Method for evaluating salt freeze-thaw cycle resistance of concrete

Abstract

The application relates to the technical field of concrete durability, and discloses a method for evaluating the salt freeze cycle resistance performance of concrete. The method acquires monitoring data of the salt solution concentration, temperature cycle and humidity cycle of the environment where a target concrete test piece is located in real time, synchronously collects monitoring data of the surface deformation and internal micro-current response of the test piece, generates environment coupling time sequence features according to the environment monitoring data, performs time-frequency domain fusion processing on the surface deformation and internal micro-current response data based on a dynamic window function, obtains concrete response feature tensors, performs space-time alignment on the environment coupling time sequence features and the concrete response feature tensors, and constructs a three-dimensional degradation feature matrix. The method can comprehensively and accurately evaluate the salt freeze cycle resistance performance of concrete, is in line with an actual environment, can capture degradation information in multiple dimensions, and can in-depth analyze the coupling relationship between the environment and degradation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of concrete durability, in particular to a method for evaluating the salt freeze-thaw cycle resistance performance of concrete. BACKGROUND

[0002] In cold and saline environments, concrete structures are subjected to long-term salt freeze-thaw cycles, which gradually deteriorate their performance, affecting the safety and service life of the structure. Currently, there are many evaluation methods for the salt freeze-thaw cycle resistance performance of concrete, but there are obvious limitations. Traditional evaluation methods mostly use static testing methods, such as setting fixed salt solution concentration, temperature and humidity conditions in the laboratory, and conducting periodic freeze-thaw tests on concrete specimens. The performance deterioration degree is determined by measuring indicators such as mass loss and strength change. However, the environment of concrete in actual engineering is dynamic, and the salt solution concentration fluctuates with factors such as precipitation and evaporation, and the temperature and humidity also show non-periodic alternating changes. Static testing is difficult to truly simulate complex natural environments, resulting in a large deviation between the evaluation results and the actual situation.

[0003] Existing methods for monitoring concrete deterioration mostly focus on surface phenomena, such as observing surface spalling and cracking, or indirectly reflecting internal damage through changes in a single physical quantity, lacking direct monitoring of changes in the internal microstructure of concrete. Under the action of salt freeze-thaw cycles, internal pores of concrete will undergo a series of changes due to salt crystallization expansion, frost heaving pressure, etc. These microscopic changes are the fundamental cause of macroscopic performance deterioration. Relying solely on surface monitoring or single physical quantity measurement cannot fully capture the deterioration process, making it difficult to achieve accurate evaluation of the salt freeze-thaw resistance performance of concrete.

[0004] Existing evaluation methods are relatively simple in data processing, usually analyzing environmental parameters and concrete performance parameters separately, ignoring the dynamic coupling relationship between the two. Changes in environmental factors will real-time affect the deterioration process of concrete, while the deterioration state of concrete will in turn change its interaction with the environment. The lack of this dynamic coupling relationship reduces the accuracy and reliability of the evaluation model, making it difficult to provide effective guidance for the maintenance and repair of concrete structures. SUMMARY

[0005] The purpose of the present application is to provide a method for evaluating the salt freeze-thaw cycle resistance performance of concrete to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides a method for evaluating the salt freeze-thaw cycle resistance performance of concrete, which comprises:

[0007] Real-time acquisition of salt solution concentration monitoring data, temperature cycle monitoring data and humidity cycle monitoring data of the environment where the target concrete test piece is located; synchronous acquisition of surface deformation monitoring data and internal micro-current response monitoring data of the target concrete test piece;

[0008] According to the salt solution concentration monitoring data, temperature cycle monitoring data and humidity cycle monitoring data, an environment coupling time sequence feature is generated; based on a dynamic window function, the surface deformation monitoring data and the internal micro-current response monitoring data are processed in time-frequency domain fusion to obtain a concrete response feature tensor;

[0009] The environment coupling time sequence feature and the concrete response feature tensor are subjected to a spatio-temporal alignment operation to construct a three-dimensional degradation feature matrix; multi-scale gradient analysis is performed on the three-dimensional degradation feature matrix to generate a surface degradation gradient field and an internal pore degradation gradient field of the target concrete test piece;

[0010] Preferably, according to the salt solution concentration monitoring data, temperature cycle monitoring data and humidity cycle monitoring data, an environment coupling time sequence feature is generated, specifically:

[0011] The salt solution concentration monitoring data, temperature cycle monitoring data and humidity cycle monitoring data are respectively subjected to standardization processing to obtain a standardized salt concentration sequence, a standardized temperature sequence and a standardized humidity sequence;

[0012] An adaptive weighted fusion algorithm is used to perform feature superposition on the standardized salt concentration sequence, the standardized temperature sequence and the standardized humidity sequence to generate an initial environment coupling sequence;

[0013] A variational decomposition model is introduced to perform modal separation on the initial environment coupling sequence to extract dominant modal components; an interaction intensity coefficient of environmental parameters is calculated according to the dominant modal components, and the environment coupling time sequence feature is reconstructed based on the interaction intensity coefficient;

[0014] Preferably, based on a dynamic window function, the surface deformation monitoring data and the internal micro-current response monitoring data are processed in time-frequency domain fusion to obtain a concrete response feature tensor, specifically:

[0015] Curvature feature extraction operation is performed on the surface deformation monitoring data to obtain a surface curvature change sequence; impedance feature analysis operation is performed on the internal micro-current response monitoring data to obtain a pore impedance change sequence;

[0016] The initial window length and the sliding step length of the dynamic window function are set, and based on the frequency energy distribution of the surface curvature change sequence and the pore impedance change sequence, the initial window length and the sliding step length are dynamically adjusted;

[0017] Perform joint time-frequency transformation on the surface curvature change sequence and the pore impedance change sequence by using the adjusted dynamic window function to construct a time-frequency joint distribution map; extract amplitude peak value, energy accumulation and frequency band concentration index in the time-frequency joint distribution map to form a concrete response feature tensor;

[0018] Preferably, the three-dimensional deterioration feature matrix is subjected to multi-scale gradient analysis to generate a surface deterioration gradient field and an internal pore deterioration gradient field of the target concrete test piece, specifically:

[0019] The three-dimensional deterioration feature matrix is divided into a plurality of cubic units according to spatial coordinates; a directional derivative operator is used to calculate the deterioration change rate of each cubic unit in the normal direction, the tangential direction and the radial direction;

[0020] A spatial gradient vector field is constructed according to the deterioration change rate; divergence analysis and curl analysis are performed on the spatial gradient vector field to output surface layer divergence distribution and internal layer curl distribution, respectively;

[0021] The surface layer divergence distribution and the internal layer curl distribution are fused to generate the surface deterioration gradient field and the internal pore deterioration gradient field;

