A high-strength concrete construction crack detection system and method

Abstract

The application belongs to the technical field of concrete crack detection, and specifically discloses a high-strength concrete construction crack detection system and method, which comprises a multi-parameter collaborative sensing module, a construction crack identification module, a crack risk analysis module and a crack early warning processing module. The application collects temperature, humidity, strain and multispectral image data in the plastic to hardening stage in real time through the multi-parameter collaborative sensing module, extracts crack features in combination with differential signal processing in different stages and a deep convolutional neural network, and significantly improves the detection accuracy of micro-cracks and internal defects. Meanwhile, the application innovatively constructs a crack evolution risk rate model by fusing multi-physical field data, realizes dynamic probability prediction before crack initiation, and automatically triggers a hierarchical treatment scheme through risk grade division and a progressive early warning mechanism, effectively distinguishes the causes of drying shrinkage and temperature cracks, significantly improves the prevention and control efficiency, and supports accurate monitoring of the whole life cycle of high-strength concrete.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of concrete crack detection, and in particular, relates to a high-strength concrete construction crack detection system and method. BACKGROUND

[0002] High-strength concrete is widely used in ultra-high-rise buildings, large-span bridges, nuclear power plants, and other harsh material performance scenarios due to its high compressive strength and good durability. Despite its high strength, micro-cracks are easily caused by temperature stress, shrinkage deformation, and other factors during the construction phase due to the large amount of cementitious material and low water-cement ratio. This highlights the necessity of crack detection during construction.

[0003] Prior art such as Chinese patent application No. 202411431893.5 discloses a civil construction monitoring intelligent sensor information detection method and system under signal mismatch. It mainly extracts features from temperature / strain signals through wavelet denoising combined with singular value decomposition algorithm, and reconstructs internal field distribution based on numerical simulation to solve the problem of accurate identification of concrete internal defects under signal mismatch, achieve high-precision monitoring and defect positioning, and improve the reliability of structure health assessment.

[0004] Prior art such as Chinese patent application No. 202311748375.1 discloses a rockfill concrete construction quality detection method. It mainly combines steady-state temperature analysis with real-time image processing through a dynamic detection mechanism triggered by temperature threshold to solve the problem of crack real-time monitoring lag caused by temperature anomalies during rockfill concrete construction period, achieve the effect of rapid identification of cracks and early warning, and ensure construction quality and structural safety.

[0005] For the above technical solutions, it is obvious that the current concrete construction crack detection still has the following deficiencies: 1. The detection mechanism for dynamic matching of plastic-hardening stage has not been established, such as fixed wavelet basis (db4) and static temperature threshold, which leads to signal feature loss and misjudgment, missing the plastic stage intervention window and insufficient sensitivity in the hardening stage.

[0006] 2. Low sensitivity to early micro-cracks and internal hidden cracks, and image recognition is easily disturbed by surface floating slurry, resulting in high missed detection rate.

[0007] 3. Lack of dynamic probability prediction model and grading disposal strategy before crack initiation, unable to actively prevent and control large-area avoidable cracks.

[0008] 4. Relies on single / dual parameters, does not integrate multiple physical field collaborative perception, ignores the inducing effect of humidity field and construction vibration on cracks, leading to cause misjudgment such as dry shrinkage misjudgment as temperature cracks. SUMMARY

[0009] In view of this, in order to solve the problems raised in the background art, a high-strength concrete construction crack detection system and method are proposed.

[0010] The purpose of the present application can be achieved by the following technical solutions: the present application provides a high-strength concrete construction crack detection system, which comprises: a multi-parameter cooperative sensing module, which collects temperature, humidity, strain and multi-spectral image data of concrete in the plastic stage to the hardening stage in real time.

[0011] The construction crack identification module is composed of an image analysis unit and a risk prediction unit.

[0012] The image analysis unit uses a deep convolutional neural network to extract crack features from multi-spectral images, and outputs the number, width and position of cracks.

[0013] The risk prediction unit, when no cracks are detected, integrates temperature, humidity and strain data, outputs the crack evolution risk rate under time series, and constructs the crack evolution risk rate growth curve.

[0014] The crack risk analysis module divides the risk level based on the number, width and position of cracks and the set risk level division rule.

[0015] The crack early warning processing module is composed of a multi-level treatment strategy library and a dynamic early warning unit.

[0016] The multi-level treatment strategy library stores the treatment scheme associated with the risk level.

[0017] The dynamic early warning unit calls the treatment scheme corresponding to the risk level in the multi-level treatment strategy library to process the cracks, and triggers the progressive early warning based on the curve, and starts the crack risk emergency response when the curve slope exceeds the threshold.

[0018] The present application also provides a high-strength concrete construction crack detection method, which comprises: A1, multi-parameter cooperative data acquisition: real-time acquisition of temperature, humidity, strain and multi-spectral image data of concrete in the plastic stage to the hardening stage.

[0019] A2, construction crack identification: using a deep convolutional neural network to extract crack features from multi-spectral images, outputting the number, width and position of cracks, and when no cracks are detected, integrating temperature, humidity and strain data, outputting the crack evolution risk rate under time series, and constructing the crack evolution risk rate growth curve.

