Modelling the damage time of the concrete cover is an important problem in estimating the durability of reinforced concrete elements. There have been many papers related to the assessment of this issue. Classic papers based on the analytical model of a thick-walled cylinder can be mentioned here [1, 2, 3], and a comprehensive review of the paper can be found in [4].
The use of computer methods allowed for a more detailed description of the problem, among others, including in the calculations additional fields related to the mass transport of aggressive substances, moisture, and heat [5], assuming an uneven and time-varying distribution of corrosion products on the circumference of the reinforcement [6, 7, 8, 9, 10, 11, 12, 13], or the description of contact steel concrete [14].
The consequence of the implementation of computational methods was the use of advanced concrete models: elastic-plastic regularized with cracking energy [15, 16] implemented in the papers [17, 10, 18, 19, 20, 21]. Elastic models with degradation, an example of which is the gradient model used in corrosion problems [22]. And finally other – often very complex – network, mesoscale, microplane, and xfem models are presented, among others, in the papers [23, 24, 25, 26, 27, 28].
In this paper, the possibility of using nonlocal gradient microplane models presented in [29] and [30] for the estimation of the cover cracking time in reinforcement corrosion was analysed. The calculations took into account the issues of contact and cohesion at the interface between reinforcement steel and concrete. The results of the calculations were compared with the results obtained on the basis of experimental tests and the results obtained for the elastic-plastic Menetrey-Willam model with fracture and hardening and softening that depend on the fracture energy.
In the paper, calculations of the propagation of damage in the concrete cover of a test reinforced concrete element subjected to reinforcement corrosion were made. The results were taken as average values from the tests presented in the paper [9], where the test sample 100x100x80 mm3 with an eccentrically placed reinforcing bar with a diameter of 20 mm was subjected to accelerated corrosion, Fig. 1.
The paper analyses the propagation of damage in the cover of a reinforced concrete element with the use of nonlocal gradient material models. In the calculations, two gradient models of concrete, the elastic damage microplane model (EDM) [29] and the coupled damage plasticity microplane model (CDPM) [30] were used. Due to the highly time-consuming calculations, a fragment of the test reinforced concrete element with a thickness of 10 mm was analysed. The plain strain condition was enforced by the applied boundary conditions. The division of the test element in the plane into finite elements with a length of 1.2 mm was assumed. In the longitudinal direction, the sample was divided into eight finite elements, Fig. 1. The impact of corrosion products on the concrete cover was captured by introducing the equivalent rate tensor of the volumetric strain [9]. It was assumed that the mechanical effects of the interaction could be represented using a steel ring to which the coordinate of increment of the volume strain tensor would be applied. The steel-concrete bonds were mapped using contact elements and a cohesion model. The diagram of the division of the test element into finite elements together with the concept of modelling corrosion effects in the form of a steel ring is shown in Fig. 1b, where the following indications have been adopted: 1) reinforcing steel, 2) concrete, 3) steel ring, 4) contact.
The analyses of concrete cover cracking were performed using microplane models [31, 32]
The mesh-independent solution was guaranteed by the formulation of a gradient (nonlocal) model [29, 30].
The EDM model is formulated in [29] and in the documentation of the ANSYS software [33]. The damage surface and the evolution equation of the damage parameter are described by the following relations:
Damage surface (modified Mises surface).
The equivalent strain corresponding to the microplane is determined in accordance with the relationship used, among others, in the papers of [34, 29, 35 and 36].
Equation of the evolution of the damage parameter
The evolution equation of the damage parameter was proposed in the form of an exponential relationship [29, 34, 35, 36]
The constitutive relationships in the CPDM model, after considering the decomposition of the strain tensor into elastic and plastic parts, will take the form [30]
The microplane yield surfaces fmic and the component cap surfaces take the form:
Microplane yield surface fmic
The Drucker-Prager yield function,
The compression cap function gc,
The tension cap function gt,
Equation of damage parameter evolution
where g1 is the Drucker-Prager yield function with hardening, gc and gt are the compression and tension cap functions.
where σ0 is the initial yield stress, fh is a hardening function, D is a hardening material constant, αf is a friction coefficient, κ is a hardening variable.
