A quick method for measuring water reduction rate

Abstract

A water-reducing rate rapid measurement method, comprising: determining a water-reducing agent to be measured, preparing a test sample, and setting a control sample; placing the test sample and the control sample in a constant-temperature test environment for stirring, and monitoring the temperature inside the test sample during the stirring process; when the temperature is balanced, performing a multi-frequency conductivity scanning, setting a scanning interval frequency band, and obtaining the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample in the full frequency band; extracting mechanism characteristics and full-spectrum characteristics, and obtaining a 12-dimensional mixed feature vector X; loading a mechanism model, and outputting a first predicted water-reducing rate R1; loading a correction model, and outputting a second predicted water-reducing rate R2 and a corresponding water-reducing rate type; setting a decision fusion criterion of the water-reducing rate, outputting a water-reducing rate R, and generating a water-reducing rate measurement result of the water-reducing agent to be measured, wherein the water-reducing rate measurement result comprises the water-reducing rate R and the water-reducing rate type. The present application is used for quantitatively calculating the water-reducing rate, and has small calculation error and high detection sensitivity.

Description

Technical Field

[0001] This invention relates to the field of water-reducing agent performance testing technology, specifically to a method for rapid calculation of water reduction rate. Background Technology

[0002] Concrete water-reducing agents are key admixtures for regulating the workability of concrete mixtures. Their water reduction rate is directly related to the fluidity, strength development, and durability of concrete. Therefore, accurate and rapid testing of the water reduction rate is a crucial aspect of concrete engineering quality control. Currently, my country generally uses the GB8076 standard for testing water reduction rates. This method requires mixing a reference concrete and concrete with the added water-reducing agent, adjusting the water content to achieve the same slump, and then calculating the water reduction rate. This method has the following significant drawbacks: (1) Complex operation: It requires precise measurement of multiple raw materials and control of the mixing process, which places high demands on the operators; (2) Time and material consumption: The single test cycle is long and the material consumption is large; (3) Large human error: Slump test is highly subjective and has poor repeatability; (4) Unable to characterize the mechanism: Only macroscopic results are obtained, which cannot reflect the mechanism of action of water-reducing agent.

[0003] To address the aforementioned issues, existing technologies have proposed some improvement methods. For example, patent document CN120404483A discloses a rapid detection method for the water-reducing performance of water-reducing polycarboxylate superplasticizers, relating to the field of polycarboxylate superplasticizer technology. This rapid detection method includes the following steps: (1) using an online viscosity testing device with automatic temperature compensation function to detect the final test viscosity η of the product after the sample reaction is completed; (2) calculating the viscosity-average molecular weight Mη and monomer conversion rate α based on the test viscosity η using the viscosity-average molecular weight calculation formula and the monomer conversion rate calculation formula; (3) comparing the viscosity-average molecular weight Mη and monomer conversion rate α with the normal performance range of the product to judge the water-reducing performance of the product. The method proposes to determine whether the water-reducing performance is qualified by testing the sample viscosity, calculating the viscosity-average molecular weight and monomer conversion rate, and comparing them with the normal performance range of the product.

[0004] However, this method does not achieve quantitative calculation of the water reduction rate. Although the method of determining the water reduction rate using cement paste fluidity, standard consistency of cement paste, and cement mortar method simplifies the operation and reduces material usage, the relative error is larger compared with the GB8076 standard method. The method of establishing a correlation model based on single-frequency conductivity cannot effectively distinguish the contribution of free ion concentration and ion mobility to conductivity. When the type of water-reducing agent changes, resulting in changes in the ion migration mechanism, the model accuracy decreases significantly. In addition, this method ignores the interfacial polarization effect, loses the high-frequency conductivity characteristics, resulting in insufficient detection sensitivity of low-dosage water-reducing agents, and lacks quantitative analysis of ion migration kinetics. The model has a narrow range of applicability and is difficult to adapt to different cement and water-reducing agent combination application scenarios.

