Comprehensive research and judgment method for negative temperature performance of MPC-UHPC based on data driving

Abstract

The invention discloses an MPC-UHPC negative temperature performance comprehensive research and judgment method based on data driving, and relates to the technical field of material performance prediction. The method comprises the following steps: preparing slurry through laboratory tests, testing a hydration temperature peak value, testing compressive strength and maximum freeze-thaw cycle times after a test piece is maintained, constructing a data set, cleaning and removing abnormal values, dividing a training set and a test set, constructing a research and judgment model to carry out three-layer research and judgment on hydration temperature, mechanical properties and durability, and respectively carrying out training and testing. And inputting real data into the model, outputting a predicted value, carrying out hydration temperature research and judgment, carrying out mechanical property and durability research and judgment if conditions are met, carrying out assignment calculation on a coding value CEC according to constraint conditions, and outputting a defined applicable scene. A multi-dimensional data set of negative temperature performance is constructed through a laboratory test, three layers of research and judgment constraints of hydration temperature, mechanical property and durability are defined in combination with an improved research and judgment model, and an applicable scene can be quickly screened and obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material performance prediction, and in particular to a MPC-UHPC negative temperature performance comprehensive research and judgment method based on data driving. BACKGROUND

[0002] UHPC is a kind of fiber reinforced cement-based composite material with excellent mechanical properties and durability. The UHPC on the market generally uses mineral admixture modified Portland cement as cementitious material. Additional thermal insulation curing measures are needed during winter negative temperature construction, which cannot meet the requirements of repair and construction in severe cold environments. Magnesium phosphate cement-based ultra-high performance concrete (MPC-UHPC) has broad application prospects in engineering construction in cold regions due to its rapid hydration rate and high early strength. However, the hydration process of MPC-UHPC at negative temperature will be inhibited, and the hydration heat release process, mechanical properties and durability will change significantly compared with normal temperature. Therefore, accurate research and judgment of the negative temperature performance is the key to guarantee the engineering quality.

[0003] The existing MPC-UHPC performance evaluation method mainly relies on traditional experimental testing, which needs to complete the judgment through the processes of preparing test pieces, simulating negative temperature curing, and developing mechanical properties and durability testing. This method has problems such as long cycle, high cost, strong dependence on experimental conditions, and difficulty in comprehensively quantifying the complex correlation between mix proportion parameters and multi-dimensional performance indicators.

[0004] With the emergence of machine learning technology, mathematical driving mix proportion design methods have received more and more attention. Compared with traditional empirical formulas, machine learning prediction models have faster speed and higher accuracy in predicting the performance of cement-based composites. However, existing research only focuses on the prediction of working performance and mechanical properties, the prediction target is single, the model structure is simple, the prediction accuracy of complex relationships between multiple variables is not enough, and the influence of different material service environments on MPC-UHPC service performance is not considered, which has poor applicability. Therefore, an intelligent research and judgment method is urgently needed to quickly and accurately predict and determine the applicability of MPC-UHPC service performance by integrating multi-layer prediction and judgment algorithms. SUMMARY

[0005] To solve the problems in the background art, the present application provides a MPC-UHPC negative temperature performance comprehensive research and judgment method based on data driving, which constructs a multi-dimensional data set of negative temperature performance through laboratory tests, defines three layers of research and judgment constraints of hydration temperature, mechanical properties and durability by combining an improved research and judgment model, and can quickly screen and obtain its applicable scenarios.

[0006] To achieve the above purpose, the present application adopts the following technical solution: a MPC-UHPC negative temperature performance comprehensive research and judgment method based on data driving, comprising the following steps:

[0007] S1: MPC-UHPC paste is prepared by laboratory test stirring in an environment of-30℃~0℃, and the mix proportion parameters include magnesium-phosphorus ratio, water-binder ratio, boron-magnesium ratio, binder-sand ratio, fly ash content, light burned magnesium content, fiber content, fiber type, phosphate type, and mixing water temperature;

