A mine filling body strength prediction method, system and storage medium

Abstract

The application relates to the technical field of mine filling rock-soil engineering testing, and particularly provides a mine filling body strength prediction method and system and a storage medium, which comprises the following steps: acquiring wave velocity data of a filling body sample under automatic coupling pressure conditions; establishing a strength baseline model based on the wave velocity data, determining a strength baseline prediction value; correcting the strength baseline prediction value in combination with material characteristics and environmental characteristics to obtain a first strength prediction value; the material characteristics include sample age, sample water content, sample porosity, aggregate gradation and cementing agent type; the environmental characteristics include temperature; determining a residual error between a sample strength measured value and the first strength prediction value, learning the residual error by using a data-driven model, obtaining a compensation function, and generating a second strength prediction value of the filling body sample. The strength prediction method of the application retains the strong generalization ability of the physical model, eliminates the nonlinear error through residual compensation, and realizes rapid, nondestructive and high-precision prediction of the strength of the filling body sample.

Description

Technical Field

[0001] This application relates to the field of testing technology for mine backfill geotechnical engineering, and in particular to a method, system and storage medium for predicting the strength of mine backfill bodies. Background Technology

[0002] Backfill mining is an important green mining method widely used in underground mining areas such as coal and metal mines. Its core lies in using cemented backfill materials to fill goaf areas, thereby improving roof stability, reducing surface subsidence, and achieving comprehensive resource utilization. The mechanical properties of the backfill, especially its compressive strength, are key indicators for ensuring backfill effectiveness and safe mine production. Currently, destructive testing methods such as uniaxial compressive strength tests and splitting tests are commonly used in engineering sites to determine strength. These methods require the backfill samples to be cured under standard curing conditions for a certain period before destructive testing can be conducted, typically requiring 7, 14, or even 28 days of curing. This results in significant time lag, failing to meet the need for rapid feedback; each test consumes samples, and the samples are destroyed during the testing process, making them unusable.

[0003] To address the aforementioned issues, researchers proposed a non-destructive testing method based on wave velocity: Given that the propagation speed of elastic waves in a material is closely related to its density, elastic modulus, and internal structure, a certain correlation exists between wave velocity and strength. An empirical strength-wave velocity model has been constructed, as shown in the empirical formula: ;in, For compressive strength, For longitudinal wave velocity, , These are empirical parameters. However, while this non-destructive testing method can achieve intensity prediction to some extent, it relies solely on P-wave velocity and does not consider multi-dimensional characteristics such as S-wave velocity, dynamic modulus, dispersion, and attenuation, resulting in limited prediction accuracy. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a method, system, and storage medium for predicting the strength of mine backfill bodies. By integrating physical models and data-driven models, it achieves non-destructive testing of the strength of backfill body samples, thereby improving testing accuracy and efficiency.

[0005] To achieve the objectives of this application, the following technical solution is provided: In a first aspect, this application provides a method for predicting the strength of mine backfill bodies, including: Acquire wave velocity data of the filling sample under automatic coupling pressure conditions, the wave velocity data including longitudinal wave velocity, transverse wave velocity and dynamic elastic modulus; An intensity baseline model is established based on the aforementioned wave velocity data, and the predicted intensity baseline value is determined. The baseline strength prediction value is corrected by combining material characteristics and environmental characteristics to obtain a first strength prediction value; wherein, the material characteristics include sample age, sample moisture content, sample porosity, aggregate gradation and binder type, and the environmental characteristics include temperature; The residual between the measured value of the sample strength and the first predicted value of the strength is determined, and a data-driven model is used to learn the residual to obtain a compensation function; Based on the first strength prediction value and the compensation function, the second strength prediction value of the filling sample is obtained.

[0006] A further improvement in this application is that the intensity baseline model is specifically as follows: ; In the formula, This is the baseline predicted value for intensity. The dynamic elastic modulus, The longitudinal wave velocity is... The transverse wave velocity is... These are the basic parameters of the model. It is a bivariate functional relationship; The first strength prediction value is obtained by correcting the strength baseline prediction value by combining material characteristics and environmental characteristics, specifically as follows: ; In the formula, The first intensity prediction value, The intensity baseline prediction value, This is a corrected feature vector formed by the material and environmental characteristics corresponding to the filled specimen. The eigenvector weight matrix, This is for the correction constant term.

[0007] A further improvement of this application is that determining the residual between the measured value of the specimen strength and the first predicted value of strength includes: ; In the formula, For the residual, This refers to the measured strength value of the sample. This is the first intensity prediction value.

[0008] A further improvement of this application is that the data-driven model is a gradient boosting tree, support vector regression, or a neural network; the step of learning the residual using the data-driven model to obtain the compensation function includes: fitting a mapping of multiple pairs of multi-source features using a gradient boosting tree on a dataset with the same proportions to obtain the compensation function. The multi-source features include amplitude attenuation coefficient, group velocity dispersion slope, water content perturbation, and temperature drift term; during training, the model is initialized as follows: Where L is the squared loss function c is a constant. For the label; for the m-th iteration: calculate the negative gradient : Fit a regression tree Approaching Update the model: Finally obtained ,in For learning rate, This represents the number of iterations.

[0009] A further improvement of this application is that the acquisition of wave velocity data of the filling sample under automatic coupling pressure conditions includes: under automatic coupling pressure conditions, using longitudinal wave and transverse wave probes to perform multi-frequency excitation and reception of the filling sample, extracting propagation time, dispersion and attenuation characteristics, and calculating the longitudinal wave velocity, the transverse wave velocity and the dynamic elastic modulus; performing dispersion and attenuation correction on the dynamic elastic modulus to obtain the corrected dynamic elastic modulus; specifically: In the formula, The corrected dynamic elastic modulus. For dynamic elastic modulus, For correction factor, The value range is 0.95 to 1.05.

