Home Research on back analysis of meso-parameters of hydraulic cemented sand and gravel based on Box-Behnken design response surface
Article Open Access

Research on back analysis of meso-parameters of hydraulic cemented sand and gravel based on Box-Behnken design response surface

  • Lixia Guo ORCID logo , Yuhang Guo , Ling Zhong ORCID logo EMAIL logo and Jiantao Zhu
Published/Copyright: April 18, 2022
Download Article (PDF)
Become an author with De Gruyter Brill
Submit Manuscript Author Information
Science and Engineering of Composite Materials
From the journal Science and Engineering of Composite Materials Volume 29 Issue 1

Abstract

Cemented sand and gravel (CSG) is a new type of dam-building material. Aiming at the cumbersome process and long calculation time of traditional methods to invert the meso-parameters, a mesophase parameter inversion method based on Box-Behnken Design response surface was proposed. By constructing a response surface simulation test scheme with different inversion parameters (elastic modulus of aggregates, mortars and interfaces, and interfacial tensile strength), the stochastic aggregate model is used to complete the numerical simulation of the damage process, and different results are obtained. The equation between the response variable (stress at different loading times) and the independent variable (inversion parameter) was verified, and the rationality of the response model was verified; the action mechanism of mesophase parameters at different stages on the mechanical properties of the specimen was analysed. The test results are brought into the response surface model, and the meso-parameters are obtained by inverse analysis. The stress–strain curve obtained by numerical simulation with this parameter has an error of 1.1% at the peak stress and 3.27% at the peak strain. The accuracy is high, the number of test groups is much smaller than other conventional inversion methods, and has feasibility of application in CSG.

Keywords: cemented sand and gravel; stress–strain; meso-parameters inversion; response surface method

1 Introduction

Cemented sand and gravel (CSG) is a composite material consisting of water, non-screened aggregates, and a small amount of cementitious material [1], and its mechanical properties determine the usability of CSG [2,3,4,5,6,7]. Both CSG materials and concrete are three-phase composites [8,9,10], consisting of aggregate phase, mortar phase, and interface phase. However, the non-screening and non-washing characteristics of CSG aggregate make its mechanical properties different from traditional concrete materials. And the aggregates are selected locally, so the properties cannot be generalised. It is necessary to study the difference between CSG and ordinary concrete materials. The meso-analysis method can accurately grasp the difference and interaction between CSG and its components, especially the mechanical properties of the aggregate and mortar, and can accurately determine the force characteristics and failure conditions of CSG. It is a method commonly used by most researchers. The acquisition of meso-component parameters is particularly important [11].

CSG has little research on its parameters and parameter acquisition as a new building material. Scholars [12] have used many uniaxial compression and loading and unloading experiments of CSG to use the linear elastic constitutive model as their constitutive model. The parameters are obtained by experimental means [13]. Batmaz [14] simulated the nonlinear mechanical characteristics of CSG by piecewise linear method and applied the method and model to actual engineering. More research on the mechanical properties of CSG is still in macroscopic experiments [15,16,17], and there are fewer meso-chromatographic analyses.

Due to the scale effect and the fractal [18], the meso-components of the materials are often obtained by back-analysis methods. Genetic algorithms [19,20,21,22,23], for example, are prone to “prematureness” and require independent coding, which does not guarantee both accuracy and simplicity. At the same time, support vector machine models [24] and particle swarm algorithms [25,26] are also more complex. The response surface method is an optimisation method that combines the experimental design method and mathematical statistics proposed by Box and Wilson [27], and has good robustness [28,29]. Pei et al. [30] combined the response surface model and the genetic algorithm to perform inversion analysis of the thermodynamic parameters; Li et al. [31] verified that the response surface method is an optimal method of mathematical modelling, which can effectively reduce the number of experiments and can investigate the interaction between influencing factors. Li [32] and Wang [33] established response surfaces of different orders and analysed the concrete meso-parameters in inversion analysis, indicating that the inversion method of geotechnical materials has high prediction accuracy.

The documents mentioned above confirm the feasibility of a response surface in engineering back analysis. Still, it has not been studied to use it to obtain the meso-component parameters of CSG materials. This article solves the non-simultaneous response surface equations obtained from the BBD simulation test scheme by numerical simulation-response surface method experiment to obtain the fine-scale parameters of the test blocks, make theoretical pavement, and case study for the experimental study of CSG meso-parameters inversion, and propose new ideas for engineering applications.

2 Parametric inverse analysis

2.1 Principles of the response surface method

Response surface methodology is an experimental improvement and model optimisation method. It works by finding representative points in the global range, testing the representative points, obtaining the resultant function, and then explicitly fitting the global range of input variables (parameters) to the response variables through regression. A large number of applications are available [34,35,36].

According to Taylor’s polynomial theory, the response variable (dependent variable) can be fitted with a polynomial as a function of the input variable (independent variable). A second-order response surface model is used on synthesis, represented by equation (1).

(1) y = β 0 + ∑ i = 1 m β i x i + ∑ i = 1 m ∑ j = 1 m β i j x i x j + ε ,

where equation y represents the response, m represents the number of independent variables, that is, the number of controlled factors, x i represents the i independent variable, and β 0 represents the constant term coefficient,   β i represents the coefficient of the first term, β i j represents the coefficient of the quadratic term and the quadratic cross term, and ε represents the random error.

2.2 Basic steps of inversion of the response surface method

BBD is a commonly used experimental design method for response surface optimisation, which is suitable for exploring the relationship between 2 and 5 factors. Therefore, this article adopts the BBD experimental design, and selects three levels of factors, namely, the maximum value, the minimum value, and the median value.

The steps of parameter inversion based on the response surface method are [37]: (1) Determine the inversion parameters and other basic parameters. (2) Determine the parameter levels of the independent variables, and design the simulation experiment scheme. (3) Complete the numerical simulation calculation process, construct the response surface function, and verify the rationality of the response surface equation. (4) The laboratory test results are used as response variables, and are brought into the response surface model function for optimisation and solution, and the parameters to be inverted are obtained. (5) Perform mesoscopic numerical simulation using the obtained parameters to obtain the simulation results. The degree of agreement with the laboratory test results is compared and analysed to verify the rationality of the inversion parameters.