[0020] A3, crack risk analysis: based on the number, width and position of cracks and the set risk level division rule, the risk level is divided.

[0021] A4, crack early warning processing: based on the risk level, calling the corresponding risk level in the multi-level treatment strategy library Disposal scheme for crack processing, and based on the crack evolution risk rate growth curve Trigger progressive early warning, and when the slope of the curve exceeds the threshold, start the crack risk emergency response.

[0022] Compared with the prior art, the beneficial effects of the present application are as follows: (1) The present application collects temperature, humidity, strain and multi-spectral image data in the plastic to hardening stage in real time through the multi-parameter cooperative sensing module, combines with the differential signal processing and deep convolutional neural network in different stages to extract crack features, significantly improves the detection accuracy of microcracks and internal defects. At the same time, the multi-physical field data are fused, the crack evolution risk rate model is innovatively constructed, the dynamic probability prediction before crack initiation is realized, and through the risk level division and progressive early warning mechanism, the graded treatment scheme is automatically triggered, the causes of drying shrinkage and temperature cracks are effectively distinguished, and the prevention and control efficiency is significantly improved.

[0023] (2) The present application uses sym8 wavelet base high-frequency filtering to capture transient temperature fluctuations in the plastic stage, and switches to db6 wavelet base low-frequency filtering to suppress noise in the hardening stage, which solves the feature loss problem caused by fixed wavelet base such as db4, significantly improves the accuracy of the intervention window in the plastic stage, and greatly improves the sensitivity in the hardening stage.

[0024] (3) The present application integrates temperature-humidity-strain-multispectral four-dimensional data, eliminates surface floating slurry interference through polarization filter, morphological filter and frequency domain fusion technology, significantly reduces internal defect projection positioning error, and through gradient layout of distributed humidity sensor combined with diffusion model compensation, can identify the coupling effect of drying shrinkage and temperature strain, facilitate cause identification, and also reduce cause misjudgment rate.

[0025] (4) The present application constructs a crack evolution risk rate model based on time series by fusing multiple physical parameters, realizes accurate risk probability prediction before crack initiation, and combines three-level risk level division and progressive early warning, so that the avoidable crack prevention and control rate is significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0027] Figure 1 The present application is a system module connection diagram.

[0028] Figure 2A flowchart for implementing the steps of the method of the present application is shown.

[0029] Figure 3 A schematic diagram of the structure of the construction crack identification module is shown.

[0030] Figure 4 A schematic diagram of the structure of the signal preprocessing unit is shown.

[0031] Figure 5 A schematic diagram of the structure of the crack early warning processing module is shown. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some 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 work fall within the scope of protection of the present application.

[0033] Please refer to Figure 1 As shown in the figure, the present application provides a high-strength concrete construction crack detection system, which comprises a multi-parameter collaborative sensing module, a construction crack identification module, a crack risk analysis module and a crack early warning processing module.

[0034] In the above, the construction crack identification module is connected with the multi-parameter collaborative sensing module and the crack risk analysis module, and the crack risk analysis module is connected with the crack early warning processing module.

[0035] The multi-parameter collaborative sensing module collects temperature, humidity, strain and multi-spectral image data in real time from the plastic stage to the hardening stage of the concrete.

[0036] Specifically, the multi-parameter collaborative sensing module comprises: a distributed optical fiber temperature sensor array embedded in the concrete along the three-dimensional orthogonal grid of the concrete structure, which collects three-dimensional distribution data of the temperature field from the plastic stage to the hardening stage in real time, denoted as temperature data.

[0037] Capacitive humidity sensor arrays are installed at the surface layer, 1 / 2 thickness layer and bottom layer of the concrete structure in a gradient distribution, which collect humidity data from the plastic stage to the hardening stage in real time.

[0038] FBG strain sensor arrays are arranged orthogonally at the beam ends and corner positions of the plate, which measure axial strain and shear strain from the plastic stage to the hardening stage in real time, and integrate to obtain strain data.

[0039] Through the multispectral imager mounted on the construction robot, visible light and near-infrared are used for double-band scanning in real time, and three-way polarization filters are combined to eliminate the interference of surface floating slurry reflection, so as to obtain the multispectral image data corresponding to the elimination of interference in the plastic stage to the hardening stage.

[0040] Understandably, high-strength concrete usually has a low water-cement ratio and is mixed with mineral admixtures, so the early strength is high, but this may also cause large self-shrinkage and be prone to cracking. Cracks can occur at different stages, such as the plastic stage, the hardening stage, or load cracks after long-term use, so the present application focuses on targeted crack detection at specific stages.

[0041] Please refer to Figure 3 As shown in the figure, the construction crack identification module is composed of an image analysis unit, a risk prediction unit and a signal preprocessing unit.

[0042] The image analysis unit uses a deep convolutional neural network to extract crack features from multispectral images, and outputs the number, width and position of cracks.

[0043] Specifically, the deep convolutional neural network includes a multi-scale feature extraction layer configured to extract crack texture features in parallel using a dilated convolution kernel (3x3, 5x5, 7x7).

[0044] The attention enhancement module is configured to strengthen the feature weights of the crack area through the SE-Net channel attention mechanism.