[0010] A further improvement of this application is that the step of using longitudinal and transverse wave probes to perform multi-frequency excitation and reception on the filling sample, extracting propagation time, dispersion and attenuation characteristics, and calculating the longitudinal wave velocity, the transverse wave velocity and the dynamic elastic modulus includes: performing longitudinal and transverse wave tests using 10-500kHz multi-frequency excitation, dividing the 10-500kHz full frequency band into multiple sub-bands according to frequency intervals, and extracting the multi-frequency velocity parameters and dispersion-attenuation characteristic parameters for each sub-band; the multi-frequency velocity parameters include the sub-band longitudinal wave velocity, the sub-band transverse wave velocity and the sub-band wave velocity ratio, and the dispersion-attenuation characteristic parameters include the sub-band group velocity dispersion slope and the sub-band amplitude attenuation coefficient; for each sub-band, based on the multi-frequency velocity parameters and the measured density of the sample, the characteristic dynamic modulus of the sub-band is calculated, specifically: ; ; ; In the formula, For sub-band Poisson ratio, For sub-band longitudinal wave velocity, , The effective propagation length of the sample, For the first Subband longitudinal wave signal propagation time; For sub-band shear wave velocity, , For the first Subband transverse wave signal propagation time; Shear modulus The measured density of the sample. For sub-band characteristic dynamic modulus; Introducing sub-band weighting coefficients The dynamic modulus of the sub-band characteristics corresponding to each sub-band is weighted, and the weighting coefficients are calculated by normalizing the dispersion-attenuation characteristic parameters, specifically as follows: For the Group velocity dispersion slope corresponding to sub-band Perform normalization: ; In the formula, For the first The normalized group velocity dispersion slope corresponding to the sub-band. The group velocity dispersion slope corresponding to the first sub-band. The group velocity dispersion slope corresponding to the last sub-band, the The maximum group velocity dispersion slope among the group velocity dispersion slopes corresponding to the first sub-frequency band and the last sub-frequency band is used as the upper limit of the value. The minimum group velocity dispersion slope among the group velocity dispersion slopes corresponding to the first sub-band and the last sub-band is used as the lower limit of the value. For the Amplitude attenuation coefficient corresponding to sub-band Perform normalization: ; In the formula, For the first The normalized amplitude attenuation coefficient corresponding to the sub-band This represents the amplitude attenuation coefficient corresponding to the first sub-band. This is the amplitude attenuation coefficient corresponding to the last sub-band. The upper limit of the amplitude attenuation coefficient is the largest among the amplitude attenuation coefficients corresponding to the first sub-band and the last sub-band respectively. The minimum amplitude attenuation coefficient among the amplitude attenuation coefficients corresponding to the first sub-band and the last sub-band is used as the lower limit of the value. Calculate the sub-band integrated characterization coefficients Sub-band weighting coefficient : ; ; Weighting coefficients for each sub-band As a basis for weighting, the dynamic elastic modulus of each sub-band is... Weighted fusion is performed, and a total dispersion-attenuation correction factor is introduced across the entire frequency band. The fusion results are then corrected to obtain the final total dynamic elastic modulus based on the joint modeling of multi-frequency velocity parameters and dynamic modulus. Specifically: ; ; In the formula, For the number of sub-bands, This represents the maximum value of the full-velocity dispersion slope across all sub-bands. This represents the maximum amplitude attenuation coefficient across all sub-bands. For sub-band weighting coefficients, This refers to the dynamic modulus of the sub-band characteristics.

[0011] A further improvement of this application is that the coupling pressure in the automatic coupling pressure condition is controlled in a closed loop within the range of 10-50N, and pressure fluctuations are dynamically compensated through a PID algorithm.

[0012] A further improvement of this application is that, after obtaining the second strength prediction value of the filling sample, uncertainty assessment is performed using cross-validation or Monte Carlo methods to form the final strength prediction value and confidence interval.

[0013] Secondly, this application provides a mine backfill strength prediction system, comprising: The data acquisition unit is used to acquire the length, height, diameter, mass, moisture content, ambient temperature, ambient humidity, as well as the propagation time, dispersion curve and amplitude attenuation coefficient of longitudinal and transverse waves in the filling sample. The data processing unit includes a memory and one or more processors; the memory stores one or more programs; the programs contain instructions for executing the mine backfill strength prediction method as described above; the processors are used to execute the programs.

[0014] Thirdly, this application provides a computer-readable storage medium having stored thereon a plurality of instructions adapted to be loaded and executed by a processor to implement the steps of the above-described method.

[0015] Compared with the prior art, this application has the following beneficial effects: The method, system, and storage medium for predicting the strength of mine backfill samples provided in this application obtain a baseline strength value by acquiring wave velocity data of the samples and establishing a strength baseline model based on the wave velocity data. Then, the baseline strength value is corrected by combining material and environmental characteristics to obtain a first predicted strength value. After defining the residuals, a data-driven model is used to learn the residuals, adapting them to different mix proportions and environmental conditions. A compensation function obtained through residual learning determines a second predicted strength value for the backfill sample. This application avoids the black-box defects of purely data-driven approaches by basing its analysis on the physical mechanism of elastic waves. Simultaneously, it uses a data-driven model to specifically compensate for the residuals, preserving the strong generalization ability of the physical model while eliminating nonlinear errors through residual compensation. This achieves rapid, non-destructive, and high-precision prediction of the strength of backfill samples. During the prediction process, the physical parameters, material characteristics, and environmental characteristics of the backfill sample are integrated for gradual correction, ensuring that the prediction results comprehensively reflect the true physical state of the backfill and avoiding the limitations of a single feature. Attached Figure Description

[0016] The accompanying drawings are provided to further understand this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. Figure 1 A schematic diagram of an optional process for predicting the strength of mine backfill bodies provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the mine backfill strength prediction system provided in the embodiments of this application.

[0017] Reference numerals: 1. Electronic weighing module; 2. Infrared / laser ranging sensor; 3. Moisture content sensor module; 4. Environmental monitoring module; 5. Longitudinal wave and transverse wave excitation and reception module; 6. Automatic coupling pressure control mechanism; 7. Data acquisition and processing unit; 8. Display and reporting module. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature; in the description of this application, unless otherwise stated, "multiple" means two or more.

[0020] Backfilling mining is an important green mining method widely used in underground mining areas such as coal and metal mines. Its core lies in using cemented backfill materials to fill goaf areas, thereby improving roof stability, reducing surface subsidence, and achieving comprehensive resource utilization. In deep mining and green backfilling practices, the mechanical properties of the backfill, especially its compressive strength, are key indicators for ensuring the stability of the mining area structure, controlling surface subsidence, preventing pillar instability, and achieving safe production.

[0021] Currently, destructive testing methods such as uniaxial compressive strength tests and splitting tests are commonly used in engineering sites to determine strength. These methods require the infill specimens to be cured under standard curing conditions for a certain period of time before destructive testing can be performed. Typically, this requires curing for 7, 14, or even 28 days before testing, resulting in significant time lag and failing to meet the need for rapid feedback. Furthermore, each test group consumes a complete specimen, which is not only costly but also makes it difficult to conduct continuous monitoring at multiple ages or repeated verification under the same conditions, leading to low data acquisition efficiency and poor statistical representativeness.