2.3 Analysis on the reasonableness of response surface model

The response surface model reflects the implicit functional relationship between the independent and response variables. Therefore, whether there is a statistical relationship in the mathematical sense between the independent variable and the response variable requires a significance test, analysis of variance, and a precision test.

Whether the constructed response surface model is reasonable or not needs to be tested. The control parameters for model testing are: R 2 and adjusted R-Squared ( R adj 2 ), F-test method, P-value test method, residual distribution test method, etc.

R 2 is used to evaluate the goodness of fit of the model and describes how well the input variables explain the output variables; the closer R 2 is to 1, the better the model is. It is calculated as follows:

(2) R 2 = 1 − ∑ t = 1 T ( x t r s − x t p ) 2 ∑ t = 1 T ( x t p − x ¯ t   ) 2 ,

where T is the number of check points in the design space, x t r s is the target value of the response surface calculation of the t-th parameter, and x t p is the target value of the finite element model calculation of the t-th parameter,   x ¯ t is the average value of the finite element model and the calculation results of the finite element model.

The adjusted R 2 ( R adj ) imposes penalties on the number of items that increase but do not improve the accuracy of the model, to increase the simplicity of the model while ensuring the reliability of the model. It is usually analysed together with the predictive determination coefficient ( R pred ). R pred reflects the predictive ability of the model and requires that the difference between the two does not exceed 0.2 as shown in the following equations (3) and (4):

(3) R adj = 1 − ( 1 − R 2 ) ( n − 1 ) T − P − 1 ,

(4) R adj − R pred < 0.2 .

The adjusted R 2 ( R adj ) increases model simplicity while ensuring model reliability by imposing a penalty on the number of terms that increase but do not improve the accuracy of the model. It is usually analysed in conjunction with the predicted R 2 ( R pred ), which reflects the predictive power of the model and requires a difference of no more than 0.2 as shown in equations (5) and (6).

(5) R adj = 1 − ( 1 − R 2 ) ( n − 1 ) T − P − 1 ,

(6) R adj − R pred < 0.2 ,

where P is the number of features, R adj lies between 0 and 1, and the larger the better, but if it is really close to 1, the over-fitting situation may occur.

Use the F-test method for hypothesis testing:

(7) F = SSR / p SSE / ( m − p − 1 ) ,

where SSR is the regression sum of squares of R, SSE is the sum of squares of the errors. p is the number of non-constant terms in the RSM function, and p and ( m − p − 1 ) are the degrees of freedom of S S R and S S E , respectively. For a given confidence level α , the RSM model is considered significant when F > F a ( p , m − p − 1 ) .

The signal-to-noise ratio, or Adeq Precision, is used to analyse the model’s interference items further when the model’s regression items are obtained. As long as the accuracy of the model is greater than 4, it is generally considered that the accuracy of the model is reasonable, and the calculation formula is shown in equation (8):

(8) Adeq precision = S N ,

where S represents the output value of the independent variable, and N represents the error value caused by the independent variable in the calculation process.

3 Analysis of calculation examples

3.1 Numerical simulation tests

3.1.1 Stochastic aggregate model

With the support of ANSYS’s APDL for secondary development, the model was created from a meso-level perspective treating the CSG material as a three-phase composite model consisting of natural sand and gravel aggregate, mortar matrix, and the interface between the mortar matrix and the aggregate, using 100 × 100 × 100 test blocks with a calculated aggregate volume rate of 70% and a secondary mix of aggregates with Fuller’s gradation, using the Walraven formula. As in equation (7), the number of aggregates in different intervals is generated by deriving the probability of occurrence of an inner truncated circle P C ( D < D 0 ) with an aggregate diameter of D < D 0 at any point, through the three-dimensional Fullerian gradation of the aggregate, from a two-dimensional cross-section.

(9) P C ( D < D 0 ) = P K 1.065 D 0 D max 0.5 − 0.053 D 0 D max 4 − 0.012 D 0 D max 6 − 0.0045 D 0 D max 8 − 0.0025 D 0 D max 10 ,

where D 0 is the diameter of the sieve hole, mm, D max is the maximum aggregate particle size, mm, D min is the minimum size of the aggregate, and P K is the percentage of aggregate volume, which is taken as 0.7.

Considering the non-sieving characteristics of the aggregate of CSG, the finite element model (FEM) 1 established by the random aggregate model is shown in Figure 1.

Figure 1 Two-dimensional random aggregate model: (a) FEM 1 and (b) mesophase material composition.
Figure 1

Two-dimensional random aggregate model: (a) FEM 1 and (b) mesophase material composition.

3.1.2 Realisation of the uniaxial compression process

The constitutive numerical simulation model adopts the linear elastic model, and the failure theory adopts the first strength theory. The uniaxial compression simulation test uses the displacement loading method, the same as the indoor test using the loading method. The analysis type is static analysis. The automatic time step is used to control the load substeps and kill the elements whose tensile stress exceeds the tensile strength of each phase material to simulate the failure process.

3.2 Establishment of response surface model and reliability analysis

Response surface design is based on design-export software, and image processing. The initial parameters were selected concerning the results of the sensitivity analysis of the relevant literature [38]. The linear elastic model was adopted according to the fine view, and the modulus of elasticity of the aggregate ( E G ), the modulus of elasticity of the mortar ( E m ), the modulus of elasticity of the interface ( E i ), and the tensile strength of the interface ( f i ) were selected as input variables. The parameters were taken according to three levels and the four factors and three levels were taken in Table 1. Since the indoor test process used displacement loading to obtain uniaxial compression stress–strain curves, and the parameters to be sought were four, the stress values at four points (corresponding after different loading steps) were used as the response quantities, and the strain values corresponding to the above response quantities are shown in Table 2. where ε 1 , ε 2 , and ε 3 are located before the peak strain and ε max is the peak strain.

