encoding the risk state vector into a structural safety warning signal using an orthogonal coding conversion algorithm, the Hadamard matrix is ​​used as the orthogonal basis. The encoding process consists of five standard steps: first, vector data is quantized, converting each dimension of the risk state vector into an integer sequence by multiplying it by 1000 and rounding it to the nearest integer; second, a Hadamard matrix is ​​generated, with the matrix order equal to the number of vector dimensions; then, an orthogonal transformation is performed, multiplying the risk state vector by the Hadamard matrix to generate an intermediate coding sequence; then, a cyclic redundancy check (CRC) is added, generating a 2-byte checksum using the CRC-16 algorithm; and finally, the code is converted into a hexadecimal string format for output. For example, a risk state vector [38, 185, 7030, 12524875] is encoded to produce the warning signal "8E3A-F5C2-9B01-D4F6." The encoding process incorporates an error control mechanism: if the checksum fails, the three most recent valid signal packets are automatically retransmitted.

[0099] Emergency processing rules are configured during the implementation process: When a particularly large risk area (projected area exceeding 8 square meters) is detected, a fast encoding mode is activated, compressing the encoding process time to under 50 milliseconds. When multiple risk areas exist simultaneously, a priority scheduling strategy is adopted, with the order of priority being ranked based on the risk intensity index. Metal roofing material parameters are accessed in real time from the material database. For example, the yield strength parameter for 304 stainless steel is set at 205 MPa, and the ultimate strength parameter is set at 515 MPa, providing a physical basis for normalized calculations.

[0100] The system implements a signal optimization mechanism: if the same risk area receives three or more consecutive warnings, the warning level is automatically raised; when the risk area is lifted, a warning release signal code is generated. During the terminal project, the warning signal transmission error rate was measured to be less than 1 in 100,000. Warning signals are transmitted to the monitoring center via an industrial wireless sensor network, using a communication protocol that complies with the IEEE802.15.4g standard.

[0101] The monitoring platform has established a full-cycle traceability system: it stores and retains all warning signal data and associated risk area snapshots for the past 48 hours. It automatically initiates a manual intervention request if three consecutive received signal verification failures occur. Implementation data shows an average end-to-end latency of 0.85 seconds from risk identification to signal output, and the accuracy of matching the warning signal with the actual structural safety status reaches 94.6%. The entire processing flow runs stably on the Industrial Internet of Things edge computing platform, with a response time that meets the real-time safety monitoring requirements of metal roofs. The resulting structural safety warning signal triggers the monitoring center's audible and visual alarms and mobile terminal push notification system, forming a complete closed-loop monitoring-warning-response system.

[0102] System maintenance unit configuration signal self-check mechanism: offline signal quality test is performed every morning, and the test data packet covers all coding combinations; when the signal-to-noise ratio of the communication environment is lower than 15 decibels, the anti-interference coding mode is automatically switched. In the actual measurement of the stadium project, the early warning system accurately predicts the buckling risk of the ridge part 32 minutes in advance, avoiding structural safety accidents. All warning signals are recorded and stored synchronously in the cloud disaster recovery system, supporting five-year historical data traceability analysis.

[0103] The metal roof temperature effect monitoring method breaks through the data decoupling limitation of traditional monitoring technology through deep coupling of multi-source sensing and intelligent algorithm:

[0104] A collaborative layout scheme of distributed temperature sensing network and piezoelectric film sensor array is adopted to establish a nonlinear correlation between temperature gradient characteristic values and dynamic constraint reaction force spectrum characteristic values, replacing the conventional direct measurement of strain / displacement. Among them, the dynamic constraint reaction force spectrum characteristic value decomposes the signal intrinsic mode through Hilbert-Huang transform, and introduces Riemannian manifold learning to construct frequency spectrum geometric invariant. This interdisciplinary technology integration breaks through the analysis paradigm of mechanical response signals, significantly improving the sensing accuracy of temperature-structure coupling characteristics.

[0105] The covariance tensor of temperature gradient characteristic values and mechanical characteristic values is constructed, the thermal force coupling main mode is extracted through high-order orthogonal decomposition, and then the temperature stress evaluation mapping model is established through least squares regression. This model converts the complex action relationship between temperature load and structure response into a quantifiable analysis of feature matrix operation, solving the problem of missing thermal force parameter correlation mechanism in traditional methods, and providing a precise mathematical model basis for temperature stress evaluation.