[0045] The transfer learning framework pre-trains a model, such as one based on 5000 groups of concrete CT scan images and synchronous surface images.

[0046] It should be noted that using a deep convolutional neural network to extract crack features from multispectral images is a relatively common technical means, and the more detailed feature extraction process will not be described in detail here.

[0047] The risk prediction unit outputs the crack evolution risk rate under the time sequence when no cracks are detected, and constructs a crack evolution risk rate growth curve by comprehensively considering temperature, humidity and strain data.

[0048] Specifically, the crack evolution risk rate under the time sequence is used to output the crack evolution risk rate under the time sequence corresponding to the plastic stage, including: B1, for the plastic stage, traverse each temperature sensor, calculate the absolute value of the temperature difference with all adjacent points, and take the maximum temperature difference absolute value as the target temperature difference.

[0049] B2, respectively, the power spectrum density integral calculation is carried out to the beam end direction stress signal and the orthogonal direction shear strain signal, the axial strain energy and the shear strain energy are output, the axial strain energy is calculated by comprehensively calculating the axial strain energy and the shear strain energy, and the total energy ratio is calculated.

[0050] Further, the specific calculation formula of the axial strain energy is: , represents the axial strain energy, represents the frequency, in the integral process, all frequency values between 10Hz and 50Hz will be traversed, represents the power spectrum density of the corresponding stress signal along the beam end direction at the frequency , reflecting the energy distribution of the frequency component.

[0051] The specific calculation formula of the shear strain energy is: , represents the shear strain energy, represents the power spectrum density of the corresponding shear strain signal along the orthogonal direction at the frequency .

[0052] The specific calculation formula for calculating the axial strain energy by comprehensively calculating the axial strain energy and the shear strain energy and calculating the total energy ratio is: , represents the interval length, represents the time length corresponding to the calculation of the axial strain energy change and the shear strain energy change, and the strain data is collected and analyzed at a period of 1 minute, the value is 60 seconds, and is a dynamic value, when the structure contraction intensifies, will be reduced, such as being set to 20 seconds. And the higher the energy ratio, the greater the accumulated strain of the structure, and the risk of plastic deformation increases.

[0053] B3, respectively, the standard deviation of the corresponding humidity data of the surface layer, the 1 / 2 thickness layer and the bottom layer of the concrete structure is calculated, and the weighted sum is calculated, and the target humidity fluctuation degree is output.

[0054] Understandably, the humidity distribution of the concrete structure may be different at different depths due to environmental exposure, water migration and other factors. The surface layer directly contacts the external environment, and the humidity fluctuation is large, the middle layer belongs to the affected area and is less affected, and the bottom layer has different humidity conditions due to the change of humidity in the placement area. Exemplarily, the weights of the standard deviations of the humidity data corresponding to the surface layer, the 1 / 2 thickness layer and the bottom layer of the concrete structure can be respectively taken as 0.4, 0.2 and 0.4.

[0055] B4. Calculate the micro crack density, bleeding channel length, aggregate dispersion, and interference fringe intensity from the multispectral image data to construct the image feature vector.

[0056] It is understandable that the micro crack density and the crack risk rate are positively correlated. The micro crack density corresponding to the plastic shrinkage crack, the area with micro crack density exceeding the threshold value, has a high probability of developing into a visible crack with a width of ≥0.1 mm in the subsequent hardening stage. The bleeding channel length corresponds to the drying shrinkage crack. The bleeding channel accelerates the local humidity drop rate, resulting in a drying shrinkage stress concentration. The aggregate dispersion corresponds to the interface crack. The interference fringe intensity corresponds to the temperature stress crack. Therefore, the micro crack density, the bleeding channel length, the aggregate dispersion, and the interference fringe intensity are selected as the four indicators for image dimension risk analysis.

[0057] Among them, the micro crack density is calculated by processing the dark lines in the polarization image, involving edge detection and morphological processing. The bleeding channel length is calculated by using Hough transform. The aggregate dispersion involves image segmentation, such as the SLIC algorithm, to distinguish the aggregate and paste regions. The interference fringe intensity can be obtained by analyzing the color difference or light intensity change, such as calculating the ΔE value in the CIELab color space.

[0058] For example, the micro crack density is obtained by Canny edge detection and morphological closing operation based on the linear polarization channel (DoLP), and the number of shrinkage dark lines per unit area is counted to obtain the micro crack density. The bleeding channel length is obtained by using Hough transform to detect linear bright bands in near-infrared polarization angle (AoP) images, and the geometric length of continuous bright bands is calculated as the bleeding channel length. The aggregate dispersion is obtained by SLIC superpixel segmentation of visible light multispectral images, and the proportion of aggregate-paste interface area is calculated as the aggregate dispersion. The interference fringe intensity is obtained by extracting the difference between the polarization degrees of the RGB three channels, and calculating the color difference in the CIELab color space to obtain the interference fringe intensity.

[0059] B5. For each acquisition time point, the crack evolution risk rate is calculated by integrating the target temperature difference, the total energy ratio, the target humidity fluctuation, and the image feature vector, and then the crack evolution risk rate corresponding to the time sequence in the plastic stage is constructed.