[0008] S2: The hydration temperature peak of the MPC-UHPC paste is tested after stirring, and the compressive strength of the molded test piece is tested at 1d, 3d, 7d, and 28d after curing, and the maximum freeze-thaw cycle number at 28d is tested, and a data set of the MPC-UHPC negative temperature performance is constructed;

[0009] S3: Data set cleaning and outlier removal processing is performed, one-hot encoding is performed on the fiber type, phosphate type, and mixing water temperature, and the data set is divided into a training set and a test set;

[0010] S4: A research and judgment model is constructed based on a stacking model, the base model includes RF, XGBoost, LightGBM, and CatBoost, and the meta-model is Bayesian Ridge Regression;

[0011] S5: The prediction error and precision coefficient of each base model are calculated through 5-fold cross-validation, the normalized weight of each base model is obtained through normalization and smoothing processing, and an Ada-stacking model with adaptive weight adjustment is formed;

[0012] S6: The three-layer research and judgment of hydration temperature, mechanical property, and durability performance are sequentially performed by the constructed Ada-stacking model, the input parameters of the model in the hydration temperature research and judgment process are the environmental temperature and the mix proportion parameters, the output parameter is the hydration temperature peak prediction value, the input parameters of the model in the mechanical property research and judgment process are the environmental temperature, the mix proportion parameters, the curing age, and the hydration temperature peak output by the hydration temperature research and judgment, the output parameter is the compressive strength prediction value, the input parameters of the model in the durability performance research and judgment process are the environmental temperature, the curing age, and the compressive strength prediction value output by the mechanical property research and judgment, and the output parameter is the maximum freeze-thaw cycle number prediction value;

[0013] S7: The training and test sets are used to train and test the three-layer research and judgment Ada-stacking model, and the R 2 value of the test set is controlled to be above 0.8;

[0014] S8: The real data of the actual material to be verified is input into the Ada-stacking model, and the hydration temperature peak prediction value, the compressive strength prediction value, and the maximum freeze-thaw cycle number prediction value are output;

[0015] S9: Perform hydration temperature research, determine the feasibility of MPC-UHPC hydration reaction in negative temperature environment according to constraint condition one, and constraint condition one is represented as:

[0016] (8)

[0017] In the formula, is the output hydration temperature peak prediction value, is the ambient temperature, if the condition is met, it is determined that the hydration condition is met, and step S10 is entered, if the condition is not met, it is determined that the hydration condition is not met, and the determination process is ended;

[0018] S10: Perform mechanical property research, determine the strength grade of MPC-UHPC in negative temperature environment according to constraint condition two, and constraint condition two is represented as:

[0019] (9)

[0020] In the formula, is the output compressive strength prediction value, when , the value is 0, when , the value is 1, and when , the value is 2;

[0021] S11: Perform durability research, determine the durability of MPC-UHPC in negative temperature environment according to constraint condition three, and constraint condition three is represented as:

[0022] (10)

[0023] In the formula, is the output maximum freeze-thaw cycle number prediction value, when , the value is 0, when , the value is 1, and when , the value is 2;

[0024] S12: According to the assignment value of mechanical property and durability research in step S10 and step S11, the encoding value CEC of MPC-UHPC negative temperature service performance research based on assignment constraint is constructed through formula (11), and the calculation formula is:

[0025] (11)

[0026] In the formula, is the assignment value determined according to constraint condition two in step S10, is the assignment value determined according to constraint condition three in step S11;

[0027] S13: The Ada-stacking model outputs a defined applicable scene according to the encoding value CEC.

[0028] Further, the matching ratio parameters specifically include: a magnesium-phosphorus ratio of 3-6, a water-binder ratio of 0.10-0.20, a boron-magnesium ratio of 0.02-0.10, a cement-sand ratio of 0.5-1.5, a fly ash content of 0%-20%, a lightly calcined magnesium content of 0%-10%, a fiber content of 0-1.5%, fiber types including steel fiber and PVA fiber, and phosphate types including ammonium dihydrogen phosphate and potassium dihydrogen phosphate, and mixing water temperatures including 0°C and 18°C.