[0005] To address the aforementioned issues, there is an urgent need to develop a water-reducing agent water-reducing rate measurement technology that can quantitatively calculate the water-reducing rate and combines high-frequency conductivity characteristics with speed, accuracy, and universality, thereby achieving a quantitative correlation between microscopic mechanisms of action and macroscopic performance indicators. Summary of the Invention

[0006] To address the technical problems of existing technologies, such as inability to quantitatively calculate water reduction, large calculation errors, loss of high-frequency conductivity characteristics, and insufficient detection sensitivity, this invention provides a rapid method for calculating water reduction rate, comprising: The water-reducing agent to be tested is identified, a test sample is prepared, and a control sample is set up. The test sample is a standard cement paste with a fixed water-cement ratio containing the water-reducing agent. The control sample is a standard cement paste with a fixed water-cement ratio without the water-reducing agent. The water-cement ratio is the mass ratio of cement to water. Set up a constant temperature test environment, place the test sample and the control sample in the constant temperature test environment and stir, and monitor the internal temperature of the test sample during the stirring process; When the internal temperature of the test sample and the control sample is in equilibrium with the test environment temperature, multi-frequency conductivity scanning is performed, the scanning interval frequency band is set, and the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample in the full frequency band are obtained. Based on the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample across the entire frequency band, mechanistic features and full-spectrum features are extracted; feature fusion is performed to obtain a 12-dimensional hybrid feature vector X; the mechanistic features include free ion concentration variation indicators and ion migration efficiency indicators; the full-spectrum features include the conductivity G(f) at the end of each frequency band and the rate of change ΔG of the conductivity of the test sample and the control sample at each frequency point. f ; A loading mechanism model is used to predict the first predicted water reduction rate R1. The 12-dimensional mixed feature vector X is input, and the first predicted water reduction rate R1 is output. Load the correction model, which is used to perform multi-task learning to predict the second predicted water reduction rate R2 and identify the water reduction rate type. Input the 12-dimensional hybrid feature vector X, and output the second predicted water reduction rate R2 and the corresponding water reduction rate type. The decision fusion criterion for water reduction rate is set. Based on the first predicted water reduction rate R1 and the second predicted water reduction rate R2, the water reduction rate R is output, and the water reduction rate calculation result of the water-reducing agent to be tested is generated. The water reduction rate calculation result includes: water reduction rate R and water reduction rate type.

[0007] Furthermore, when the correction model performs multi-task learning, it uses the first predicted water reduction rate R1 as a prior feature to enable the correction model to learn under the guidance of physical laws.

[0008] Furthermore, the multi-frequency range includes frequencies from 1 Hz to 100 kHz; the interval frequency band includes a low-frequency band, a mid-frequency band, and a high-frequency band; the low-frequency band is from 1 Hz to 10 Hz; the mid-frequency band is from 10 Hz to 1 kHz; and the high-frequency band is from 1 kHz to 100 kHz.

[0009] Furthermore, the free ion concentration change index is the rate of change ΔG of conductivity between the test sample and the control sample in the 10Hz frequency band. 10Hz The expression is: ; in, The conductivity of the sample was tested in the 10Hz frequency band. The conductivity is that of the control sample in the 0Hz frequency band.

[0010] Furthermore, the ion mobility efficiency index is the rate of change ΔG of the conductivity of the test sample and the control sample in the 100kHz frequency band. 100kHz , represented as: ; in, The conductivity of the sample was tested at a frequency of 100 kHz. The conductivity is that of the control sample at a frequency of 100 kHz.

[0011] Furthermore, the calculation expression for the mechanism model is as follows: ; Where R1 is the first predicted water reduction rate; α, β, and γ are model parameters, α∈[5.2, 8.6], β∈[0.8, 1.3], and γ∈[0.3, 1.1]; K is the ion mobility efficiency factor.

[0012] Furthermore, the calculation expression for the ion mobility efficiency factor K is as follows: .