[0060] Specifically, the calculation process of the crack evolution risk rate is as follows: the target temperature difference, the total energy ratio, and the target humidity fluctuation are normalized respectively, and the normalized target temperature difference, the total energy ratio, and the target humidity fluctuation are output respectively, denoted as , , .

[0061] The seepage channel length and the micro trace density are normalized by a Min-Max normalization method, and the aggregate dispersion and the interference stripe intensity are standardized by a Z-Score standardization method, a weighted sum of the normalized seepage channel length, the micro trace density, the aggregate dispersion and the interference stripe intensity is calculated, and a comprehensive image feature value is output, denoted as .

[0062] Based on , , and , a crack evolution risk rate is obtained by a weighted sum, denoted as , , , , and represent the weights corresponding to the target temperature difference, the total energy ratio, the target humidity fluctuation degree and the comprehensive image feature value, respectively.

[0063] Understandably, the weights of the seepage channel length, the micro trace density, the aggregate dispersion and the interference stripe intensity and the specific values of , , and are set by historical data fitting or expert experience.

[0064] Further, the crack evolution risk rate in the time sequence is output, including: C1, based on the crack evolution risk rate in the time sequence corresponding to the plastic stage, setting a hardening stage risk compensation factor.

[0065] Further, the hardening stage risk compensation factor is set, including: selecting the maximum crack evolution risk rate from the crack evolution risk rate in the time sequence corresponding to the plastic stage.

[0066] The duration of the crack evolution risk rate greater than the set reference disturbance value is counted, denoted as the crack evolution disturbance duration.

[0067] The maximum crack evolution risk rate and the crack evolution disturbance duration are normalized and input into a sigmoid function, and the hardening stage risk compensation factor is output.

[0068] Specifically, in the calculation of the concrete hardening stage risk compensation factor, the core reason for using the sigmoid function is that it can convert the nonlinear and asymmetric risk characteristics into continuous compensation values in the 0-1 interval, which meets the threshold effect and risk control requirements of crack evolution.

[0069] C2, based on the temperature, humidity, strain and multi-spectral image data of the hardening stage, calculate the temperature rise rate, strain acceleration value, humidity gradient value and polarization interference fringe density of the hardening stage in each time window.

[0070] Understandably, in the hardening stage of the concrete, the temperature rise rate can be calculated by time series analysis of temperature data, and the temperature difference between adjacent time points is divided by the time interval. The strain acceleration value needs to be second-order time differentiated for strain data, and the numerical difference method is used to calculate the change amplitude of the strain change rate per unit time. The humidity gradient value is based on multi-position humidity sensor data, and the ratio of the humidity difference and the distance between adjacent points in space is calculated. The polarization interference fringe density is quantified by multi-spectral image analysis, and the number or reciprocal of the interference fringe spacing per unit area is counted. These parameters respectively reflect the dynamic characteristics of the thermal dynamics, mechanical deformation, water migration and microstructure evolution in the hardening process. Therefore, the temperature rise rate, strain acceleration value, humidity gradient value and polarization interference fringe density are analyzed.

[0071] C3, normalize the temperature rise rate, strain acceleration value, humidity gradient value and polarization interference fringe density, and obtain the hardening stage crack evolution risk rate in each time window by weighted summation.

[0072] It should be added that the time window can be set to 30 minutes.

[0073] It should also be added that the normalization process is to uniformly scale the parameters of different dimensions such as temperature rise rate, strain acceleration value, humidity gradient value and polarization interference fringe density to [0, 1], eliminating the influence of dimension difference on comprehensive evaluation, and usually using extreme value normalization. The weight setting needs to be determined according to the contribution of each parameter to the crack risk, and it is recommended that the temperature rise rate and strain acceleration be given higher weight because they are directly related to thermal stress and deformation mutation, and the values are set to 0.3, the humidity gradient reflects the shrinkage stress caused by water migration, and the value is set to 0.25. The polarization interference fringe density as a microstructure evolution index has a lower weight, which is set to 0.15.

[0074] C4, based on the hardening stage risk compensation factor, the hardening stage crack evolution risk rate is modified, and based on the modified crack evolution risk rate value in each time window, the crack evolution risk rate in the time sequence is output.

[0075] Specifically, the larger the hardening stage risk compensation factor, the larger the hardening stage crack evolution risk rate, and exemplarily, based on the hardening stage risk compensation factor, the hardening stage crack evolution risk rate is modified, which satisfies the following modification formula: , represents the modified crack evolution risk rate value, represents the hardening stage crack evolution risk rate, represents a hardening stage risk compensation factor.

[0076] The application constructs a time series-based crack evolution risk rate model by fusing multiple physical parameters in real time, realizes accurate risk probability prediction before crack initiation, and combines three-level risk grade division and progressive early warning, so that the avoidable crack prevention rate is significantly improved.

[0077] Please refer to Figure 4 As shown in the figure, the signal preprocessing unit is specifically configured as follows: the time series data processing channel performs the following phased differential processing on temperature, humidity and strain data: in the plastic stage, the temperature data is subjected to high-frequency band-pass filtering using the sym8 wavelet basis, the strain data is subjected to dynamic wide band-pass filtering, and the humidity data is subjected to baseline drift correction.

