Professional Documents
Culture Documents
A correction was made to the e-First version of this paper on 9 June 2016 prior to final issue publication. The current online and print versions are identical and both contain the correction.
of the ageing degradation effects. Analysis results from FEMs in shape. The ring beam forms a transition component
can be used to better understand the instrumentation data between the upper dome and the perimeter wall, and it acts
collected from existing CANDU 6 CCBs and provide guidance as a support for the upper dome, the top of the perimeter
for the development of novel inspection and monitoring wall, and the lower dome. There is a hinge connection
approaches for CANDU 6 reactors. The modelling assumptions between the perimeter wall and the base slab.
and simplifications, approach and validation are the focus of
this paper. A brief discussion is given on the preliminary Figure 2 presents a schematic layout of P-T cables in a
studies of temperature load and ageing degradation effects. CANDU 6 CCB and shows that the base slab is post-tensioned
with 2 layers of cables, the perimeter wall has both vertical
and horizontal (circumferential) cables, and the upper dome
2. Configuration of CANDU 6 Containment Buildings
has 3 layers of cables [2]. Four vertical buttresses, of thick-
Figure 1 presents a schematic configuration of a typical ness equal to that of the ring beam, are attached to the
CANDU 6 CCB, which includes a base slab, a perimeter wall, perimeter wall at right angles from each other. The ends
a ring beam, an inner dome, and an upper dome [2]. The base (anchor heads) of the horizontal P-T cables in the perimeter
CNL Nuclear Review Downloaded from pubs.cnl.ca by 210.6.150.96 on 01/26/22
slab, perimeter wall, ring beam, and upper dome form the wall are placed in the buttresses. The vertical P-T cables in
containment pressure boundary and are all prestressed. In the wall are anchored in the ring beam (top) and the base
addition, CANDU 6 CCBs include local details, such as the slab (bottom). The anchor heads of the cables in the upper
buttresses for anchorage of circumferential P-T cables, pene- dome are contained in the ring beam.
trations, temporary openings, and interior structures. As
shown in Figure 1, the base slab sits on the sub-base, which In addition to P-T cables, reinforcing bars have been used in
in turn is placed on bedrock. The perimeter wall is cylindrical CANDU 6 CCBs. For example, the upper dome contains
For personal use only.
38
CNL NUCLEAR REVIEW DEVELOPMENT OF FINITE ELEMENT MODELS FOR THE STUDY OF AGEING EFFECTS IN CANDU 6
VOL 5, NUMBER 1, June 2016 CONCRETE CONTAINMENT BUILDINGS – YUQING DING AND SHAHZMA JAFFER
reinforcing bars in both the circumferential and radial direc- understand the effect of different loading conditions and deg-
tions, and the perimeter wall has reinforcing bars in both the radation mechanisms on the performance of CCBs. The predic-
horizontal and vertical directions. The reinforcement is tions from the FEMs developed based on these simplifications
placed at both the bottom and top faces of the upper and and assumptions can be used to understand the global dis-
inner domes and at both the inside and outside faces of the placements and local strain and stress data, which were usu-
perimeter wall. The arrangement of reinforcement in the ring ally required for condition assessments according to the
beam and the bottom of the perimeter wall near the hinge Canadian and French experiences [3, 4]. Those measurements
connection between the perimeter wall and the base slab is were made to verify the elastic behaviour of the containment
very complex. in the course of the pressure loading and evaluate if the mean
strains and displacements are reversible [4]. Instruments for
Performing experiments to study the long-term behaviour of the global and local measurements for the concrete are
complex concrete structures such as CANDU 6 CCBs would mainly installed in or on the surface of the upper dome and
be extremely difficult and potentially impractical. As an perimeter wall; thus, the response of the upper dome and
alternative to true experiments, FEMs can be developed to perimeter wall is the major interest of the modelling.
facilitate the study of the behaviour of complex concrete
structures subjected to different loading conditions (includ- 4. Modelling Approaches
ing extreme loading assumed for accidents) and time-
dependent ageing degradation. Therefore, FEMs have been Formally approved versions, ANSYS Mechanical 12.0.1 [5]
developed for a better understanding of the overall behav- and LS-DYNA 971 [6], were used for the model development
iour of a CANDU 6 CCB during long-term operations and for so that the simulation results from independent FEMs can be
the study of the ageing degradation effects. compared with each other for validation. Both the ANSYS and
LS-DYNA codes have special features for the modelling of
concrete and have been widely used for the analysis of pre-
3. Modelling Assumptions and Simplifications
stressed concrete structures. The approaches used in the
The following assumptions and simplifications have been ANSYS and LS-DYNA models for the modelling of concrete,
made in modelling the CANDU 6 CCB for the study of the reinforcing bars, and P-T cables are summarized in Table 1
overall behaviour of the concrete structure: and discussed in the sections that follow.
