A method for early intensity detection of UHPC based on terahertz-infrared dual modes

Abstract

This invention discloses a method for early intensity detection of UHPC based on terahertz-infrared dual-mode, belonging to the field of intelligent construction monitoring technology. The method includes: acquiring zero-point data, emissivity parameters, and synchronous trigger signals to form a unified calibration parameter set; obtaining terahertz and infrared feature sets on a detection grid; aligning, normalizing, and fusing the two types of features in time and space dimensions to establish a latent variable model integrating physical mechanisms and data features; calculating the distribution of UHPC hydration degree; constructing an early intensity prediction model based on curing temperature to obtain the early intensity distribution at each age; calculating quality scores based on multi-source quality assessment indicators, identifying retesting areas, and generating standardized quality control reports and construction control instructions; collecting measured intensity data after construction, integrating quality scores and risk classification information, and iteratively updating the intensity prediction model and latent variable model parameters to achieve high-precision early intensity prediction, quality assessment, and long-term adaptive construction control.

Description

Technical Field

[0001] This invention relates to the field of intelligent construction monitoring technology, and in particular to a method for early intensity detection of UHPC based on terahertz-infrared dual-mode. Background Technology

[0002] Ultra-high performance concrete (UHPC), due to its high strength, good density, and outstanding durability, has been increasingly used in bridge, tunnel, and high-rise building projects. Construction pace and structural safety are directly constrained by the rate of early strength formation; therefore, the visualization and quantitative assessment of early strength have long been a focus of engineering and academic circles. Currently, there are various non-destructive testing methods. Ultrasonic testing focuses on the correlation between wave velocity and density, infrared thermography observes the amplitude and phase changes of the stimulated temperature field, and electromagnetic methods utilize the evolution of dielectric and conductive behavior to reflect the internal state of the material. In recent years, with advancements in hardware and algorithms for terahertz imaging and infrared thermography, information such as pore water content and product formation during hydration processes can be obtained with greater detail under non-contact conditions. Therefore, multimodal fusion analysis that couples different physical quantities is gradually becoming an important development direction for non-destructive testing of materials.

[0003] However, existing methods still have two limitations: First, most methods rely on a single sensing method and cannot simultaneously acquire the electromagnetic and thermophysical parameters of the material, resulting in limited accuracy of early strength prediction and difficulty in achieving refined characterization of spatial distribution; Second, existing methods usually lack deep integration with quality assessment, risk classification and construction control, and cannot achieve closed-loop management from data acquisition and strength prediction to quality control, making it difficult to meet the comprehensive requirements of real-time performance, adaptability and intelligence in UHPC projects. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a UHPC early strength detection method based on terahertz-infrared dual modes to solve the problems of insufficient prediction accuracy and lack of linkage with construction quality control in existing UHPC early strength detection methods due to single-mode detection.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for early intensity detection of UHPC based on terahertz-infrared dual-mode, which includes acquiring zero-point time, emissivity parameters and synchronous trigger signals to form a unified calibration parameter set;

[0008] Based on a unified calibration parameter set, time-domain waveforms of the same polarization and cross polarization are collected in the detection grid and frequency-domain transformation and inversion processing are performed to obtain the terahertz feature set.

[0009] Based on a unified calibration parameter set, a sweep frequency phase-locked loop and pseudo-random thermal excitation are applied to the infrared thermal imager to collect temperature rise data and calculate thermal diffusivity and thermal inertia parameters to obtain an infrared feature set.

[0010] By fusing the terahertz feature set with the infrared feature set, a latent variable model is constructed to obtain the latent variable distribution of the hydration degree of UHPC;

[0011] By combining latent variable distribution with maintenance temperature information, an intensity prediction model is established to obtain the early intensity distribution;

[0012] Based on the early intensity distribution and combined with multi-source quality assessment indicators, the quality score is calculated, the retest area is determined, and construction suggestions are obtained through the early intensity distribution, quality score and retest area, and standardized quality control reports and equipment control instructions are generated.

[0013] After construction is completed, measured strength is collected, and the parameters of the strength prediction model and the latent variable model are updated based on the measured strength, mass fraction and risk classification information.

[0014] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual-mode described in this invention, the acquisition time zero point, emissivity parameters, and synchronization trigger signal form a unified calibration parameter set, and the specific steps are as follows:

[0015] A grid of detection areas was divided on the surface of the UHPC specimen, and a terahertz sensor, an infrared thermal imager, and a pseudo-random binary sequence heat source were placed above the grid of detection areas to collect time zero point, emissivity parameters, and synchronous trigger signals to obtain a unified calibration parameter set.

[0016] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual modes described in this invention, the steps for acquiring same-bias and cross-bias time-domain waveforms on the detection grid based on a unified calibration parameter set, performing frequency domain transformation and inversion processing, and obtaining the terahertz feature set are as follows:

[0017] Based on a unified calibration parameter set, the terahertz sensor is controlled to acquire co-polarized and cross-polarized terahertz time-domain waveforms, and frequency domain conversion and Fresnel inversion are performed to obtain the real part, imaginary part, loss tangent and group delay of the dielectric constant.

[0018] The polarization anisotropy index is obtained by the ratio of the energy difference to the energy sum of the co-polarized and cross-polarized terahertz time-domain waveforms. First-order compensation is then performed based on the polarization anisotropy index to obtain the terahertz feature set.

[0019] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual-mode described in this invention, the specific steps of applying sweep frequency phase-locking and pseudo-random thermal excitation to the infrared thermal imager based on a unified calibration parameter set are as follows:

[0020] Based on a unified calibration parameter set, the infrared thermal imager and the heat source are aligned in time through synchronous triggering, and the sampling frequency, phase-locked band and heat source output power are set at the same time.

