Concrete compressive strength prediction method and device based on Stacking ensemble learning model, computer program product and storage medium
By using a Stacking ensemble learning model to predict concrete compressive strength by integrating multiple factors, the problem of large prediction bias in traditional methods is solved, enabling rapid and accurate concrete strength assessment and supporting timely decision-making and quality control in engineering construction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-10
- Publication Date
- 2026-03-10
AI Technical Summary
Existing technologies cannot quickly and accurately predict the compressive strength of modern concrete. Traditional methods, such as the Balromy formula, fail to fully consider the influence of various components, resulting in large prediction deviations and making it difficult to meet the immediate and accurate requirements of modern engineering construction.
A Stacking-based ensemble learning model was adopted to comprehensively consider multiple factors such as water-cement ratio, sand ratio, paste-aggregate ratio, and 28-day compressive strength of cementitious materials. The Stacking ensemble learning model was used to predict the compressive strength of concrete. Machine learning models such as GBDT, DT, KNN, RF, Ridge, and GA-BP were used to generate preliminary prediction results, which were then fused using an SVM model.
It enables rapid and accurate prediction of the 28-day compressive strength of concrete, reduces testing cycles and resource consumption, provides real-time and accurate data support, and improves the efficiency and quality control of engineering construction.
Smart Images
Figure CN121637636A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of concrete technology, specifically to a method, apparatus, computer program product, and storage medium for predicting the compressive strength of concrete based on a Stacking ensemble learning model. Background Technology
[0002] Concrete, a fundamental material widely used in civil engineering, building construction, and transportation infrastructure, has compressive strength as one of its core mechanical properties. The compressive strength of concrete directly affects the safety and durability of engineering structures; therefore, the testing and control of compressive strength are of paramount importance during concrete production and construction.
[0003] Currently, concrete compressive strength testing generally follows a standard testing procedure: first, the concrete mixture is prepared according to the design mix proportions; then, the concrete is poured into standard molds and cured for 28 days before the compressive strength test is conducted. This process not only consumes a large amount of concrete materials and labor costs, but also, due to the 28-day curing period, the time required to obtain test results is relatively long, making it difficult to provide timely and efficient support for technical decision-making in the early stages of concrete production or construction. Especially during engineering construction, when production schedules are tight, traditional compressive strength testing often cannot meet the needs for immediate and accurate quality control.
[0004] To optimize this process, the prediction of concrete compressive strength has become a key research focus. Traditionally, the Bolomey Formula has been widely used for predicting concrete compressive strength. This formula estimates the compressive strength of concrete using two variables: the water-cement ratio and the strength of the cementitious materials. However, the Bolomey Formula only considers these two factors and fails to adequately account for the influence of various components in modern concrete on compressive strength. The composition of modern concrete is complex, including not only traditional components such as cement, sand, and aggregates, but also mineral admixtures and additives. The types and proportions of these components have a direct and significant impact on the compressive strength of concrete. Therefore, the Bolomey Formula may produce significant prediction bias in practical applications and is difficult to accurately reflect the true strength of modern concrete.
[0005] Furthermore, with the diversification of concrete mix proportions, traditional prediction methods based solely on the water-cement ratio and cementitious material strength are no longer sufficient to meet the demands of accurate prediction of modern concrete performance. Therefore, it is particularly important to accurately predict concrete compressive strength by considering more influencing factors in light of the actual characteristics of modern concrete. A more comprehensive and scientific prediction model can not only improve concrete production efficiency and reduce resource waste, but also better control concrete quality during construction, ensuring the safety and durability of projects. Summary of the Invention
[0006] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is: how to provide a prediction method with high efficiency and accuracy that can meet the needs of rapid assessment of concrete compressive strength in modern engineering construction.
[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A method for predicting the compressive strength of concrete based on a Stacking ensemble learning model is proposed. The method acquires characteristic parameters of concrete, including water-cement ratio, sand ratio, paste-aggregate ratio, 28-day compressive strength of cementitious materials, coarse aggregate gradation parameters, coarse aggregate porosity, fine aggregate gradation parameters, fineness modulus, and mortar allowance coefficient. These concrete characteristic parameters are then input into a pre-trained Stacking ensemble learning model, which includes a base model layer and a meta-model layer. The base model layer integrates at least two different machine learning models to generate initial prediction results, while the meta-model layer fuses all initial prediction results. Finally, the method outputs the predicted 28-day compressive strength of concrete.