[0078] In the hardening stage, the temperature data is switched to the db6 wavelet basis for low-frequency band-pass filtering, the strain data is subjected to narrow band-pass filtering, and the humidity data is subjected to phase compensation based on the diffusion model established based on the second Fick's law.

[0079] The image data processing channel, in the plastic stage, uses multi-angle polarization fusion technology to suppress the reflection interference of the liquid surface of the concrete, and eliminates the artifacts caused by bubbles and floating slurry in the image through morphological filtering, and in the hardening stage, multi-spectral frequency domain fusion technology is enabled to enhance the edge features of micro-cracks, and combined with the strain distribution thermograph, the projection area of the internal defects of the concrete on the surface is located.

[0080] Specifically, high-frequency band-pass filtering based on the sym8 wavelet basis is used, and the passband frequency range is set to 50-200Hz, to suppress disturbance noise with a frequency greater than 500Hz generated in the pouring process due to vibration, pumping and other operations, and to retain high-frequency characteristic signals reflecting the early state changes of the concrete. Through dynamic wide band-pass filtering, the pulse characteristics of early shrinkage of the concrete are captured. The frequency range of the dynamic wide band-pass is dynamically adjusted according to the time after the concrete is poured. The initial range is 30-300Hz, and as the plastic stage progresses, the upper frequency gradually decreases to 200Hz, to accurately capture the early shrinkage characteristics. Baseline drift correction is performed by monitoring the environmental humidity and the internal humidity change trend of the concrete in real time, establishing a reference baseline for humidity change, and correcting the humidity measurement value using linear regression algorithm to ensure the accuracy of the humidity data.

[0081] Further, the pulse characteristics of early shrinkage of the concrete are captured through dynamic wide band-pass filtering, and the frequency range of the dynamic wide band-pass is dynamically adjusted according to the time after the concrete is poured. The specific adjustment process is as follows: 1) within 0-1 hours after the concrete is poured, the initial frequency range is set to 30-300Hz.

[0082] At this stage, the concrete has just been cast and is in a state of high fluidity, with cement particles in a dispersed and suspended state, and water is sufficient and relatively evenly distributed. Due to the strong fluidity of the concrete, its internal structure has not yet begun to coagulate and solidify, so the frequency distribution of the shrinkage characteristic signal is relatively wide.

[0083] In order to comprehensively capture these early shrinkage signals caused by different factors, a wider frequency band is used for filtering. In actual operation, the strain signal of the concrete is collected at a sampling frequency of 1000 times per second to ensure that high-frequency shrinkage pulse characteristics can be captured. After the collected signal is converted by an analog-to-digital converter, it enters a dynamic wide-band pass filter. The filter filters the signal according to the preset 30-300Hz frequency range, removes noise signals below 30Hz and above 300Hz, and thus retains various possible early shrinkage signals.

[0084] 2) During the 1-3 hours after casting, reduce the upper frequency at a rate of 50Hz per hour.

[0085] Understandably, as time goes on, the concrete gradually loses its fluidity, and the high-frequency component of the shrinkage characteristic signal decreases. Reducing the upper frequency can more accurately focus on the effective shrinkage characteristic frequency band. For example, 2 hours after casting, the frequency range is adjusted to 30-250Hz. At this time, the system will automatically adjust the parameters of the dynamic wide-band pass filter according to the advancement of time, reducing the upper frequency from 300Hz to 250Hz. This can more accurately focus on the effective shrinkage characteristic frequency band and reduce the interference of high-frequency noise on the shrinkage characteristic signal.

[0086] 3) During the 3-6 hours after casting, continue to reduce the upper frequency at a rate of 25Hz per hour until the upper frequency reaches 200Hz.

[0087] At this stage, the plasticity of the concrete gradually decreases, the internal structure further solidifies and stabilizes, and the shrinkage characteristics become more stable, which can slowly increase at a rate.

[0088] 4) When the upper frequency reaches 200Hz, the frequency range is finally fixed at 30-200Hz to accurately capture the early shrinkage characteristics at this stage.

[0089] At this stage, the strain signal of the concrete will continue to be filtered with this fixed frequency range to ensure that the characteristic information related to early shrinkage can be accurately extracted.

[0090] It should be noted that the low-frequency band-pass filtering based on the db6 wavelet base is switched to, and the cut-off frequency is set to be less than 10 Hz, so as to facilitate subsequent extraction of the trend characteristics of the concrete hydration temperature rise, and separation of the long-period and low-frequency temperature change signal generated by the cement hydration reaction. The passband frequency range of the narrow frequency band-pass filtering is 2-5 Hz, the response component caused by the creep is separated, and the strain fluctuation caused by other interference factors is removed. The diffusion model established based on the second Fick's law considers the pore structure, humidity gradient and other factors of the hardened concrete body, compensates the hysteresis effect in the humidity measurement, and accurately reflects the internal humidity change of the concrete.