[0021] The input power of the heat source is continuously modulated at a linearly increasing frequency. The frame sequence of the infrared thermal imager is acquired synchronously, and the frequency at each instant is demodulated by phase lock to obtain the amplitude and phase spectrum of the temperature response and the corresponding spatial distribution. Then, pseudo-random binary thermal excitation is superimposed.

[0022] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual-mode described in this invention, the specific steps for acquiring temperature rise data and calculating thermal diffusivity and thermal inertia parameters to obtain an infrared feature set are as follows:

[0023] Temperature rise data is collected, and thermal diffusivity and thermal inertia parameters are estimated grid by grid through cross-correlation deconvolution and multi-frequency phase fitting. The thermal diffusivity and thermal inertia parameters are then summarized to obtain an infrared feature set.

[0024] As a preferred embodiment of the terahertz-infrared dual-mode UHPC early intensity detection method described in this invention, the specific steps for fusing the terahertz feature set and the infrared feature set to construct a latent variable model and obtain the latent variable distribution of UHPC hydration degree are as follows:

[0025] Based on the time zero point and synchronous trigger signal of the unified calibration parameter set, the terahertz feature set and the infrared feature set are time-aligned. The infrared thermal imager pixels and the terahertz sensor measurement points are mapped to the detection area grid according to the calibration to achieve spatial alignment. Through normalization and noise suppression methods, the terahertz feature set and the infrared feature set are processed into feature input data with unified scale and accuracy.

[0026] Based on the feature input data, a latent variable model integrating physical mechanisms and data features is established to calculate the latent variable distribution of hydration degree of UHPC on each detection grid.

[0027] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual-mode described in this invention, the step of establishing an intensity prediction model by combining latent variable distribution with curing temperature information to obtain the early intensity distribution includes the following specific steps:

[0028] The latent variable distribution is mapped to the real-time collected maintenance temperature data in both spatial and temporal dimensions to form fused input data;

[0029] Based on the fused input data, an early intensity prediction model is constructed to calculate the early intensity distribution of each detection grid at different ages.

[0030] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual-mode described in this invention, the steps for calculating the mass score and determining the retest area based on the early intensity distribution and multi-source quality assessment indicators are as follows:

[0031] By combining early intensity distribution with multi-source quality assessment indicators, a comprehensive analysis of each detection grid is performed to calculate the corresponding quality score;

[0032] By using quality score and quality reliability thresholds, areas that need to be remeasured are identified, and the spatial location of the remeasurement areas is obtained.

[0033] As a preferred embodiment of the UHPC early intensity detection method based on terahertz-infrared dual-mode described in this invention, the steps for obtaining construction suggestions and generating standardized quality control reports and equipment control instructions based on early intensity distribution, mass fraction, and retest areas are as follows:

[0034] The early intensity distribution, quality score, and candidate spatial location of the retest area are compared with the construction quality control standard grid by grid to obtain the construction adjustment requirements, spatial location and quality score information of the retest area for each test grid. The early intensity distribution of each test grid at different ages is statistically summarized to obtain a standardized quality control report.

[0035] By using standardized quality control reports to adjust the curing temperature, spray pressure, and retesting schedule according to the construction needs, the equipment control instructions are transformed into executable equipment control commands in a parameterized form.

[0036] As a preferred embodiment of the terahertz-infrared dual-mode UHPC early strength detection method described in this invention, the following steps are taken: After construction is completed, measured strength is collected; based on the measured strength, mass fraction, and risk classification information, the parameters of the strength prediction model and the latent variable model are updated to achieve long-term adaptive detection.

[0037] After construction is completed, the measured strength data of each detection grid is collected, and the quality score and risk classification information are integrated to obtain a comprehensive quality assessment dataset.

[0038] Based on the comprehensive quality assessment dataset, the parameters of the intensity prediction model and the latent variable model are iteratively updated.

[0039] The beneficial effects of this invention are as follows: By constructing a latent variable model of hydration degree based on the fusion of terahertz and infrared features, the electromagnetic and thermophysical parameters of the material are synergistically mapped to the hydration process, achieving high-precision prediction of the early strength distribution of UHPC at different spatial locations and ages; by introducing multi-source quality assessment indicators and mass fraction calculation methods, the reliability of the prediction results is quantitatively assessed and the retest area is intelligently determined; by combining early strength distribution, mass fraction, and risk classification information to generate standardized quality control reports and equipment control instructions, automated quality control and long-term adaptive optimization of the construction process are achieved. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of an early intensity detection method for UHPC based on terahertz-infrared dual modes.

[0042] Figure 2 The flowchart for obtaining the terahertz feature set.

[0043] Figure 3 The flowchart for obtaining the infrared feature set.

[0044] Figure 4 A flowchart for integrating modeling, intensity prediction, and quality control output. Detailed Implementation

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0046] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0048] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for early intensity detection of UHPC based on terahertz-infrared dual-mode, including the following steps:

[0049] S1. Acquire zero-point time, emissivity parameters, and synchronous trigger signals to form a unified calibration parameter set.

[0050] A grid of detection areas was divided on the surface of the UHPC specimen, and a terahertz sensor, an infrared thermal imager, and a pseudo-random binary sequence heat source were placed above the grid of detection areas to collect time zero point, emissivity parameters, and synchronous trigger signals to obtain a unified calibration parameter set.

[0051] Furthermore, after setting up a detection grid on the surface of the UHPC specimen, the terahertz sensor, infrared thermal imager, and pseudo-random binary sequence heat source are precisely placed on the surface of the UHPC specimen using the position coordinates of the detection grid. This ensures that the spatial positions of the terahertz sensor, infrared thermal imager, and pseudo-random binary sequence heat source strictly correspond to the detection grid. After the terahertz sensor, infrared thermal imager, and pseudo-random binary sequence heat source are placed, the zero point of time is obtained based on the time trigger signal of the terahertz sensor, infrared thermal imager, and pseudo-random binary sequence heat source. The zero point of time is recorded with millisecond-level accuracy by a synchronous trigger controller.