[0008] As an optimization, the coarse aggregate gradation parameters include the individual sieve residue values of coarse aggregate within a preset particle size range, and the fine aggregate gradation parameters include the cumulative sieve residue values of fine aggregate within a preset particle size range. The preset particle size range covers the typical sieve range of continuous particle size coarse and fine aggregates.
[0009] As an optimization, the machine learning model includes GBDT model, DT model, KNN model, RF model, Ridge model and GA-BP model, and the meta-model layer uses SVM model to fuse the primary prediction results.
[0010] As an optimization, the pre-training method of the Stacking ensemble learning model includes: collecting several concrete feature parameters and corresponding 28-day compressive strength data of concrete; normalizing the above data and dividing it into training set and test set; training the base model layer through parameter optimization; the meta-model layer adopts the SVM model; retraining the results obtained from each machine learning model; and obtaining the optimal parameters through hyperparameter optimization.
[0011] A concrete compressive strength prediction device includes a data input module, a processing module, and an output module. The data input module receives water-cement ratio, sand ratio, paste-aggregate ratio, 28-day compressive strength of cementitious materials, coarse aggregate gradation parameters, coarse aggregate porosity, fine aggregate gradation parameters, fineness modulus, and mortar allowance coefficient. The processing module embeds the Stacking ensemble learning model described above. The output module displays the predicted 28-day compressive strength of concrete.
[0012] A computer program product includes a computer program that, when executed by a computer, implements the method described above.
[0013] A computer-readable storage medium having a computer program stored thereon, which, when executed by a computer, implements the method described above.
[0014] In concrete quality control, strength is one of the most critical indicators. Predicting concrete strength typically relies on multiple factors. The traditional Borromie formula only includes two independent variables: the water-cement ratio and the strength of the cementitious materials. However, modern concrete, as a heterogeneous material, is affected by various factors beyond just the water-cement ratio and cementitious material strength. Based on these considerations, this invention aims to establish a multi-factor method for predicting concrete strength that comprehensively considers factors such as water-cement ratio, sand ratio, paste-aggregate ratio, 28-day compressive strength of the cementitious materials, coarse aggregate gradation parameters, coarse aggregate porosity, fine aggregate gradation parameters, fineness modulus, and mortar allowance coefficient. This application employs a multi-factor model, utilizing existing concrete test data to predict concrete strength, thereby reducing practical workload and improving the efficiency of strength data acquisition.
[0015] With the development of computer science, ensemble learning methods have gradually become important tools for solving complex problems. Stacking ensemble learning is a method that improves prediction accuracy by combining the prediction results of multiple base learners. Stacking can effectively combine the advantages of different models, reduce the bias of individual models, and thus improve the accuracy of concrete strength prediction.
[0016] This application utilizes the Stacking ensemble learning algorithm to establish a concrete strength prediction model, using concrete mix proportions and other relevant factors as independent variables and concrete strength as the dependent variable. This method enables more efficient prediction of concrete strength, providing timely and efficient data support for technical decisions during construction, optimizing resource utilization, and improving production efficiency.
[0017] Compared with the prior art, the present invention has the following advantages: (1) The traditional concrete strength test process is cumbersome and consumes a lot of time and resources. The method of this application can avoid the lengthy test cycle and high material consumption. This method relies on various parameters and raw material indicators in the design of concrete mixtures. No additional concrete strength test is required, thereby reducing a lot of experimental workload and improving work efficiency. (2) Traditional concrete compressive strength tests require a standard curing period of 28 days, which results in a long time to obtain concrete strength data and affects the timeliness of construction decisions. By using the method of this application, relevant data can be input in real time at the construction site, and the concrete strength level can be predicted quickly, significantly shortening the time to obtain strength data and providing real-time and accurate data support for construction and quality control. Attached Figure Description
[0018] Figure 1 This is a flowchart of the prediction process of the Stacking ensemble learning model in this invention; Figure 2 This is a diagram of the APP interface based on the computer program of this invention. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.