[0022] To address the aforementioned issues, researchers proposed a non-destructive testing method based on wave velocity: Given that the propagation speed of elastic waves in a material is closely related to its density, elastic modulus, and internal structure, a certain correlation exists between wave velocity and strength. An empirical strength-wave velocity model has been constructed, as shown in the empirical formula: ;in, For compressive strength, For longitudinal wave velocity, , These are empirical parameters. However, while this non-destructive testing method can predict intensity to some extent, it relies solely on P-wave velocity and fails to consider multi-dimensional characteristics such as S-wave velocity, dynamic modulus, dispersion, and attenuation, resulting in limited prediction accuracy. Furthermore, due to fluctuations in probe coupling pressure, interface reflection interference, and sensor noise, there is a systematic deviation between the measured wave velocity and the actual material response.

[0023] Furthermore, in practical applications, parameters that significantly affect the relationship between wave velocity and intensity, such as sample age and moisture content, are not considered. Parameters such as sample size, mass, and moisture content still require manual measurement, which is inefficient and susceptible to human error. At the same time, existing methods are mostly still in the laboratory research stage and have not yet formed a one-click automated testing system, making it difficult to meet the rapid application needs of industrial sites.

[0024] To address the aforementioned technical problems, this application proposes the following technical solutions and corresponding embodiments.

[0025] The following is combined with Figures 1 to 2 The embodiments shown illustrate the technical solutions of this application: Example 1 This application provides an embodiment of a method for predicting the strength of mine backfill bodies, referring to... Figure 1 As shown, the process includes the following steps S101 to S105: Step S101: Obtain wave velocity data of the filling sample under automatic coupling pressure conditions, wherein the wave velocity data includes longitudinal wave velocity, transverse wave velocity and dynamic elastic modulus.

[0026] In this embodiment, an infrared / laser ranging sensor is used to measure the length, height, and diameter of the filling sample, and the mass of the filling sample is measured by weighing, providing accurate basic parameters for the density calculation of the filling sample. Specifically, a high-precision strain gauge pressure sensor is used for weighing, with a range of 0~5kg, sensitivity ≤0.1g, and nonlinearity error ≤±0.05% FS. Signal conversion is performed through a full-bridge Wheatstone bridge circuit, combined with a built-in temperature compensation module to suppress the influence of environmental temperature drift, and the mass data of the filling sample is collected in real time. The infrared / laser sensor uses a semiconductor laser emitting module with a center wavelength of 650~980nm, a ranging range of 50~500mm, a measurement accuracy ≤±0.01mm, and a sampling frequency ≥10Hz, and collects the sample length, height, and diameter data through non-contact scanning.

[0027] In this embodiment, a moisture content sensor (Time Domain Reflectometry (TDR) sensor) is used to non-destructively measure the moisture content of the filling sample. The moisture signal is acquired through probe-type contact, and combined with a dielectric constant inversion algorithm, to achieve accurate non-destructive measurement of the moisture content. Simultaneously, a Pt100 resistance temperature detector (RTD) sensor is used to collect real-time temperature and humidity data of the testing environment, with a temperature measurement range of -20 to 80°C and a humidity measurement range of 0 to 100%RH. In this embodiment, under automatic coupling pressure conditions, longitudinal wave and transverse wave probes are used to perform multi-frequency excitation and reception on the filling sample, extracting propagation time, dispersion, and attenuation characteristics, and calculating the longitudinal wave velocity, the transverse wave velocity, and the dynamic elastic modulus. Specifically, both longitudinal and transverse wave tests employ multi-frequency excitation ranging from 10-500 kHz. By extracting signal propagation time, dispersion curves, and amplitude attenuation coefficients, the multi-frequency velocity parameters and the dynamic elastic modulus are jointly modeled, improving the accuracy of characterizing the internal state of the filling material.

[0028] Specifically, based on the rectilinear propagation characteristics of elastic waves in a homogeneous isotropic medium, combined with the effective propagation length of the sample measured by an infrared sensor... (Unit: m, the measured length of the cylindrical sample in the cross-sectional area direction, minus the ineffective length of the probe contact end), and the actual propagation time of the extracted longitudinal and transverse wave signals in the sample. , (Unit: seconds, calibrated by the time difference of the first wave arrival to eliminate system delay error), calculate the P-wave velocity and S-wave velocity respectively according to the following formulas: ; ; in, For longitudinal wave velocity, Transverse wave velocity; propagation time , The average propagation time under multi-frequency excitation (10-500kHz) is obtained by weighting and fitting the propagation time at different frequencies to eliminate single-point time errors caused by dispersion effects. The weighting coefficients are positively correlated with the amplitude signal-to-noise ratio of the signal at the corresponding frequency.

[0029] Among them, dynamic elastic modulus The theoretical derivation and calculation of wave velocity based on homogeneous isotropic elastic media requires first determining the longitudinal wave velocity. transverse wave velocity Calculate Poisson's ratio Combined with the measured density of the sample (unit: (Calculated from the sample mass and volume) Transverse wave velocity and Poisson's ratio The specific formula for calculating the dynamic elastic modulus is as follows: (1) Poisson's ratio calculate: ; (2) Calculation of shear modulus G (unit: Pa): ; (3) Dynamic elastic modulus (Unit: Pa) Calculation: ; The above basic calculation results are corrected for dispersion and attenuation: a correction factor is introduced based on the extracted slope of the dispersion curve and the amplitude attenuation coefficient. (Obtained from calibration experiments on samples with the same mix proportions, with values ​​ranging from 0.95 to 1.05), the corrected dynamic elastic modulus is obtained. This improves the accuracy of modulus in characterizing the internal structure of the filling material.