[0013] Furthermore, the expression for the decision fusion criterion is: ; in, These are the weighting coefficients. .

[0014] Furthermore, the correction model employs the Gradient Boosting Decision Tree (GBDT) algorithm.

[0015] Furthermore, the water-reducing agent to be tested includes solid water-reducing agents and liquid water-reducing agents; when preparing the test sample, if a solid water-reducing agent is added, the solid water-reducing agent is added to the cement slurry after the dosage is calculated; if a liquid water-reducing agent is added, the mass of the water-reducing agent is calculated according to the solid content of the liquid water-reducing agent and then added to the cement slurry.

[0016] The beneficial effects of this invention are as follows: This invention combines multi-frequency conductivity spectroscopy detection with ion migration kinetic analysis. By constructing the ion migration efficiency factor K through the ratio of low-frequency free ion concentration to high-frequency ion mobility, it achieves a quantitative correlation between microscopic ion behavior and macroscopic water reduction rate, breaking through the mechanistic limitations of traditional single-frequency detection. By loading a mechanism model and a correction model, it enables the identification of water-reducing agent types and the calculation of water reduction rate, solving the problem of poor model adaptability under different application systems. By establishing a standardized test benchmark through fixed water-cement ratio and constant temperature control, combined with broadband scanning technology, it effectively eliminates systematic errors caused by environment and equipment, improves the repeatability and reliability of the calculation results, and is suitable for rapid evaluation of water-reducing agent research and development, quality control, and engineering applications. Attached Figure Description

[0017] Figure 1 This is a flowchart of the rapid water reduction rate calculation method provided by the present invention. Detailed Implementation

[0018] The technical solution of the present invention is further described below, but the scope of protection is not limited to what is described.

[0019] This invention provides a method for rapid calculation of water reduction rate, including: Step S100: Determine the water-reducing agent to be tested, prepare the test sample, and set up a control sample; the test sample is a standard cement paste with a fixed water-cement ratio containing the water-reducing agent; the control sample is a standard cement paste with a fixed water-cement ratio without the water-reducing agent; the water-cement ratio is the mass ratio of cement to water. The water-reducing agent includes solid water-reducing agents and liquid water-reducing agents. When preparing test samples, if a solid water-reducing agent is added, it is added to the cement slurry after calculating the dosage; if a liquid water-reducing agent is added, its mass is calculated based on the solid content of the liquid water-reducing agent before being added to the cement slurry. The dosage is the recommended dosage provided by the water-reducing agent manufacturer.

[0020] For example, the water-cement ratio is 0.4, the mass of cement is 450g, the mass of water is 180g, and the water-reducing agent is added precisely using a micro-syringe according to the recommended dosage, with the dosage control error not exceeding 0.02%.

[0021] Step S200: Set up a constant temperature test environment, place the test sample and the control sample in the constant temperature test environment and stir, and monitor the internal temperature of the test sample during the stirring process.

[0022] The isothermal testing environment, such as a testing environment where the temperature is maintained at 25±0.5°C, specifically includes the following steps: transferring the test sample and control sample to isothermal testing containers, placing them in a 25°C isothermal water bath, and starting medium-speed stirring; placing identical temperature sensors in the test sample and control sample respectively, and monitoring the temperature change inside the sample in real time during stirring; when the temperature stabilizes within the range of 25±0.1°C and remains stable for more than 30 seconds, it is determined that temperature equilibrium has been completed. Then, multi-frequency conductivity spectroscopy is initiated for scanning.

[0023] To minimize the initial temperature difference between the test sample and the control sample and the water bath temperature difference, the materials used for the test sample and the control sample were placed in an environment of 25°C ± 0.5°C for 24 hours in advance.