[0029] Further, the specific process of the step S5 includes:

[0030] The weight proportions of the four base models are adjusted, the training set is divided into five mutually exclusive subsets, one of the mutually exclusive subsets is selected as a validation set, five-fold cross-validation is performed on each base model, and the prediction errors of the four base models on the validation set are calculated according to formula (1) , as follows:

[0031] (1)

[0032] In the formula, represents the number of the base model, represents the number of the fold of cross-validation, represents the number of samples in the fold validation set, is the measured value of the th sample, is the predicted value of the th sample;

[0033] The accuracy coefficients of each base model on the 1st-5th fold are calculated according to formula (2) , as follows:

[0034] (2)

[0035] In the formula, is to prevent the denominator from being 0;

[0036] The initial weights of the four base models on the 1st-5th fold are obtained by normalizing according to formula (3) , and the smoothed weights are obtained by smoothing each fold weight of each base model according to formula (4) , as follows:

[0037] (3)

[0038] (4)

[0039] wherein, when takes [0, 1), when takes 1;

[0040] The formula (5) is used to normalize the, and the normalized weight of each base model is obtained The meta-model training set is calculated according to formula (6) The meta-model is trained, and the meta-model test set is calculated according to formula (7) The meta-model is tested, and is expressed as follows:

[0041] (5)

[0042] (6)

[0043] (7) wherein,

[0044] is the prediction value of the i-th base model to the test set sample , and is the test set sample. Further, in the step S13, the coding value CEC corresponds to the definition of the applicable scene:

[0045] CEC=00, the applicable scene corresponds to the environment of compressive strength requirement ≤60MPa and freeze-thaw cycle ≤250 times;

[0046] CEC=01, the applicable scene corresponds to the environment of compressive strength requirement ≤60MPa and 250 times < freeze-thaw cycle ≤500 times;

[0047] CEC=02, the applicable scene corresponds to the environment of compressive strength requirement ≤60MPa and freeze-thaw cycle >500 times;

[0048] CEC=10, the applicable scene corresponds to the environment of 60MPa < compressive strength requirement ≤100MPa and freeze-thaw cycle ≤250 times;

[0049] CEC=20, the applicable scene corresponds to the environment of compressive strength requirement >100MPa and freeze-thaw cycle ≤250 times;

[0050] CEC=20, the applicable scene corresponds to the environment of compressive strength requirement >100MPa and freeze-thaw cycle ≤250 times;

[0051] ​​​CEC=11, the applicable scene corresponds to an environment of 60MPa < compressive strength requirement <= 100MPa and 250 times < freeze-thaw cycle <= 500 times;

[0052] CEC=12, the applicable scene corresponds to an environment of 60MPa < compressive strength requirement <= 100MPa and freeze-thaw cycle > 500 times;

[0053] CEC=21, the applicable scene corresponds to an environment of compressive strength requirement > 100MPa and 250 times < freeze-thaw cycle <= 500 times;

[0054] CEC=22, the applicable scene corresponds to an environment of compressive strength requirement > 100MPa and freeze-thaw cycle > 500 times.

[0055] Compared with the prior art, the beneficial effects of the present application are: the method of the present application automatically judges the rationality of the MPC-UHPC negative temperature mix proportion by combining the hydration feasibility, mechanical property grade and durability grade, can help the engineering personnel to quickly screen out the economic and suitable material mix proportion according to the specific needs of the actual engineering, avoid the cost waste caused by performance surplus or the safety hidden danger brought by performance deficiency, cover the multi-dimensional prediction and judgment of negative temperature construction feasibility, mechanical property and frost resistance grade classification and applicable scene, compared with the common single mechanical property prediction, meet the complex and changeable engineering actual needs, realize the accurate prediction of the MPC-UHPC negative temperature service performance combined with multi-dimensional experimental data, and determine the applicability of the material according to the specification requirements, provide efficient and reliable technical support for engineering application, unlike the traditional trial and error type research and development mode, through data driving, the research and development cycle of MPC-UHPC in negative temperature environment can be greatly shortened and the trial and error cost can be reduced, and a new design idea is provided for the research and development and application of negative temperature MPC-based composite material. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a flowchart of the judgment method of the present application. DETAILED DESCRIPTION