[0052] The emissivity parameter of the data at the zero time point is calibrated using the temperature radiation measurement function of the infrared thermal imager. After aligning the acquired time series based on the zero time point, several stable temperature points are sequentially set on the standard blackbody temperature stage and the standard blackbody temperature series is recorded. At each stable temperature point, the ambient reflection temperature is kept constant and the brightness temperature series of the infrared thermal imager is acquired. The emissivity parameter is obtained by minimizing the mean square error between the measured brightness temperature and the standard blackbody temperature. Using the emissivity parameter and the zero time point, the synchronous trigger signal of the terahertz sensor, the infrared thermal imager, and the pseudo-random binary sequence heat source is used to trigger control under the same time reference, ensuring that all acquisition actions are performed under a unified time reference. Finally, a unified calibration parameter set is generated based on the zero time point, the emissivity parameter, and the synchronous trigger signal.

[0053] S2. Based on a unified calibration parameter set, collect time-domain waveforms of the same bias and cross bias in the detection grid, and perform frequency domain transformation and inversion processing to obtain the terahertz feature set.

[0054] Based on a unified calibration parameter set, the terahertz sensor is controlled to acquire co-polarized and cross-polarized terahertz time-domain waveforms, and frequency domain conversion and Fresnel inversion are performed to obtain the real part, imaginary part, loss tangent, and group delay of the dielectric constant.

[0055] Furthermore, by controlling the terahertz sensor to acquire co-polarized and cross-polarized terahertz time-domain waveforms on the detection grid through a unified calibration parameter set, a terahertz time-domain waveform dataset is obtained. The terahertz time-domain waveform dataset is then converted into a terahertz frequency-domain waveform dataset using the fast Fourier transform method. The frequency resolution is determined by the sampling time interval within the unified calibration parameter set. For example, when the sampling time interval is 1 nanosecond, the frequency resolution can reach 1 gigahertz. Based on the terahertz frequency-domain waveform dataset, the real part of the dielectric constant, the imaginary part of the dielectric constant, the loss tangent, and the group delay are calculated using the Fresnel inversion method.

[0056] First-order compensation based on polarization anisotropy index is used to obtain terahertz feature set.

[0057] Under the same detection grid and time conditions, the polarization anisotropy index is obtained by the ratio of the energy difference to the energy sum of the same-polarization terahertz time-domain waveform and the cross-polarization terahertz time-domain waveform. The polarization anisotropy index is used to perform first-order compensation for the real part of the dielectric constant, the imaginary part of the dielectric constant, the loss tangent, and the group delay. First, it is determined whether the value of the polarization anisotropy index exceeds the polarization anisotropy index compensation trigger threshold of 0.2. When the value of the polarization anisotropy index is greater than 0.2, the error changes from negligible to significant, and the polarization anisotropy index is denoted as... For the real part of the dielectric constant and the imaginary part of the dielectric constant Apply a linear correction factor, expressed as:

[0058] ;

[0059] in, This represents the real part of the corrected dielectric constant. This represents the imaginary part of the corrected dielectric constant. The scaling factor represents the correction coefficient.

[0060] It should be noted that the polarization anisotropy index compensation trigger threshold of 0.2 is derived from the statistical results of terahertz time-domain polarization response experiments on multiple UHPC specimens. In the polarization anisotropy index distributions measured under different curing ages and moisture contents, when the polarization anisotropy index compensation trigger threshold is below 0.2, the deviations of the real and imaginary parts of the dielectric constant are both less than 5%, which can be considered as requiring no additional compensation. However, when the polarization anisotropy index compensation trigger threshold exceeds 0.2, the propagation differences of electromagnetic waves in the anisotropic direction are significantly amplified, and the calculation errors of the dielectric constant and group delay increase significantly. Therefore, the polarization anisotropy index compensation trigger threshold is set to 0.2. The method involves collecting the relative deviations between the real and imaginary parts of the dielectric constant and their reference values ​​under different polarization anisotropy indices, fitting the linear relationship between the relative deviation and the portion of the polarization anisotropy index exceeding a threshold, and taking the slope of the fitted line as the scaling factor of the linear correction coefficient.

[0061] After correcting the real and imaginary parts of the dielectric constant, the loss tangent is recalculated using the corrected real and imaginary parts of the dielectric constant. Group latency The expression is:

[0062] ;

[0063] ;

[0064] in, This represents the path length of a terahertz wave propagating through a material. It represents the speed of light.

[0065] The real part of the dielectric constant, the imaginary part of the dielectric constant, the loss tangent, and the group delay after first-order compensation of the polarization anisotropy index are uniformly constructed into a terahertz feature set.

[0066] S3. Apply sweep frequency phase-locked loop and pseudo-random binary thermal excitation to the infrared thermal imager based on a unified calibration parameter set, collect temperature rise data and calculate thermal diffusivity and thermal inertia parameters to obtain the infrared feature set.

[0067] Based on a unified calibration parameter set, the infrared thermal imager and the heat source are aligned in time through synchronous triggering, and parameters such as sampling frequency, phase-locked band, and heat source output power are set simultaneously.

[0068] Furthermore, by using a unified calibration parameter set to synchronize the trigger signals of the infrared thermal imager and the heat source, both respond and output time synchronization results under the same time reference. Based on the time zero point and the synchronization trigger signal in the unified calibration parameter set, the frame-by-frame sampling rate of the infrared thermal imager is set to obtain the sampling frequency setting result. For example, the sampling frequency can be set to 60 frames per second. By analyzing the periodic characteristics of the synchronization trigger signal, frequency reference information is obtained. Using the frequency reference information, the phase detection bandwidth of the infrared thermal imager is limited within a set range to obtain the phase-locked band setting result. For example, the phase-locked band can be set to 0.05 to 0.5 Hz. The thermal flux response amplitude of the infrared thermal imager under initial excitation is calculated using the emissivity parameter to obtain the initial thermal flux calibration data. Using the initial thermal flux calibration data, the power amplitude output by the heat source is adjusted to obtain the heat source output power setting result. For example, the heat source output power can be set to 100 watts, thereby completing the unified setting of the sampling frequency, phase-locked band, and heat source output power.