[0022] Preferably, the method further comprises:

[0023] According to the surface deterioration gradient field and the internal pore deterioration gradient field, a key evaluation area of the target concrete test piece is divided;

[0024] The deterioration accumulation, the gradient change rate and the spatial correlation degree of each key evaluation area are calculated to generate a regional deterioration comprehensive index;

[0025] Based on the regional deterioration comprehensive index, the key evaluation areas are sorted to determine the regional detection priority;

[0026] Preferably, the method further comprises:

[0027] According to the regional detection priority, a high-resolution detection point set is generated;

[0028] Micro-damage image data of the high-resolution detection point set is obtained based on an ultrasonic flaw detection device; crack morphology segmentation and pore topology analysis are performed on the micro-damage image data to obtain a micro-damage feature set;

[0029] Preferably, the method further comprises:

[0030] The micro-damage feature set is mapped to the surface deterioration gradient field and the internal pore deterioration gradient field to update the deterioration accumulation and the gradient change rate of the key evaluation area;

[0031] According to the updated degradation cumulative amount and the gradient change rate, a regional degradation comprehensive index is reconstructed;

[0032] Based on the reconstructed regional degradation comprehensive index, the region detection priority is corrected;

[0033] Preferably, the method further comprises:

[0034] According to the surface degradation gradient field and the internal pore degradation gradient field, a time-varying stress distribution model is constructed;

[0035] An incremental iteration algorithm is used to calculate the equivalent fatigue load spectrum of the time-varying stress distribution model;

[0036] Based on the equivalent fatigue load spectrum, the stress amplitude variation period of the critical damage path is analyzed;

[0037] Preferably, the method further comprises:

[0038] According to the stress amplitude variation period, the cumulative damage degree of the critical damage path is calculated;

[0039] Combined with the standard damage threshold curve of the concrete material, the remaining service cycle number of the critical damage path is predicted;

[0040] Based on the remaining service cycle numbers of all critical damage paths, the salt-frost cycle resistance performance evaluation result of the target concrete specimen is generated;

[0041] Preferably, the method further comprises:

[0042] The salt-frost cycle resistance performance evaluation result is packaged into a performance evaluation data packet according to a preset communication protocol;

[0043] The performance evaluation data packet is transmitted to a remote monitoring center through a wireless sensor network;

[0044] The performance evaluation data packet is decoded and verified at the remote monitoring center, and a final performance evaluation report is output.

[0045] Compared with the prior art, the beneficial effects of the present application are:

[0046] By real-time acquisition of the salt solution concentration, temperature cycle and humidity cycle monitoring data of the environment where the target concrete specimen is located, dynamic changes in environmental factors can be accurately captured, breaking the limitations of traditional static testing in simulating complex natural environments, making the evaluation process more in line with actual engineering environmental conditions. The surface deformation and internal micro-current response monitoring data of the concrete specimen are collected synchronously, realizing multi-dimensional monitoring of concrete degradation, focusing on both surface macroscopic changes and internal microscopic structure changes, avoiding the information partiality caused by single monitoring method, and fully reflecting the degradation state of concrete under the action of salt-frost cycles.

[0047] The environment coupling time sequence feature is generated according to the environment monitoring data, and the dynamic correlation and change rule between the salt solution concentration, temperature and humidity can be clearly presented, so as to provide a basis for analyzing the comprehensive influence of the environmental factors on the concrete deterioration. The concrete response feature tensor is obtained by performing time-frequency domain fusion processing on the surface deformation and internal micro-current response monitoring data based on a dynamic window function, the monitoring information in different time and frequency ranges can be effectively integrated, the dynamic response rule in the concrete deterioration process is revealed, and the richness and accuracy of the deterioration feature description are improved.

[0048] The environment coupling time sequence feature and the concrete response feature tensor are subjected to spatio-temporal alignment operation, a three-dimensional deterioration feature matrix is constructed, the dynamic coupling relationship between the environmental factors and the concrete deterioration response is fully considered, the two are accurately matched in the time and space dimensions, the internal connection between the environmental change and the concrete deterioration can be in-depth analyzed, and the defects in the separate analysis of the environmental parameters and the performance parameters in the traditional method are overcome. The surface deterioration gradient field and the internal pore deterioration gradient field are generated by performing multi-scale gradient analysis on the three-dimensional deterioration feature matrix, the distribution and development trend of the concrete deterioration can be intuitively shown from different scales, the spatial change gradient of the surface deterioration can be presented, the micro gradient of the internal pore deterioration can be reflected, and a comprehensive perspective for understanding the mechanism of the concrete deterioration is provided. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 A working principle diagram of the concrete salt freeze-thaw cycle resistance performance evaluation method is shown in the figure.

[0050] Figure 2 A flowchart for generating the environment coupling time sequence feature is shown in the figure.

[0051] Figure 3 A flowchart for generating the deterioration gradient field by performing multi-scale gradient analysis is shown in the figure.

[0052] Figure 4 A flowchart for updating the deterioration index and correcting the priority is shown in the figure.

[0053] Figure 5 A flowchart for constructing the time-varying stress distribution model and analyzing the stress amplitude change is shown in the figure. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0055] Please refer to Figure 1 The application provides a method for evaluating the salt freeze-thaw cycle resistance of concrete, the method comprising:

[0056] The salt solution concentration monitoring data, temperature cycle monitoring data, and humidity cycle monitoring data of the environment in which the target concrete test piece is located are obtained in real time by an embedded sensor network. Simultaneously, a laser displacement sensor collects surface deformation monitoring data of the test piece, and an embedded microelectrode array collects internal microcurrent response monitoring data of the test piece. The salt solution concentration monitoring data, temperature cycle monitoring data, and humidity cycle monitoring data are fused to generate environment-coupled time series features. The surface deformation monitoring data and internal microcurrent response monitoring data are fused in the time-frequency domain by a dynamic window function, and a concrete response feature tensor is output. The environment-coupled time series features and the concrete response feature tensor are aligned in time and space to construct a three-dimensional degradation feature matrix. The matrix is analyzed by a multi-scale gradient to generate a surface degradation gradient field and an internal pore degradation gradient field, and the salt freeze-thaw cycle resistance is quantitatively characterized.

[0057] Embodiment 1: Please refer to Figure 2 The salt solution concentration monitoring data, temperature cycle monitoring data, and humidity cycle monitoring data are input into a data processing module. The salt solution concentration monitoring data is collected by a conductivity sensor, the temperature cycle monitoring data is collected by a platinum resistance temperature sensor, and the humidity cycle monitoring data is collected by a capacitive humidity sensor. Each monitoring data is processed by a z-score standardization method to eliminate dimensional differences and generate a standardized salt concentration sequence, a standardized temperature sequence, and a standardized humidity sequence. The calculation process of the standardized salt concentration sequence is as follows: subtract the mean value from the original sequence and divide by the standard deviation; the same processing logic is used for the standardized temperature sequence and the standardized humidity sequence.