[0024] Step S300: When the internal temperature of the test sample and the control sample is in equilibrium with the test environment temperature, perform multi-frequency conductivity scanning, set the scanning interval frequency band, and obtain the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample across the entire frequency band. The multi-frequency range includes frequencies from 1 Hz to 100 kHz; the interval frequency band includes a low-frequency band, a mid-frequency band, and a high-frequency band; the low-frequency band is from 1 Hz to 10 Hz; the mid-frequency band is from 10 Hz to 1 kHz; and the high-frequency band is from 1 kHz to 100 kHz.

[0025] The low-frequency band (1Hz~10Hz) mainly reflects free ions in the slurry, such as Ca. 2+ OH - The concentration change. When water-reducing agent molecules are adsorbed on the surface of cement particles, the particle dispersibility is improved, and the free ions originally wrapped in the particle agglomerates are released, resulting in a significant increase in the conductivity of the low-frequency band; The high-frequency band (1kHz~100kHz) mainly reflects ion mobility and the interfacial polarization effect between cement particles and the paste. After the water-reducing agent molecules form a stable adsorption layer on the particle surface, the interfacial resistance decreases, and the ion migration path becomes smoother. The rate of change of conductivity in the high-frequency band is directly related to the degree of particle dispersion.

[0026] The multi-frequency conductivity scanning is achieved using a multi-frequency conductivity spectrum, which employs a wideband AC impedance testing system to continuously scan the conductivity of the test sample within a frequency range of 1 Hz to 100 kHz. The scanning interval frequency band is set, and scanning is performed in segments with the following step sizes: 1 Hz to 10 Hz (low frequency band): 1 Hz step; 10 Hz to 1 kHz (mid frequency band): 10 Hz step; 1 kHz to 100 kHz (high frequency band): 100 Hz step. The conductivity G(f) at each frequency point across the entire frequency band is recorded.

[0027] A full-band scan was performed on the control sample, and the conductivity value G0(f) at each frequency point was recorded as the baseline data.

[0028] During the multi-frequency conductivity scan, a pair of platinum black electrodes representing the multi-frequency conductivity spectrum are inserted into the sample, ensuring a fixed electrode spacing of 2.0 cm and an insertion depth of 1.5 cm. The containers for the test and control samples are made of high-density polyethylene, with an inner diameter of 100 mm ± 2 mm, a height of 150 mm ± 5 mm, and a thickness of 3 mm to 5 mm; a 1.5 cm depth line is marked on the inner wall of the container; the electrode insertion depth is ensured to be consistent each time; a rubber ring is added to the bottom to prevent slippage in the water bath.

[0029] The electrodes are made of platinum black to reduce electrode polarization interference. The electrode spacing is fixed at 2 cm and the insertion depth into the slurry is 1.5 cm to ensure that the electrode configuration is completely consistent for each test and to reduce system error.

[0030] Step S400: Based on the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample across the entire frequency band, extract mechanistic features and full-spectrum features; perform feature fusion to obtain a 12-dimensional hybrid feature vector X; the mechanistic features include free ion concentration change index and ion migration efficiency index; the full-spectrum features include the conductivity G(f) at the end of each frequency band and the rate of change ΔG of the conductivity of the test sample and the control sample at each frequency point. f ; The free ion concentration change index is the rate of change ΔG of conductivity between the test sample and the control sample in the 10Hz frequency band. 10Hz The expression is: (1) in, The conductivity of the sample was tested in the 10Hz frequency band. The conductivity is that of the control sample in the 0Hz frequency band.

[0031] The ion mobility efficiency index is the rate of change ΔG of the conductivity of the test sample and the control sample in the 100kHz frequency band. 100kHz , represented as: (2) in, The conductivity of the sample was tested at a frequency of 100 kHz. The conductivity is that of the control sample at a frequency of 100 kHz.

[0032] The extraction of the aforementioned mechanistic features is based on the core mechanism of the water-reducing agent's "adsorption-dispersion" action; a quantitative correlation between changes in conductivity and water reduction rate is established; the water-reducing agent molecules disperse cement particles through electrostatic repulsion and steric hindrance effects, a process accompanied by an increase in free ion concentration and a decrease in ion migration resistance, wherein an increase in ion concentration corresponds to an increase in low-frequency conductivity, and a decrease in ion migration resistance corresponds to an increase in high-frequency conductivity.