[0106] The early warning threshold range is based on solar radiation intensity and environmental wind speed parameters, and the reference threshold is dynamically corrected through fuzzy rule base. This mechanism adjusts the safety boundary in real time by combining meteorological parameters, realizing a fundamental change from static threshold to environmental self-adaptation. Especially through the orthogonal coding conversion algorithm, the risk state vector is converted into an anti-interference warning signal, ensuring the communication reliability in complex industrial environments.

[0107] The above technical solutions realize three-level integration through cross-scale signal processing (micro-vibration sequence→manifold features), cross-dimension modeling (three-dimensional temperature field→thermal force tensor), and dynamic decision optimization (fuzzy rules→orthogonal coding), forming a complete technical closed loop. The core breakthrough is to upgrade the temperature effect monitoring from the passive mode of "single-point data acquisition-fixed threshold judgment" to the active protection system of "multi-source coupling analysis-environmental adaptive decision", overcoming the inherent defects of existing technologies that cannot quantify the dynamic response of temperature-structure.

[0108] Example 2: Figure 2A structural schematic diagram of a temperature effect monitoring system of a metal roof is given, and the temperature effect monitoring system of the metal roof comprises the following modules:

[0109] A temperature sensing module is configured to acquire three-dimensional temperature field data of the metal roof through a distributed temperature sensing network.

[0110] A gradient extraction module is configured to extract temperature gradient characteristic values of the metal roof from the three-dimensional temperature field data.

[0111] A mechanical characteristic module is configured to collect micro-vibration signal sequences of constraint points of the metal roof through a piezoelectric thin film sensor array, generate instantaneous frequency-energy spectrum by processing the micro-vibration signal sequences through a Hilbert-Huang transform, and extract dynamic constraint counterforce spectrum characteristic values from the instantaneous frequency-energy spectrum through a Riemannian manifold learning method.

[0112] A thermal force coupling module is configured to establish a coupling relationship model between the temperature gradient characteristic values and the dynamic constraint counterforce spectrum characteristic values, and generate temperature stress evaluation values of the metal roof.

[0113] A threshold dynamic module is configured to dynamically adjust a warning threshold range according to solar radiation intensity and environmental wind speed parameters.

[0114] A safety warning module is configured to output a structural safety warning signal of the metal roof when the temperature stress evaluation values exceed the warning threshold range.

[0115] The calculations involved in the embodiments are all de-dimensioned numerical calculations, and the preset parameters and threshold values in the calculations are set by a person skilled in the art according to actual conditions.

[0116] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized in the form of a computer program product, wholly or partially.

[0117] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application of the technical solutions and the constraints of the invention. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0118] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0119] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the above-described device embodiment is only a logical function division, and there can be another division manner for actual implementation, for example, multiple devices or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different parts can be indirect couplings or communication connections through some interfaces, devices or modules, and can be in electrical, mechanical or other forms.

[0120] The above describes only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any modification or replacement within the technical range disclosed by the present application can be easily thought by any person skilled in the art, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0121] Finally: the above described is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for monitoring the temperature effect of a metal roof, characterized in that: The steps include: S1: Obtain three-dimensional temperature field data of the metal roof through a distributed temperature sensing network; S2: Extract the temperature gradient characteristic value of the metal roof from the three-dimensional temperature field data; S3: The micro-vibration signal sequence of the metal roof constraint points is collected through a piezoelectric film sensor array. The micro-vibration signal sequence is processed using the Hilbert-Huang transform to generate an instantaneous frequency-energy spectrum. The dynamic constraint reaction force spectrum eigenvalues ​​are extracted from the instantaneous frequency-energy spectrum using the Riemannian manifold learning method, including: The micro-vibration signal sequence is decomposed into intrinsic mode function components by Hilbert-Huang transform; Perform Hilbert spectrum analysis on the intrinsic mode function components to generate the instantaneous frequency-energy spectrum; Map the instantaneous frequency-energy spectrum to the Riemannian manifold space to construct the tangent vector field; Compute the eigenvalues ​​of the geodesic distance matrix of the tangent vector field; The eigenvector corresponding to the maximum geodesic distance matrix eigenvalue is selected as the eigenvalue of the dynamic constraint reaction force spectrum; S4: Establish a coupling relationship model between the temperature gradient eigenvalue and the dynamic constraint reaction spectrum eigenvalue to generate the temperature stress assessment value of the metal roof, including: Construct the covariance tensor of the temperature gradient eigenvalue and the dynamic constraint reaction force spectrum eigenvalue; Perform orthogonal decomposition on the covariance tensor to obtain the eigenmode matrix; The mapping relationship between the characteristic mode matrix and the temperature stress value is established through the least squares regression algorithm; The real-time eigenmode matrix is ​​input into the mapping relationship to generate the temperature stress evaluation value of the metal roof; S5: Dynamically adjust the warning threshold range according to solar radiation intensity and ambient wind speed parameters; S6: When the temperature stress evaluation value exceeds the warning threshold range, a structural safety warning signal of the metal roof is output.