[0091] Further, a multi-angle polarization fusion technology is adopted to acquire multi-spectral images from different polarization angles, the weighted average algorithm is used to fuse the images at different angles, the influence of surface reflected light on the image quality is reduced, the weight can be set corresponding to the experience data, the morphological filtering adopts the morphological operation combined with open and close operation, the size of the structural element is determined according to the image resolution and the approximate size of the bubbles and floating slurry, and is generally 3*3 to 7*7 pixels, and small-size noise and artifacts are removed. The multi-spectral frequency domain fusion technology converts the images of different spectral bands to the frequency domain, adjusts the weight of each band image in the frequency domain, highlights the feature information of the micro-cracks in the frequency domain, converts the processed frequency domain image back to the spatial domain, and enhances the clarity of the micro-crack edge. The strain distribution thermograph is generated according to the data collected by the strain sensor, the projection area of the internal defect of the concrete on the surface is located by using the image registration technology to align the thermograph and the multi-spectral image, and the projection area of the internal defect on the surface is located in the multi-spectral image by analyzing the strain abnormal area in the thermograph.

[0092] In the embodiment of the present application, the sym8 wavelet base high-frequency filtering is adopted in the plastic stage to capture transient temperature fluctuations, and the db6 wavelet base low-frequency filtering is switched to in the hardening stage to suppress noise, and the phased differentiated signal processing is combined, so that the feature loss problem caused by the fixed wavelet base such as db4 is solved, the intervention window recognition accuracy in the plastic stage is significantly improved, the sensitivity in the hardening stage is also greatly improved, the temperature-humidity-strain-multipolar four-dimensional data are integrated, the surface floating slurry interference is eliminated through the polarization filter, the morphological filtering and the frequency domain fusion technology, the internal defect projection positioning error is obviously reduced, the coupling effect of the drying shrinkage and the temperature strain can be recognized through the gradient layout of the distributed humidity sensor combined with the diffusion model compensation, the cause identification is facilitated, and the cause misjudgment rate is also reduced.

[0093] Exemplarily, in order to further analyze, the signal-to-noise ratio, early warning time, crack misjudgment rate, strain sensitivity, hardware survival rate, cross-modal response time and long-term monitoring accuracy attenuation are taken as the comparison technical indexes for comparison, and specific comparison results are shown in Table 1.

[0094] Table 1 shows the effect of the technical solution mode comparison table

[0095]

[0096] The crack risk analysis module divides the risk level based on the number, width and position of the cracks and the set risk level division rule.

[0097] Specifically, the set risk level division rule is implemented as follows: the apparent image of the concrete is extracted from the multispectral image data, and the key stress area, the secondary key stress area and the non-key area of the concrete are defined accordingly.

[0098] If the number of cracks and the maximum crack width are less than the corresponding first trigger threshold, and the crack positions are all in the non-key area, the first level risk level is triggered.

[0099] If the number of cracks or the maximum crack width exceeds the corresponding first trigger threshold and is less than or equal to the corresponding second trigger threshold, or the crack position is in the secondary key area, the second level risk level is triggered.

[0100] If the number of cracks or the maximum crack width exceeds the corresponding second trigger threshold, or the crack position is in the key stress area, the third level risk level is triggered.

[0101] It should be noted that the first trigger threshold corresponding to the number of cracks and the maximum crack width can be 2 and 0.1mm respectively, and the second trigger threshold corresponding to the number of cracks and the maximum crack width can be 5 and 0.3mm respectively. The non-key area is, for example, a non-load-bearing wall or a decorative surface layer. The secondary key stress area is, for example, the edge of the floor slab or the non-joint area of the shear wall. The key stress area is, for example, the beam-column joint, the root of the load-bearing wall or the prestressed anchorage area.

[0102] Please refer to Figure 5 As shown, the crack early warning processing module is composed of a multi-level treatment strategy library and a dynamic early warning unit.

[0103] The multi-level treatment strategy library stores the treatment scheme associated with the risk level.

[0104] The dynamic early warning unit calls the treatment scheme corresponding to the risk level in the multi-level treatment strategy library to process the cracks, and triggers a progressive early warning based on the curve. When the slope of the curve exceeds the threshold, the crack risk emergency response is started.

[0105] Specifically, the specific trigger of the progressive early warning is as follows: the real-time slope of the crack evolution risk rate growth curve is calculated by the central difference method, denoted as the risk change slope The real-time crack evolution risk rate in the curve is denoted as , represents the time point number, .

[0106] Locating the starting time point and the ending time point from the crack evolution risk rate growth curve, forming a time interval, calculating the integral value of the crack evolution risk rate growth curve in the time interval, and recording the cumulative risk amount as .

[0107] Based on , and and pre-set warning trigger rules trigger corresponding warnings, the warning trigger rules are as follows: if there are and , and , trigger a level I warning, and are the set first risk rate threshold and the second risk rate threshold, and are the set first risk change slope and the second risk amount threshold.

[0108] If there are or or , trigger a level II warning, is a set third risk rate threshold, and are the set second risk change slope and the second risk amount threshold.

[0109] If there are or or , trigger a level II warning.

[0110] When the risk rate change slope continuously exceeds the set value in multiple time windows, the warning level is automatically upgraded.