[0033] The conductivity G(f) at the end frequency points of each frequency band is as follows: conductivity values ​​G(f) at 1Hz, 10Hz, 1kHz, 10kHz, and 100kHz. The rate of change ΔG of conductivity of the test sample and the control sample at each frequency point f For example: the rate of change ΔG of the conductivity of the control sample relative to the frequency points of 1Hz, 10Hz, 1kHz, 10kHz, and 100kHz. f .

[0034] The 12-dimensional hybrid feature vector X is represented as X=[K, ΔG] 10Hz ΔG 100kHz , G (1Hz), ΔG 1Hz , ..., G (100kHz), ΔG 100kHz ]; Step S500: Load the mechanism model, which is used to predict the first predicted water reduction rate R1. Input the 12-dimensional mixed feature vector X and output the first predicted water reduction rate R1. The calculation expression for the mechanism model is as follows: (3) Where R1 is the first predicted water reduction rate; α, β, and γ are model parameters, α∈[5.2, 8.6], β∈[0.8, 1.3], and γ∈[0.3, 1.1]; K is the ion mobility efficiency factor.

[0035] The formula for calculating the ion mobility efficiency factor K is as follows: (4) The aforementioned mechanism model eliminates the influence of single-frequency fluctuations by using the ratio of low-frequency to high-frequency conductivity, and achieves direct coupling between microscopic mechanisms and macroscopic water-reducing performance by introducing an ion migration efficiency factor K.

[0036] Step S600: Load the correction model. The correction model is used to perform multi-task learning to predict the second predicted water reduction rate R2 and identify the water reduction rate type. Input the 12-dimensional mixed feature vector X and output the second predicted water reduction rate R2 and the corresponding water reduction rate type. The correction model uses the Gradient Boosting Decision Tree (GBDT) algorithm.

[0037] When the correction model performs multi-task learning, the first predicted water reduction rate R1 is used as a priori feature to enable the correction model to learn under the guidance of physical laws.

[0038] The multitasking includes: (1) Regression, the second prediction of water reduction rate R2, the loss function is the mean squared error; optimize the tree structure through the first-order gradient; when predicting, first standardize the input 12-dimensional mixed feature vector X according to the statistics of the training set, accumulate the predicted value tree by tree, and then use the mean and standard deviation of the water reduction rate of the training set to perform inverse standardization output; at the same time, calculate the confidence interval of the predicted value and provide feature contribution analysis.

[0039] (2) Classification: Identifying the type of water-reducing agent, with cross-entropy as the loss function. The classification task uses a softmax output layer to convert the original scores of the decision tree into a category probability distribution. The final type is determined by the maximum probability value, and the confidence level is output simultaneously. Based on the pre-defined category labels of different water-reducing agents, the type of the sample to be tested is inferred from the category label corresponding to the maximum probability value. The correction model uses a multi-class cross-entropy loss function during training, which is jointly optimized with the mean squared error loss of the regression task through weighted summation.

[0040] The label formats for the different water-reducing agent categories are as follows; the specific content can be expanded according to actual needs: 1—High-performance water-reducing agent; 2—High-efficiency water-reducing agent; 3—Ordinary water-reducing agents; 4—Air-entraining water-reducing agent.

[0041] Step S700: Set the decision fusion criteria for water reduction rate. Based on the first predicted water reduction rate R1 and the second predicted water reduction rate R2, output the water reduction rate R and generate the water reduction rate calculation result of the water-reducing agent to be tested. The water reduction rate calculation result includes: water reduction rate R and water reduction rate type.

[0042] The expression for the decision fusion criterion is: (5) in, These are the weighting coefficients. .