[0069] It should be noted that the sampling frequency is set to 60 frames per second because the infrared thermal imager can capture the details of temperature changes over time without motion blur at this sampling frequency. If the sampling frequency is too low, the temporal information of temperature changes will be lost, affecting the accuracy of the phase-locked loop results. If the sampling frequency is too high, a large amount of redundant data will be generated, increasing the storage and computational burden. The phase-locked loop bandwidth is set to 0.05 to 0.5 Hz because this bandwidth can ensure that the amplitude and phase information of the thermal diffusion process at different frequencies are sufficiently separated, which is convenient for subsequent feature extraction. If the bandwidth is too narrow, it may not be able to cover the complete frequency characteristics of the material under different thermal responses. If the bandwidth is too wide, it may introduce additional noise in the low signal-to-noise ratio band. The heat source output power is set to 100 watts because at the sampling frequency, a sufficiently high signal-to-noise ratio can be obtained while ensuring that the material surface is not overheated. If the sampling frequency is too low, the infrared thermal imager may not be able to detect a sufficiently significant temperature rise change. If the power is too high, it may cause local overheating of the material or generate nonlinear thermal effects, affecting the accuracy of feature calculation.

[0070] A linear sweep frequency phase-locked signal is applied to obtain the temperature response at different frequencies, and then a pseudo-random binary thermal excitation is superimposed to enhance the time-domain information.

[0071] Furthermore, based on the sampling frequency setting results, phase-locked band setting results, and heat source output power setting results, a linear sweep frequency point sequence is first generated according to the phase-locked band setting results (e.g., 10 points are uniformly selected from 0.05 to 0.50 Hz). Then, the dwell time for each frequency point is calculated according to the sampling frequency setting results (e.g., each point is sampled for ≥3 whole cycles, dwell time = 3 / frequency point). The heating drive amplitude and duty cycle are calculated according to the heat source output power setting results. Subsequently, the heat source is started with a synchronous trigger signal of a unified calibration parameter set, so that the heat source outputs a sinusoidal power modulation signal at each frequency point in the order of preheating stage, stable excitation stage, and frequency point switching stage. At the same time, the infrared thermal imager records the temperature frame sequence according to the sampling frequency setting results and marks the current frequency point index. The time stamp is used to immediately transition to the next frequency point after each frequency point ends until the frequency sweep is completed. During this period, phase-locked demodulation is performed on the temperature frame sequence to obtain the amplitude and phase, and the data is collected in the order of frequency points to form a linear frequency sweep temperature response dataset. While maintaining the application of the linear frequency sweep phase-locked signal, the time interval and sequence length of the pseudo-random binary sequence are determined according to the integer multiple of the sampling period of the sampling frequency setting result and the number of periods to be covered by the lowest frequency in the phase-locked frequency band setting result. The amplitude ratio of the superimposed excitation is set according to the heat source output power setting result, so that the pseudo-random binary thermal excitation and the linear frequency sweep phase-locked signal are applied to the heat source synchronously in time. The temperature response data after the enhancement of time domain information is collected to obtain the temperature response dataset superimposed with pseudo-random binary thermal excitation.

[0072] Temperature rise data is collected, and thermal diffusivity and thermal inertia parameters are obtained through cross-correlation deconvolution and multi-frequency phase fitting, thus acquiring an infrared feature set.

[0073] Furthermore, temperature rise data was collected by superimposing a pseudo-random binary thermal excitation temperature response dataset. The impulse response information of the heat conduction process was extracted based on the cross-correlation deconvolution method, and the cross-correlation deconvolution results were obtained. Phase delay curves at different frequencies were extracted from the cross-correlation deconvolution results to obtain a multi-frequency phase delay dataset. Based on the analysis of one-dimensional heat conduction theory in the frequency domain, the thermal diffusivity, which represents the diffusion rate of temperature change inside the material, and the thermal inertia parameter, which represents the temperature hysteresis characteristics of the material under the action of heat input, were fitted by minimizing the mean square error between the measured phase delay and the theoretical phase delay. The thermal diffusivity and the thermal inertia parameter were combined to obtain an infrared feature set.

[0074] S4. Fuse the terahertz feature set with the infrared feature set to construct a latent variable model and obtain the latent variable distribution of the hydration degree of UHPC.

[0075] The terahertz feature set and the infrared feature set are aligned in time and space, and the terahertz feature set and the infrared feature set are processed into feature input data with uniform scale and accuracy through normalization and noise suppression methods.

[0076] Furthermore, the terahertz and infrared feature sets are synchronized and aligned in the time dimension using timestamp information, and the time alignment result is output. The time alignment result is then combined with the spatial coordinate information of the terahertz and infrared feature sets to complete spatial registration in the spatial dimension and obtain the temporal-spatial alignment result. A normalization method is used to unify the numerical range of the terahertz and infrared feature sets to between 0 and 1 to obtain the normalized terahertz and infrared feature sets. Finally, a noise suppression method is used to filter the random noise in the normalized terahertz and infrared feature sets to obtain feature input data with uniform scale and accuracy.

[0077] Based on the feature input data, a latent variable model integrating physical mechanisms and data features is established to calculate the latent variable distribution of hydration degree of UHPC on each detection grid.