[0058] The standardized sequences are processed by an adaptive weighted fusion algorithm. The algorithm first calculates the variance of each sequence, and the reciprocal of the variance is used as the basis for distributing the weight coefficient. The proportion of the reciprocal of the variance of the standardized salt concentration sequence in the sum of the total reciprocal of the variance is used as the salt concentration weight coefficient, and the weight coefficients of the standardized temperature sequence and the standardized humidity sequence are generated according to the same rule. The salt concentration weight coefficient, the temperature weight coefficient, and the humidity weight coefficient are multiplied by the corresponding standardized sequences, and the three sets of product results are superimposed to generate an initial environment-coupled sequence. The initial environment-coupled sequence is a one-dimensional array in the time dimension, and the sampling frequency is consistent with the original monitoring data.

[0059] The variational decomposition model loads the initial environment coupling sequence to perform modal decomposition. The model presets the number of intrinsic modes as 5, and the bandwidth parameter is constrained by a penalty factor. After iterative convergence, 5 sets of modal components are output. The energy proportion of each modal component is calculated, and the modal component with an energy proportion exceeding 70% is selected as the dominant modal component. The calculation process of the interaction intensity coefficient is as follows: the amplitude extreme point sequence of the salt concentration, temperature, and humidity in the dominant modal component is extracted, the mutual information entropy value of the three sequences is calculated, and a 3*3 interaction intensity coefficient matrix is generated. The matrix and the dominant modal component perform tensor product operation to reconstruct the environment coupling time sequence feature. The environment coupling time sequence feature includes timestamp, coupling strength value, and interaction mode encoding three-dimensional data.

[0060] The surface deformation monitoring data is collected by a laser displacement sensor at an interval of 0.1 seconds. The original point cloud data is fitted into a continuous surface by cubic spline interpolation, and the Gaussian curvature change is calculated at each node of the surface to output the surface curvature change sequence. The sequence value represents the curvature change rate per unit area, and a negative value represents a concave deformation, and a positive value represents a convex deformation. The internal micro-current response monitoring data is collected by a pre-embedded four-probe electrode array, and the current excitation frequency is 1 kHz. Based on Ohm's law, the current-voltage monitoring value is converted into impedance modulus to generate a pore impedance change sequence. The sampling interval of the sequence is synchronized and aligned with the surface curvature change sequence.

[0061] The initial parameters of the dynamic window function are set as a window length of 10 seconds and a sliding step of 2 seconds. The frequency energy distribution analysis module loads the surface curvature change sequence to perform fast Fourier transform to detect the energy peak position in the 0.5-5 Hz frequency band; at the same time, the pore impedance change sequence is loaded to perform wavelet packet decomposition to calculate the energy entropy value of the 2-4 scale sub-band. If the energy peak value exceeds the threshold value and the energy entropy is lower than the set value, the window length is shortened to 5 seconds; if the energy peak value is dispersed and the energy entropy is increased, the window length is extended to 15 seconds. The sliding step is dynamically adjusted to 1-3 seconds according to the frequency band energy concentration.

[0062] The adjusted window function processes the surface curvature change sequence and the pore impedance change sequence. The window function slides along the time axis, and at each step, the short-time Fourier transform is performed on the data within the window. The transform result of the surface curvature change sequence generates a time-frequency spectrum A, and the pore impedance change sequence generates a time-frequency spectrum B. The two spectra are aligned and superimposed according to the time-frequency coordinates to construct a time-frequency joint distribution graph. The horizontal axis of the distribution graph is the time axis (unit: seconds), and the vertical axis is the frequency axis (unit: Hz), and the color intensity represents the amplitude intensity.

[0063] The amplitude peak value index extraction logic is to scan the amplitude matrix of the distribution map and record the local maximum value point coordinate set exceeding 20 dB. The energy accumulation amount index calculation logic is to integrate the amplitude intensity value of each frequency point along the time axis within the frequency band range of 0.1-10 Hz. The frequency band concentration index calculation logic is to extract the main energy frequency band (the continuous frequency band with an amplitude sum ratio of more than 80%) and calculate the proportion of the half-power bandwidth of the main energy frequency band in the full frequency band. The three indexes are converted into 64-bit floating point arrays, and combined into a 6×N-dimensional concrete response feature tensor after alignment according to the time stamp, where N is the total number of time points.

[0064] The sensor acquisition module and the processing module are connected through a real-time data bus. The analog signals of the conductivity sensor, the temperature sensor, and the humidity sensor are input into the standardization processing unit after being converted by the 24-bit ADC. The digital signals of the laser displacement sensor are transmitted to the curvature calculation unit through the gigabit Ethernet. The monitoring data of the microelectrode array are input into the impedance conversion unit after being processed by the isolation amplifier. The adjustment instruction of the dynamic window function is transmitted by the frequency domain analysis unit through shared memory. The construction of the time-frequency joint distribution map is completed in the GPU parallel computing unit, and the amplitude peak value detection is realized by using the CUDA parallel reduction algorithm.

[0065] The environmental coupling time sequence feature and the concrete response feature tensor are temporarily stored in a distributed time sequence database. The database stores data blocks according to time slicing, and each data block contains a millisecond-level timestamp, a data check code, and a compressed feature vector. The feature vector is double-compressed by using Delta encoding and Huffman encoding, and the compressed data block is transmitted to the next processing node through the message queue.

[0066] Example 2: refer to Figure 3 The analytical operation of the three-dimensional deterioration feature matrix is performed in the space computing engine. The coordinate dimension of the matrix is X×Y×Z, wherein the X-axis represents the length direction of the test piece (0-100 mm), the Y-axis represents the width direction (0-100 mm), and the Z-axis represents the depth direction (0-50 mm). The space division module decomposes the matrix into 8000 cubic units at a step length of 5 mm, and each unit contains 8 vertex data values. The center point coordinates of the unit are generated by arithmetic average calculation of the vertex coordinates, and the spatial resolution error is controlled within ±0.1 mm.

[0067] The multi-directional derivative calculation is performed at the center point of the cubic unit. The normal directional derivative is processed by a Sobel operator: the adjacent three layers of data are selected along the Z-axis direction, and a three-dimensional convolution operation is performed with a convolution kernel of [[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]]. The tangential directional derivative is processed by a Prewitt operator: a two-dimensional convolution is performed in the XY plane with a convolution kernel of [[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]]. The radial directional derivative is processed by a central difference method: 12 adjacent points are selected at equal angles along the radial direction based on the current center point, and the weighted difference of the data values thereof is calculated. The derivative calculation results of the three directions are respectively stored as normal change rate, tangential change rate, and radial change rate floating point arrays.