where
where
where 〈ɛI〉 is the McAulay bracket, dmic is the damage parameter assigned to the microplane,
Studies that analysed the discrepancies of the calculated results obtained for the elastic-plastic hardening and softening model, different types of contact, and different variants of corrosion product modelling were conducted and presented, among others, in [21]. In this paper, it was assumed that the impact of corrosion products on the concrete cover can be described by imposing a deformation field on a steel ring on the perimeter of the reinforcement bar. According to [9] the evolution tensor of the corrosion volumetric strains rate was accepted in the form
Contact interactions were analysed according to the algorithm presented in the paper [21]. The contact area of steel and concrete was mapped using contact elements in which the Coulomb friction model was included along with the possibility of separation of the contact surface and the cohesion model, which were implemented in the APDL language [33]. The paper analysed the options of contact interactions: rigid contact (B), characterised by the lack of slippage and separation; flexible contact (NSS), allowing contact slippage in the tangential direction; standard contact (S), allowing contact slippage and separation of contact surfaces; and the cohesion model (CZM), allowing for the description of contact degradation of contact surfaces. The cohesive degradation of the steel-concrete contact zone was explained using a bilinear cohesive model of the material [14], depending on the contact normal stresses σn and tangential stresses τt. The cohesion model used in the calculations carried out in the paper is applied simultaneously with the Coulomb friction model (the shear stress is the maximum value determined on the basis of both models) [21].
The computer simulation was carried out using the parameters adopted in the paper [9] in the ANSYS software. It was accepted that the corrosion products are a mixture of iron oxides Fe(OH)2 and Fe(OH)3, the effective electrochemical equivalent of iron keff=0.006271 (g/
The value of increments of corrosion volumetric strains was determined for the averaged function of the corrosion current intensity [9]. The course in the changes in the increments of volumetric strains as a function of the averaged current intensity for the entire cross-section of the reinforcing bar and the parameter χ =1 is shown in Fig. 2 (calculations were carried out assuming the parameter χ =0.4).
In the FEM model, the elements of ANSYS software CPT215 were used for the concrete modelling and SOLID185 for the reinforcement bar modelling. Contact interactions were modelled using elements TARGE170 and CONTA174. The computational analysis also considered the issue of cohesion, which was associated with the contact elements in the computational system. The corrosive effects were included considering the increases in corrosive volumetric strains in the steel ring, Fig. 1b. Calculations were made for the cover described with EDM and CDPM materials in a nonlocal gradient formulation. For steel, the Huber-Mises-Hencky elastic-plastic model without hardening was assumed. The initial elastic material parameters of the concrete along with the basic strength parameters of the steel are summarized in Table 1 and Table 2, respectively.
Initial elastic and strength material parameters of concrete
Description | Value |
---|---|
Modulus of elasticity, E (GPa) | 38.28 |
Poisson’s ratio, ν (−) | 0.2 |
Uniaxial tensile strength, ft (MPa) | 3.99 |
Uniaxial compressive strength, fc (MPa) | 56.4 |
Biaxial compressive strength, fbc=1.15 fc (MPa) | 64.86 |
Initial elastic and strength material parameters of steel
Description | Value |
---|---|
Modulus of elasticity, Es (GPa) | 200 |
Poisson’s ratio, νs (1) | 0.3 |
Yield strength, fy (MPa) | 235 |
The determination of inelastic parameters requires experimental research. Due to the impossibility of making such tests, the inelastic parameters of concrete were determined on the basis of empirical relationships characterising the parameters of the model as well as numerical tests of cyclic loading and unloading of homogeneous elements in compression/tension, Fig. 3b and nonhomogeneous tensile test elements with a notch, Fig. 3a. Numerical tests were related to the calculations made using the elastic-plastic model of the material with hardening/softening dependent on the fracture energy Menetrey-Willam (MW) model with HSD2 implemented in the ANSYS software [21, 33]. The material parameters of the model are listed in Table 1 and Table 3. The mesh of the test elements was densified in the stress concentration region, Fig. 3a.