[0111] The embodiment of the application collects temperature, humidity, strain and multispectral image data in the plastic to hardening stage in real time through the multi-parameter collaborative sensing module, combines differential signal processing and deep convolutional neural network to extract crack features, and significantly improves the detection accuracy of microcracks and internal defects. At the same time, the multi-physical field data are fused, the crack evolution risk rate model is innovatively constructed, the dynamic probability prediction before crack initiation is realized, and through the risk level division and the progressive warning mechanism, the graded treatment scheme is automatically triggered, the causes of drying shrinkage and temperature cracks are effectively distinguished, and the prevention and control efficiency is significantly improved.

[0112] Please refer to Figure 2 , the application also provides a high-strength concrete construction crack detection method, which comprises the following steps: A1, multi-parameter collaborative data acquisition: real-time acquisition of temperature, humidity, strain and multispectral image data of concrete in the plastic stage to the hardening stage.

[0113] A2, construction crack identification: a deep convolutional neural network is used to extract crack features from multispectral images, output the number, width and position of cracks, and when no cracks are detected, integrate temperature, humidity and strain data to output the crack evolution risk rate in time series, and construct the crack evolution risk rate growth curve.

[0114] A3, crack risk analysis: based on the number, width and position of cracks and the set risk level division rule, the risk level is divided.

[0115] A4, crack early warning processing: based on the risk level, the corresponding risk level disposal scheme in the multi-level disposal strategy library is called to process the crack, and the progressive early warning is triggered based on the crack evolution risk rate growth curve, and when the curve slope exceeds the threshold, the crack risk emergency response is started.

[0116] The above is only an example and description of the concept of the present application, and those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the concept of the present application or exceed the scope defined by the present application, which shall belong to the protection scope of the present application.

Claims

1. A high-strength concrete construction crack detection system, characterized by, Comprise: Multi-parameter collaborative sensing module, real-time acquisition of temperature, humidity, strain and multi-spectral image data of concrete in plastic stage to hardening stage; Construction crack identification module, composed of image analysis unit and risk prediction unit; The image analysis unit adopts deep convolutional neural network to extract crack features from multi-spectral images, and outputs the number, width and position of cracks; The risk prediction unit, when no cracks are detected, integrates temperature, humidity and strain data to output crack evolution risk rate in time sequence, and constructs crack evolution risk rate growth curve; Crack risk analysis module, based on crack number, width and position, and set risk grade division rule to divide risk grade; Crack early warning processing module, composed of multi-level treatment strategy library and dynamic early warning unit; The multi-level treatment strategy library stores the treatment scheme corresponding to the risk level; Dynamic early warning unit, call the corresponding risk level in the multi-level treatment strategy library to process the crack, and trigger the progressive warning based on the curve, when the curve slope exceeds the threshold, start the crack risk emergency response; The construction crack identification module further comprises a signal preprocessing unit, which is specifically configured as follows: Time series data processing channel, the following stage differentiation processing is performed on temperature, humidity and strain data: In plastic stage, temperature data is filtered by sym8 wavelet basis for high frequency band, strain data is filtered by dynamic wide band, and humidity data is corrected by baseline drift; In hardening stage, temperature data is switched to db6 wavelet basis for low frequency band filtering, strain data is filtered by narrow band, and humidity data is compensated by phase based on Fick's second law; Image data processing channel, in plastic stage, adopt multi-angle polarization fusion technology to suppress the reflection interference of concrete liquid surface, and eliminate the artifacts caused by bubbles and floating slurry in image by morphological filtering, in hardening stage, enable multi-spectral frequency domain fusion technology to enhance the edge features of micro cracks, and combine with strain distribution thermograph to locate the projection area of internal defects on the surface of concrete.

2. A high strength concrete construction crack detection system as claimed in claim 1, wherein: The multi-parameter collaborative sensing module comprises: Real-time acquisition of three-dimensional distribution data of temperature field from plastic stage to hardening stage by distributed optical fiber temperature sensor array embedded in concrete along three-dimensional orthogonal grid of concrete structure, recorded as temperature data; Install a gradient distributed array of capacitive humidity sensors at the surface, 1 / 2 thickness layer and bottom layer of the concrete structure, and real-time collect humidity data from plastic stage to hardening stage; Orthogonal arrangement of FBG strain sensor array at the end of concrete beam and plate corner position, real-time measurement of axial strain and shear strain from plastic stage to hardening stage, and integration of strain data; Through the multi-spectral imager erected on the construction robot, real-time dual-band scanning is carried out by visible light and near infrared, and three-way polarization filter is used to eliminate the reflection interference of surface floating slurry, and multi-spectral image data after interference elimination from plastic stage to hardening stage is obtained.

3. A high strength concrete construction crack detection system as claimed in claim 1, wherein: The output of crack evolution risk rate in time sequence is used to output the crack evolution risk rate in time sequence corresponding to the plastic stage, which comprises: For the plastic stage, the temperature sensor is traversed, the absolute value of the temperature difference with all adjacent points is calculated, and the maximum temperature difference absolute value is taken as the target temperature difference; The power spectrum density integral calculation is performed on the beam end direction stress signal and the orthogonal direction shear strain signal respectively, and the axial strain energy and shear strain energy are output, and the total energy ratio is calculated by comprehensively calculating the axial strain energy and shear strain energy; The standard deviation of the corresponding humidity data of the surface layer, 1 / 2 thickness layer and bottom layer of the concrete structure is calculated respectively, and the weighted sum calculation is performed, and the target humidity fluctuation degree is output; The micro mark density, bleeding channel length, aggregate dispersion and interference fringe intensity are counted from the multispectral image data, and the image feature vector is constructed; For each collection time point, the crack evolution risk rate is calculated by comprehensively considering the target temperature difference, total energy ratio, target humidity fluctuation degree and image feature vector, and then the crack evolution risk rate under the time sequence corresponding to the plastic stage is constructed.