Table 1

Table of factor levels

Factors Levels
−1 0 1
E g (MPa) 100 200 300
E m (MPa) 50 75 100
E i (MPa) 10 30 50
f i (MPa) 0.1 0.3 0.5
Table 2

Strain values

Strain Name ε 1 ε 2 ε 3 ε max
Strain value 0.00004 0.0041 0.0201 Selected based on numerical simulation results

Other parameters include aggregate Poisson’s ratio ( μ G ), mortar Poisson’s ratio ( μ m ), interface Poisson’s ratio ( μ i ), aggregate tensile strength ( f g ), and mortar tensile strength ( f m ). These four quantities are less sensitive to the overall intensity of CSG. Referring to the relevant literature and based on the parameters to be inverted, the values of other parameters are shown in Table 3.

Table 3

Values of basic parameters

Parameter Type μ G (MPa) μ m (MPa) μ i (MPa) f g (MPa) f m (MPa)
Value 0.16 0.2 0.16 0.6 0.5

The response surface simulation test design is shown in Table 4, and 29 sets of simulation test programs are obtained.

Table 4

Response surface design test parameter combination

Run Factor 1 Factor 2 Factor 3 Factor 4
A:   E G (MPa) B:   E m (MPa) C:   E i (MPa) D:   f i (MPa)
1 200 75 10 0.1
2 200 50 10 0.3
3 200 75 30 0.3
4 200 100 30 0.5
5 100 100 30 0.3
6 300 75 50 0.3
7 300 50 30 0.3
8 100 75 30 0.5
9 100 75 10 0.3
10 200 100 10 0.3
11 200 50 30 0.5
12 300 75 10 0.3
13 200 50 30 0.1
14 200 50 50 0.3
15 200 100 50 0.3
16 300 75 30 0.5
17 200 75 30 0.3
18 200 75 50 0.1
19 100 50 30 0.3
20 300 75 30 0.1
21 200 75 30 0.3
22 200 100 30 0.1
23 100 75 50 0.3
24 200 75 10 0.5
25 200 75 50 0.5
26 300 100 30 0.3
27 200 75 30 0.3
28 200 75 30 0.3
29 100 75 30 0.1

Performing uniaxial compression numerical simulation experiments based on the parameters in the table above, 29 sets of peak stress and stress–strain curves are obtained, and a quadratic polynomial with cross terms as the response surface function is used. For convenience, the response surface function based on the coded factor is expressed as follows:

(10) σ 1 = ( 37.65967353 + 4.487211221 A + 8.37310231 B + 3.339521452 C + 1.558910891 A B + 1.016089109 A C + 1.160891089 B C − 1.463403131 A 2 − 0.695952636 B 2 − 1.882215012 C 2 ) × 10 − 4 ,

(11) σ 2 = ( 0.386008563 + 0.045992574 A + 0.085825908 B + 0.034232673 C + 0.015980198 A B + 0.010418317 A C + 0.011898515 B C − 0.01499796 A 2 − 0.007129148 B 2 − 0.019291276 C 2 ) × 10 − 2 ,

(12) σ 3 = 1.893762376 + 0.192879538 A + 0.386815182 B + 0.172962046 C + 0.145825083 D + 0.072079208 A B + 0.089331683 A D + 0.058391089 B C + 0.095519802 B D − 0.072326733 A 2 − 0.049641089 B 2 − 0.086460396 C 2 − 0.146596535 D 2 ,

(13) σ max = 5.802389064 + 1.202444719 C + 1.282436469 D + 0.900346535 A D + 1.189727723 B C + 1.513094059 B D − 1.22575 C D − 1.105250028 B 2 − 1.284709187 D 2 , ,

where f 1 , f 2 , f 3 , f max are the corresponding stress values at ε 1 , ε 2 , ε 3 , and ε m a x , respectively. A is the modulus of elasticity of the aggregate, B is the modulus of elasticity of the mortar, C is the modulus of elasticity of the interface, and D is the tensile strength of the interface.

In the process of obtaining the four response surface equations, the regression equation coefficients ( R squared ), regression equation prediction coefficients ( R pred ), regression equation correction coefficients ( R pred ) and p-values (p-test) and F-values (F-test) were analysed to remove unreasonable terms, adjust out-of-fit terms, compare R adj with R pred , etc. Polynomial processing was carried out to correct the response surface model and determine the model’s overall reliability and accuracy, as shown in Table 5.

Table 5

Overall parameter levels of the model

Sum of square Mean square error F value p value R adj − R pred Prediction accuracy
σ 1 1.26957 × 10−5 1.41064 × 10−6 1,648.049555 <0.0001 0.002946756 14.65581
σ 2 0.133388289 0.014820921 1,647.783261 <0.0001 0.002946681 26.6141
σ 3 3.130542779 0.260878565 96.93932557 <0.0001 0.037642069 21.351527
σ max 78.61966947 9.827458683 15.63771354 <0.0001 0.098638087 23.59744

It can be seen from the table that the p-value and F-value are reasonable, and the four response surface models are overall significant, so the parameter selection is reliable. In addition, the difference between R adj and R pred is less than 0.2, indicating that the response surface model has good predictive ability, and all the prediction accuracy is greater than 4. Therefore, it is considered that the model accuracy meets the expected requirements, and the four response surface models are overall reliable. The residual normal distribution probability diagram of the response surface model is shown in Figure 2.

Figure 2 The normal distribution of the residuals of the regression model.
Figure 2

The normal distribution of the residuals of the regression model.

The residual normal distribution plot shows that the data fall on a straight line, which means that the errors are distributed regularly. In addition, the distribution of the residuals shows that the residuals are normally random, which shows the accuracy of the model [37].

The level of the parameters reflects the sensitivity of the meso-parameters, which is consistent with the initial assumptions and the verification conclusions of the cited literature. It further illustrates the rationality of the response surface model and judges that the control items are reasonable.

3.3 Response surface model analysis

According to the response surface model (equations (10)–(14)), a 1D graph is drawn corresponding to the intermediate level value of each parameter and the influence of the primary term of each variable on the overall stress value is analysed; a 3D response surface diagram of the interaction of factors is drawn, and the interaction of various parameters is observed visually (Figures 3 and 4).