[0078] Furthermore, within a unified spatiotemporal grid, the a priori hydration degree calculated from the hydration kinetic equation is combined with the thermal diffusivity and thermal inertia parameters from the infrared feature set. A weighted mapping method is used to obtain the latent variable of the hydration degree. The hydration degree is then obtained using the Arrhenius maturity and S-type hydration law, expressed as:

[0079] ;

[0080] in, Indicates time The dynamic prior degree of hydration at time t, Represents the hydration rate coefficient. Indicates the curing temperature. Indicates the temperature correction factor. Indicates the reaction order.

[0081] It should be noted that, The value range of is [0,1]. =0 corresponds to the initial unhydrated state. =1 corresponds to the fully hydrated state. It was obtained through isothermal calorimetry experiments or fitting of early hydration exothermic curves. It is obtained by fitting early hydration experimental data and is used to adjust the curve shape and control the characteristics of the acceleration and deceleration phases of the hydration process. The temperature data of the UHPC is collected by a temperature sensor. The actual temperature of the environment at any given time is used to determine the temperature correction factor, which is determined by combining the difference between the maintenance temperature and the reference temperature with the activation energy and the gas constant.

[0082] The a priori hydration degree of the dynamics is calculated point-by-point within a predetermined time range to form a complete time series curve, thus obtaining the time prior trajectory of the hydration degree. The real part of the dielectric constant and the thermal diffusivity collected from the terahertz sensor and infrared thermal imager are used as observations to construct a monotonic function mapping between the latent variables and the observations, expressed as:

[0083] ;

[0084] ;

[0085] in, Indicates the latent variable in the degree of hydration The real part of the complex permittivity is given by [the following]. Represents the solid-state equivalent dielectric constant. This represents the equivalent dielectric constant of water in a saturated state within the pores. Indicates pore saturation. Indicates the latent variable in the degree of hydration Thermal diffusivity This indicates the lower limit of thermal diffusivity for early-age UHPC specimens. This indicates the upper limit of thermal diffusivity for mature UHPC specimens. This represents the thermal diffusivity growth coefficient, with a value range of [0.1, 1].

[0086] It should be noted that, This was obtained through calibration measurements of a fully hydrated UHPC specimen in the terahertz frequency band. This method involves observing the transmission waveform of a liquid water sample against a reference waveform under terahertz time-domain spectroscopy, extracting amplitude and phase information, and then using the Fresnel inversion method to obtain the real part of the complex permittivity as the equivalent permittivity. This indicates that when the latent variable of hydration degree is The volume fraction of liquid water in the pores was obtained through an exponential decay method. It was obtained through infrared thermal measurements at an early age of UHPC. It was obtained through infrared thermal measurement during the long-term UHPC period. It is a latent variable controlling the thermal diffusivity with the degree of hydration. The growth rate was obtained by nonlinear fitting of discrete measurements of the thermal diffusivity of UHPC specimens at different ages. When <0.1, the thermal diffusivity varies with the degree of hydration. The growth trend is too gradual to reflect the enhanced heat conduction process in the early hydration stage. When the value is greater than 1, the thermal diffusivity growth curve is too steep, leading to overestimation in the early stage and rapid saturation in the later stage, which does not match the actual thermal conductivity of UHPC.

[0087] By minimizing the weighted deviation between the kinetic prior hydration degree, the real part of the dielectric constant, and the thermal diffusivity, the optimal solution of the latent variable of hydration degree is obtained. The optimal solution of the latent variable of hydration degree is then mapped to the time prior trajectory of hydration degree, the real and imaginary parts of the terahertz dielectric constant, the thermal diffusivity, and the thermal inertia parameters on a unified spatial grid and time axis, forming a joint feature input that integrates kinetic, thermal, and electrical information. Based on the joint feature input, a latent variable model that integrates physical mechanisms and data features is constructed. The latent variable model simultaneously includes physical constraint terms of the hydration kinetic equation, fitting terms of terahertz and infrared observation data, and spatial-temporal smoothing regularization terms, so that the latent variable model can take into account the reliability of observation data and spatiotemporal continuity while ensuring physical consistency.

[0088] The a priori degree of hydration given by the hydration kinetic equation As initial latent variables for hydration degree, the real part spectrum of dielectric constant, bandpass average thermal diffusivity, and water state index from the feature input data (compensated by the first-order polarization anisotropy index) are input into the latent variable model according to a unified spatial grid and time axis. An objective function incorporating physical and data consistency is constructed, and the distribution of the hydration degree latent variable in each detection grid of the UHPC is solved within interval constraints. The expression is:

[0089]

[0090] in, Indicates the latent variable in the degree of hydration Under constraints, the weighted deviation between the latent variable model output and the observed data, Represents the feature input observation vector, Represents a physically consistent observation prediction vector. This represents the evidence quality weight matrix. Represents spatial gradient, This represents the first-order difference between adjacent ages. Represents the sum of squares of the positive parts. Indicates the intensity of controlling spatial smoothness. This indicates the intensity of age-related monotonicity.

[0091] It should be noted that, It is obtained by band-averaging the real part of the dielectric constant, the bandpass average thermal diffusivity, and the water state index, which are compensated for by the first order of polarization anisotropy index, and splicing them together with a unified grid and time axis. It is obtained by mapping the hydration kinetic equations and the thermoelectric properties of materials to a space of the same dimension as the characteristic input observation vector, and then predicting the dielectric constant, thermal diffusivity, and water state index component by component. At each detection grid and time point, three quality assessment indicators—prediction uncertainty, mechanistic residual, and signal-to-noise ratio—are first extracted from each component of the feature input observation vector. Then, the prediction uncertainty, mechanistic residual, and signal-to-noise ratio are mapped to positive quality scores between 0 and 1 using an exponential decay function and a normalization function, respectively. It was obtained by calculating the discrete first derivatives in the horizontal and vertical directions of the latent variable of hydration degree on a unified spatial grid using the central difference method. and The optimal value is obtained by performing parameter sensitivity analysis on representative specimens to achieve the best balance between smoothness, monotonicity, and observational fitting accuracy of the latent variable distribution of hydration degree.