4. A high strength concrete construction crack detection system as claimed in claim 3, wherein: The output time sequence crack evolution risk rate comprises: Based on the crack evolution risk rate under the time sequence corresponding to the plastic stage, a hardening stage risk compensation factor is set; Based on the temperature, humidity, strain and multispectral image data of the hardening stage, the temperature rise rate, strain acceleration value, humidity gradient value and polarization interference fringe density of the hardening stage in each time window are calculated; The temperature rise rate, strain acceleration value, humidity gradient value and polarization interference fringe density are normalized, and the hardening stage crack evolution risk rate in each time window is obtained by weighted sum; Based on the hardening stage risk compensation factor, the hardening stage crack evolution risk rate is corrected, and based on the corrected crack evolution risk rate value in each time window, the time sequence crack evolution risk rate is output.

5. A high strength concrete construction crack detection system as claimed in claim 4, wherein: The setting of the hardening stage risk compensation factor comprises: The maximum crack evolution risk rate is selected from the crack evolution risk rate under the time sequence corresponding to the plastic stage; The duration of the crack evolution risk rate greater than the set reference interference value is counted, which is recorded as the crack evolution interference duration; The maximum crack evolution risk rate and the crack evolution interference duration are normalized and input into the sigmoid function to output the hardening stage risk compensation factor.

6. A high strength concrete construction crack detection system as defined in claim 1, wherein: The set risk level division rule is specifically implemented as follows: The concrete apparent image is extracted from the multispectral image data, and the key stress area, secondary key stress area and non-key area of the concrete are defined according to the apparent image; If the number of cracks and the maximum crack width are less than the corresponding first trigger threshold, and the crack position is located in the non-key area, the first level risk is triggered; If the number of cracks or the maximum crack width exceeds the corresponding first trigger threshold and is less than or equal to the corresponding second trigger threshold, or the crack position is located in the secondary key area, the second level risk is triggered; If the number of cracks or the maximum crack width exceeds the corresponding second trigger threshold, or the crack position is located in the key stress area, the third level risk is triggered.

7. A high strength concrete construction crack detection system as defined in claim 1, wherein: The specific trigger of the progressive early warning is as follows: The real-time slope of the crack evolution risk rate growth curve is calculated by the central difference method, denoted as the risk change slope The real-time fissile risk rate in the curve is denoted as , The time point number is denoted as ; The starting time point and the ending time point are located from the crack evolution risk rate growth curve, a time interval is formed, the integral value of the crack evolution risk rate growth curve in the time interval is calculated, and is recorded as a cumulative risk amount, and is recorded as ; Based on , and and pre-set early warning trigger rules trigger corresponding early warnings.

8. A method of detecting construction cracks in high-strength concrete, characterized by: The method comprises: A1, multi-parameter collaborative data acquisition: real-time acquisition of temperature, humidity, strain and multispectral image data of concrete from plastic stage to hardening stage; A2, construction crack identification: a deep convolutional neural network is used to extract crack features from multispectral images, output the number, width and position of cracks, and when no cracks are detected, the temperature, humidity and strain data are integrated to output the crack evolution risk rate in time series, and a crack evolution risk rate growth curve is constructed; A3, crack risk analysis: based on the number, width and position of cracks and the set risk level division rules, the risk level is divided; A4, crack early warning processing: based on the risk level, the corresponding risk level disposal scheme in the multi-level disposal strategy library is called to process the cracks, and the progressive early warning is triggered based on the crack evolution risk rate growth curve, and when the curve slope exceeds the threshold, the crack risk emergency response is started; The crack evolution risk rate in time series is used to output the crack evolution risk rate in time series corresponding to the plastic stage, including: for the plastic stage, each temperature sensor is traversed, the absolute value of the temperature difference with all adjacent points is calculated, and the maximum temperature difference absolute value is taken as the target temperature difference; the power spectral density integral calculation is performed on the beam end direction stress signal and the orthogonal direction shear strain signal respectively, and the axial strain energy and shear strain energy are output, and the total energy ratio is calculated by integrating the axial strain energy and shear strain energy; the standard deviations of the corresponding humidity data of the surface layer, 1 / 2 thickness layer and bottom layer of the concrete structure are calculated respectively, and the weighted sum is calculated, and the target humidity fluctuation degree is output; the micro mark density, bleeding channel length, aggregate dispersion and interference stripe intensity are counted from the multispectral image data, and the image feature vector is constructed; for each acquisition time point, the target temperature difference, total energy ratio, target humidity fluctuation degree and image feature vector are integrated to calculate the crack evolution risk rate, and then the crack evolution risk rate in time series corresponding to the plastic stage is constructed.

Images