Inelastic parameters of the MW model with HSD2 [21]
Description | Value | Description | Value |
---|---|---|---|
Fracture energy, Gft (N/m) | 151 | Ωci (1)*) | 0.33 |
Dilation angle, ψ (Deg) | 20 | Ωcu (1)*) | 0.85 |
κcm (1)*) | 0.00151 | Ωcr (1)*) | 0.2 |
κcu (1)*) | 0.00175 | Ωtr (1)*) | 0.1 |
parameter characterizing the hardening/softening curve of the HSD2 model under compression and tension
The formulation of the gradient elastic damage model (EDM) requires the definition of 7 parameters (plus E and ν). The first three k0, k1, k2 and the parameter defining the damage threshold
In the element, according to Fig. 3a, the displacement of the element edge Δu = 0.2 mm was induced, assuming variable parameters characterizing the shape of the degradation function βmic. The results of the tensile test characterizing the response of the test element are presented in Fig. 4a, where the following denotations were assumed: βmic = 50 (F1), βmic = 100 (F2), βmic = 300 (F3) for c=5 mm2 and βmic = 50 (F5), βmic = 100 (F6), βmic = 300 (F7) for c=8 mm2, model MW with HSD2 (F4). The βmic parameter was finally assumed by iterative analysis of the response of the notched tensile sample, the deformation map for the EDM model, as well as by analysing the results and the convergence of the solution of the test sample analysed in the paper, in which reinforcement corrosion occurs. For the purposes of further calculations, the values βmic = 100, c= 8 mm2 (model F6) were assumed, for which the calculations in the test sample were also characterized by high stability.
In the case of an elastic-plastic gradient material model with degradation (CDPM), it is required to define 13 material parameters (plus E and ν). The parameters of the CDPM model
Other parameters require computational tests. In the first stage, a cyclic axial compression test was performed, which allowed the determination of the parameters D, βc, γc0. In the next one, the axial stretching of the inhomogeneous notched sample was analysed which allowed the determination of the values RT, βt, γt0, as well as the gradient parameters c and m.
In the first stage of the calculations, in the cyclic compression and tensile test of the unnotched concrete sample, Fig. 3b, the parameters D, βc, γc0 were determined. In the absence of experimental studies, the parameters were accepted as starting values for the calculations [30]
Computer calculations of the cyclic compression test were made for the material parameters listed in Table 4. Table 5 presents the variable test parameters for which the response of the system was tested to select its best values.
The material parameters of the CDPM model assumed in the cyclic compression test to determine the values D, βc, γc0, [30]
Description of the variable | Value |
---|---|
Abscissa of the intersection point between the compression cap and the Drucker-Prager yield function,
|
−50 |
The ratio between the major and minor axes of the cap, R (−) | 2 |
Tension cap hardening constant, RT (−) | 1 |
Tension damage thresholds, γt0 · 105 (−) | 9.38 |
Nonlocal interaction range parameter, c (−) | 10 |
Over-nonlocal averaging parameter, m (−) | 2.5 |
Tension damage evolution constant,
|
|
Hardening material constant, D* (MPa) | 10000 |
Compression damage thresholds,
|
0.0001 |
Compression damage evolution constant,
|
1000 |
Variable material parameters of the CDPM model assumed for the cyclic compression test to determine the value D, βc, γc0, [30]
Description | M1 | M2 | M3 | M4 | M5 | M6 | M7 | M8 |
---|---|---|---|---|---|---|---|---|
D/D* | 4 | 4 | 4 | 1 | 8 | 1 | 1 | 1 |
|
1 | 1 | 1 | 1 | 1 | 0.7 | 1 | 1.3 |
|
1 | 2 | 3 | 2 | 2 | 2 | 2 | 2 |
|
1 | 2 | 3 | 2 | 2 | 1.33 | 1.33 | 1.33 |
The calculation results obtained for the variable (modified) material parameters βt, βc, D, γc0 are shown in Fig. 5 where the following indications have been introduced: a) M1, M2, M3 are the results of the calculations for changing parameters βt and βc acc. to relation (18); b) M4, M5 are the results of the calculation for the change parameter D; c) M6, M7, M8 are the results of the calculations obtained with the changing parameter γc0 (other parameters were assumed according to Table 4). The test results obtained using the MW model with HSD2 are marked with the E2 index. The parameters obtained for the M7 test were assumed to be optimal: D=10000 MPa, βc = 2000, βt = 2000, γc0 = 0.0001.
The second stage of selection of material parameters was performed for the test element with a notch, Fig. 3a. In the tensile test, the values of the parameters βt and c, Table 6, were adopted while the remaining constants were assumed based on Table 4 and Table 5.