[0057] The technical solutions in the present application will be described clearly and completely in the embodiments of the present application combined with the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0058] As shown in Figure 1 , a data-driven MPC-UHPC negative temperature performance comprehensive judgment method comprises the following steps:

[0059] S1: Prepare the MPC-UHPC slurry by laboratory test stirring in the environment of-30℃~0℃, the mix proportion parameters of which include magnesium-phosphorus ratio, water-binder ratio, boron-magnesium ratio, binder-sand ratio, fly ash content, light burned magnesium content, fiber content, fiber type, phosphate type and mixing water temperature. Specifically, the magnesium-phosphorus ratio is 3~6, the water-binder ratio is 0.10~0.20, the boron-magnesium ratio is 0.02~0.10, the binder-sand ratio is 0.5~1.5, the fly ash content is 0%~20%, the light burned magnesium content is 0%~10%, the fiber content is 0~1.5%, the fiber type includes steel fiber and PVA fiber, the phosphate type includes ammonium dihydrogen phosphate and potassium dihydrogen phosphate, and the mixing water temperature includes 0℃ and 18℃.

[0060] S2: After stirring, test the hydration temperature peak of the MPC-UHPC slurry using a temperature sensor, then shape the test piece and perform curing, after the curing is completed, test the compressive strength of the MPC-UHPC at four ages of 1d, 3d, 7d and 28d, and test the maximum freeze-thaw cycle number of the MPC-UHPC at the age of 28d, and build a data set of the MPC-UHPC negative temperature performance.

[0061] S3: After cleaning and outlier removal processing of the data set, the fiber type, phosphate type and mixing water temperature are one-hot encoded, the steel fiber is encoded as 0, the PVA fiber is encoded as 1, the ammonium dihydrogen phosphate is encoded as 0, the potassium dihydrogen phosphate is encoded as 1, the 0℃ mixing water is encoded as 0, and the 18℃ mixing water is encoded as 1, then the data set is divided into training set and test set in the ratio of 6:4~9:1.

[0062] S4: Build a research and judgment model, the base model of which is a stacking model, the base models of the stacking model include RF, XGBoost, LightGBM and CatBoost, and the meta-model is Bayesian Ridge Regression.

[0063] S5: Adjust the weight proportion of the four base models in the stacking model, first divide the training set into five mutually exclusive subsets, select one of the mutually exclusive subsets as the validation set, perform 5-fold cross-validation on each base model, calculate the prediction error of the four base models on the validation set according to formula (1) , and calculate the accuracy coefficient of each base model on the 1st~5th fold according to formula (2) , normalize according to formula (3), and get the initial weight of the four base models on the 1st~5th fold , and smooth the weight of each base model according to formula (4) , the normalized weight of each base model is obtained by using formula (5) to normalize , , , , ,

[0064] Each formula is as follows:

[0065] (1)

[0066] In the formula, represents the number of base models, represents the number of cross-validation folds, represents the number of samples in the th fold of the validation set, is the measured value of the th sample, is the predicted value of the th sample;

[0067] (2)

[0068] In the formula, is to prevent the denominator from being 0;

[0069] (3)

[0070] (4)

[0071] In the formula, when , take [0, 1), when , take 1;

[0072] (5)

[0073] (6)

[0074] (7)

[0075] In the formula, is the predicted value of the th base model for the test set sample , is the test set sample.

[0076] According to the above improved process, the stacking model with adaptive weight adjustment is obtained, which is defined as the Ada-stacking model.​

[0077] S6: Through the constructed Ada-stacking model, the hydration temperature, mechanical property and durability are judged in turn. The input parameters of the model in the hydration temperature judgment process are the environmental temperature, magnesium-phosphorus ratio, water-binder ratio, boron-magnesium ratio, binder-sand ratio, fly ash content, light burned magnesium content, fiber content, fiber type, phosphate type and mixing water temperature, and the output parameter is the peak value prediction of the hydration temperature of MPC-UHPC paste. The input parameters of the model in the mechanical property judgment process are the environmental temperature, magnesium-phosphorus ratio, water-binder ratio, boron-magnesium ratio, binder-sand ratio, fly ash content, light burned magnesium content, fiber content, fiber type, phosphate type, mixing water temperature, curing age and the peak value of hydration temperature output by the hydration temperature judgment, and the output parameter is the compressive strength prediction of MPC-UHPC. The input parameters of the model in the durability judgment process are the environmental temperature, curing age and the compressive strength prediction output by the mechanical property judgment, and the output parameter is the maximum freeze-thaw cycle prediction of MPC-UHPC.