[0043] The The calibration method, based on historical data, includes: during the initialization phase, using no fewer than 50 sets of calibration samples, calculating the theoretical model prediction error MAE1 and the corrected model prediction error MAE2 respectively, according to the formula... Calculate the initial value; during the operation phase, calculate the difference Δ=|R between the predicted values ​​of the mechanistic model and the correction model. 1- R2 | Dynamic Adjustment: When Δ≤0.8%, take =0; When Δ>2%, take =1, and trigger an exception review; When 0.8% < Δ ≤ 2%, Obtained by linear interpolation between 0.1 and 0.6; The system continuously accumulates new data and updates it regularly. This enables adaptive optimization.

[0044] The technical solution of the present invention also includes establishing an incremental learning mechanism. While obtaining the measurement results using the technical solution of the present invention, the measurement is performed according to the GB8076 standard method to obtain the standard measurement results. The comparison results between the measurement results of the present solution and the standard measurement results are continuously collected. When 50 sets are accumulated, incremental training is automatically started to update the parameters of the GBDT algorithm in the correction model.

[0045] The parameters of the mechanistic model are self-calibrated. Specifically, the parameters (α, β, γ) of the mechanistic model are inferred from the output of the updated parameters of the correction model, so that the mechanistic model can continue to evolve with the accumulation of data and maintain its interpretability and timeliness.

[0046] When changing the test samples in each batch (changing the type of cement or water-reducing agent), 3 to 5 groups of water reduction rates measured according to the GB8076 standard method are randomly selected to verify the mechanism model and the correction model or fine-tune the parameters.

[0047] It automatically saves conductivity, mechanism characteristics, full spectrum characteristics, test results, confidence scores, time parameters, etc., which facilitates traceability and quality analysis.

[0048] Specific examples are as follows: Example 1: Select one high-performance water-reducing agent as the test sample. The test sample is a liquid. Calculate the pure substance dosage based on its solid content. The dosage is 1%, and it is added to cement. The test environment temperature was stabilized at 25℃±0.5℃. 450g±5g of cement mixed with water-reducing agent and 180g±1g of water were accurately weighed. The cement and water were placed in an environment of 25℃±0.5℃ for 24 hours in advance.

[0049] According to GB8076, test samples were prepared using a cement paste mixer. The mixture was first stirred at low speed for 120 seconds, paused for 15 seconds, and then stirred at high speed for 120 seconds. After stirring stopped, the mixture was immediately poured into a high-density polyethylene container. At the same time, a small paddle mixer was inserted into the paste container and stirred at a medium speed to ensure that the paste temperature reached 25℃±0.1℃ within 5 minutes.

[0050] When the test sample is monitored to be maintained at 25℃±0.1℃, the platinum black electrode is inserted into the prepared slurry, and the broadband AC impedance analyzer is started to perform a frequency scan from 1Hz to 100kHz. The conductivity data at each frequency point is recorded, and the conductivity values ​​at 10Hz and 100kHz are extracted. G(10Hz) = 3.21mS / cm, G(100kHz) = 11.48mS / cm.

[0051] A control sample without the high-performance water-reducing agent was prepared in the same manner. The key conductivity data extracted included G0 (10Hz) = 2.01mS / cm and G0 (100kHz) = 9.16mS / cm.

[0052] Input mechanism model and correction model.

[0053] The specific process of inputting the mechanism model to predict the first predicted water reduction rate R1 includes: First, calculate ΔG according to (Equation 1) and (Equation 2). 10Hz and ΔG 100kHz ΔG 10Hz =59.7%, ΔG 100kHz =25.3%; Secondly, the ion mobility efficiency factor K is calculated according to (Equation 4), K = 59.7 / 25.3 = 2.360; In this example, the parameters of the mechanistic model are: α=8.6, β=1.3, γ=0.3. Substituting these values ​​into the mechanistic model (Equation 3), the first predicted water reduction rate R1=8.6×2.360 can be calculated. 1.3 +0.3 = 26.6%.