[0092] By using interval constraints Minimize The distribution of latent variables of hydration degree in UHPC on each detection grid was obtained.

[0093] S5. By combining the latent variable distribution with the maintenance temperature information, an intensity prediction model is established to obtain the early intensity distribution.

[0094] The latent variable distribution is mapped to the real-time collected maintenance temperature data in both spatial and temporal dimensions to form fused input data.

[0095] Furthermore, the latent variable distribution of hydration degree is synchronized with the timestamp information of the real-time collected maintenance temperature data in the time dimension to obtain the time synchronization registration result. The spatial coordinate information of the latent variable distribution of hydration degree and the real-time collected maintenance temperature data is spatially mapped in the spatial dimension to obtain the spatial mapping result. The latent variable distribution of hydration degree and maintenance temperature data are registered point by point on the time axis according to a unified age sequence. Finally, a unified fusion input dataset is obtained in both spatial and temporal dimensions.

[0096] Based on the fused input data, an early intensity prediction model is constructed to calculate the early intensity distribution of each detection grid at different ages.

[0097] Furthermore, the latent variable distribution of hydration degree in the fused input data is registered with the curing temperature data according to a unified spatial grid and time axis. At each detection grid and age point, curing temperature data is first collected in real time by a curing temperature sensor. Substitute it into the temperature correction factor The maturity index is obtained by integrating time according to age, and is expressed as:

[0098] ;

[0099] in, Indicates age period The maturity index within, It is an integral variable, representing any moment within the age period.

[0100] Will Linearly normalize to the [0,1] interval to obtain the maturity normalization value. An early intensity prediction model is constructed based on the hydration kinetic equation and maturity, and its expression is:

[0101] ;

[0102] in, Indicates spatial location and age The predicted early intensity Indicates the maximum compressive strength. Represents the distribution of latent variables indicating the degree of hydration. Represents the balance parameters. Represents the shape factor.

[0103] It should be noted that, It is obtained by conducting compressive strength tests on UHPC standard specimens at long-term ages (e.g., 28 days or 56 days) and taking the average of the results from multiple sets of specimens. It is obtained by fitting measured early intensity data at different ages with the early intensity prediction formula and solving it using the least squares method. The intensity growth curves at different ages were fitted with shape characteristics and obtained from the steepness of the curves using a nonlinear regression method.

[0104] S6. Based on the early intensity distribution and combined with multi-source quality assessment indicators, calculate the quality score and determine the retest area.

[0105] By combining early intensity distribution with multi-source quality assessment indicators, a comprehensive analysis of each detection grid is performed to calculate the corresponding quality score.

[0106] Furthermore, by combining early intensity distribution with prediction uncertainty in multi-source quality assessment indicators... Mechanism residual and signal-to-noise ratio information Synchronous registration is performed on both the spatial grid and temporal dimensions, and the synchronization registration results are output. Based on the synchronization registration results, the comprehensive quality score of each detection grid at different times is calculated using a quality gating factor, and the comprehensive quality score result is output. The expression is:

[0107] ;

[0108] in, Indicates the detection grid and age The overall quality score is as follows: It is the minimum of the three positive scores for prediction uncertainty, mechanistic residual, and signal-to-noise ratio, representing the quality level of the most unfavorable observation index. , It is the median of the three positive scores for prediction uncertainty, mechanistic residuals, and signal-to-noise ratio, representing the quality level of typical observation indicators. , This represents the positive score corresponding to the prediction uncertainty. , This represents the positive score corresponding to the mechanism residual. , This represents the positive score corresponding to the signal-to-noise ratio. , This represents the variance of the three positive scores: prediction uncertainty, mechanistic residual, and signal-to-noise ratio. This represents the conservatism sensitivity coefficient.

[0109] It should be noted that, By first calculating different values ​​at representative detection grids and age points... The error between the overall quality score and the actual quality label or subsequent verification results is then considered. Within the feasible interval where the discrepancy is greater than 0, the optimal approach is chosen by minimizing the validation error or striking a balance between making the predictions more conservative when the discrepancy is large and fully utilizing the information when the discrepancy is small. As a sensitivity coefficient for conservatism.

[0110] The overall quality score is weighted and normalized according to the spatial grid to obtain the quality score corresponding to each detection grid.

[0111] By using quality score and quality reliability thresholds, areas that need to be remeasured are identified, and the spatial location of the remeasurement areas is obtained.

[0112] Furthermore, the quality score is compared with a pre-set quality confidence threshold grid by grid, all detection grids with quality scores lower than the quality confidence threshold are marked, and the positions of the detection grids with quality scores lower than the quality confidence threshold are extracted in spatial coordinates to obtain the spatial location of the retest area.

[0113] It should be noted that the quality reliability judgment threshold is determined by statistically analyzing the distribution characteristics of all quality scores on a representative detection grid, and combining this with the results of subsequent actual strength testing or construction quality acceptance. The receiver operating characteristic curve method is used to select the critical value that achieves the best balance between the false negative rate and the false negative rate, which is then used as the quality reliability judgment threshold.

[0114] S7. Based on early intensity distribution, mass fraction, and retesting areas, obtain construction suggestions and generate standardized quality control reports and equipment control instructions.

[0115] After completing the calculation of early strength distribution, mass fraction, and identification of retest areas, the construction adjustment requirements for each test grid are obtained by combining the construction quality control standards. Based on the spatial location and mass fraction information of the retest areas, the early strength distribution of each test grid at different ages is statistically summarized, generating a standardized quality control report that includes early strength distribution curves, mass fraction statistics tables, and spatial location maps of the retest areas. According to the construction adjustment requirements, the curing temperature adjustment, spray pressure control, and retest plan arrangement are converted into executable equipment control instructions in parameterized form through the standardized quality control report, enabling the construction equipment to automatically adjust the temperature, pressure, and curing time according to the early strength distribution and mass fraction of different test grids.