• Local details in the CCB, including buttresses, penetra- 4.1. Modelling of concrete
tions, temporary openings, and embedded parts, were Concrete is a quasi-brittle material and behaves differently in
not considered in the modelling because their impact tension and compression states. As shown in Table 1,
TABLE 1. Summary of modelling approaches used in ANSYS cylindrical, and spherical coordinate systems were defined
and LS-DYNA models. for the input of reinforcement in the base slab, perimeter
wall, and domes, respectively.
ANSYS LS-DYNA
Concrete SOLID65 element with a Winfrith concrete model A discrete modelling approach was used in LS-DYNA models
tension “cutoff” criterion with strain softening in by modelling the concrete and reinforcing bars separately.
tension
Rebars Smeared reinforcement Discrete modelling with Both the discrete and smeared reinforcement models are
(volume ratios of rebars in truss elements widely used in the modelling of reinforced concrete struc-
concrete elements)
tures. The smeared reinforcement approach makes it easy
Cables Truss elements for cables in Same as ANSYS model
to construct a reinforced concrete model, which has a simple
perimeter wall, equivalent
reinforcement arrangement. However, it is difficult to obtain
layers with shell elements
for other cables the stress and location of reinforcement in a smeared
reinforcement model. The discrete representation can
CNL Nuclear Review Downloaded from pubs.cnl.ca by 210.6.150.96 on 01/26/22
elements [5]. Cracking is treated with SOLID65 elements as LS-DYNA models. This assumption is valid when the reinforc-
ing bars and grouted P-T cables in CCBs are still in a good
a “smeared band” of cracks, rather than discrete cracks, and
occurs as soon as stresses in the concrete exceed the tensile shape and the bond slips can be ignored. The perfect bonding
strength of the material. The constitutive stress–strain curve effect was modelled by establishing coupling between the
nodes of reinforcing bars or P-T cables and the correspond-
assumed for tension is linearly elastic up to the maximum
tensile strength. It was assumed that the concrete cannot ing coincident concrete nodes in the ANSYS model. An
undertake any tension once the tensile stress reaches the embedded technology was used in the LS-DYNA model to
simulate the perfect bond by constraining the velocity and
ultimate tensile strength of the concrete material. The use
of a strength-based criterion for crack initiation is based on acceleration of nodes of concrete and reinforcement or P-T
the assumption that the concrete structure is initially in a cables using an LS-DYNA keyword, *ALE_COUPLING_NODAL_
CONSTRAINT [6, 8].
perfect crack-free condition. In compression, a perfectly plas-
tic response was assumed after the peak compressive stress
is reached. 4.3. Modelling of prestressing cables
The arrangement of P-T cables in CANDU 6 CCBs is very com-
A Winfrith concrete model [6] with strain softening in ten- plex. To simplify the modelling work, truss elements were
sion was used in the LS-DYNA model, as indicated in used to model the horizontal and vertical P-T cables in the
Table 1. The Winfrith concrete model provides a basic plas- perimeter wall and the shell elements were used to model
ticity model that includes the third stress invariant for con- the equivalent layers representing the P-T cables in the
sistently treating both triaxial compression and triaxial upper dome, ring beam, and base slab in both the ANSYS
extension [7]. and LS-DYNA models. As discussed, a perfect bond was
assumed between the concrete and P-T cables and modelled
using different approaches in the ANSYS and LS-DYNA
4.2. Modelling of reinforcing bars models.
As summarized in Table 1, a smeared reinforcement model
was used in the ANSYS model. In the ANSYS model, the 5. Description of ANSYS and LS-DYNA Models
reinforcement layers were defined using the smeared
reinforcement option of the SOLID65 [5] element. The Figure 3 shows a 3-D view of the ANSYS model; a quarter of
reinforcement was assumed to be smeared and uniformly the CANDU 6 CCB without local details (e.g., buttresses) was
spread throughout the element. The amount of reinforce- modelled. The ANSYS and LS-DYNA models share the mesh
ment was defined by specifying a volume ratio and the orien- of concrete with 64 220 solid elements used in each model.
tation angles of the reinforcing bars. Local Cartesian, In addition, each model includes 5531 truss elements and
40
CNL NUCLEAR REVIEW DEVELOPMENT OF FINITE ELEMENT MODELS FOR THE STUDY OF AGEING EFFECTS IN CANDU 6
VOL 5, NUMBER 1, June 2016 CONCRETE CONTAINMENT BUILDINGS – YUQING DING AND SHAHZMA JAFFER
5.1.1. Concrete
Concrete material properties vary by a large amount depend-
ing on the concrete composition and construction process.