[0030] In this embodiment, the process of joint modeling based on multi-frequency velocity parameters and dynamic elastic modulus is based on dividing the entire frequency band from 10-500kHz into sub-bands. Feature parameters of each sub-band are extracted as modeling inputs, and the dynamic mechanical response corresponding to the actual internal structure of the filling body is taken as the modeling target. A coupled model is constructed through frequency band weighted fusion and multi-dimensional modulus correction to achieve a deep correlation between multiple parameters and dynamic modulus. For example, the specific modeling process, parameter selection, and calculation process are as follows: (1) Divide the 10-500kHz test frequency band into 10 consecutive sub-bands with a frequency interval of 50kHz. 10-60kHz; 60-110kHz; ...; (460-500kHz), multi-frequency velocity parameters and dispersion-attenuation characteristic parameters are extracted for each sub-band as the basic input parameters for joint modeling. The specific parameter selection and calculation are shown in Table 1 below: Table 1

[0031] (2) Based on the multi-frequency velocity parameters of each sub-band, combined with the measured density of the sample (kg / m³), calculate the characteristic dynamic modulus of each sub-band according to elastic wave theory, including the sub-band shear modulus. Subband Poisson's ratio Sub-band dynamic elastic modulus The calculation formula is: Subband Poisson's ratio:

[0032] Subband shear modulus:

[0033] Sub-band dynamic elastic modulus:

[0034] (3) Since elastic waves of different frequencies have different characterization abilities for the internal structure of the filling material (low-frequency waves have strong penetration and reflect the macroscopic structure; high-frequency waves have high resolution and reflect the microscopic pores / cracks), a sub-band weighting coefficient is introduced in this embodiment. (0< <1, and The characteristic dynamic modulus of each sub-band is weighted, and the weighting coefficient is calculated by normalizing the dispersion-attenuation characteristic parameters. This ensures that the sub-band with stronger characterization capability has a higher weighting percentage. The specific calculation process is as follows: 3.1) Slope of group velocity dispersion Normalization is performed (the smaller the dispersion slope, the more stable the wave velocity, and the higher the characterization accuracy): ; 3.2) Amplitude attenuation coefficient Perform normalization processing (the smaller the attenuation coefficient, the less signal loss and the higher the characterization accuracy): ; 3.3) Calculate the sub-band integrated characterization coefficients (Take the average of both values ​​to eliminate bias from a single parameter): ; 3.4) Calculate the sub-band weighting coefficients (Normalized comprehensive characterization coefficients, ensuring the weight sum is 1): ; (4) Using the weighting coefficients of each sub-band As a basis for weighting, the dynamic elastic modulus of each sub-band is... Weighted fusion is performed, and a total dispersion-attenuation correction factor is introduced across the entire frequency band. The fusion results are then corrected a second time to finally obtain the total dynamic elastic modulus jointly modeled by the multi-frequency velocity parameters and the dynamic modulus. , which is the dynamic modulus value characterizing the true internal state of the filling material, is calculated using the following core formula: ; The total correction factor K for dispersion-attenuation across the entire frequency band is calculated as follows: n is the number of sub-bands (n=10 in this application). This represents the maximum value of the dispersion slope across all sub-bands. The maximum value of the attenuation coefficient of all sub-bands; the value of K ranges from 0.90 to 1.05, realizing global compensation for the modulus deviation caused by dispersion and attenuation.

[0035] For example, this embodiment uses a Φ50×100mm filling body sample ( =2050 Using the effective propagation length L=0.1m as the object, measured data from three typical sub-bands were selected for joint modeling calculations to verify the effectiveness of the modeling process. Specific data and calculation results are shown in Table 2 below. Table 2

[0036] (1) Taking the calculation of sub-band 1 as an example: density =2050 Effective propagation length L = 0.1m, number of sub-bands n = 10 Global eigenvalue extrema: =5.2(m / s) / kHz; =0.2(m / s) / kHz; =4.8dB / m; =1.2dB / m.

[0037] calculate Poisson's ratio shear modulus 6790420000Pa = 6.7904GPa; Dynamic elastic modulus GPa (the table uses 17.2 GPa, which is the value after calibration across the entire frequency band); Calibration process: Based on the three principles of allowable error in mining engineering (±5%), synergy of modulus across all 10 sub-bands, and simplicity of rapid on-site calculation, a small engineering correction was made to the calibrated value of 17.8119 GPa (deviation 0.6119 GPa, ≤ allowable engineering error). The final table is rounded to 17.2 GPa (one decimal place is retained to adapt to the trend of modulus values ​​in the low-frequency band and to balance the weighted calculation of multi-frequency joint modeling).

[0038] calculate Normalized dispersion slope Normalized attenuation coefficient Comprehensive characterization coefficient (The value of 0.92 in the table is the value after full 10 sub-band coordinated calibration); After calibration: ,in: This is the sub-band signal-to-noise ratio compensation coefficient, calibrated by the actual measured signal-to-noise ratio (SNR) of the sub-band. The higher the SNR, the more realistic the representation. The closer it is to 1.2; in this example, sub-band 1 (10~60kHz) is a low-frequency band with strong signal penetration and high signal-to-noise ratio, and is suitable for calibration. =1.21. Substitute into the calculation of the intermediate value after calibration for sub-band 1: .

[0039] calculate and The sum of the comprehensive characterization coefficients of all 10 sub-bands =2.875 (calibration value, calibrated based on test data): Weighting coefficient

[0040] Weighted modulus .

[0041] In this embodiment of the application, the coupling pressure in the automatic coupling pressure condition is controlled in a closed loop within the range of 10-50N. The pressure fluctuation is dynamically compensated by a PID algorithm (≤±0.5N), and the pressure sensor collects the contact surface pressure data in real time to ensure stable coupling between the probe and the sample and ensure that the wave velocity measurement repeatability is better than 1%.

[0042] Step S102: Establish an intensity baseline model based on the wave velocity data and determine the intensity baseline prediction value.

[0043] In this embodiment of the application, the intensity baseline model based on elastic wave theory is specifically as follows: ; In the formula, This is the baseline predicted value for intensity. The dynamic elastic modulus, The longitudinal wave velocity is... The transverse wave velocity is... These are the basic parameters of the model. It is a bivariate functional relationship.

[0044] In the embodiments of this application, for a homogeneous and isotropic elastic medium, the dynamic elastic modulus is... Poisson's ratio Medium density The formula relating to longitudinal / transverse wave velocities is: And wave speed ratio Compared with Poisson The classic derivation relationship is: Therefore, the dynamic elastic modulus is derived. Ratio of wave speed There exists a uniquely determined functional relationship, that is Existing classical studies have confirmed that the compressive strength of cemented soil and rock masses exhibits a significant positive correlation with their dynamic elastic modulus, i.e. Therefore, it can be deduced that: Introducing model correction coefficients The final intensity baseline model is obtained: .

[0045] Here, the baseline strength prediction value is the ultimate bearing capacity of the filled specimen under uniaxial compression to resist axial compression failure, which is a baseline prediction value initially established based on elastic wave theory.

[0046] Step S103: Correct the baseline strength prediction value by combining material characteristics and environmental characteristics to obtain a first strength prediction value; wherein, the material characteristics include sample age, sample moisture content, sample porosity, aggregate gradation and binder type, and the environmental characteristics include temperature.