Variable material parameters of the CDPM model assumed for the tensile test to determine the values βt and c [30]
Description | N1 | N2 | N3 | N4 |
---|---|---|---|---|
βt | 2000 | 2000 | 2000 | 3000 |
c (mm2) | 12 | 8 | 5 | 5 |
The values of the searched model parameters βt and c, just as the first stage, were selected on the basis of the results of the tensile test of the notched sample. Graphs showing the system response in Fig. 3a as a result of forcing the displacement of the element edge are presented in Fig. 6 where N1–N4 are the test parameters of the model, E4 is the calculation result obtained using the MW model with HSD2.
In Fig. 6: N1–N4 are the test parameters of the model, and E4 is the calculation result obtained using the MW model with HSD2. The distribution of the principal tensile strains ɛ1 of the notched test element in the CDPM model as a function of the parameter βt (1) and c (mm2): a) N1, βt = 2000, c = 12; b) N2, βt = 2000, c = 8; c) N3, βt = 2000, c = 5; d) N4, βt = 3000, c = 5, was shown in Fig. 7.
The parameter βt and c were finally adopted in an iterative way by analysing the response of the notched tensile sample, the principal strain maps for the CDPM model shown in Fig. 7 as well as by analysing the results and convergence of the solution of the test sample analysed in the paper, where reinforcement corrosion occurred. For the purposes of further calculations, the values of the parameters βt = 2000, c= 8 (mm2) (model N2) were adopted, for which the calculations in the analysed test sample were also characterized by high stability.
The contact between the reinforcing steel and concrete was modelled for different contact models: a) rigid (B), b) no separation with slip (NSS), c) standard model with slip and separation (S), and d) cohesive model with degradation of the contact layer (CZM).
Material parameters of cohesion ch, maximum allowable shear stresses τmax and parameters related to normal stiffness Kn and tangent stiffness Kt were formulated as functions of the initial value
The material parameters assumed in the contact and cohesive models used in the calculations were defined in Table 7 and Table 8.
Contact model parameters: bonded, no separation with sliding, CZM, standard
Parameter | Value | |||
---|---|---|---|---|
Coefficient of friction, μ (1) | 1.0 | 0.2 | 0.2 | 0.2 |
Cohesion coefficient, ch(MPa) | 0.0 | 0.375 | 0.375 | 0.375 |
Normal contact stiffness,
|
1.0 | 1.0 | 1.0 | 1.0 |
Tangent contact stiffness,
|
1.0 | 2.0 | 5.0 | 5.0 |
Maximum allowable shear stress, τmax (MPa) | 1E20 | 18.77 | 18.77 | 18.77 |
Contact Type | B | NSS | CZM | S |
Supplementary parameters of the cohesive model (CZM)
Parameter | Value |
---|---|
Maximum normal contact stress, σmax (MPa) | 3.99 |
Critical crack energy in the normal direction, Gcn (N/m) | 151 |
Maximum tangential contact stress, τt,max (MPa) | 2.26 |
Critical fracture energy in the tangential direction, Gct (N/m) | 113 |
Artificial damping parameter, η (1) | 0.0001 |
The calculation results for the EDM and CDPM models were compared with the mean displacement values obtained for the MW model with HSD2 and the solution obtained from the ATENA programme (elastic-plastic MW model with cracking [9]). The solution was also compared with the results of experimental studies of element edge displacements and the width of the crack, published in the paper [9]. The variants of materials and types of contact elements analysed in the paper are presented in Table 9.
List of materials and contact models analysed in the paper, along with denotations
Contact Type | MW with HSD2 | EDM | CDPM |
---|---|---|---|
B | M1 | E1 | C1 |
NSS | M2 | E2 | C2 |
S | M3 | E3 | C3 |
CZM | M4 | E4 | C4 |
In Table 9, the models used are marked with the first letter of the model name and the number that identifies the type of contact. The results of experimental calculations [9] are marked with the symbols: a) EX1 for the measured displacements, and b) EX2 for the values of the measured crack widths. The average calculation results obtained from the ATENA software for the rigid contact between steel and concrete are marked with the A index.