[0068] The construction of the spatial gradient vector field is based on the change rate arrays. The normal change rate array is bound with the Z coordinate to generate a surface layer data set, and the tangential and radial change rate arrays are bound with the XYZ coordinates to generate an internal layer data set. The surface layer data set is input into a divergence calculation module: a 0.1 mm grid is established in the XY plane, and the divergence value of each grid point is calculated by the Gauss divergence theorem: a 5 mm radius circular domain is centered at the point, and the area integral of the boundary normal change rate flux of the circular domain is calculated. The internal layer data set is input into a curl calculation module: a 0.2 mm grid is established in three-dimensional space, and the tangential and radial change rate components are processed by the curl operator ▽× at each grid point.

[0069] The divergence distribution map generation process is: the divergence values of all grid points are smoothed by bilinear interpolation and mapped to a 256-level gray scale. The curl distribution map generation process is: the modulus of the curl vector of each grid point is extracted, and after median filtering, it is mapped to an RGB color spectrum. The areas in the divergence distribution map that exceed the threshold value ±0.05 are marked as red highlight areas, and the areas in the curl distribution map with a modulus greater than 0.1 are marked as blue highlight areas. The surface degradation gradient field is output as a three-dimensional point cloud containing divergence values and color encoding, and the internal pore degradation gradient field is output as a four-dimensional tensor containing curl vectors and moduli.

[0070] The key evaluation region division is based on the positions of the divergence extreme points and the curl extreme points. The first 50 maximum positive divergence points and the first 50 maximum negative divergence points are selected in the divergence distribution map, and the first 100 maximum curl modulus points are selected in the curl distribution map. A spherical space region with a radius of 10 mm is generated around each extreme point, and the overlapping parts of the regions are merged, finally generating no more than 200 key evaluation regions. Each region is assigned a unique region identifier, and its spatial boundary coordinates and the included cubic unit numbers are recorded.

[0071] The degradation accumulation calculation is unfolded in the time dimension. For each key evaluation region, the rate of change data of all the cubic cells contained in it in 200 cycles of salt spray is extracted. The surface area cumulative divergence value calculation logic: the area of the normal rate of each unit in the region at the end of each cycle is integrated. The internal region cumulative rotation value calculation logic: the volume of the rotation modulus of each unit in the region at the end of each cycle is integrated. The integral result is converted into a scalar value as the degradation accumulation respectively.

[0072] The gradient change rate analysis adopts time series derivative. For each key evaluation region, record the change value of its degradation accumulation in the next 10 cycles, and calculate the average change in the unit time interval. The time interval is standardized to 1 cycle of salt spray, and the output value is the change rate floating point number. The spatial correlation degree calculation adopts Pearson correlation coefficient: select the degradation accumulation sequence of the target region and its adjacent 8 regions (the distance does not exceed 15 mm), and calculate the ratio of the covariance and the standard deviation.

[0073] The synthesis of the region degradation comprehensive index is based on three parameters: the degradation accumulation is normalized to the [0, 1] interval after logarithmic conversion; the gradient change rate is normalized after hyperbolic tangent function compression; the spatial correlation degree is directly normalized by taking the absolute value. The three normalized values are linearly superimposed according to the 0.5:0.3:0.2 weight, to generate the region degradation comprehensive index. The index values of all key evaluation regions are stored as a floating point array, and the array index is bound to the region identifier.

[0074] The region detection priority ranking is realized through the heap sorting algorithm. A maximum heap data structure is constructed, and the region degradation comprehensive index is used as the sorting key value. The heap is initialized to insert all key evaluation region data, and the descending order priority list is output after the sinking operation is executed. Each entry in the list contains three items of data: region identifier, spatial center coordinates, and degradation comprehensive index value. The list data is packaged in JSON format and stored in non-volatile memory.

[0075] The hardware platform of the spatial computing engine is equipped with an FPGA accelerator. Three-dimensional convolution operation is realized through programmable logic gate array parallel computing: XYZ three-way convolution of Sobel operator is executed synchronously, and the single calculation period is reduced to 5μs. The Gaussian divergence calculation adopts multi-core DSP cooperative processing: each calculation core is responsible for a 10×10 grid block, and the boundary data is exchanged through the ring bus. The Pearson correlation coefficient calculation is completed through the coprocessor: the coprocessor has a built-in 128-bit floating point operation unit, and the single correlation coefficient calculation takes 0.8ms. The heap sorting algorithm is executed on the GPU: 1024 CUDA cores are used to compare node data in parallel, and the sorting of million-level data can be completed within 50ms.

[0076] The data storage adopts a hierarchical structure. The original three-dimensional deterioration feature matrix is stored in the DDR4 memory pool, and the spatial gradient vector field is transferred to the NVMe solid state disk. The spatial boundary information of the key evaluation area is encoded as an octree structure, and the deterioration cumulative amount time series is compressed and stored using the Delta-of-Delta algorithm. The final output surface deterioration gradient field data volume is 2.5 MB / cycle, the internal pore deterioration gradient field data volume is 3.7 MB / cycle, and the area detection priority list data volume is 150 KB.

[0077] Example 3: refer to Figure 4 The area detection priority list is loaded into the spatial positioning system. The system analyzes the top 10% high-priority areas in the list, and each area corresponds to a spherical range in three-dimensional space. A local coordinate system is established in the XY plane based on the spherical center coordinates. A grid point array is generated in the coordinate system with a 0.5 mm interval, and the array range covers the boundary extension of 2 cm of the spherical projection. The grid point height value is automatically compensated according to the specimen surface curvature, and finally a high-resolution detection point set containing 800-1200 detection points is formed. Each point records the absolute coordinates (X, Y, Z), the area attribution identifier, and the priority weight coefficient.

[0078] The ultrasonic flaw detection equipment selects a phased array system of model UT-9000, and the probe center frequency is 50 MHz. The equipment motion control module receives the detection point set and automatically plans the optimal scanning path: grouping by area, and scanning the points in the same area in a zigzag path. The probe contact pressure is constantly maintained at 0.5 N, and the coupling agent uses standard glycerol solution. During scanning, an ultrasonic pulse with a pulse width of 100 ns is emitted at each point, and the received echo signal sampling rate is 500 MHz. After time domain gain compensation, the original echo data outputs a micro-damage image of 128x128 pixels, and the resolution of a single image reaches 10 μm.