Figure 3 Analysis of single factor impact: (a) σ 1 \hspace{.25em}{\sigma }_{1} analysis of single factor impact and (b) σ 2 \hspace{.25em}{\sigma }_{2} analysis of single factor impact.
Figure 3

Analysis of single factor impact: (a) σ 1 analysis of single factor impact and (b) σ 2 analysis of single factor impact.

Figure 4 Comparative analysis of the influence of interaction terms on   σ 1 and  σ 2 \text{&#x00A0;}{\sigma }_{1}\hspace{.5em}\text{and&#x00A0;}{\sigma }_{2} . (a) The influence of AB interaction term on   σ 1 \text{&#x00A0;}{\sigma }_{1} . (b) The influence of AB interaction term on   σ 2 \text{&#x00A0;}{\sigma }_{2} . (c) The influence of AC interaction term on   σ 1 \text{&#x00A0;}{\sigma }_{1} . (d) The influence of AC interaction term on   σ 2 \text{&#x00A0;}{\sigma }_{2} . (e) The influence of BC interaction term on   σ 1 \text{&#x00A0;}{\sigma }_{1} . (f) The influence of BC interaction term on   σ 2 \text{&#x00A0;}{\sigma }_{2} .
Figure 4

Comparative analysis of the influence of interaction terms on   σ 1 and  σ 2 . (a) The influence of AB interaction term on   σ 1 . (b) The influence of AB interaction term on   σ 2 . (c) The influence of AC interaction term on   σ 1 . (d) The influence of AC interaction term on   σ 2 . (e) The influence of BC interaction term on   σ 1 . (f) The influence of BC interaction term on   σ 2 .

As can be seen from the figure, σ 1 and σ 2 have similar single-factor action plots and interaction term action plots, indicating that the stress–strain relationship is stable in the early stage of the test, and it can be observed that the stress magnitude of the model is basically determined only by the relationship with E G , E m , and E i at small strains, and the degree of influence is: E m > E G > E i ; that is, at small strains, the stress magnitude of CSG mainly related to E G , E m , and E i is positively correlated, and the relationship with f i is not significant, indicating that the relative deformation between different components is small at this stage, and the interface is subject to little tension, and the material stress is mainly determined by the modulus of elasticity of each component, while under small strains, although the different modulus of elasticity of each component will cause the deformation difference, and thus produce tensile stress, the effect of tensile stress is not obvious because the tensile stress between materials does not exceed the tensile strength of each component. However, because the parameter taking interval is relatively small and the coefficients of each variable are also small, the increase in σ is not obvious as E G , E m , and E i increase (Figure 5).

Figure 5 Single factor impact analysis diagram of σ 3 {\sigma }_{3} .
Figure 5

Single factor impact analysis diagram of σ 3 .

During the transition from σ 1 and σ 2 to σ 3 , the degree of σ affected by E m > E G > E i remains unchanged, but σ 3 is more affected by f i ; a 3D response surface diagram of the interaction of factors is drawn, and the interaction of various parameters are visually observe as shown in Figure 6.

Figure 6 Comparative analysis of the influence of interaction terms on   σ 3 \text{&#x00A0;}{\sigma }_{3} . (a) The influence of AB interaction term on   σ 3 \text{&#x00A0;}{\sigma }_{3} . (b) The influence of AD interaction term on   σ 3 \text{&#x00A0;}{\sigma }_{3} . (c) The influence of BC interaction term on   σ 3 \text{&#x00A0;}{\sigma }_{3} . (d) The influence of BD interaction term on   σ 3 \text{&#x00A0;}{\sigma }_{3} .
Figure 6

Comparative analysis of the influence of interaction terms on   σ 3 . (a) The influence of AB interaction term on   σ 3 . (b) The influence of AD interaction term on   σ 3 . (c) The influence of BC interaction term on   σ 3 . (d) The influence of BD interaction term on   σ 3 .

During the transition from σ 1 and σ 2 to σ 3 , the development trends of the interaction terms AB and BC remain unchanged, and the overall trends of the interaction terms AD and BD are the same. σ 3 increases with the increase in f i and E i . σ increases with the increase in the four parameters as a whole, that is, as a whole, the elastic modulus of the model increases with the increase in E G , E m , E i , and f i .

It can be seen visually on Figures 7 and 8 that the two terms E i and f i have the greatest influence on the model σ max and are positively correlated, with f i having a greater influence than E i ; this is inversely proportional to E g and E m because the damage criterion uses the maximum tensile stress damage criterion and material damage is caused by insufficient interface deformation and small tensile strength leading to interface damage first. In contrast, the aggregate and mortar damage is minimal (due to the general literature) [39,40].

Figure 7 Single factor impact analysis diagram of σ max {\sigma }_{\text{max}} .
Figure 7

Single factor impact analysis diagram of σ max .

Figure 8 Comparative analysis of the influence of interaction terms on σ max {\sigma }_{\text{max}} . (a) The influence of AD interaction term on σ max {\sigma }_{\text{max}} . (b) The influence of BC interaction term on σ max {\sigma }_{\text{max}} . (c) The influence of BD interaction term on σ max {\sigma }_{\text{max}} . (d) The influence of CD interaction term on σ max {\sigma }_{\text{max}} .
Figure 8

Comparative analysis of the influence of interaction terms on σ max . (a) The influence of AD interaction term on σ max . (b) The influence of BC interaction term on σ max . (c) The influence of BD interaction term on σ max . (d) The influence of CD interaction term on σ max .

3.4 Indoor test verification analysis

According to the test procedure, three 100 × 100 × 100 test blocks were made. The cement used is ordinary Portland cement (P·O42.5), the aggregate is sand and gravel from a river bed, and the fly ash is purchased from a power plant class I ash. Standard grade II aggregates are used. The test coordination ratio is shown in Table 6. The specimens were cured for 28 days for the uniaxial compression test. In the test, the universal testing machine and displacement control method were used, the rate was 0.6 mm/min. Three groups of specimens were made for uniaxial compression test, and the test results were in accordance with the specification [41], so the test results were considered reliable. According to the specification, a stress–strain curve is selected as the inversion curve, as shown in Figure 10.