[0054] Input the correction model to predict the second predicted water reduction rate R2 and identify the water reduction rate type. Output the second predicted water reduction rate R2=26.4% and the water-reducing agent classification probability [0.92, 0.04, 0.03, 0.01]. According to the classification rules, the water-reducing agent can be identified as category label 1, which is a high-performance water-reducing agent.

[0055] According to the decision fusion criterion, |R1-R2|=0.2%<0.8%, the final water reduction rate R of the sample to be tested is calculated to be 26.4%.

[0056] Example 2: Select one high-efficiency water-reducing agent as the test sample. The test sample is a powder and is prepared according to the same method as in Example 1. Under constant temperature conditions of 25℃±0.5℃, use a wideband AC impedance testing system to scan the conductivity spectrum from 1Hz to 100kHz and extract the conductivity values ​​at 10Hz and 100kHz. G(10Hz) = 2.56mS / cm, G(100kHz) = 10.47mS / cm.

[0057] The control sample was set up the same as in Example 1, and the extracted conductivity data were also the same. Input mechanism model and correction model.

[0058] The specific process of inputting the mechanism model to predict the first predicted water reduction rate R1 includes: First, calculate ΔG according to (Equation 1) and (Equation 2). 10Hz and ΔG 100kHz ΔG 10Hz =27.4%, ΔG 100kHz =14.3%; Secondly, the ion mobility efficiency factor K is calculated according to (Equation 4), K = 27.4 / 14.3 = 1.916; In this example, the parameters of the mechanistic model are α=7.8, β=1.1, and γ=0.8. Substituting these values ​​into the mechanistic model (Equation 3), the first predicted water reduction rate R1=7.8×1.916 can be calculated. 1.1 +0.8 = 16.7%.

[0059] Input the correction model to predict the second predicted water reduction rate R2 and identify the water reduction rate type. Output the second predicted water reduction rate R2=17.9% and the water-reducing agent classification probability [0.11, 0.86, 0.02, 0.01]. According to the classification rules, the water-reducing agent can be identified as category label 2, which is a high-efficiency water-reducing agent.

[0060] According to the decision fusion criterion, |R1-R2|=1.2%∈(0.8%,2.0%), the final water reduction rate of the sample to be tested should be calculated according to λR1+(1-λ)R2. In this embodiment, λ=0.27, that is, 0.27×16.7+0.73×17.9=17.6%.

[0061] This invention combines multi-frequency conductivity spectroscopy detection with ion migration kinetics analysis. By constructing a characteristic factor based on the ratio of low-frequency free ion concentration to high-frequency ion mobility, it achieves a quantitative correlation between microscopic ion behavior and macroscopic water reduction rate, overcoming the mechanistic limitations of traditional single-frequency detection. A nonlinear coupling model incorporating a ratio term is designed, and machine learning algorithms are introduced to optimize model parameters, enhancing the model's generalization ability and enabling water-reducing agent type identification and water reduction rate prediction, thus solving the problem of poor model adaptability under different application systems. A standardized testing benchmark is established by fixing the water-cement ratio and maintaining constant temperature. Combined with broadband scanning technology, systematic errors caused by the environment and equipment are effectively eliminated, improving the repeatability and reliability of the measurement results.