[0079] The micro-damage image processing flow is divided into two parallel channels. The crack morphology segmentation channel adopts the U-Net network architecture: the input image is down-sampled by 5 layers of convolution to extract features, and the resolution is restored by 4 layers of deconvolution. The output layer uses a sigmoid activation function to generate a binary mask, and the connected component analysis algorithm extracts the connected domain in the mask. Each connected domain calculates the crack length (main axis extension distance), crack width (orthogonal direction average distance), and fractal dimension (box counting method to calculate boundary complexity). The pore topology analysis channel adopts morphological processing: after median filtering the input image, Otsu threshold segmentation separates the pore region. The skeletonization algorithm extracts the pore network, calculates the pore equivalent diameter (area equivalent circle diameter), connectivity (number of adjacent pore connections), and tortuosity (skeleton path length to straight line distance ratio). All parameters are bound and stored according to the detection point coordinates to form a micro-damage feature set.

[0080] The feature mapping update operation is performed in a three-dimensional spatial database. The crack length data is associated with the surface deterioration gradient field: the current divergence value is read at the corresponding spatial coordinates, and a divergence-crack length linear regression model is established. The model slope parameter is dynamically adjusted according to the new data points, and the updated slope value is used to recalculate the divergence coefficient matrix. The pore tortuosity data is associated with the internal pore deterioration gradient field: the corresponding spatial coordinates are read at the corresponding spatial coordinates, and a spinor-tortuosity gradient descent optimization model is constructed. The model takes the mean square error as the loss function, and the learning rate is set to 0.01. After 50 iterations, the updated spinor coefficient matrix is output.

[0081] The deterioration accumulation amount recalculation process: in each key evaluation area, based on the updated divergence coefficient matrix, the normal change rate of all grid points in the area is calculated. The integral is calculated by the Gauss integral method, and 20 integral points are selected to calculate the weighted sum. The gradient change rate recalculation process: extract the deterioration accumulation amount sequence of the last 20 times of salt freezing cycle, and calculate the moving average (window width 5 cycles) after first-order difference. The spatial correlation degree recalculation process: in the spherical coordinate space, all regions within a distance of 15 mm from the center of the target region are retrieved, and the eigenvalues of the covariance matrix of the updated deterioration accumulation amount sequence are calculated.

[0082] The area deterioration comprehensive index reconstruction process: the updated deterioration accumulation amount is normalized by double logarithmic transformation; the gradient change rate is compressed to the [0, 1] interval by Sigmoid function; the spatial correlation degree is normalized by taking the maximum eigenvalue of the covariance matrix. The three parameters are weighted and summed according to the original weight coefficient 0.5:0.3:0.2 to generate a new area deterioration comprehensive index. The new index value replaces the corresponding value in the original priority list to form an unsorted intermediate data set.

[0083] The area detection priority correction is realized by sorting algorithm. Bubble sort optimization strategy is adopted: first traversal compares all adjacent region index values, and exchanges the reverse order pair; the second traversal only compares the regions with a historical priority difference of more than 30%. The spatial topological relationship of the region is maintained during the sorting process: when the index difference of adjacent regions is less than 0.05, the original relative order is maintained. The final output of the new region detection priority list records the region change log with a new and old priority change of more than 10%.

[0084] The hardware system consists of a motion control platform, an ultrasonic acquisition unit, and a data processing server. The six-axis robot carries the ultrasonic probe to perform scanning, with a repeat positioning accuracy of ±2 μm. Real-time monitoring of the motion trajectory is performed using a laser interferometer, and path re-planning is triggered when the position deviation exceeds 5 μm. The ultrasonic acquisition unit is equipped with a 1 GB cache area, and image data is transmitted through a PCIe3.0 interface. The data processing server is configured with dual Xeon processors, and the crack segmentation task is assigned to four Tesla T4 GPUs for execution, while the topology analysis task is processed by multiple threads on the CPU. The feature mapping update module uses a memory computing architecture, with a dedicated buffer area in 128 GB DDR4 memory to store gradient field update data.

[0085] The time synchronization mechanism ensures data consistency. The ultrasonic scanning trigger signal, the robot position feedback signal, and the salt freeze cycle state signal are synchronized at the microsecond level through the PTP protocol. Each detection point's microscopic damage image is attached with a timestamp (in the format of Unix timestamp + millisecond offset), which is time-correlated with the degradation gradient field data of the corresponding cycle number. Feature update operations are performed during the intermittent period of salt freeze cycles, and the system automatically detects the stable stage of environmental parameters to start the calculation task.

[0086] The data storage uses a hierarchical index structure. The original microscopic damage images are stored in directories according to region numbers, with a single region data volume of about 1.2 TB. The microscopic damage feature set is indexed in four dimensions: spatial coordinates (X, Y, Z) + timestamp, and the feature values are stored in the form of a floating-point array. The update log records the modified coordinate range, old coefficient value, and new coefficient value of each feature mapping, and the log file is stored using binary differential compression. The final output of the new priority list adds a version identifier, which is incrementally related to the historical version.

[0087] Example 4: refer to Figure 5 The construction of the time-varying stress distribution model is based on the spatial fusion of the surface degradation gradient field and the internal pore degradation gradient field. The divergence distribution data of the surface degradation gradient field (dimension X × Y × 1) and the curl modulus data of the internal pore degradation gradient field (dimension X × Y × Z) perform tensor expansion operations: the surface data is copied along the Z axis to 50 layers to form a pseudo-three-dimensional tensor; the internal data remains in the original dimension. After multiplying the two data sets by the spatial coupling coefficient matrix, they are spliced along the depth direction into a four-dimensional stress feature body (dimension X × Y × Z × 2), which contains divergence values and curl modulus in the fourth dimension. This feature body is converted into a continuous field through cubic B-spline interpolation, and is defined as the time-varying stress distribution model.

[0088] The incremental iterative algorithm processes the time-varying stress distribution model in stages. In the initialization stage, the concrete material parameters are loaded: the initial value of the elastic modulus is 30 GPa, the Poisson's ratio is 0.2, and the damage factor is 0.01. In the first iteration, the equivalent stress distribution is calculated: at the spatial grid point (i, j, k), the divergence value σs and the curl modulus σ r , the principal stress is calculated by using the plane strain assumption:

[0089]

[0090] wherein: represents the equivalent stress value of the first iteration, in MPa (mega pascal); σ s represents the surface divergence value, dimensionless, representing the concrete surface deformation gradient; σ r represents the internal curl modulus, dimensionless, representing the directionality of the concrete internal pore structure change; K represents the morphology-stress conversion coefficient, in MPa (mega pascal), used to convert the dimensionless σ s and σ r to the dimensionally equivalent stress value, and the value of K is calibrated through the same condition standard salt freeze-thaw cycle test.