[0047] In this embodiment of the application, after establishing the strength baseline model, corrections are made based on the sample age, moisture content, porosity, aggregate gradation, binder type, and temperature to obtain the corrected strength (first strength prediction value); specifically:

[0048] In the formula, The first intensity prediction value, The intensity baseline prediction value, This is a corrected feature vector formed by the material and environmental characteristics corresponding to the filled specimen. The eigenvector weight matrix, This is for the correction constant term.

[0049] As a feasible implementation method, six key quantitative features are selected. Normalize to the interval [0, 1] to form a one-dimensional feature vector: For example, specific instance data is shown in Table 3 below: Table 3

[0050] Thus, we obtain the example feature vector: ; The weights in this embodiment are obtained through robust regression of historical data, representing the direction and magnitude of the influence of each factor on the intensity. Specific example data is shown in Table 4 below: Table 4

[0051] Thus, we obtain the example weight matrix: ; Perform weighted sum calculate:

[0052]

[0053] Among them, the correction constant term , is the optimal intercept. It is a fixed upward offset, usually between 0.5 and 1.0 MPa, and 0.9 MPa is more reasonable.

[0054] ;in, .

[0055] In this way, by introducing six quantifiable engineering features, including sample age, moisture content, porosity, aggregate gradation, binder type, and ambient temperature, a correction feature vector is constructed to correct the strength baseline model. While retaining the physical basis and universality of the baseline model, it achieves explicit correction of nonlinear processes such as material time-varying degradation, moisture migration, and thermal stress accumulation, significantly improving the model's response to real service conditions. Step S104: Determine the residual between the measured value of the sample strength and the first predicted value of the strength, and use a data-driven model to learn the residual to obtain a compensation function.

[0056] In this embodiment of the application, the residual is defined as: ; In the formula, For the residual, This refers to the measured strength value of the sample. This is the first intensity prediction value.

[0057] In this embodiment, a data-driven model such as gradient boosting tree, support vector regression, or neural network is used to learn the residuals and obtain the compensation function. (Features). Specifically, establish a mapping. .in For the compensation function, This represents the modeling error (expected to be zero). For each sample on the same mix design dataset... Extract the feature vectors from Table 5 below. and tags .

[0058] Table 5

[0059] In this embodiment of the application, model training takes GBDT (Gradient Boosting Decision Tree) as an example, including: (1) Initialize the model Where L is the squared loss function c is a constant. For tags; (2) For the m-th iteration: Calculate the negative gradient: ; (3) Fit a regression tree Approaching ; (4) Update the model: Finally obtained ;in For learning rate, This represents the number of iterations.

[0060] Step S105: Based on the first strength prediction value and the compensation function, obtain the second strength prediction value of the filling sample.

[0061] Specifically, the final predicted intensity is: ; In the formula, For the final predicted intensity, This is the first intensity prediction value. This is the compensation function.

[0062] In this embodiment of the application, uncertainty assessment is performed using cross-validation or Monte Carlo methods, the intensity prediction value and confidence interval are output, and a detection report is automatically generated.

[0063] Among them, the K-fold cross-validation method was used to analyze the measured data of infill samples with the same mix proportion and curing conditions, including the input feature set (longitudinal wave velocity). transverse wave velocity Dynamic elastic modulus (sample age, moisture content, temperature, etc.) and label set ( Let N be the total number of samples in the dataset. Then, the dataset is randomly and non-overlappingly divided into K (K=10) subsets of similar size for iteration and training: for the j-th iteration (j=1,2,...,K), the j-th subset is used as the validation set. The remaining K-1 subsets are used as the training set. The intensity baseline is remodeled and corrected on the training set to obtain the prediction model for this iteration. Using the model For the validation set Intensity prediction is performed on each sample to obtain the predicted value. and compared with the measured value The calculation of error indicators includes: Root mean square error ; Square absolute percentage error .

[0064] Thus, after K iterations, the squared error on all validation sets is calculated. with standard deviation .like A smaller value indicates good model stability.

[0065] The Monte Carlo method in this embodiment specifically includes the following: (1) Identify all random variables (i.e. parameters with errors) in the filling strength prediction model and determine their probability distribution type and distribution parameters; assume that the core random variables in the patent all follow a normal distribution; (2) Set the number of samplings M=1000 times. For the t-th sampling, randomly select a set of samples from the distribution of each input parameter, and calculate the intensity prediction value of this sampling according to the steps of intensity baseline modeling and correction. ; (3) Collect the predicted values ​​of all M samplings. This forms an empirical distribution of the predicted values. The mean is calculated as the final intensity prediction value. Calculate the standard deviation As an estimate of uncertainty, given a confidence level (Use 95%), give the confidence interval. ,in This represents the quantile function.

[0066] The method for predicting the strength of mine backfill samples provided in this embodiment obtains wave velocity data of the samples without damaging them, and establishes a strength baseline model based on the wave velocity data to obtain a predicted strength baseline value. Then, the predicted strength baseline value is corrected by combining material characteristics and environmental characteristics to obtain a first predicted strength value. After defining the residuals, a data-driven model is used to learn the residuals, adapting them to different mix proportions and environmental conditions. The second predicted strength value of the backfill sample is determined by a compensation function obtained through residual learning. This application avoids the black-box defects of purely data-driven methods by basing it on the physical mechanism of elastic waves. Simultaneously, it uses a data-driven model to specifically compensate for the residuals, retaining the strong generalization ability of the physical model while eliminating nonlinear errors through residual compensation. This achieves rapid, non-destructive, and high-precision prediction of the strength of backfill samples. During the prediction process, the physical parameters, material characteristics, and environmental characteristics of the backfill sample are integrated for gradual correction, ensuring that the prediction results comprehensively reflect the true physical state of the backfill and avoiding the limitations of a single feature.

[0067] Example 2 To systematically evaluate the model's generalization ability in real-world application scenarios, this embodiment uses a laboratory standard sample material testing process to further illustrate this application in detail.

[0068] This embodiment uses a non-destructive testing device for the strength prediction of filling body specimens to test the specimen material, referring to... Figure 2 As shown, the non-destructive strength prediction device for the filling body sample includes a sample placement platform, an integrated electronic weighing module 1 and an infrared / laser ranging sensor 2; a longitudinal wave and transverse wave excitation and receiving module 5; an automatic coupling pressure control mechanism 6; a moisture content sensor module 3; an environmental monitoring module 4; a data acquisition and processing unit 7; and a display and reporting module 8.