[0078] S7: The Ada-stacking model for the three-level judgment of hydration temperature, mechanical property and durability is trained and tested by using the training set and the test set respectively. The R 2 value of the test set is taken as the evaluation index, and the prediction accuracy of the Ada-stacking model for the peak value of hydration temperature, compressive strength and maximum freeze-thaw cycle of MPC-UHPC in negative temperature environment is obtained, and the R 2 value is controlled to be above 0.8.

[0079] S8: Enter the prediction process, input the real data of the actual MPC-UHPC material to be verified into the Ada-stacking model, including the environmental temperature, magnesium-phosphorus ratio, water-binder ratio, boron-magnesium ratio, binder-sand ratio, fly ash content, light burned magnesium content, fiber content, fiber type, phosphate type and mixing water temperature, and the Ada-stacking model outputs the peak value prediction of the hydration temperature, the compressive strength prediction and the maximum freeze-thaw cycle prediction corresponding to the actual MPC-UHPC material to be verified.

[0080] S9: Enter the judgment process, first, the hydration temperature judgment is carried out, and the feasibility of MPC-UHPC hydration reaction in negative temperature environment is judged according to constraint condition one, and constraint condition one is represented as:

[0081] (8)

[0082] In the formula, is the peak value prediction of the hydration temperature output by step S8, is the environmental temperature.

[0083] If the conditions are met, it is determined that the hydration conditions are met, and the output is "hydration conditions are met, and early hydration reaction can be carried out", and step S10 is entered. If the conditions are not met, it is determined that the hydration conditions are not met, and the output is "material failure, and stirring cannot be carried out", and it is determined that the process is ended, and no subsequent research and judgment steps are performed.

[0084] S10: Mechanical property research and judgment is performed, the strength grade of MPC-UHPC in a negative temperature environment is determined according to constraint condition two, and constraint condition two is represented as:

[0085] (9)

[0086] In the formula, is the compressive strength prediction value output by step S8.

[0087] When , the value is 0, when , the value is 1, and when , the value is 2.

[0088] S11: Durability performance research and judgment is performed, the durability performance of MPC-UHPC in a negative temperature environment is determined according to constraint condition three, and constraint condition three is represented as:

[0089] (10)

[0090] In the formula, is the maximum freeze-thaw cycle number prediction value output by step S8.

[0091] When , the value is 0, when , the value is 1, and when , the value is 2.

[0092] S12: According to the values corresponding to the mechanical property and durability performance research and judgment in steps S10 and S11, the coding value CEC of MPC-UHPC negative temperature service performance research and judgment based on the value constraint is constructed through formula (11), and the calculation formula is:

[0093] (11)

[0094] In the formula, is the value determined according to constraint condition two in step S10, is the value determined according to constraint condition three in step S11.

[0095] S13: According to the coding value CEC in step S12, there are nine conditions, as shown in the following table:

[0096]

[0097] Finally, the Ada-stacking model outputs the corresponding applicable scenario according to the meaning of the encoding value CEC.

[0098] Embodiment

[0099] Five groups of MPC-UHPC materials are given below, and their applicable scenarios are determined by the method of the application, which correspond to Embodiments 1-5. The mixing ratio parameters of each group of materials are shown in the following table:

[0100]