[0020] like Figure 1 As shown in the figure, the concrete compressive strength prediction method based on the Stacking ensemble learning model in this specific embodiment obtains concrete characteristic parameters, including water-cement ratio, sand ratio, paste-aggregate ratio, 28-day compressive strength of cementitious materials, coarse aggregate gradation parameters, coarse aggregate porosity, fine aggregate gradation parameters, fineness modulus, and mortar allowance coefficient; the concrete characteristic parameters are input into a pre-trained Stacking ensemble learning model, which includes a base model layer and a meta-model layer. The base model layer integrates at least two different machine learning models to generate primary prediction results, and the meta-model layer is used to fuse all primary prediction results; the predicted value of the 28-day compressive strength of concrete is output.
[0021] In this specific embodiment, the coarse aggregate gradation parameters include the individual sieve residue values of coarse aggregate within a preset particle size range, and the fine aggregate gradation parameters include the cumulative sieve residue values of fine aggregate within a preset particle size range. The preset particle size range covers the typical sieve range of continuous particle size coarse and fine aggregates.
[0022] In this specific embodiment, the machine learning model includes GBDT model, DT model, KNN model, RF model, Ridge model and GA-BP model, and the meta-model layer uses SVM model to fuse the primary prediction results.
[0023] In this specific embodiment, the pre-training method of the Stacking ensemble learning model includes: collecting several concrete feature parameters and corresponding 28-day compressive strength data of concrete; normalizing the above data and dividing it into training set and test set; training the base model layer through parameter optimization; the meta-model layer adopts the SVM model; retraining the results obtained from each machine learning model; and obtaining the optimal parameters through hyperparameter optimization.
[0024] A concrete compressive strength prediction device includes a data input module, a processing module, and an output module. The data input module receives water-cement ratio, sand ratio, paste-aggregate ratio, 28-day compressive strength of cementitious materials, coarse aggregate gradation parameters, coarse aggregate porosity, fine aggregate gradation parameters, fineness modulus, and mortar allowance coefficient. The processing module embeds the Stacking ensemble learning model described above. The output module displays the predicted 28-day compressive strength of concrete.
[0025] A computer program product includes a computer program that, when executed by a computer, implements the method described above.
[0026] A computer-readable storage medium having a computer program stored thereon, which, when executed by a computer, implements the method described above.
[0027] In practice, 12 sets of concrete experiments were conducted to test and verify the concrete strength. The concrete parameters are shown in Tables 1, 2, and 3.
[0028] Table 1
[0029] Table 2
[0030] Table 3 Table 2 shows the individual sieve residues for each grade of coarse aggregate, and Table 3 shows the cumulative sieve residues for each grade of fine aggregate. Additionally, the cement grade is P.O42.5; the fly ash is Class F; tap water is used; the fine aggregate is Zone II manufactured sand; and the coarse aggregate is continuous crushed stone with a particle size of 5~25mm.
[0031] The parameter settings for each machine learning model and meta-model are as follows: GBDT model hyperparameters: NumLearningCycles = 43, MaxNumSplits = 3020, MinLeafSize = 48, NumVariablesToSample = 13; DT model hyperparameters: MinLeafSize is 2, MaxNumSplits is 137, and NumVariablesToSample is 18; The hyperparameter K of the KNN model is 4; RF model hyperparameters: NumLearningCycles = 10, MinLeafSize = 22, MaxNumSplits = 23; The hyperparameter λ (regularization parameter) of the Ridge model is 1.3183e-07; The hyperparameters of the GA-BP model are: S1 = 26, epochs = 179, learning rate = 0.000791, batch size = 5.067187e+01, and lambda = 0.1498. Stacking ensemble learning model hyperparameters: BoxConstraint = 0.20471, Epsilon = 0.013849, KernelFunction = polynomial, PolynomialOrder = 2, Standardize = True.