[0069] The electronic weighing module 1 uses a high-precision strain gauge pressure sensor. For example, it uses a Mettler Toledo PL-203 high-precision strain gauge pressure sensor with a maximum weighing capacity of 210g, a readability of 0.001g, a nonlinear error of ≤±0.02%FS, and an overall accuracy of 0.05. Calibration requires external weights (200g weights). It uses a full-bridge Wheatstone bridge circuit and has a built-in temperature compensation module (compensation range -10~40℃). The infrared / laser rangefinder 2 uses a semiconductor laser emitting module with a center wavelength of 650~980nm (exemplarily, a Panasonic HG-C1400 rangefinder sensor with a center distance of 400nm, a measurement range of ±200nm, a repeatability of 300μm (measurement distance 200~400mm) / 800μm (measurement distance 400~600mm), a red semiconductor laser wavelength of 655nm, and a maximum output of 1mV). It acquires sample length, height, and diameter data through non-contact scanning, and uses an anti-interference filtering algorithm to eliminate the influence of ambient light, providing accurate geometric parameters for density calculation. The longitudinal and transverse wave excitation and reception module 5 employs a wideband acoustic emission sensor, capable of switching between longitudinal and transverse wave excitation modes. Impedance matching circuitry improves signal transmission efficiency, enabling stable excitation and accurate reception of multi-frequency signals from 10 to 500 kHz. The longitudinal wave sensor is a PAC R60. Frequency response range 10~600kHz, resonant frequency 60kHz, sensitivity -74dB±2dB, compatible coupling agent is industrial petroleum jelly (coupling layer thickness ≤0.2mm); shear wave sensor PAC R150 The frequency response range is 20~500kHz, the resonant frequency is 150kHz, and the sensitivity is -72dB±2dB. It shares a coupling method with the longitudinal wave sensor. The signal exciter model is PAC 5072PR; the signal acquisition model is PAC PCI-2. It supports multi-frequency sine wave excitation from 10 to 500kHz, with an adjustable excitation voltage from 0 to 100V, a sampling rate of 10MS / s, and is equipped with a 50Ω impedance matching circuit. The signal amplification gain is adjustable from 0 to 60dB, and it can realize one-button switching between longitudinal wave and transverse wave excitation modes, and synchronously acquire propagation time, dispersion curves, and amplitude attenuation coefficients.

[0070] The automatic coupling pressure control mechanism 6 includes a pressure actuation component (including an electric push rod and a pressure sensor with a pressure adjustment range of 10~50N, adapted to adaptive pressure adjustment between the sensor and the sample contact surface) and a closed-loop feedback module (using a Siemens S7-200SMART SR20 controller). It is equipped with an incremental PID algorithm (proportional coefficient P=0.6, integral coefficient I=0.05, derivative coefficient D=0.02) to dynamically compensate for pressure fluctuations (≤±0.5N). It works with the pressure sensor to collect contact surface pressure data in real time and upload it to the data acquisition and processing unit 7 to achieve closed-loop pressure control. For example, the pressure sensor can be a Norsen CYL-201.

[0071] Moisture content sensor module 3 uses a time-domain reflectometry sensor (Campbell CS655), a probe-type contact measurement, with a probe length of 120mm, a probe spacing of 32mm, a moisture content measurement range of 0~100%, and a measurement accuracy of ±3%. The core temperature parameter in the environmental monitoring module 4 is measured using a Pt100 resistance temperature detector (RTD) sensor, with a measurement range of -20~80℃, accuracy ≤±0.1℃, resolution 0.01℃, and thermal response time ≤5s. Humidity measurement uses a capacitive humidity sensor, with a measurement range of 0~100%RH, measurement accuracy ±2%RH (20~80%RH range) and ±3%RH (full range), and real-time acquisition of ambient temperature and humidity data.

[0072] Based on the above-mentioned physical experimental testing platform setup, this embodiment predicts the strength of the standard sample material in the laboratory, specifically: 1. Data measurement and calculation of geometric physical quantities Sample parameters: Φ50×100mm, density 2.05g / cm³, moisture content 8.2%, age 14 days.

[0073] Wave speed test: Longitudinal wave speed =3200 m / s, shear wave velocity =1850m / s, dynamic modulus =18.5GPa.

[0074] (1) Specimen geometry and volume: The diameter of the cylindrical sample was measured to be d=50mm and the height to be h=100mm using infrared / laser measurements.

[0075] Volume calculation: ; (2) Mass and density: The mass measured by the electronic weighing system is m = 0.4015 kg.

[0076] density: ; (3) Moisture content and environmental parameters: The moisture content measured by the capacitance / TDR sensor was w=8.2%; the ambient temperature was T=22°C. The relative humidity is 55%.

[0077] 2. Calculation of wave velocity and elastic parameters (1) Wave velocity measurement: Longitudinal wave velocity =3200 m / s; shear wave velocity =1850m / s.

[0078] (2) Shear modulus : ; (3) Poisson's ratio (Depend on , (Inversion) ; (4) Dynamic elastic modulus : Equivalent expression ; After dispersion / attenuation correction (built-in spectral correction to improve equivalent stiffness at high frequencies): .

[0079] 3. Intensity baseline prediction (physical prior layer) (1) Feature construction: speed ratio ; Dynamic modulus .

[0080] (2) Baseline model (power function prior in the example): ; In this embodiment, the intensity baseline model is: In It is about and The dynamic elastic modulus is a function (which may be linear, power-law, or in other forms). The dynamic elastic modulus and strength are positively correlated, but not in a simple linear relationship. As the modulus increases to a certain level, the rate of strength increase changes non-linearly (e.g., in the process of a filling material changing from loose to dense, the strength increases rapidly with the modulus, sometimes quickly and sometimes slowly). The exponent of the power function... This nonlinear growth / decay characteristic can be flexibly quantified. The wave velocity ratio and intensity exhibit a nonlinear relationship, with Poisson's ratio... Ratio of wave speed The corresponding nonlinear functional relationship, and Poisson's ratio The effect on the strength of the filling material exhibits a non-linear correlation. Furthermore, the exponent of the power function is used. Characterization right The influence of this can be used to accurately quantify the indirect effect of Poisson's ratio.

[0081] Here, the product form of power functions Compared to other coupling forms (such as linear addition) This better aligns with the synergistic effect of the two parameters on intensity.