[0101] In each embodiment, the magnesium oxide, light-burned magnesium, ammonium dihydrogen phosphate, potassium dihydrogen phosphate and borax used in step S1 are purchased from Shanghai Qiren Chemical Co., Ltd., the fly ash is purchased from Jiangsu Jiupin Energy-saving and Environmental Protection Material Co., Ltd., and the steel fiber and PVA fiber are purchased from Ganzhou Daye Metal Fiber Co., Ltd. In step S2, the hydration temperature peak data obtained by experiment are 1026 groups, the compressive strength data at 1d, 3d, 7d and 28d ages are 2763 groups in total, and the maximum freeze-thaw cycle number data of MPC-UHPC at 28d age are 1962 groups. In step S3, the division ratio of the data set is 8:2. The Ada-stacking model used in steps S4-S6 is consistent with the description above. In step S7, the Ada-stacking model outputs the predicted value of the hydration temperature peak of MPC-UHPC under negative temperature environment, the test set R 2 The values are 0.989, 0.984 and 0.974, respectively, all of which are above 0.8. Each embodiment is described from the prediction process as follows:

[0102] Embodiment 1

[0103] S8: Input group 1 data into the Ada-stacking model, and the model outputs the predicted value of the actual material hydration temperature peak as 56℃, the predicted value of the compressive strength at 28d age as 127.6MPa, and the predicted value of the maximum freeze-thaw cycle number as 673 times;

[0104] S9: According to constraint condition one , it is determined that the hydration condition is met, and the output is “the hydration condition is met, and the early hydration reaction can be carried out”;

[0105] S10: According to constraint condition two , the value is assigned as 2;

[0106] S11: According to constraint condition three , the value is assigned as 2;

[0107] S12: CEC=22;

[0108] S13: The output applicable scenario is "environment with compressive strength requirement > 100 MPa and freeze-thaw cycle > 500 times".

[0109] Example 2

[0110] S8: Input group 2 data into the Ada-stacking model, the model outputs the actual material hydration temperature peak prediction value to be verified as 37°C, the 28d age compressive strength prediction value as 86.3MPa, and the maximum freeze-thaw cycle prediction value as 682 times;

[0111] S9: According to constraint condition one, it is determined that , it is determined that the hydration condition is met, and the output is "hydration condition is met, early hydration reaction can be carried out";

[0112] S10: According to constraint condition two, it is determined that , the value is 1;

[0113] S11: According to constraint condition three, it is determined that , the value is 2;

[0114] S12: CEC=12;

[0115] S13: The output applicable scenario is "environment with compressive strength requirement > 100 MPa and freeze-thaw cycle > 500 times".

[0116] Example 3

[0117] S8: Input group 3 data into the Ada-stacking model, the model outputs the actual material hydration temperature peak prediction value to be verified as 5°C, the 28d age compressive strength prediction value as 3.1MPa, and the maximum freeze-thaw cycle prediction value as 0 times;

[0118] S9: According to constraint condition one, it is determined that , it is determined that the hydration condition is not met, the output is "material failure, cannot be stirred", and the determination process is ended.

[0119] Example 4

[0120] S8: Input group 4 data into the Ada-stacking model, the model outputs the actual material hydration temperature peak prediction value to be verified as 29°C, the 28d age compressive strength prediction value as 106.3MPa, and the maximum freeze-thaw cycle prediction value as 463 times;

[0121] S9: According to constraint condition one, it is determined that , it is determined that the hydration condition is met, and the output is "hydration condition is met, early hydration reaction can be carried out";

[0122] S10: According to constraint condition two, it is determined that , assigned value 2;

[0123] S11: determining according to constraint condition three , assigned value 1;

[0124] S12: CEC=21;

[0125] S13: outputting the applicable scenario as “environment of compressive strength requirement > 100 MPa and 250 times < freeze-thaw cycle ≤ 500 times”.

[0126] Example 5

[0127] S8: inputting group 5 data into the Ada-stacking model, and the model outputting the actual material hydration temperature peak prediction value to be verified as 27℃, the 28d age compressive strength prediction value as 54.42 MPa, and the maximum freeze-thaw cycle prediction value as 272 times;

[0128] S9: determining according to constraint condition one , determining as having hydration condition, and outputting “hydration condition meets, early hydration reaction can be carried out”;

[0129] S10: determining according to constraint condition two , assigned value 0;

[0130] S11: determining according to constraint condition three , assigned value 1;

[0131] S12: CEC=01;

[0132] S13: outputting the applicable scenario as “environment of compressive strength requirement ≤ 60 MPa and 250 times < freeze-thaw cycle ≤ 500 times”.

