Abstract:The conventional design method of concrete mix ratio relies on a large number of tests for trial mixing and optimization, and the workload is massive. It is challenging to cope with today's diverse raw materials and the concrete's specific performance to fit modern concrete development. To innovate the design method of concrete mix ratio and effectively use the various complex novel raw materials, the traditional mix ratio test method can be replaced with the intelligent optimization algorithm, and the concret… Show more
“…In order to improve the convergence rate of the network and avoid the deviation adjustment of weight caused by dimensional differences [44], normalization was used for data preprocessing. The scale transformation of original data was conducted according to Equation (2), which is realized by [y, ps] = mapminmax (x, y min , y max ) in MATLAB, where ps represents the mapping relationship; x and y are the data before and after normalization; y max , y min is the maximum and minimum of normalized boundary, respectively; and x max , x min is the maximum and minimum of input before normalization, respectively.…”
Section: Data Collection and Preprocessingmentioning
confidence: 99%
“…Furthermore, abundant soft computing methods have been successfully applied to various materials. These advanced techniques include regression tree, random forest, support vector machine, extreme learning machine, genetic programming, and Gaussian process regression [2,45,46,58]. With the help of deep learning, more work is foreseeable in the future to broaden the application under different experimental conditions, to concrete with different material compositions, and eventually to structural components [20,54,59].…”
Section: Limitations and Future Workmentioning
confidence: 99%
“…As a complex multi-phase composite material, concrete exhibits significant discreteness in fatigue life [1]. Moreover, since a nonlinear mapping relationship exists between fatigue life and its influencing factors [2], fatigue life estimation has become the focus of concrete fatigue research.…”
Concrete tensile properties usually govern the fatigue cracking of structural components such as bridge decks under repetitive loading. A fatigue life reliability analysis of commonly used ordinary cement concrete is desirable. As fatigue is affected by many interlinked factors whose effect is nonlinear, a unanimous consensus on the quantitative measurement of these factors has not yet been achieved. Benefiting from its unique self-learning ability and strong generalization capability, the Bayesian regularized backpropagation neural network (BR-BPNN) was proposed to predict concrete behavior in tensile fatigue. A total of 432 effective data points were collected from the literature, and an optimal model was determined with various combinations of network parameters. The average relative impact value (ARIV) was constructed to evaluate the correlation between fatigue life and its influencing parameters (maximum stress level Smax, stress ratio R, static strength f, failure probability P). ARIV results were compared with other factor assessment methods (weight equation and multiple linear regression analyses). Using BR-BPNN, S-N curves were obtained for the combinations of R = 0.1, 0.2, 0.5; f = 5, 6, 7 MPa; P = 5%, 50%, 95%. The tensile fatigue results under different testing conditions were finally compared for compatibility. It was concluded that Smax had the most significant negative effect on fatigue life; and the degree of influence of R, P, and f, which positively correlated with fatigue life, decreased successively. ARIV was confirmed as a feasible way to analyze the importance of parameters and could be recommended for future applications. It was found that the predicted logarithmic fatigue life agreed well with the test results and conventional data fitting curves, indicating the reliability of the BR-BPNN model in predicting concrete tensile fatigue behavior. These probabilistic fatigue curves could provide insights into fatigue test program design and fatigue evaluation. Since the overall correlation coefficient between the prediction and experimental results reached 0.99, the experimental results of plain concrete under flexural tension, axial tension, and splitting tension could be combined in future analyses. Besides utilizing the valuable fatigue test data available in the literature, this work provided evidence of the successful application of BR-BPNN on concrete fatigue prediction. Although a more accurate and comprehensive method was derived in the current study, caution should still be exercised when utilizing this method.