[0091] The calibration steps of the morphology-stress conversion coefficient K are as follows: (1) specimen preparation: select 3 groups of standard cubic specimens (size 100mmx100mmx100mm) of the same strength grade, the same aggregate type, and the same mix proportion as the target concrete specimen, with 3 parallel specimens in each group; (2) salt freeze-thaw cycle test: 3 groups of specimens are subjected to 20, 50, and 100 times of salt freeze-thaw cycle treatment, respectively, with the test conditions being: salt solution concentration 5% (mass fraction), freeze-thaw temperature range-20℃-20℃, and single cycle freeze-thaw time 4h each; (3) data acquisition: after each cycle, the surface divergence value σ s and the internal curl modulus σ r of the specimen are collected according to the method described above, while the equivalent stress σ eq,实测 (unit: MPa) inside the specimen is directly measured by using a pre-embedded stress sensor; (4) fitting calculation: for each group of specimens, take σ as the independent variable x (dimensionless) and σ eq,实测 as the dependent variable y (unit: MPa), and obtain y=Kx by linear regression fitting, and the slope is the K value of the type of concrete; (5) value selection principle: the K value of the target concrete specimen is matched and selected from the above calibrated K values according to its actual strength grade and aggregate type, and if there are multiple sets of calibration results, the arithmetic mean is taken as the final K value.

[0092] σ s and σ r are physically related to the equivalent stress, in which, in the salt freeze-thaw cycle of concrete, the internal pore water freezes and expands to generate frost heaving pressure, and the external salt solution penetrates to aggravate the damage to the pore structure, both of which cause surface deformation and internal stress concentration: the surface divergence value σ sThe larger the absolute value, the more significant the gradient of local depressions or bulges on the surface, corresponding to a more concentrated frost heave pressure in the internal pores, and a greater stress on the specimen; the internal curl modulus σ r The larger the value, the stronger the squeezing effect caused by the directional expansion of pores due to salt crystallization, and the higher the degree of local stress concentration. The physical meaning of the morphology-stress conversion coefficient K is "the stress response value corresponding to a unit change in morphology". Its calibration process is based on experimental data under the same conditions to ensure that the dimensionless morphology / structure index and the dimensionless mechanical stress index are quantitatively correlated. The calculation results are consistent with the actual mechanical laws of concrete salt-freezing deterioration.

[0093] Example of K-value calibration: The K-values ​​of different types of concrete calibrated by the above steps are shown in the table below. The calibration results show a fluctuation range of ≤0.1MPa, and the stability meets the requirements of engineering calculations.

[0094] Table 1: Calibration Results of K-value in Salt Freezing Cycle Test for Concrete of Different Strength Grades and Aggregate Types

[0095]

[0096] The second round of iterative updates to material parameters: when When the stress exceeds 20 MPa, the elastic modulus is reduced according to the exponential decay model; when the local stress gradient is greater than 1 MPa / mm, the damage factor increases linearly by 0.005. The stress redistribution calculation is repeated in the third to fifth iterations: based on the updated material parameters, the equilibrium equation is resolved, with the iteration step size fixed at 0.1 salt-freezing cycles. The convergence criterion is that the maximum relative error between adjacent iterations is less than 0.1%.

[0097] The process of generating the equivalent fatigue load spectrum: extracting the final equivalent stress distribution after convergence. The extreme stress values ​​of the cross section are taken every 5 mm along the depth direction. The load spectrum entries are defined as three-dimensional tuples (position coordinates, stress amplitude). Mean stress σ m Position coordinates are a three-dimensional vector (X, Y, Z); stress amplitude Within the cross section The difference between the maximum and minimum values; mean stress σ m Within the cross section The arithmetic mean of the values. All cross-sectional entries are sorted by spatial location to form a load spectrum sequence, with one load spectrum instance corresponding to each salt-freezing cycle.

[0098] The criterion for identifying critical damage paths is: within 10 consecutive salt-freezing cycles, the stress amplitude fluctuation rate at the same spatial location exceeds 30% of its mean. The fluctuation rate is calculated using the following formula:

[0099]

[0100] wherein: δ represents fluctuation ratio (dimensionless), represents stress amplitude of the t-th cycle (unit: MPa), t is the cycle number of salt spray test, is the maximum value of stress amplitude in the last 10 cycles, is the minimum value of stress amplitude in the last 10 cycles. The path satisfying δ>0.3 is marked as critical damage path, and its starting position coordinates, fluctuation cycle number T (i.e. the number of cycles satisfying the condition continuously), amplitude range [A min ,A max ] (unit: MPa) are recorded.

[0101] The cumulative damage degree is calculated independently for each critical damage path. A piecewise linear cumulative model is adopted: the stress amplitude in the load spectrum is divided into several segments, and the cumulative damage degree is calculated for each segment. The cumulative damage degree is calculated independently for each critical damage path. A piecewise linear cumulative model is adopted: the stress amplitude in the load spectrum is divided into several segments, and the cumulative damage degree is calculated for each segment. p The actual cycle number n p is counted from the starting time of the path marking. The cumulative damage degree D p is defined as:

[0102]

[0103] wherein: M represents the number of load spectrum segments contained in the path, is the actual cycle number experienced by the m-th segment, is the allowable cycle number of the m-th segment. The allowable cycle number is obtained by standard curve interpolation: when , take N p =10 6 ; when , take N p =10 3 ; the intermediate value is interpolated linearly.

[0104] The remaining service cycle number prediction is based on the damage evolution trend. The cumulative damage degree sequence of the last 5 salt spray cycles is extracted, and the damage degree increment per cycle is calculated:

[0105]

[0106] wherein: represents the current cumulative damage degree, represents the cumulative damage degree before 5 cycles (all dimensionless).

[0107] The remaining service cycle number of path p is calculated by the following model:

[0108]

[0109] wherein: the remaining cycle number prediction value (unit: times), the accumulated damage degree of path p at the current time, the damage degree increment of path p in the last unit time, The numerator represents the damage degree change rate. The minimum value of all critical paths is taken as the final predicted life value of the target concrete specimen. The minimum value of all critical paths is taken as the final predicted life value of the target concrete specimen.

[0110] The system hardware consists of a multi-core processor, a distributed storage array, and a real-time clock module. The time-varying stress model calculation is distributed to 4 physical cores: Core 1 processes the surface field data expansion, Core 2 processes the internal field data extraction, Core 3 performs tensor multiplication, and Core 4 is responsible for interpolation conversion. Incremental iteration tasks are parallel on 8 logical cores: each core is responsible for the calculation of spatial regions, and the iteration intermediate results are exchanged through shared memory. The cumulative damage degree update adopts a pipeline architecture: the first pipeline retrieves the standard threshold curve, the second pipeline calculates the ratio, and the third pipeline performs cumulative summation.