Table 6

Mix proportions used in the uniaxial compression test

Cement (kg/m3) Flash (kg/m3) Sand (kg/m3) Water (kg/m3) Aggregate (kg/m3) Sand rate
70 20 434 90 1,736 0.2

The measured values are as follows in Table 7.

Table 7

Measured values of stress and strain

Strain ε 1 ε 2 ε 3 ε max
Stress (MPa) 1.16 × 10−4 0.48033 1.82484 5.26245

Bringing the measured values in the above table as the inversion target into the response surface model, which requires a guaranteed rate of more than 97%, the four response surface model equations are solved, and the meso-parameters obtained are shown in Table 8.

To verify the accuracy of parameter inversion, two groups of finite element models with the same gradation as in Figure 1 are re-established, as shown in Figure 9. The abovementioned inversion mesoscopic parameters are used to bring into three finite element models to simulate the failure process of the indoor uniaxial compression test.

Figure 9 Two-dimensional random aggregate model: (a) FEM 2 and (a) FEM 3.
Figure 9

Two-dimensional random aggregate model: (a) FEM 2 and (a) FEM 3.

The stress–strain relationship curve is obtained and compared with the laboratory test results, as shown in Figure 10.

Figure 10 Comparison of stress–strain curves.
Figure 10

Comparison of stress–strain curves.

It can be seen from Figure 10 that the stress–strain curves of the three numerical models are in good agreement with the indoor test curves, and the results are reliable.

The error calculation results are shown in Table 9.

Table 8

Inversion calculation results

A:   E G (MPa) B:   E m (MPa) C:   E i (MPa) D:   f i (MPa)
237.37 62.77 43.49 0.45
Table 9

Calculation of errors in numerical model calculation results

Test group Peak strain Peak stress
Value Error with indoor test (%) Value (MPa) Error with indoor test (%)
Indoor test 0.055 5.273
FEM 1 0.0532 3.273 5.331 1.100
FEM 2 0.0516 6.182 5.232 0.778
FEM 3 0.0541 1.636 5.456 3.471

The analysis found that the strength of CSG was significantly lower than that of ordinary concrete materials [42]. The maximum error of peak strain is 6.182%, and the maximum error of peak stress is 3.471%, which appear in different finite element models, and the overall error is small. The optimal experimental group is FEM 1, the inversion model has high accuracy, and the inversion results can be considered reliable. Compared with other inversion methods [43,44], the above four-parameter inversion analysis method of response surface only uses 29 sets of numerical simulation schemes, which greatly reduces the number of calculation sets and improves the calculation efficiency.

4 Conclusion

This article uses the indoor test-response surface method-finite element test to complete the material parameter analysis of the mesophase of cementitious sand and gravel materials. The reliability of the inversion parameters is verified, and the specific conclusions are as follows:

  1. Based on the response surface theory, the mesoscopic phase parameter inversion method proposed in this article is feasible. Compared with the traditional inversion method, the experimental design of this method is simple, and the number of inversion groups is small. Provides a reference for the interdisciplinary application of response surface methods.

  2. It can be seen from the response surface model that in the process of reaching the peak stress, the relationship of each component to the mechanical properties of the material is also changing. The initial stress is mainly positively correlated with E g , E m , and E i , and f i is not significant. The peak stress is most affected by f i and E i , which is a positive correlation, and has a negative correlation with E g and E m as a whole. With the increase in strain, each phase has different elastic modulus, so different deformation occurs, and tensile failure occurs.

  3. The parameter inversion results are brought into the finite element model. Compared with the laboratory test, the stress–strain curve is generally consistent. The optimal group has an error of 1.1% at the peak stress and 3.27% at the peak strain, with high accuracy and reliable inversion results. The rationality of the response surface method inversion is verified, which lays a foundation for the subsequent theoretical research based on this method.

Acknowledgments

We would like to submit the enclosed manuscript entitled “Research on Back Analysis of Meso-parameters of Hydraulic Cemented Sand and Gravel Based on Box-Behnken Design Response Surface,” which we wish to be considered for publication in “Science and Engineering of Composite Materials.” No conflict of interest exists in the submission of this manuscript, and the manuscript is approved by all authors for publication. I would like to declare on behalf of my co-authors that the work described was original research that has not been published previously, and not under consideration for publication elsewhere, in whole or in part. The data is authentic and reliable, and the code is usable. All the authors listed have approved the manuscript that is enclosed.

  1. Funding information: National Natural Science Foundation of China: (5210090341); Natural Science Foundation of Henan Province: (202300410270); Fund of Innovative Education Program for Graduate Students at North China University of Water Resources and Electric Power, China (grading no.YK-2021-28).

  2. Conflict of interest: The authors declare no conflict of interest.

  3. Data availability statement: All data, models, and code generated or used during the study appear in the published article.

References

[1] Jiang M, Zhang W, Sun Y, Utili S. An investigation on loose cemented granular materials via DEM analyses. Granul Matter. 2013;15:65–84.10.1007/s10035-012-0382-8Search in Google Scholar

[2] D’Addetta GA, Kun F, Ramm E. On the application of a discrete model to the fracture process of cohesive granular materials. Granul Matter. 2002;4:77–90.10.1007/s10035-002-0103-9Search in Google Scholar

[3] Desrues J, Viggiani G. Strain localization in sand: an overview of the experimental results obtained in Grenoble using stereophotogrammetry. Int J Numer Anal Met. 2004;28:279–321.10.1002/nag.338Search in Google Scholar

[4] Drescher A, Vardoulakis I. Geometric softening in triaxial tests on granular material. Geotechnique. 1982;32:291–303.10.1680/geot.1982.32.4.291Search in Google Scholar

[5] Han C. Localization of deformation in sand. University of Minnesota, ProQuest Dissertations Publishing; 1991.Search in Google Scholar

[6] Jiang M, Zhu H, Li X. Strain localization analyses of idealized sands in biaxial tests by distinct element method. Front Archit Civ Eng China. 2010;4:208–22.10.1007/s11709-010-0025-2Search in Google Scholar