[0116] S8. After construction is completed, the measured strength is collected. Based on the measured strength, quality score and risk classification information, the parameters of the strength prediction model and the latent variable model are updated to achieve long-term adaptive detection.

[0117] After construction is completed, the measured strength data of each detection grid is collected, and the quality score and risk classification information are integrated to obtain a comprehensive quality assessment dataset.

[0118] Furthermore, non-destructive testing equipment such as ultrasonic rebound hammers and ground-penetrating radar were deployed at each testing grid location. Surface and internal strength parameters of each testing grid at different ages were collected point-by-point according to a unified spatial grid sequence. Damage compressive strength tests were conducted on selected representative testing grids using standard cubic or cylindrical core drilling methods to obtain measured compressive strength values. The strength parameters obtained from non-destructive testing were regressed and calibrated with the compressive strength values ​​from the damage tests to generate measured strength data for each testing grid after construction. The measured strength data was then integrated with the quality scores obtained during construction for each testing grid in a unified time and spatial dimension to obtain integrated strength-quality results. These results were then used to determine the strength and quality of each testing grid. The grid location compares the early strength distribution with the measured strength data after construction. The strength margin index is obtained by comparing the ratio of the difference between the measured strength data after construction and the early strength distribution to the measured strength data. The strength margin index and the quality score are uniformly normalized. Combined with the construction quality standards, a graded judgment rule is adopted: below the first threshold is judged as high risk, between the two thresholds is judged as medium risk, and above the second threshold is judged as low risk. The risk level results of all detection grids are summarized according to spatial grid and age to obtain risk classification information. The integrated strength quality results are integrated with the risk classification information to obtain a comprehensive quality assessment dataset.

[0119] It should be noted that the first and second thresholds are determined by collecting measured strength data, early strength distribution, and quality scores, statistically analyzing their relationship with actual engineering quality accidents or performance degradation events, and using receiver operating characteristic curves. Under the condition of ensuring a balance between the false negative rate and the false positive rate, two critical values ​​that can optimally distinguish between high risk, medium risk, and low risk are determined. The critical values ​​are adjusted in accordance with the requirements of engineering safety specifications, for example, by shifting them to the safety side with a certain margin, to form the first and second thresholds for risk classification and determination.

[0120] Based on the comprehensive quality assessment dataset, the parameters of the intensity prediction model and the latent variable model are iteratively updated.

[0121] Furthermore, the prediction error between the current parameters of the intensity prediction model and the comprehensive quality assessment dataset is obtained through the comprehensive quality assessment dataset. The prediction error is then used to update the parameters of the intensity prediction model using the gradient descent method, and the updated intensity prediction model parameter results are obtained. The current parameters of the latent variable model are compared with the distribution of latent variables of hydration degree in the comprehensive quality assessment dataset to obtain the latent variable deviation of the current parameters of the latent variable model at each detection grid and age. The prediction error and the latent variable deviation are weighted and summed in the spatial and temporal dimensions according to a unified weight coefficient to obtain the joint error result. Based on the joint error result, the latent variable model parameters are updated through an iterative optimization method for early intensity prediction and quality assessment in subsequent cycles.

[0122] This embodiment also provides a computer device applicable to the early intensity detection method of UHPC based on terahertz-infrared dual mode, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the early intensity detection method of UHPC based on terahertz-infrared dual mode as proposed in the above embodiment.

[0123] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0124] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the UHPC early intensity detection method based on terahertz-infrared dual-mode as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0125] In summary, this invention achieves high-precision prediction of early intensity distribution of UHPC at different spatial locations and ages by constructing a latent variable model of hydration degree based on the fusion of terahertz and infrared features as input data; it also achieves quantitative evaluation of the reliability of prediction results and intelligent determination of retest areas by introducing multi-source quality assessment indicators and quality score calculation methods; and it realizes automated quality control and long-term adaptive optimization of the construction process by combining early intensity distribution, quality score, and risk classification information to generate standardized quality control reports and equipment control instructions.

[0126] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting early intensity of UHPC based on terahertz-infrared dual modes, characterized in that: include, The zero-point acquisition time, emissivity parameters, and synchronization trigger signals are used to form a unified calibration parameter set. Based on a unified calibration parameter set, time-domain waveforms of the same polarization and cross polarization are collected in the detection grid and frequency-domain transformation and inversion processing are performed to obtain the terahertz feature set. Based on a unified calibration parameter set, a sweep frequency phase-locked loop and pseudo-random thermal excitation are applied to the infrared thermal imager to collect temperature rise data and calculate thermal diffusivity and thermal inertia parameters to obtain an infrared feature set. By fusing the terahertz feature set with the infrared feature set, a latent variable model is constructed to obtain the latent variable distribution of the hydration degree of UHPC; By combining latent variable distribution with maintenance temperature information, an intensity prediction model is established to obtain the early intensity distribution; Based on the early intensity distribution and combined with multi-source quality assessment indicators, the quality score is calculated, the retest area is determined, and construction suggestions are obtained through the early intensity distribution, quality score and retest area, and standardized quality control reports and equipment control instructions are generated. After construction is completed, the measured strength is collected, and the parameters of the strength prediction model and the latent variable model are updated based on the measured strength, mass fraction and risk classification information. The specific steps for fusing the terahertz feature set and the infrared feature set to construct a latent variable model and obtain the latent variable distribution of the hydration degree of UHPC are as follows: based on the time zero point and synchronization trigger signal of the unified calibration parameter set, the terahertz feature set and the infrared feature set are time-aligned; the infrared thermal imager pixels and the terahertz sensor measurement points are mapped to the detection area grid according to the calibration to achieve spatial alignment; and the terahertz feature set and the infrared feature set are processed into feature input data with unified scale and accuracy through normalization and noise suppression methods. On a unified time axis and detection area grid, the a priori degree of hydration is calculated using the hydration kinetic equation. Based on the feature input data, the dynamic prior hydration degree is used as a prior constraint, and the real part of the dielectric constant and thermal diffusivity are used as observations to construct a latent variable model consisting of physical constraint terms, observation fitting terms, and spatiotemporal smoothing terms. By minimizing the weighted bias between the kinetic prior hydration degree and the observed values, the optimal solution for the latent variable of hydration degree corresponding to each detection grid is obtained, thus acquiring the distribution of the latent variable of hydration degree of UHPC on each detection grid.

2. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 1, characterized in that: The acquisition time zero point, emissivity parameters, and synchronization trigger signal are used to form a unified calibration parameter set. The specific steps are as follows: A grid of detection areas was divided on the surface of the UHPC specimen, and a terahertz sensor, an infrared thermal imager, and a pseudo-random binary sequence heat source were placed above the grid of detection areas to collect time zero point, emissivity parameters, and synchronous trigger signals to obtain a unified calibration parameter set.

3. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 2, characterized in that: The process involves acquiring time-domain waveforms of the same-bias and cross-bias types on a detection grid based on a unified calibration parameter set, performing frequency domain transformation and inversion processing, and obtaining a terahertz feature set. The specific steps are as follows: Based on a unified calibration parameter set, the terahertz sensor is controlled to acquire co-polarized and cross-polarized terahertz time-domain waveforms, and frequency domain conversion and Fresnel inversion are performed to obtain the real part, imaginary part, loss tangent and group delay of the dielectric constant. The polarization anisotropy index is obtained by the ratio of the energy difference to the energy sum of the co-polarized and cross-polarized terahertz time-domain waveforms. First-order compensation is then performed based on the polarization anisotropy index to obtain the terahertz feature set.

4. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 3, characterized in that: The specific steps for applying swept-frequency phase-locked loop and pseudo-random thermal excitation to the infrared thermal imager based on a unified calibration parameter set are as follows: Based on a unified calibration parameter set, the infrared thermal imager and the heat source are aligned in time through synchronous triggering, and the sampling frequency, phase-locked band and heat source output power are set at the same time. The input power of the heat source is continuously modulated at a linearly increasing frequency. The frame sequence of the infrared thermal imager is acquired synchronously, and the frequency at each instant is demodulated by phase lock to obtain the amplitude and phase spectrum of the temperature response and the corresponding spatial distribution. Then, pseudo-random binary thermal excitation is superimposed.

5. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 4, characterized in that: The specific steps for collecting temperature rise data, calculating thermal diffusivity and thermal inertia parameters, and obtaining an infrared feature set are as follows: Temperature rise data is collected, and thermal diffusivity and thermal inertia parameters are estimated grid by grid through cross-correlation deconvolution and multi-frequency phase fitting. The thermal diffusivity and thermal inertia parameters are then summarized to obtain an infrared feature set.

6. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 5, characterized in that: The method involves establishing an intensity prediction model by combining latent variable distribution with maintenance temperature information to obtain early intensity distribution. The specific steps are as follows: The latent variable distribution of hydration degree corresponding to each detection grid and the real-time collected maintenance temperature data are registered in the spatial and temporal dimensions to obtain fused input data; At each detection grid and age point, a maturity index is calculated based on the fused input data, and the maturity index is normalized to obtain the maturity normalization value. Using the latent variable distribution of hydration degree and maturity normalization as independent variables, and the maximum compressive strength, equilibrium parameter and shape factor as model parameters, an early strength prediction model is constructed, and the predicted early strength value is calculated for each detection grid and each age period. The predicted early intensity values ​​of all detection grids at different ages are combined to form an early intensity distribution.

7. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 6, characterized in that: The process involves calculating a quality score based on early intensity distribution and combining it with multi-source quality assessment indicators to determine the retest area. The specific steps are as follows: The prediction uncertainty, mechanism residual, and signal-to-noise ratio of each detection grid at different ages are extracted and mapped to obtain positive scores. The overall quality score is calculated based on the positive scores of the same detection grid at the same age. The comprehensive quality scores of each detection grid at different ages are weighted and normalized according to the spatial grid to obtain the quality score corresponding to each detection grid. The distribution characteristics of all quality scores are statistically analyzed, and the quality reliability judgment threshold is determined based on the characteristic curve of the distribution characteristics. By using quality score and quality reliability thresholds, areas that need to be remeasured are identified, and the spatial location of the remeasurement areas is obtained.

8. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 7, characterized in that: The process involves obtaining construction recommendations based on early intensity distribution, mass fraction, and retested areas, and generating standardized quality control reports and equipment control instructions. The specific steps are as follows: The early intensity distribution, quality score, and candidate spatial location of the retest area are compared with the construction quality control standard grid by grid to obtain the construction adjustment requirements, spatial location and quality score information of the retest area for each test grid. The early intensity distribution of each test grid at different ages is statistically summarized to obtain a standardized quality control report. By using standardized quality control reports to adjust the curing temperature, spray pressure, and retesting schedule according to the construction needs, the equipment control instructions are transformed into executable equipment control commands in a parameterized form.

9. The UHPC early intensity detection method based on terahertz-infrared dual-mode as described in claim 8, characterized in that: After the construction is completed, measured strength is collected. Based on the measured strength, quality score, and risk classification information, the parameters of the strength prediction model and the latent variable model are updated to achieve long-term adaptive monitoring. The specific steps are as follows: After construction is completed, the measured strength data of each detection grid is collected, and the quality score and risk classification information are integrated to obtain a comprehensive quality assessment dataset. Based on the comprehensive quality assessment dataset, the parameters of the intensity prediction model and the latent variable model are iteratively updated.

Images