[0111] The data storage adopts a three-layer cache structure: the original gradient field data is stored in the DDR4 memory pool (capacity 128GB); the equivalent fatigue load spectrum is stored in the NVMe solid state disk (read-write bandwidth 3GB / s); and the critical path information is stored in the non-volatile memory (access delay 200ns). The timing control is synchronized through the PTP protocol: the stress calculation period is bound to the freeze-thaw cycle event, and the calculation task is started in the temperature stabilization window after the cycle ends. The life prediction result is output as a structured array, containing path coordinates, remaining life value, damage degree change rate, etc.

[0112] Example 5: Anti-freeze-thaw cycle performance evaluation result data packet encapsulation in edge computing node. The node receives the structured array output by the life prediction module, which contains three data areas: the path coordinate area records the three-dimensional position information of the critical damage path in the format of floating-point triplets; the remaining life value area stores the integer value of the remaining service cycle number of each path; and the damage trend area contains the damage degree change rate and the last detection timestamp. Before encapsulation, data verification is performed: the path coordinates are limited within the boundary box of the concrete specimen, the remaining life value is greater than zero, and the timestamp conforms to the ISO8601 specification.

[0113] The MQTT protocol encapsulation process configures a fixed topic path: the first level topic is a project number, and the second level topic is a test piece identification code. The packet body structure strictly follows a predefined format: the first 4 bytes are fixed as a packet header identifier with a hexadecimal value of A5A5; 16 bytes of device ID are generated by splicing a device MAC address and a device type code; 32 bytes of timestamp include date, time zone, and millisecond information; 128 bytes of evaluation result main body are encoded in TLV format, in which the path coordinates occupy 60 bytes, the remaining life value occupies 40 bytes, and the auxiliary parameters occupy 28 bytes; the last 4 bytes of CRC check code are calculated using the IEEE standard. The complete packet size is constant at 184 bytes, and pre-allocated memory buffer is used to avoid dynamic memory allocation.

[0114] The LoRa wireless transmission module uses the E32-433T30D model and operates in the 433 MHz frequency band. Channel detection is performed before transmission: 8 preset channels are scanned, and a channel with a background noise lower than -120 dBm is selected to establish a connection. The data packet segmentation transmission strategy: each data packet is treated as an independent transmission unit, and the transmission rate is adaptively adjusted. The initial rate is set to 5 kbps, and it is reduced to 1.2 kbps after 3 consecutive transmission failures. Each data packet is attached with a 2-byte sequence number, and the receiving end reassembles the data stream according to the sequence number. The transmission interval is fixed at 2 seconds, and the module has a built-in retransmission counter. If an ACK signal is not received within the timeout, retransmission is triggered, with a maximum of 3 retries.

[0115] The remote monitoring center deploys a message broker server. The server is configured with a dual-machine hot standby architecture, with the main node processing real-time data streams and the standby node synchronizing memory states. The message broker implements MQTT protocol version 3.1.1 and sets up three subscription queues: the high-priority queue handles data packets with header identifier matching A5A5, the medium-priority queue handles device status messages, and the low-priority queue handles log information. CRC check is performed immediately after the data packet arrives: the last 4 bytes are extracted and compared with the calculated value. If the check fails, the data packet is discarded and a NAK signal is sent through the management channel.

[0116] The effective data packet enters the analysis pipeline. The device ID field separates the MAC address and device type code, which are routed to the corresponding project database. The timestamp field is converted into a UTC+8 time zone standard time object. The evaluation result main body is parsed using a state machine mode: the type tag of the TLV structure is identified, the path coordinate type tag is parsed into a three-dimensional float array, the remaining life tag is parsed into an integer array, and the auxiliary parameter tag is parsed into a key-value pair dictionary. During the parsing process, array length out-of-bounds, value overflow, and other exceptions are automatically detected, and abnormal data is transferred to the isolated storage area.

[0117] The report generation engine loads the parsed data. The three-dimensional degradation cloud rendering module reads the path coordinate array to generate Voronoi graph elements with coordinate points as seeds. The color mapping rule for each element is: display green if the remaining life is greater than 1000 cycles, display yellow if it is between 500 and 1000 cycles, and display red if it is less than 500 cycles. The life prediction curve module extracts time series data, with the horizontal axis representing the detection time point and the vertical axis representing the remaining life value. Adjacent data points are connected using a cubic Bezier curve. Chart elements are described in SVG vector format with a resolution of 300 dpi.

[0118] The PDF document synthesis stage calls the template engine. The main template includes fixed areas such as enterprise identification, detection date, test piece information, etc. The dynamic insertion area reserves placeholders: the three-dimensional cloud placeholder is replaced by the rendered PNG raster image, and the life curve placeholder is replaced by the vector graphics. Document metadata settings include author information, creation tool version number, and security encryption fingerprint. The final report document is digitally signed: use the SHA-256 algorithm to generate a digest, and use the RSA private key to encrypt a 256-byte signature block and append it to the end of the file.

[0119] The disaster recovery mechanism implements dual-path protection. The network transmission layer establishes a redundant channel: automatically switches to the 4G backup link after the main channel is interrupted, and the transmission rate is switched to 1 Mbps. The data center sets up a local cache: the edge node retains a 24-hour data copy locally after sending a data packet. When the monitoring center sends a data retransmission request, the original data packet is extracted from the cache area and retransmitted. Report storage uses off-site dual backup: the main data center stores the original report, and the backup center stores the encrypted compressed archive file.

[0120] The device linkage example occurs at the end of the detection cycle. When the minimum remaining life value of a certain test piece critical path is less than 50 cycles, the system automatically triggers the alarm protocol: sends instruction code 0x0F to the field device through the MQTT management topic. After receiving the instruction, the field device controls the mechanical arm to spray a red marker ring on the surface of the test piece, with the center of the marker ring being the surface projection point of the lowest life path. At the same time, it returns a device response code to the monitoring center, including operation time, position coordinates, execution status, etc.

[0121] Data center physical configuration example: the main server uses Dell PowerEdge R750, equipped with dual Xeon Gold 6338 processors. Three-dimensional rendering tasks are assigned to 4 NVIDIA A100 graphics cards, each with 80GB of video memory. The report storage array uses an all-flash architecture, with 24 3.84TB SSDs forming a RAID60 array. The network interface is configured with dual 10GbE fiber channels, and multi-path load balancing is achieved through BGP protocol. The air conditioning system is set up with temperature and humidity linkage: when the server cabinet temperature exceeds 35℃, the environmental humidity setting value is automatically reduced.