[0032] The predicted values and prediction accuracy obtained by the Stacking ensemble learning model are shown in Table 4:
[0033] Table 4 Analysis of Table 4 shows that the absolute error between the predicted and measured values of concrete compressive strength remains within 6 MPa, with an average absolute error of 1.66 MPa. Regarding relative error, the average error is 5.02%. Furthermore, the ratio of predicted to measured values (Ypredicted / Ymeasured) falls between 0.887 and 1.156, indicating that the prediction method can reliably generate results close to the measured values. These results fully validate the effectiveness and reliability of the prediction method presented in this application. The high correlation between predicted and measured values demonstrates that this method can effectively reflect the actual strength level of concrete and possesses high prediction accuracy, providing effective support for practical engineering projects.
[0034] In summary, this invention can efficiently predict the 28-day compressive strength of concrete by relying solely on relevant parameters in the concrete mix design. This prediction method not only quickly determines whether the strength of the concrete mix meets design requirements, but also provides a scientific basis for flexibly adjusting the mix design during actual production, thereby improving the construction quality of concrete and facilitating the smooth progress of engineering projects.
[0035] Figure 2 The App developed based on the computer program design in this application can complete complex data processing with just one interface, and can perform predictive analysis more intuitively and efficiently.
[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described with reference to preferred embodiments, those skilled in the art should understand that various changes in form and detail can be made without departing from the spirit and scope of the invention as defined in the appended claims.
Claims
1. A method for predicting compressive strength of concrete based on a stacking ensemble learning model, characterized in that: The concrete characteristic parameters include water-binder ratio, sand ratio, paste-aggregate ratio, 28d compressive strength of cementitious material, coarse aggregate grading parameter, coarse aggregate void ratio, fine aggregate grading parameter, fineness modulus and mortar excess coefficient; the concrete characteristic parameters are input into a pre-trained Stacking integrated learning model, the Stacking integrated learning model comprises a base model layer and a meta model layer, the base model layer integrates at least two different machine learning models, and is used for generating primary prediction results, and the meta model layer is used for fusing all the primary prediction results; and a 28d compressive strength prediction value of the concrete is output.
2. The method of predicting the compressive strength of concrete based on the stacking ensemble learning model according to claim 1, characterized in that: The coarse aggregate grading parameter comprises a residual value of coarse aggregate in a preset particle size interval, and the fine aggregate grading parameter comprises a cumulative residual value of fine aggregate in a preset particle size interval, and the preset particle size interval covers a typical screening range of continuous particle size coarse aggregate and fine aggregate. 3.The method of predicting the compressive strength of concrete based on a stacking ensemble learning model according to claim 1, characterized in that: The machine learning model comprises a GBDT model, a DT model, a KNN model, an RF model, a Ridge model and a GA-BP model, and the meta model layer adopts an SVM model to fuse the primary prediction results. 4.The method of predicting the compressive strength of concrete based on a stacking ensemble learning model according to claim 1, characterized in that: The pre-training method of the Stacking integrated learning model comprises the following steps: collecting a plurality of concrete characteristic parameters and corresponding 28d compressive strength data of the concrete, normalizing the data and dividing the data into a training set and a test set; training the base model layer through parameter optimization, the meta model layer adopts an SVM model, the results obtained by each machine learning model are retrained, and optimal parameters are obtained through hyperparameter optimization.
5. A concrete compressive strength prediction device characterized by comprising: a concrete compressive strength prediction program according to any one of claims 1 to 4. The system comprises a data input module, a processing module and an output module; the data input module is used for receiving water-binder ratio, sand ratio, paste-aggregate ratio, 28d compressive strength of cementitious material, coarse aggregate grading parameter, coarse aggregate void ratio, fine aggregate grading parameter, fineness modulus and mortar excess coefficient; the processing module is embedded with the Stacking integrated learning model according to any one of claims 1 to 4; and the output module displays a 28d compressive strength prediction value of the concrete.
6. A computer program product, characterised in that: The computer program is executed by a computer to realize the method according to any one of claims 1 to 4.
7. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by a computer to realize the method according to any one of claims 1 to 4.