[0062] The above-disclosed embodiments are merely specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for rapid calculation of water reduction rate, characterized in that, include: Identify the water-reducing agent to be tested, prepare the test sample, and set up the control sample; The test sample was a standard cement paste with a fixed water-cement ratio containing the water-reducing agent. The control sample was a standard cement paste with a fixed water-cement ratio without the water-reducing agent; the water-cement ratio was the mass ratio of cement to water. Set up a constant temperature test environment, place the test sample and the control sample in the constant temperature test environment and stir, and monitor the internal temperature of the test sample during the stirring process; When the internal temperature of the test sample and the control sample is in equilibrium with the test environment temperature, multi-frequency conductivity scanning is performed, the scanning interval frequency band is set, and the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample in the full frequency band are obtained. Based on the conductivity G(f) of the test sample and the conductivity G0(f) of the control sample across the entire frequency band, mechanistic features and full-spectrum features are extracted. Feature fusion was performed to obtain a 12-dimensional hybrid feature vector X; the mechanistic features included free ion concentration variation index and ion migration efficiency index; the full-spectrum features included conductivity at the end frequencies of each frequency band and the rate of change ΔG of conductivity between the test sample and the control sample at each frequency point. f ; A loading mechanism model is used to predict the first predicted water reduction rate R1. The 12-dimensional mixed feature vector X is input, and the first predicted water reduction rate R1 is output. Load the correction model, which is used to perform multi-task learning to predict the second predicted water reduction rate R2 and identify the water reduction rate type. Input the 12-dimensional hybrid feature vector X, and output the second predicted water reduction rate R2 and the corresponding water reduction rate type. The decision fusion criterion for water reduction rate is set. Based on the first predicted water reduction rate R1 and the second predicted water reduction rate R2, the water reduction rate R is output, and the water reduction rate calculation result of the water-reducing agent to be tested is generated. The water reduction rate calculation result includes: water reduction rate R and water reduction rate type.

2. The rapid water reduction rate calculation method as described in claim 1, characterized in that, When the correction model performs multi-task learning, the first predicted water reduction rate R1 is used as a priori feature to enable the correction model to learn under the guidance of physical laws.

3. The rapid water reduction rate calculation method as described in claim 2, characterized in that, The multi-frequency range includes frequencies from 1 Hz to 100 kHz; the interval frequency band includes a low-frequency band, a mid-frequency band, and a high-frequency band; the low-frequency band is from 1 Hz to 10 Hz; the mid-frequency band is from 10 Hz to 1 kHz; and the high-frequency band is from 1 kHz to 100 kHz.

4. The rapid water reduction rate calculation method as described in claim 3, characterized in that, The free ion concentration change index is the rate of change ΔG of conductivity between the test sample and the control sample in the 10Hz frequency band. 10Hz The expression is: ； in, The conductivity of the sample was tested in the 10Hz frequency band. The conductivity is that of the control sample in the 0Hz frequency band.

5. The rapid water reduction rate calculation method as described in claim 4, characterized in that, The ion mobility efficiency index is the rate of change ΔG of the conductivity of the test sample and the control sample in the 100kHz frequency band. 100kHz , is represented as: ； in, The conductivity of the sample was tested at a frequency of 100 kHz. The conductivity is that of the control sample at a frequency of 100 kHz.

6. The rapid water reduction rate calculation method as described in claim 5, characterized in that, The calculation expression for the mechanism model is as follows: ； Where R1 is the first predicted water reduction rate; α, β, and γ are model parameters, α∈[5.2, 8.6], β∈[0.8, 1.3], and γ∈[0.3, 1.1]; K is the ion mobility efficiency factor.

7. The rapid water reduction rate calculation method as described in claim 6, characterized in that, The formula for calculating the ion mobility efficiency factor K is as follows: 。 8. The method for rapid calculation of water reduction rate as described in claim 1, characterized in that, The expression for the decision fusion criterion is: ； in, These are the weighting coefficients. .

9. The rapid water reduction rate calculation method as described in claim 1, characterized in that, The correction model uses the Gradient Boosting Decision Tree (GBDT) algorithm.

10. The method for rapid calculation of water reduction rate as described in claim 1, characterized in that, The water-reducing agent to be tested includes solid water-reducing agents and liquid water-reducing agents. When preparing the test sample, if a solid water-reducing agent is added, the solid water-reducing agent is added to the cement slurry after the dosage is calculated. If a liquid water-reducing agent is added, the water-reducing agent is added to the cement slurry after the mass of the water-reducing agent is calculated based on the solid content of the liquid water-reducing agent.

Images