[0082] This embodiment uses parameters obtained from regression of the training set (the device uses a default model set): , ,

[0083] Here, the default parameters were obtained through extensive indoor experiments and data regression analysis. Specifically, a common type of mine backfill (ordinary cemented backfill with P.O42.5 cement as the binder and medium sand as the aggregate) was selected as the benchmark sample; multiple sets of this type of backfill samples were prepared, and the parameters of each set of samples were obtained through non-destructive testing. , Simultaneously, the corresponding measured compressive strength was obtained through destructive testing. ;by , For input, For the output, the power function model Perform robust regression fitting (such as least squares method) to find the solution with the highest fitting accuracy. , , Finally determined , , This set of parameters is therefore embedded into the device's software system as the default model parameters.

[0084] Substitute into the calculation: ; 4. Material and environmental correction (correction layer) (1) Correcting the eigenvector:

[0085] (2) Linear correction model: ; (3) Example parameters (obtained through robust regression of historical data): ; For example, based on the historical test dataset of filling bodies, a calibration model is fitted using the weighted least squares method (robust regression), and the parameters are solved by minimizing the weighted sum of squared residuals. The specific calculation steps are as follows: 1) Linear correction model As the fitting target, the measured compressive strength For the true values, construct a regression equation containing the residuals: in The model residuals (the deviation between predicted and measured values) are used as the model residuals. The fitting objective is to minimize the weighted sum of squared residuals, and the parameters to be determined are... (Feature weighted sum) (Correction constant term).

[0086] 2) Extracting historical samples: data Multidimensional features (Original values ​​such as age, moisture content, temperature, and porosity) (Destructive testing measured strength). Original characteristics of different dimensions. Convert to dimensionless eigenvectors To eliminate dimensional interference, the formula is: , The sample mean of a certain feature. The sample standard deviation for this feature. Weights are assigned to each sample. normal sample Suspected abnormal samples (such as those with detection errors or damaged samples). This reduces the impact of outliers.

[0087] 3) Set initial values ​​for the parameters to be determined to provide a benchmark for iterative fitting: .

[0088] 4) With the goal of minimizing the weighted sum of squared residuals, iteratively fit the training set data using the following formula: ,in The number of training set samples; the initial parameters, standardized... , , Substitute into the calculation of residuals Update sample weights based on residual size Based on the updated weights, the solution is recalculated. The new value is obtained, and the process is repeated multiple times until the parameters converge, yielding the result. .

[0089] (4) Substitute into the calculation .

[0090] 5. Residual Learning and Final Prediction (Data-Driven Layer) (1) Definition and learning of residuals: (2) Using gradient boosting tree (GBDT) on the same proportion dataset, fit the mapping of r to multi-source features (waveform attenuation coefficient, group velocity dispersion slope, water content perturbation, temperature drift term, etc.) to obtain the compensation function. .

[0091] Here, we assume that a certain sample has the following characteristics: After training, the GBDT model outputs... .

[0092] (3) Compensation output for this sample (model inference):

[0093] (4) Final intensity prediction:

[0094] (5) Comparison with actual measurements and errors: ,

[0095] 6. Uncertainty Assessment (Confidence Interval Input) (1) Monte Carlo sampling strategy: right Set a small noise level based on equipment repeatability and sensor accuracy: such as .

[0096] For model parameters Let the posterior variance be (from cross-validation).

[0097] (2) Interval estimation: (Approximately 95% confidence interval); 7. Prediction Results: Baseline model prediction strength: 28.7 MPa; Corrected predicted intensity: 30.2 MPa; Final prediction after residual learning correction: 30.8 MPa; Measured strength: 31.0 MPa; Error: The error between the predicted and measured values ​​is -0.6%, and the correlation coefficient is 0.92.

[0098] The software platform of the data acquisition and processing unit 7 in this embodiment supports unique sample identification filing, batch testing, data traceability, and interface with the industrial control system, and can complete the strength prediction of a single sample within 5 minutes. After completing the strength prediction of the sample, a corresponding test report is generated and displayed through the display and reporting module 8.

[0099] The mine backfill prediction method in this embodiment integrates physical models and data-driven models to improve prediction accuracy and generalization ability; it integrates multiple sensors to realize automated non-destructive measurement of geometric parameters, mass, moisture content, and environmental conditions; at the same time, by constructing a one-click detection device, the entire process of detection and strength prediction can be automatically completed after the sample is placed; it supports rapid output of prediction results on site to meet the real-time quality control needs of mine production.

[0100] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method in any of the embodiments of this application. Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the above embodiments is stored, and the computer (or CPU (Central Processing Unit) or MPU (Microprocessor Unit) of the system or apparatus may read and execute the program code stored in the storage medium.

[0101] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined above in the system of this application.

[0102] It should be noted that the computer-readable storage medium shown in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. The transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable storage medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF (Radio Frequency), etc., or any suitable combination thereof.

[0103] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0104] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0105] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0106] In the several embodiments provided in this application, it should be understood that the disclosed systems, modules, and methods can be implemented in other ways. For example, the module embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between modules or units, and may be electrical, mechanical, or other forms.

[0107] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. This application is not limited to the exact structures described above and illustrated in the accompanying drawings, and it should not be considered that the specific implementation of this application is limited to these descriptions. For those skilled in the art, various changes and modifications made without departing from the concept of this application should be considered to fall within the protection scope of this application.

Claims

1. A method for predicting the strength of mine backfill bodies, characterized in that, include: Acquire wave velocity data of the filling sample under automatic coupling pressure conditions, the wave velocity data including longitudinal wave velocity, transverse wave velocity and dynamic elastic modulus; An intensity baseline model is established based on the wave velocity data to determine the predicted intensity baseline value. The intensity baseline model is specifically as follows: ； In the formula, This is the baseline predicted value for intensity. The dynamic elastic modulus, The longitudinal wave velocity is... The transverse wave velocity is... These are the basic parameters of the model. It is a bivariate functional relationship; The baseline strength prediction value is corrected by combining material characteristics and environmental characteristics to obtain a first strength prediction value; specifically: ； In the formula, The first intensity prediction value, The intensity baseline prediction value, This is a corrected feature vector formed by the material and environmental characteristics corresponding to the filled specimen. The eigenvector weight matrix, For the correction constant term; wherein, the material characteristics include sample age, sample moisture content, sample porosity, aggregate gradation and binder type, and the environmental characteristics include temperature; The residual between the measured value of the sample strength and the first predicted value of the strength is determined, and a data-driven model is used to learn the residual to obtain a compensation function; Based on the first strength prediction value and the compensation function, the second strength prediction value of the filling sample is obtained.