[7] Mandel J. Conditions de stabilité et postulat de Drucker. Proceedings of IUTAM Symposium on Rheology and Soil Mechanics; 1966. p. 58–68.10.1007/978-3-662-39449-6_5Search in Google Scholar

[8] Pei LZ, Cai ZY. Research review of the cement sand and gravel (CSG) dam. Front Struct Civ Eng. 2012;6:19–24.10.1007/s11709-012-0145-ySearch in Google Scholar

[9] Dong L, Liu J, Xiuli D. A theoretical study on the influence of aggregate gradation on the tensile strength of 2-dimensional model concrete. Eng Mech. 2017;34:64–72.Search in Google Scholar

[10] Xu Y, Chen S. A method for modeling the damage behavior of concrete with a three-phase mesostructure. Constr Build Mater. 2016;102:26–38.10.1016/j.conbuildmat.2015.10.151Search in Google Scholar

[11] Dong L, Liu J, Xiuli D. Analysis theory for size effect of concrete in material-level based on properties and scales of meso structures. J Hydraulic Eng. 2019;102:26–38.Search in Google Scholar

[12] Hirose T, Fujisawa T, Yoshida H, Kawasaki H, Hirayama D, Sasaki T. Concept of CSG and its material properties. In Proceedings of the 4th International Symposium on Roller Compacted Concrete Dams. 2018. p. 465–73.Search in Google Scholar

[13] Jiang MJ, Sun YG, Li LQ, Zhu HH. Contact behavior of idealized granules bonded in different interparticle distances: an experimental investigation. Mech Mater. 2012;55:1–15.10.1016/j.mechmat.2012.07.002Search in Google Scholar

[14] Batmaz S. Cindere dam–107 m high roller compacted hardfill dam (RCHD) in Turkey. In Proceedings 4th International Symposium on Roller Compacted Concrete Dams. 2003. p. 121–6.Search in Google Scholar

[15] Gu C, Li B, Xu G, Yu H. Back analysis of mechanical parameters of roller compacted concrete dam. Sci China Technol Sci. 2010;53:848–53.10.1007/s11431-010-0053-0Search in Google Scholar

[16] Jia JS, Ma FL, Li XY, Chen ZP. Study on material characteristics of cement-sand-gravel dam and engineering application. J Hydraul Eng. 2006;37:578–82.Search in Google Scholar

[17] Yang J, Cai X, Guo XW, Zhao JL. Effect of cement content on the deformation properties of cemented sand and gravel material. Appl Sci. 2019;9:2369.10.3390/app9112369Search in Google Scholar

[18] Ren Q, Yin Y, Shen L. Fractal study of random distribution of concrete aggregate and its effect on damage characteristics. J Hydraul Eng. 2020;51:1267–77 and 88.Search in Google Scholar

[19] Ince R. Prediction of fracture parameters of concrete by artificial neural networks. Eng Fract Mech. 2004;71:2143–59.10.1016/j.engfracmech.2003.12.004Search in Google Scholar

[20] Wang ZH, Zhu YM, Chen FJ, Yu SP. Back analysis of concrete thermal parameters and its application based on genetic algorithms. J Hunan Univ Sci Technol (Nat Sci Ed). 2007;2:57–61.Search in Google Scholar

[21] Li SJ, Liu YX, Chen CL, Li ZG, He X, Zhou Y. Inversion procedure for mechanical parameters of concrete dam with hybrid genetic algorithm. J Dalian Univ Technol. 2004;44:195–9.Search in Google Scholar

[22] Jia SP. Inverse analysis for geotechnical engineering based on genetic algorithm and numerical implementation in ABAQUS. Hydrogeol Eng Geol. 2012;39:31–5.Search in Google Scholar

[23] Sun QC, Li SJ, Zhao HB, Zheng MZ, He ZY. Probabilistic back analysis of rock mechanical parameters based on displacement and relaxation depth. Chin J Rock Mech Eng. 2019;38:1884–94.Search in Google Scholar

[24] Liu DH, Zhao MQ. Meso-structural parameters inversion of PFC model for compaction of asphaltic concrete in core wall. J Hohai Univ. 2020;48:53–9.Search in Google Scholar

[25] Jia Y, Chi S. Back-analysis of soil parameters of the Malutang II concrete face rockfill dam using parallel mutation particle swarm optimization. Comput Geotech. 2015;65:87–96.10.1016/j.compgeo.2014.11.013Search in Google Scholar

[26] Li H, Wang G, Wei B, Zhong Y, Zhan L. Dynamic inversion method for the material parameters of a high arch dam and its foundation. Appl Math Model. 2019;71:60–76.10.1016/j.apm.2019.02.008Search in Google Scholar

[27] Box GEP, Wilson KB. On the experimental attainment of optimum conditions. J R Stat Soc Ser B Stat Methodol. 1992;270–310.10.1007/978-1-4612-4380-9_23Search in Google Scholar

[28] Zhang YY, Shen HM. Thin plate’s vibration based on MATLAB software. J Sichuan Univ (Eng Sci Ed). 2012;44:36–40.Search in Google Scholar

[29] Jiang BK, Li YH. Response analysis of rotating viscoelastic sandwich beam. J Sichuan Univ. 2012;44:167–70.Search in Google Scholar

[30] Pei L, Chen JK, Wu ZY, Li YL, Zhang H. Response surface genetic algorithm of back analysis of concrete thermal parameters. Mater Res Innov. 2015;19:S8-840–5.10.1179/1432891715Z.0000000001817Search in Google Scholar

[31] Li L, Zhi Z, Sai Z, Shuixiu P. Study on technological parameters of the treatment of landfill leachate by MAP method using response surface methodology. Chin J Eng. 2010;4:1289–95.Search in Google Scholar

[32] Li D. Parameter inversion of micro constitutive model for concrete materials based on macro experimental data. Dalian University of Technology; 2015.Search in Google Scholar

[33] Wang S. Parameter inversion of micro constitutive model for concrete materials based on macro experimental data. Dalian University of Technology; 2015.Search in Google Scholar