“…In order to improve the convergence rate of the network and avoid the deviation adjustment of weight caused by dimensional differences [44], normalization was used for data preprocessing. The scale transformation of original data was conducted according to Equation (2), which is realized by [y, ps] = mapminmax (x, y min , y max ) in MATLAB, where ps represents the mapping relationship; x and y are the data before and after normalization; y max , y min is the maximum and minimum of normalized boundary, respectively; and x max , x min is the maximum and minimum of input before normalization, respectively.…”
Section: Data Collection and Preprocessingmentioning
confidence: 99%
“…Furthermore, abundant soft computing methods have been successfully applied to various materials. These advanced techniques include regression tree, random forest, support vector machine, extreme learning machine, genetic programming, and Gaussian process regression [2,45,46,58]. With the help of deep learning, more work is foreseeable in the future to broaden the application under different experimental conditions, to concrete with different material compositions, and eventually to structural components [20,54,59].…”
Section: Limitations and Future Workmentioning
confidence: 99%
“…As a complex multi-phase composite material, concrete exhibits significant discreteness in fatigue life [1]. Moreover, since a nonlinear mapping relationship exists between fatigue life and its influencing factors [2], fatigue life estimation has become the focus of concrete fatigue research.…”
Concrete tensile properties usually govern the fatigue cracking of structural components such as bridge decks under repetitive loading. A fatigue life reliability analysis of commonly used ordinary cement concrete is desirable. As fatigue is affected by many interlinked factors whose effect is nonlinear, a unanimous consensus on the quantitative measurement of these factors has not yet been achieved. Benefiting from its unique self-learning ability and strong generalization capability, the Bayesian regularized backpropagation neural network (BR-BPNN) was proposed to predict concrete behavior in tensile fatigue. A total of 432 effective data points were collected from the literature, and an optimal model was determined with various combinations of network parameters. The average relative impact value (ARIV) was constructed to evaluate the correlation between fatigue life and its influencing parameters (maximum stress level Smax, stress ratio R, static strength f, failure probability P). ARIV results were compared with other factor assessment methods (weight equation and multiple linear regression analyses). Using BR-BPNN, S-N curves were obtained for the combinations of R = 0.1, 0.2, 0.5; f = 5, 6, 7 MPa; P = 5%, 50%, 95%. The tensile fatigue results under different testing conditions were finally compared for compatibility. It was concluded that Smax had the most significant negative effect on fatigue life; and the degree of influence of R, P, and f, which positively correlated with fatigue life, decreased successively. ARIV was confirmed as a feasible way to analyze the importance of parameters and could be recommended for future applications. It was found that the predicted logarithmic fatigue life agreed well with the test results and conventional data fitting curves, indicating the reliability of the BR-BPNN model in predicting concrete tensile fatigue behavior. These probabilistic fatigue curves could provide insights into fatigue test program design and fatigue evaluation. Since the overall correlation coefficient between the prediction and experimental results reached 0.99, the experimental results of plain concrete under flexural tension, axial tension, and splitting tension could be combined in future analyses. Besides utilizing the valuable fatigue test data available in the literature, this work provided evidence of the successful application of BR-BPNN on concrete fatigue prediction. Although a more accurate and comprehensive method was derived in the current study, caution should still be exercised when utilizing this method.
“…To address these issues, machine learning techniques are used to predict the compressive strength of concrete. In fact, with the development of artificial intelligence, various machine learning algorithms such as artificial neural network (ANN), support vector machines (SVM), random forest, and extreme learning machine (ELM) have been applied to predict the mechanical properties of concrete [12][13][14][15][16][17][18][19][20][21][22][23]. Ly et al [24] employed an optimal deep neural network model on a database of 223 experimental data to predict the 28 days compressive strength of rubber concrete and achieved a high prediction accuracy of R � 0.9874.…”
Support vector regression (SVR) has been applied to the prediction of mechanical properties of concrete, but the selection of its hyperparameters has been a key factor affecting the prediction accuracy. To this end, hybrid machine learning combines the SVR model and grid search (GS), namely, the GS-SVR model was proposed to predict the compressive strength of concrete and sensitivity analysis in this work. The hybrid model was trained and tested on a total of 98 datasets retrieved from literature, and the model performance was compared with the original SVR model under the same datasets. The obtained results in terms of R of 0.981, MSE of 3.44, RMSE of 1.85, MAE of 1.17, and MAPE of 0.05 demonstrate that the GS-SVR model proposed can be a candidate method for compressive strength prediction in subsequent related studies. Additionally, a graphical user interface (GUI) was developed to conveniently provide some initial estimates of the outcomes before performing extensive laboratory or fieldwork. Finally, the effect of each variable on the compressive strength in a random environment was analyzed.
“…However, the laboratory test method has many disadvantages, such as low efficiency and high cost [37][38][39][40][41][42][43]. To find a more efficient and low-cost method to predict the performance of concrete, many researchers choose to use machine learning models to predict the properties of concrete [44][45][46][47][48][49][50][51]. Salimbahrami et al studied the compressive strength prediction methods of recycled concrete based on the artificial neural network (ANN) and support vector machine (SVM), and the research results show that machine learning models have good prediction effects on the compressive strength of recycled concrete [52].…”
Concrete production by replacing cement with green materials has been conducted in recent years considering the strategy of sustainable development. This study researched the topic of compressive strength regarding one type of green concrete containing blast furnace slag. Although some researchers have proposed using machine learning models to predict the compressive strength of concrete, few researchers have compared the prediction accuracy of different machine learning models on the compressive strength of concrete. Firstly, the hyperparameters of BP neural network (BPNN), support vector machine (SVM), decision tree (DT), random forest (RF), K-nearest neighbor algorithm (KNN), logistic regression (LR), and multiple linear regression (MLR) are tuned by the beetle antennae search algorithm (BAS). Then, the prediction effects of the above seven machine learning models on the compressive strength of concrete are evaluated and compared. The comparison results show that KNN has higher R values and lower RSME values both in the training set and test set; that is, KNN is the best model for predicting the compressive strength of concrete among the seven machine learning models mentioned above.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.