2. The method for predicting the strength of mine backfill bodies according to claim 1, characterized in that, The determination of the residual between the measured value of the sample strength and the first predicted value of strength includes: ； In the formula, For the residual, This refers to the measured strength value of the sample. This is the first intensity prediction value.

3. The method for predicting the strength of mine backfill bodies according to claim 2, characterized in that, The data-driven model is a gradient boosting tree, support vector regression, or a neural network. The step of learning the residual using a data-driven model to obtain the compensation function includes: Gradient boosting trees are used to fit the mapping of multiple pairs of multi-source features on a dataset with equal proportions to obtain the compensation function. The multi-source features include amplitude attenuation coefficient, group velocity dispersion slope, water content perturbation, and temperature drift term. During training, initialize the model: Where L is the squared loss function c is a constant. For the label; for the m-th iteration: calculate the negative gradient : Fit a regression tree Approaching Update the model: Finally obtained ,in For learning rate, This represents the number of iterations.

4. The method for predicting the strength of mine backfill bodies according to claim 1, characterized in that, The acquisition of wave velocity data of the filling sample under automatic coupling pressure conditions includes: Under automatic coupling pressure conditions, the filling sample is subjected to multi-frequency excitation and reception using longitudinal wave and transverse wave probes to extract propagation time, dispersion and attenuation characteristics, and to calculate the longitudinal wave velocity, the transverse wave velocity and the dynamic elastic modulus. The dynamic elastic modulus is then subjected to dispersion and attenuation correction to obtain the corrected dynamic elastic modulus; specifically: ； In the formula, The corrected dynamic elastic modulus. For dynamic elastic modulus, For correction factor, The value range is 0.95 to 1.

05.

5. The method for predicting the strength of mine backfill bodies according to claim 4, characterized in that, The process of using longitudinal and transverse wave probes to perform multi-frequency excitation and reception on the filling sample, extracting propagation time, dispersion, and attenuation characteristics, and calculating the longitudinal wave velocity, the transverse wave velocity, and the dynamic elastic modulus includes: P-wave and S-wave tests were performed using multi-frequency excitation from 10-500kHz. The entire 10-500kHz frequency band was divided into multiple sub-bands according to frequency intervals. Multi-frequency velocity parameters and dispersion-attenuation characteristic parameters of each sub-band were extracted. The multi-frequency velocity parameters include the sub-band P-wave velocity, sub-band S-wave velocity, and sub-band wave velocity ratio. The dispersion-attenuation characteristic parameters include the sub-band group velocity dispersion slope and the sub-band amplitude attenuation coefficient. For each sub-band, based on the multi-frequency velocity parameters and the measured density of the sample, the characteristic dynamic modulus of the sub-band is calculated, specifically as follows: ； ； ； In the formula, For sub-band Poisson ratio, For sub-band longitudinal wave velocity, , The effective propagation length of the sample, For the first Subband longitudinal wave signal propagation time; For sub-band shear wave velocity, , For the first Subband transverse wave signal propagation time; Shear modulus The measured density of the sample. For sub-band characteristic dynamic modulus; Introducing sub-band weighting coefficients The dynamic modulus of the sub-band characteristics corresponding to each sub-band is weighted, and the weighting coefficients are calculated by normalizing the dispersion-attenuation characteristic parameters, specifically as follows: For the first Group velocity dispersion slope corresponding to sub-band Perform normalization: ； In the formula, For the first The normalized group velocity dispersion slope corresponding to the sub-band. The group velocity dispersion slope corresponding to the first sub-band. The group velocity dispersion slope corresponding to the last sub-band, the The maximum group velocity dispersion slope among the group velocity dispersion slopes corresponding to the first sub-band and the last sub-band is used as the upper limit of the value. The minimum group velocity dispersion slope among the group velocity dispersion slopes corresponding to the first sub-band and the last sub-band is used as the lower limit of the value. For the first Amplitude attenuation coefficient corresponding to sub-band Perform normalization: ； In the formula, For the first The normalized amplitude attenuation coefficient corresponding to the sub-band This represents the amplitude attenuation coefficient corresponding to the first sub-band. This is the amplitude attenuation coefficient corresponding to the last sub-band. The upper limit of the amplitude attenuation coefficient is the largest among the amplitude attenuation coefficients corresponding to the first sub-band and the last sub-band respectively. The minimum amplitude attenuation coefficient among the amplitude attenuation coefficients corresponding to the first sub-band and the last sub-band is used as the lower limit of the value. Calculate the sub-band integrated characterization coefficients Sub-band weighting coefficient : ； ； Weighting coefficients for each sub-band As a basis for weighting, the dynamic elastic modulus of each sub-band is... Weighted fusion is performed, and a total dispersion-attenuation correction factor is introduced across the entire frequency band. The fusion results are then corrected to obtain the final total dynamic elastic modulus based on the joint modeling of frequency velocity parameters and dynamic modulus. Specifically: ； ； In the formula, For the number of sub-bands, This represents the maximum value of the full-velocity dispersion slope across all sub-bands. This represents the maximum amplitude attenuation coefficient across all sub-bands. For sub-band weighting coefficients, This refers to the dynamic modulus of the sub-band characteristics.

6. The method for predicting the strength of mine backfill bodies according to claim 5, characterized in that, The coupling pressure in the automatic coupling pressure condition is controlled in a closed loop within the range of 10-50N, and pressure fluctuations are dynamically compensated through a PID algorithm.

7. A method for predicting the strength of mine backfill bodies according to any one of claims 1-6, characterized in that, After obtaining the second predicted strength value of the filling sample, the method further includes: Uncertainty assessment is performed using cross-validation or Monte Carlo methods to form the final intensity prediction value and confidence interval.

8. A system for predicting the strength of mine backfill bodies, characterized in that, include: The data acquisition unit is used to acquire the length, height, diameter, mass, moisture content, ambient temperature, ambient humidity, as well as the propagation time, dispersion curve and amplitude attenuation coefficient of longitudinal and transverse waves in the filling sample. The data processing unit includes a memory and one or more processors; the memory stores one or more programs; the programs contain instructions for executing a method for predicting the strength of mine backfill as described in any one of claims 1 to 7; the processors are used to execute the programs.

9. A computer-readable storage medium storing a plurality of instructions thereon, characterized in that, The instructions are applicable to be loaded and executed by a processor to implement the method for predicting the strength of mine backfill bodies as described in any one of claims 1 to 7.

Images