2. The method for monitoring the temperature effect of a metal roof according to claim 1, characterized in that: The three-dimensional temperature field data of the metal roof is obtained through a distributed temperature sensing network, including: The distributed temperature sensing network includes multiple temperature measurement fiber optic subsystems arranged at angles to each other. The three-dimensional temperature field data is collected by using the distributed fiber optic temperature measurement technology of the temperature measurement fiber optic subsystem to collect discrete temperature point data on the metal roof surface. Spatial correlation analysis is performed on the discrete temperature point data, and the three-dimensional temperature field data is reconstructed using an improved interpolation algorithm.

3. The method for monitoring the temperature effect of a metal roof according to claim 2, characterized in that: The temperature measurement optical fiber layout subsystem covers the entire metal roof area.

4. The method for monitoring the temperature effect of a metal roof according to claim 1, characterized in that: Extract the temperature gradient characteristic values ​​of the metal roof from the 3D temperature field data, including: The temperature difference values ​​at different spatial positions of the metal roof are calculated, the physical boundary constraints of the metal roof are identified based on the three-dimensional temperature field data, the temperature difference values ​​are corrected in combination with the physical boundary constraints, and the characteristic vectors are extracted from the corrected temperature difference values ​​as temperature gradient characteristic values ​​through a principal component analysis algorithm.

5. The method for monitoring the temperature effect of a metal roof according to claim 1, characterized in that: The micro-vibration signal sequence is obtained by collecting the vibration response of the metal roof constraint point through a piezoelectric film sensor array.

6. The method for monitoring the temperature effect of a metal roof according to claim 1, characterized in that: Dynamically adjust the warning threshold range based on solar radiation intensity and ambient wind speed parameters, including: Calling the benchmark threshold range in the historical operating condition database; Construct a fuzzy logic rule base for solar radiation intensity and ambient wind speed parameters; Match the correction factor in the fuzzy logic rule base according to the real-time solar radiation intensity and ambient wind speed parameters; The dynamic warning threshold range is generated by multiplying the baseline threshold range by the correction factor.

7. The method for monitoring the temperature effect of a metal roof according to claim 1, characterized in that: When the temperature stress assessment value exceeds the warning threshold range, the structural safety warning signal of the metal roof is output, including: Mapping the temperature stress assessment values ​​to the three-dimensional grid coordinates of the metal roof; Identify the grid coordinate areas that exceed the warning threshold range and generate risk area identification; Construct risk state vector based on risk area identification; The risk state vector is encoded into a structural safety warning signal through the orthogonal coding conversion algorithm.

8. A temperature effect monitoring system for a metal roof, used to implement a temperature effect monitoring method for a metal roof according to any one of claims 1 to 7, characterized in that: Includes the following modules: Temperature sensing module, used to obtain three-dimensional temperature field data of the metal roof through a distributed temperature sensing network; Gradient extraction module, used to extract the temperature gradient characteristic value of the metal roof from the three-dimensional temperature field data; The mechanical feature module is used to collect the micro-vibration signal sequence of the metal roof constraint points through a piezoelectric film sensor array, process the micro-vibration signal sequence using the Hilbert-Huang transform to generate an instantaneous frequency-energy spectrum, and extract the dynamic constraint reaction force spectrum eigenvalues ​​from the instantaneous frequency-energy spectrum using the Riemannian manifold learning method; Thermomechanical coupling module is used to establish a coupling relationship model between the temperature gradient eigenvalue and the dynamic constraint reaction spectrum eigenvalue to generate the temperature stress assessment value of the metal roof; Threshold dynamic module, used to dynamically adjust the warning threshold range according to solar radiation intensity and ambient wind speed parameters; The safety warning module is used to output a structural safety warning signal of the metal roof when the temperature stress evaluation value exceeds the warning threshold range.

Images