[0133] It is apparent to a person skilled in the art that the present application is not limited to the details of the above-described exemplary embodiments, but can be implemented in other embodiments without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and not restrictive, the scope of the application being defined by the appended claims rather than that of the above description, and it is intended to embrace all changes and modifications that fall within the meaning and scope of equivalents of the claims. Any reference signs in the claims should not be construed as limiting the claims to the figures in which the reference signs are used.

[0134] Furthermore, it should be understood that although the specification is described in terms of embodiments, not every embodiment includes every feature described. The specification can include implicit combinations of explicitly mentioned features and / or explicit combinations of implicitely mentioned features. Each embodiment depends on the explicit combinations of features and / or the implicit combinations of features made specifically within that embodiment, and each such embodiment can be combined with every other such embodiment to create further embodiments.

Claims

1. A data-driven comprehensive evaluation method for the negative temperature performance of MPC-UHPC, characterized in that: Includes the following steps: S1: MPC-UHPC slurry was prepared by stirring in the laboratory under an environment of -30℃~0℃. The mixing parameters included magnesium-phosphorus ratio, water-cement ratio, boron-magnesium ratio, mortar ratio, fly ash content, light calcined magnesium content, fiber content, fiber type, phosphate type, and mixing water temperature. S2: After stirring, the peak hydration temperature of the MPC-UHPC slurry was tested. After curing the molded specimens, the compressive strength at four ages (1d, 3d, 7d, and 28d) and the maximum number of freeze-thaw cycles at 28d were tested to construct a dataset of MPC-UHPC negative temperature performance. S3: Perform dataset cleaning and outlier removal, perform unique thermal encoding of fiber type, phosphate type and mixing water temperature, and divide the dataset into training set and test set. S4: The judgment model is built based on the stacking model. The base models include RF, XGBoost, LightGBM and CatBoost, and the meta-model is Bayesian Ridge Regression. S5: The prediction error and accuracy coefficient of each base model are calculated through 5-fold cross-validation. After normalization and smoothing, the normalized weights of each base model are obtained, forming an Ada-stacking model with adaptive weight adjustment. S6: The Ada-stacking model is used to conduct three-level assessments of hydration temperature, mechanical properties, and durability. The input parameters of the model for hydration temperature assessment are ambient temperature and mix proportion parameters, and the output parameter is the predicted peak value of hydration temperature. The input parameters of the model for mechanical property assessment are ambient temperature, mix proportion parameters, curing age, and the peak value of hydration temperature output from the hydration temperature assessment, and the output parameter is the predicted value of compressive strength. The input parameters of the model for durability assessment are ambient temperature, curing age, and the predicted value of compressive strength output from the mechanical property assessment, and the output parameter is the predicted value of the maximum number of freeze-thaw cycles. S7: Train and test the three-layer judgment Ada-stacking model using the training and test sets respectively, and control the R of the test set. 2 Values ​​above 0.8; S8: Input the actual data of the material to be verified into the Ada-stacking model, and output its peak hydration temperature, compressive strength and maximum number of freeze-thaw cycles; S9: Conduct a hydration temperature assessment. Based on constraint one, determine the feasibility of MPC-UHPC undergoing a hydration reaction in a negative temperature environment. Constraint one is expressed as: （8） In the formula, This is the predicted peak value of the output hydration temperature. If the ambient temperature meets the condition, it is determined that the hydration condition is met, and the process proceeds to step S10. If the condition does not meet the condition, it is determined that the hydration condition is not met, and the determination process ends. S10: Conduct mechanical property assessment. Determine the strength grade of MPC-UHPC under negative temperature conditions based on constraint condition two. Constraint condition two is expressed as follows: （9） In the formula, The output is the predicted compressive strength value, when When, the value is assigned to 0, when When, assign a value of 1, when When the time comes, the value is assigned to 2; S11: Conduct a durability performance assessment. Determine the durability of MPC-UHPC under negative temperature conditions based on constraint condition three, which is expressed as follows: （10） In the formula, The predicted maximum number of freeze-thaw cycles is the output value when... When, the value is assigned to 0, when When, assign a value of 1, when When the time comes, the value is assigned to 2; S12: Based on the assigned values ​​corresponding to the mechanical and durability performance assessments in steps S10 and S11, the coded value CEC for the MPC-UHPC negative temperature service performance assessment based on the assigned value constraints is constructed using formula (11). The calculation formula is as follows: （11） In the formula, The value is determined based on constraint condition two in step S10. The value is determined according to constraint condition three in step S11; S13: The applicable scenarios defined by the Ada-stacking model based on the CEC output of the encoded value.