[0122] Transmission process example record: 2025-08-19 09:23:17, device D-CT-2107 transmits the 382nd cycle evaluation package. The data package sequence number is 0x1A8F, and the transmission time through channel 3 is 2.8 seconds. The monitoring center completes the analysis at 09:23:21, and detects that the remaining life value of path No. 3 has dropped to 48 times. The alarm instruction is issued at 09:23:25, and the field device sprays a 20mm diameter marker ring at the coordinates at 09:23:31, while returning a response code indicating successful operation. The complete report file is generated and stored at 09:24:03, and the file size is recorded as 8.7MB.

[0123] It should be noted that the relational terms herein, such as first and second, are used solely to distinguish one from another entity or action without necessarily requiring or implying any actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0124] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives, and variations can be made in the embodiments without departing from the spirit and scope of the present application as defined by the appended claims and their equivalents.

Claims

1. A method for evaluating salt-frost cycle resistance of concrete, characterized by, The method comprises the following steps: Real-time acquisition of salt solution concentration monitoring data, temperature cycle monitoring data and humidity cycle monitoring data of the environment in which the target concrete test piece is located; synchronous acquisition of surface deformation monitoring data and internal micro-current response monitoring data of the target concrete test piece; According to the salt solution concentration monitoring data, temperature cycle monitoring data and humidity cycle monitoring data, an environment coupling time sequence feature is generated; Based on a dynamic window function, the surface deformation monitoring data and the internal micro-current response monitoring data are processed in time-frequency domain fusion to obtain a concrete response feature tensor; The environment coupling time sequence feature and the concrete response feature tensor are subjected to a spatio-temporal alignment operation to construct a three-dimensional degradation feature matrix; Multi-scale gradient analysis is performed on the three-dimensional degradation feature matrix to generate a surface degradation gradient field and an internal pore degradation gradient field of the target concrete test piece; The environment coupling time sequence feature is generated according to the salt solution concentration monitoring data, the temperature cycle monitoring data and the humidity cycle monitoring data, specifically as follows: The salt solution concentration monitoring data, the temperature cycle monitoring data and the humidity cycle monitoring data are respectively subjected to standardization processing to obtain a standardized salt concentration sequence, a standardized temperature sequence and a standardized humidity sequence; An adaptive weighted fusion algorithm is used to perform feature superposition on the standardized salt concentration sequence, the standardized temperature sequence and the standardized humidity sequence to generate an initial environment coupling sequence; A variational decomposition model is introduced to perform modal separation on the initial environment coupling sequence to extract dominant modal components; the interaction intensity coefficient of the environmental parameters is calculated according to the dominant modal components, and the environment coupling time sequence feature is reconstructed based on the interaction intensity coefficient; The surface deformation monitoring data and the internal micro-current response monitoring data are processed in time-frequency domain fusion based on a dynamic window function to obtain a concrete response feature tensor, specifically as follows: Curvature feature extraction is performed on the surface deformation monitoring data to obtain a surface curvature change sequence; impedance feature analysis is performed on the internal micro-current response monitoring data to obtain a pore impedance change sequence; The initial window length and the sliding step length of the dynamic window function are set, and the initial window length and the sliding step length are dynamically adjusted based on the frequency energy distribution of the surface curvature change sequence and the pore impedance change sequence; The surface curvature change sequence and the pore impedance change sequence are subjected to joint time-frequency transformation using the adjusted dynamic window function to construct a time-frequency joint distribution map; the amplitude peak value, the energy accumulation and the frequency band concentration degree index in the time-frequency joint distribution map are extracted to form the concrete response feature tensor; The three-dimensional degradation feature matrix is divided into a plurality of cubic units according to the spatial coordinates; the degradation change rate of each cubic unit in the normal direction, the tangential direction and the radial direction is calculated using a directional derivative operator; A spatial gradient vector field is constructed according to the degradation change rate; divergence analysis and curl analysis are performed on the spatial gradient vector field to output the surface layer divergence distribution and the internal layer curl distribution, respectively; ​ Fusing the surface layer divergence distribution and the internal layer curl distribution generates the surface deterioration gradient field and the internal pore deterioration gradient field.

2. The method for evaluating salt freeze-thaw cycle resistance of concrete according to claim 1, characterized by, Further comprising: According to the surface deterioration gradient field and the internal pore deterioration gradient field, the key evaluation area of the target concrete specimen is divided; Calculate the deterioration accumulation, gradient change rate and spatial correlation degree of each key evaluation area to generate the regional deterioration comprehensive index; Based on the regional deterioration comprehensive index, the key evaluation area is sorted to determine the regional detection priority.

3. The method for evaluating salt freeze-thaw cycle resistance of concrete according to claim 2, characterized by, Further comprising: According to the regional detection priority, a high-resolution detection point set is generated; Based on the ultrasonic flaw detection equipment, micro-damage image data of the high-resolution detection point set is obtained; Performing crack morphology segmentation and pore topology analysis on the micro-damage image data to obtain a micro-damage feature set.

4. The method for evaluating salt freeze-thaw cycle resistance of concrete according to claim 3, characterized by, Further comprising: Map the micro-damage feature set to the surface deterioration gradient field and the internal pore deterioration gradient field to update the deterioration accumulation and gradient change rate of the key evaluation area; According to the updated deterioration accumulation and gradient change rate, the regional deterioration comprehensive index is reconstructed; Based on the reconstructed regional deterioration comprehensive index, the regional detection priority is corrected.

5. The method for evaluating salt freeze-thaw cycle resistance of concrete according to claim 1, wherein Further comprising: According to the surface deterioration gradient field and the internal pore deterioration gradient field, a time-varying stress distribution model is constructed; An incremental iteration algorithm is used to calculate the equivalent fatigue load spectrum of the time-varying stress distribution model; Based on the equivalent fatigue load spectrum, the stress amplitude variation period of the critical damage path is analyzed.

6. The method for evaluating salt freeze-thaw cycle resistance of concrete according to claim 5, wherein Further comprising: According to the stress amplitude variation period, the cumulative damage degree of the critical damage path is calculated; Combined with the standard damage threshold curve of concrete material, the remaining service cycle number of the critical damage path is predicted; Based on the remaining service cycle number of all critical damage paths, the anti-salt freeze cycle performance evaluation result of the target concrete specimen is generated.

7. The method for evaluating salt freeze-thaw cycle resistance of concrete according to claim 6, characterized by, Further comprising: The anti-salt freeze cycle performance evaluation result is packaged into a performance evaluation data packet according to a preset communication protocol; The performance evaluation data packet is transmitted to the remote monitoring center through the wireless sensor network; The performance evaluation data packet is decoded and verified at the remote monitoring center to output the final performance evaluation report.

Images