[34] Roussouly N, Petitjean F, Salaun M. A new adaptive response surface method for reliability analysis. Probabil Eng Mech. 2013;32:103–15.10.1016/j.probengmech.2012.10.001Search in Google Scholar

[35] Lü Q, Low BK. Probabilistic analysis of underground rock excavations using response surface method and SORM. Comput Geotech. 2011;38:1008–21.10.1016/j.compgeo.2011.07.003Search in Google Scholar

[36] Wang Y, Wang C. The application of response surface methodology. J Cent Univ Nat (Nat Sci Ed). 2005;3.Search in Google Scholar

[37] Zhouhong Z, Minglin G, Zhanghua X. Finite element model validation of the continuous rigid frame bridge based on structural health monitoring Part I: FE model updating based on the response surface method. Chin Civ Eng J. 2011;44:90–8.Search in Google Scholar

[38] Guo L, Zhang Y, Zhong L, Zhang F, Wang M, Li S. CSG elastic modulus model prediction considering meso-components and its effect. Sci Eng Compos Mater. 2020;27:27–271.10.1515/secm-2020-0027Search in Google Scholar

[39] Scrivener KL, Crumbie AK, Laugesen P. The interfacial transition zone (ITZ) between cement paste and aggregate in concrete. Interf Sci. 2004;12:411–21.10.1023/B:INTS.0000042339.92990.4cSearch in Google Scholar

[40] Zhou S, Chen L, Liu M. Analysis on effect of aggregate on mechanical property of concrete. J Build Mater. 2016;19:143–8.Search in Google Scholar

[41] China Academy of Building Research, Tsinghua University, Tongji University, et al. Design specifications for concrete structures. AQSIQ; 2010: 440P; B5.Search in Google Scholar

[42] Baktheer A, Chudoba R. Experimental and theoretical evidence for the load sequence effect in the compressive fatigue behavior of concrete. Mater Struct. 2021;54:1–23.10.1617/s11527-021-01667-0Search in Google Scholar

[43] Tian M, Zhou J. New algorithm for parameter inversion in geotechnical engineering. Chin J Rock Mech Eng. 2005;24:1492–6.Search in Google Scholar

[44] Jin CY, Ma ZY, Zhang YL, Sha RH, Chen QF. Application of neural network to back analysis of mechanical parameters and initial stress field of rock masses. Rock Soil Mech. 2006;27:1263–6.Search in Google Scholar
Received: 2022-01-14
Revised: 2022-03-05
Accepted: 2022-03-28
Published Online: 2022-04-18

© 2022 Lixia Guo et al., published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

  1. Regular Articles
  2. Experimental investigations of a novel pressure microfoam preparation device for dust removal
  3. Influence of hydrothermal aging on the mechanical performance of foam core sandwich panels subjected to low-velocity impact
  4. Experimental study on surface wrapping strengthening of EPS particles and its concrete performance
  5. Modification of mechanical properties of Shanghai clayey soil with expanded polystyrene
  6. A new EPS beads strengthening technology and its influences on axial compressive properties of concrete
  7. A novel superabsorbent material based on soybean straw: Synthesis and characterization
  8. Use of line laser scanning thermography for the defect detection and evaluation of composite material
  9. Research on back analysis of meso-parameters of hydraulic cemented sand and gravel based on Box-Behnken design response surface
  10. Hot deformation behavior and microstructure of a 0.5 wt% graphene nanoplatelet reinforced aluminum composite
  11. Analysis of electromagnetic characteristics of the proposed composite four-rail electromagnetic launcher
  12. Preparation and characterization of a graphene hybridizing polyurethane damping composite
  13. Effects of layup parameters and interference value on the performance of CFRP–metal interference fit joints
  14. Vibration and noise reduction of pipelines using shape memory alloy
  15. Finite element analysis of behavior and ultimate strength of composite column
  16. Dynamic response of functionally graded plate under harmonic load with variable gradient parameters
  17. Deformation behavior of rubber composite based on FEA and experimental verification
  18. Effects of Z-pin on moisture absorption property and damage mode under flexural load for carbon fiber composite
  19. Design and testing of a smart rubber stave for marine water-lubricated bearings
  20. Study of carbon nano-modifier of fly ash in cement concrete mixtures of civil engineering
  21. Analysis of multiple impact tests’ damage to three-dimensional four-directional braided composites
  22. Theoretical analysis of aluminum honeycomb sandwich panel supported by reinforced concrete wall under low-speed impact load
  23. Effects of local fiber discontinuity on the fatigue strength parameter at the fiber inclusion corner in fiber-reinforced composites
  24. Experimental investigation on compressive properties of three-dimensional five-directional braided composites in hygrothermal environment
  25. Failure process of steel–polypropylene hybrid fiber-reinforced concrete based on numerical simulations
  26. A simple method for measuring the monofilament diameter of continuous filament yarn with high bending stiffness via synthetic laser imaging
  27. Span length effect on flexural properties of composite laminate reinforced with a plain weave carbon fiber fabric in a polymer matrix
  28. Mechanical properties improving and microstructure characterization of inorganic artificial stone binder
  29. Effect of thermal treatment process on the structure of C/SiO2 composite aerogels
  30. Mechanical and corrosion resistance analysis of laser cladding layer
  31. Wear and corrosion mechanisms of Ni–WC coatings modified with different Y2O3 by laser cladding on AISI 4145H steel
  32. Damage and failure analysis of composite stiffened panels under low-velocity impact and compression after impact with damp-heat aging
  33. In-situ CT characterization of 2D woven SiCf/SiC composite loading under compression
  34. Effect of the manufacturing process on the equivalency qualification of glass fiber reinforced polymer
  35. Study of concrete properties based on crushed stone sand mixture and fiber of fly ash of thermal power plants
  36. Establishment of wear mechanism distribution diagram of ZTAp-reinforced iron matrix composites
  37. Calculation method of elastic modulus for carbon fiber-reinforced plastics considering inhomogeneous interphase
  38. An experimental study on the failure and enhancement mechanism of bolt-strengthening GFRP T-joint subjected to tensile loading
  39. The viability of cell that encapsulated in calcium alginate hydrogel beads
  40. Discussion of ceramic bar reinforced TWIP steel composite structure
  41. A theoretical framework underlying an accelerated testing method and its application to composites under constant strain rates and fatigue loading
  42. Theoretical analysis of interfacial design and thermal conductivity in graphite flakes/Al composites with various interfacial coatings
  43. Multiscale heat conduction and fractal oxidation behaviors of needle-punched carbon/carbon composites
  44. Numerical simulation of composite grid sandwich structure under low-velocity impact
  45. Wear properties of Al/TiO2 composites fabricated via combined compo-casting and APB process
  46. Review Articles
  47. Application of melanin as biological functional material in composite film field
  48. Review on research progress of cemented sand and gravel dam
  49. Communication
  50. Fabrications and microstructure analysis of cobalt-based coatings by an easy-coating and sintering process
  51. Letter to the Editor
  52. Investigation on mechanical and conductive behaviors of nano-graphite-based concrete
https://doi.org/10.1515/secm-2022-0008
Keywords for this article
cemented sand and gravel; stress–strain; meso-parameters inversion; response surface method
Creative Commons
BY 4.0