2. The data-driven comprehensive evaluation method for the negative temperature performance of MPC-UHPC according to claim 1, characterized in that: In step S1, the specific proportion parameters include: a magnesium-to-phosphorus ratio of 3-6, a water-to-binder ratio of 0.10-0.20, a boron-to-magnesium ratio of 0.02-0.10, a binder-to-mortar ratio of 0.5-1.5, a fly ash content of 0%-20%, a light-burned magnesium content of 0%-10%, a fiber content of 0-1.5%, and the fiber types include steel fiber and PVA fiber. The phosphate types include ammonium dihydrogen phosphate and potassium dihydrogen phosphate, and the mixing water temperature includes 0℃ and 18℃.

3. The data-driven comprehensive evaluation method for the negative temperature performance of MPC-UHPC according to claim 1, characterized in that: The specific process of step S5 includes: The weight ratios of the four base models are adjusted, the training set is divided into five mutually exclusive subsets, one of which is selected as the validation set. Five-fold cross-validation is performed on each base model, and the prediction error of the four base models on the validation set is calculated according to formula (1). , means as follows: （1） In the formula, The number representing the base model. The number of folds represents the cross-validation factor. Representing the The number of samples in the validation set. For the first Measured values ​​of a sample For the first Predicted values ​​for each sample; The accuracy coefficients of each base model in the first to fifth folds are calculated according to formula (2). , means as follows: （2） In the formula, for To prevent the denominator from being 0; According to formula (3) After normalization, the initial weights of the four base models at folds 1 to 5 are obtained respectively. According to formula (4), the weights of each fold of each base model are smoothed to obtain the smoothed weights. , means as follows: （3） （4） In the formula, when hour Take [0,1), when hour Take 1; Use formula (5) to Normalization is performed to obtain the normalized weights for each base model. The meta-model training set is calculated according to formula (6). The meta-model is trained, and the meta-model test set is calculated according to formula (7). The meta-model was tested and is represented as follows: （5） （6） （7） In the formula, For the first Each base model on the test set samples The predicted value, This is a sample set for testing.

4. The data-driven comprehensive evaluation method for the negative temperature performance of MPC-UHPC according to claim 1, characterized in that: In step S13, the applicable scenarios corresponding to the meaning of the encoded value CEC are defined as follows: CEC=00 is applicable to environments where the compressive strength requirement is ≤60MPa and the freeze-thaw cycle requirement is ≤250. CEC=01 is applicable to environments where the compressive strength requirement is ≤60MPa and the number of freeze-thaw cycles is 250-500. CEC=02 is applicable to environments where the compressive strength requirement is ≤60MPa and the freeze-thaw cycle requirement is >500 cycles. CEC=10 is applicable to environments where the compressive strength requirement is between 60MPa and 100MPa and the freeze-thaw cycle requirement is between 250 and 1000. CEC=20 is applicable to environments where the compressive strength requirement is >100MPa and the freeze-thaw cycle requirement is ≤250. CEC=11, applicable scenarios are those with compressive strength requirements of 60MPa to 100MPa and freeze-thaw cycles of 250 to 500. CEC=12, applicable scenarios are those with a compressive strength requirement of 60MPa < compressive strength requirement ≤ 100MPa and a freeze-thaw cycle requirement of > 500 times; CEC=21, applicable scenarios corresponding to compressive strength requirements >100MPa and 250 < freeze-thaw cycles ≤500 cycles; CEC=22, applicable to environments with compressive strength requirements >100MPa and freeze-thaw cycles >500.

Images