The results of the calculations of displacement of the edge of element ΔLAB, Fig. 1a by using the EDM and CDPM gradient models that are available in the ANSYS program, were presented graphically together with the comparative calculations and the results of the experimental tests in Fig. 8. The notation placed in the figure was explained in Table 9. The results of the experimental tests are marked with the following symbols: EX0 is the course of the change in the average value of the current intensity [9], EX1 is the course of the change of the averaged function of the edge displacements [9], EX2 is the course of the change in the average function of the width of the crack in the element [9]. The average results obtained using FEM from the ATENA software, assuming a rigid contact between steel and concrete, are marked with index A.
The analysis of the research results shows high compatibility of the results obtained using MW models with HSD2 and those obtained using the nonlocal gradient CDPM model. These results also show compliance with the final values of the crack width and displacements of the edges of the element. However, reference to the results of experimental tests is for illustrative purposes only, because obtaining these test results is subject to specific uncertainties related to the nature of the reinforcement corrosion processes. The method of transferring the results of experimental research assumptions to computer calculations is also characterized by uncertainty (uncertain parameter of the intensity of the impact of corrosion products χ). The results obtained using the ATENA software for the MW model with cracking, included in the paper for comparison, give lower displacement values, although it can be seen that the shape of the curve describing the history of displacements is similar to the previously obtained results. The mentioned model was additionally calculated for a coarser mesh of finite elements.
However, a strong deviation (about 30%) of the results obtained using the EDM gradient model from the results obtained with the MW models with HSD2 and CDPM as well as from the results obtained with the ATENA software [9] is very characteristic.
The results of the calculation (approximately) at the end of the time step t = 388 h for both the analysed EDM and CDPM models are compared in Table 10. The reference value was the result of the experiment included in the paper [9]. The calculations were made according to the dependencies.
Comparison of the calculation results and the percentage deviation from the results obtained for the MW model with HSD2
Model | LAB (mm) | |ΔAB|(mm) |
|
---|---|---|---|
EX1 | 101.22 | - | - |
EDM E1 | 100.94 | 0.27 | 29 |
EDM E2 | 100.96 | 0.25 | 26 |
EDM E3 | 101.01 | 0.20 | 20 |
EDM E4 | 100.98 | 0.23 | 24 |
CDPM C1 | 101.22 | 0.01 | 1 |
CDPM C2 | 101.20 | 0.02 | 2 |
CDPM C3 | 101.20 | 0.02 | 2 |
CDPM C4 | 101.20 | 0.02 | 2 |
MW M1 | 101.29 | 0.07 | 6 |
MW M2 | 101.26 | 0.04 | 3 |
MW M3 | 101.31 | 0.09 | 7 |
MW M4 | 101.30 | 0.08 | 6 |
In addition, graphic images of the main tensile strains for the following models are presented: a) CDPM in Fig. 9, b) EDM in Fig. 10, c) MW with HSD2 in Fig. 11.
The results obtained from the calculations well reflect the behaviour of the reinforced concrete covers under the conditions of accelerated reinforcement corrosion. The convergence of the results of calculations aimed at determining the displacements of the edge of the reinforced concrete element subjected to reinforcement corrosion in the case of the MW models with HSD2 and CDPM is high (the deviation of the results is approximately 5%). The use of gradient models allows for obtaining fully mesh-independent (objective) results of computer calculations; however, it is very computationally expensive and requires a dense FE mesh. In the test calculations analyzed in the paper, in the case of irregular finite elements, the solution did not converge. Convergence problems were also conditioned by the type of contact used. In the case of contact B, the NSS solution converged very well. In the case of S-type and CZM contact, a high dependence of the stability of the solution on the proportion of the dimensions of the finite elements could be observed, both in the case of the EDM and CDPM models. The deviation of the solution obtained using the EDM gradient model from the results obtained for the MW and CDPM models may result from the selection of the gradient parameter c.
In the case of the EDM model (as in the CDPM), the value of the gradient parameter c=8 mm2 was assumed. The adoption of lower values of the gradient parameter allowed for a more precise map of total deformations (cracks), but it required the use of a denser FE mesh, which resulted in a significant increase in calculation time. In the EDM model, the use of higher values of the gradient parameter c did not represent the principal maps of the total tensile strains in an acceptable manner. In the CDPM model, the influence of the gradient parameter on the values of the calculation results was less than in the EDM model.
From the point of view of the damage pattern distribution the results from CDPM and MW with HSD2 models are very close to the damage pattern. From the applicability point of view (because of the rate of the solution of the problem) the MW with HSD2 model is most promising.