Articles in the same Issue

  1. Regular Articles
  2. Experimental investigations of a novel pressure microfoam preparation device for dust removal
  3. Influence of hydrothermal aging on the mechanical performance of foam core sandwich panels subjected to low-velocity impact
  4. Experimental study on surface wrapping strengthening of EPS particles and its concrete performance
  5. Modification of mechanical properties of Shanghai clayey soil with expanded polystyrene
  6. A new EPS beads strengthening technology and its influences on axial compressive properties of concrete
  7. A novel superabsorbent material based on soybean straw: Synthesis and characterization
  8. Use of line laser scanning thermography for the defect detection and evaluation of composite material
  9. Research on back analysis of meso-parameters of hydraulic cemented sand and gravel based on Box-Behnken design response surface
  10. Hot deformation behavior and microstructure of a 0.5 wt% graphene nanoplatelet reinforced aluminum composite
  11. Analysis of electromagnetic characteristics of the proposed composite four-rail electromagnetic launcher
  12. Preparation and characterization of a graphene hybridizing polyurethane damping composite
  13. Effects of layup parameters and interference value on the performance of CFRP–metal interference fit joints
  14. Vibration and noise reduction of pipelines using shape memory alloy
  15. Finite element analysis of behavior and ultimate strength of composite column
  16. Dynamic response of functionally graded plate under harmonic load with variable gradient parameters
  17. Deformation behavior of rubber composite based on FEA and experimental verification
  18. Effects of Z-pin on moisture absorption property and damage mode under flexural load for carbon fiber composite
  19. Design and testing of a smart rubber stave for marine water-lubricated bearings
  20. Study of carbon nano-modifier of fly ash in cement concrete mixtures of civil engineering
  21. Analysis of multiple impact tests’ damage to three-dimensional four-directional braided composites
  22. Theoretical analysis of aluminum honeycomb sandwich panel supported by reinforced concrete wall under low-speed impact load
  23. Effects of local fiber discontinuity on the fatigue strength parameter at the fiber inclusion corner in fiber-reinforced composites
  24. Experimental investigation on compressive properties of three-dimensional five-directional braided composites in hygrothermal environment
  25. Failure process of steel–polypropylene hybrid fiber-reinforced concrete based on numerical simulations
  26. A simple method for measuring the monofilament diameter of continuous filament yarn with high bending stiffness via synthetic laser imaging
  27. Span length effect on flexural properties of composite laminate reinforced with a plain weave carbon fiber fabric in a polymer matrix
  28. Mechanical properties improving and microstructure characterization of inorganic artificial stone binder
  29. Effect of thermal treatment process on the structure of C/SiO2 composite aerogels
  30. Mechanical and corrosion resistance analysis of laser cladding layer
  31. Wear and corrosion mechanisms of Ni–WC coatings modified with different Y2O3 by laser cladding on AISI 4145H steel
  32. Damage and failure analysis of composite stiffened panels under low-velocity impact and compression after impact with damp-heat aging
  33. In-situ CT characterization of 2D woven SiCf/SiC composite loading under compression
  34. Effect of the manufacturing process on the equivalency qualification of glass fiber reinforced polymer
  35. Study of concrete properties based on crushed stone sand mixture and fiber of fly ash of thermal power plants
  36. Establishment of wear mechanism distribution diagram of ZTAp-reinforced iron matrix composites
  37. Calculation method of elastic modulus for carbon fiber-reinforced plastics considering inhomogeneous interphase
  38. An experimental study on the failure and enhancement mechanism of bolt-strengthening GFRP T-joint subjected to tensile loading
  39. The viability of cell that encapsulated in calcium alginate hydrogel beads
  40. Discussion of ceramic bar reinforced TWIP steel composite structure
  41. A theoretical framework underlying an accelerated testing method and its application to composites under constant strain rates and fatigue loading
  42. Theoretical analysis of interfacial design and thermal conductivity in graphite flakes/Al composites with various interfacial coatings
  43. Multiscale heat conduction and fractal oxidation behaviors of needle-punched carbon/carbon composites
  44. Numerical simulation of composite grid sandwich structure under low-velocity impact
  45. Wear properties of Al/TiO2 composites fabricated via combined compo-casting and APB process
  46. Review Articles
  47. Application of melanin as biological functional material in composite film field
  48. Review on research progress of cemented sand and gravel dam
  49. Communication
  50. Fabrications and microstructure analysis of cobalt-based coatings by an easy-coating and sintering process
  51. Letter to the Editor
  52. Investigation on mechanical and conductive behaviors of nano-graphite-based concrete
Sign up now to receive a 20% welcome discount
Institutional Access
How does access work?
Have an idea on how to improve our website?
Please write us.
© 2025 De Gruyter Brill
Downloaded on 28.10.2025 from https://www.degruyterbrill.com/document/doi/10.1515/secm-2022-0008/html
Scroll to top button