Ageing degradation of concrete could lead to significant
changes in concrete material properties. The concrete
material properties used in the FEMs have a significant
impact on the simulation results. The concrete material prop-
erties used for the design calculations of a typical CANDU 6
CCB were used for the modelling. It is recommended that
For personal use only.
42
CNL NUCLEAR REVIEW DEVELOPMENT OF FINITE ELEMENT MODELS FOR THE STUDY OF AGEING EFFECTS IN CANDU 6
VOL 5, NUMBER 1, June 2016 CONCRETE CONTAINMENT BUILDINGS – YUQING DING AND SHAHZMA JAFFER
As shown in Table 1, truss elements were used to directly A modified ANSYS model was developed using the equivalent
model the horizontal and vertical P-T cables in the perimeter prestressing loads calculated for the horizontal and vertical
wall, and equivalent layers of shell elements were created to cables in the perimeter wall instead of equivalent tempera-
represent the P-T cables in the upper dome, ring beam, and ture loads used in the baseline case. There is little difference,
base slab. Equivalent temperature loads were applied to the <4%, between the predicted stresses and strains at the
horizontal and vertical cables in the perimeter wall to simu- outside surfaces from the modified case and the baseline
late the prestressing effects in both the ANSYS and case. The result of comparison demonstrated that both
LS-DYNA baseline cases. An equivalent load approach was approaches can be used to simulate the prestressing effect.
used to simulate the prestressing effect provided by the P-T The approach using equivalent temperature loads was used
in case studies to make it easy to check the stress or strain in • The upper dome of a CANDU 6 CCB contains 3 tendon
a single P-T cable in the perimeter wall and study the effect of layers with an angle of 60° between each other.
the breakage of a single cable or a group of cables with mod- However, there are only 2 orthogonal tendon layers in
els modified from the baseline case. the dome of the test structure.
• The overall geometry of the test structure was scaled
The ANSYS model uses an implicit static solution. The dead 1:14 from the CANDU 6 CCB. However, the same scale
load and prestressing load were applied in one load step, could not be used for all construction details to maintain
and the internal pressure load was applied to the inside sur- realistic material properties. Different scale factors were
face of the model in a separate load step. The automatic time- used to define the dimensions and construction details
stepping feature of ANSYS was invoked to determine the of the test structure. For example, the scaling ratio of
appropriate sizes or numbers to break the load step into 1:14 will lead to in a thickness of 2.97 cm (1.17 in) for
substeps. the upper dome. However, the dome in the test structure
was constructed with a thickness of 12.7 cm (5 in) to
An explicit dynamic solution was applied in the LS-DYNA realistically represent construction details.
CNL Nuclear Review Downloaded from pubs.cnl.ca by 210.6.150.96 on 01/26/22
model. The total simulation time was set as 2 seconds for • Reinforcement ratios of the test structure are much
the baseline case with an internal pressure of 20.7 psig and higher than those of the CCB modelled in this study.
5 seconds for the case with an internal pressure of 103.5 psig. • The values of prestress applied in the test structure are
The dead load and prestressing load were applied during the different from those in the standard CANDU 6 CCB.
first second. The internal pressure was applied in the rest of
the analysis. The kinetic energy of the deforming structure Because of the differences listed above, the test structure is
was found to be well below 5% of its internal energy not a perfectly scaled replica of the standard CANDU 6 CCB,
throughout the majority of analysis; hence, the LS-DYNA but only a parallel model to predict the sequence of cracking
model was considered suitable for a quasi-static analysis. and the failure mode of the standard CANDU 6 CCB [14].
For personal use only.
• The test structure only includes the elements that resist In the BOSOR5 model, the CCB was modelled as axisymmet-
the overpressure loading, namely the base slab, perim- ric with a series of shell segments defined with respect to a
eter wall, ring beam, and upper dome. The test structure reference surface. The shell segments were layered to re-
does not include the inner dome as well as the weight of present different materials through the thickness of a com-
water stored within the inner dome. ponent. Each layer of reinforcing or prestressing steel was
44
CNL NUCLEAR REVIEW DEVELOPMENT OF FINITE ELEMENT MODELS FOR THE STUDY OF AGEING EFFECTS IN CANDU 6
VOL 5, NUMBER 1, June 2016 CONCRETE CONTAINMENT BUILDINGS – YUQING DING AND SHAHZMA JAFFER
CNL Nuclear Review Downloaded from pubs.cnl.ca by 210.6.150.96 on 01/26/22
For personal use only.
FIGURE 7. Comparison of deformations from numerical models and experiments (A) pressure versus vertical deflection of the
dome; (B) pressure versus horizontal defection at the middle of the perimeter wall.
modelled as a separate layer. Significant simplifications were models. Three types of loads including the deadweight of
made for the ring beam, inner dome, and hinge connection of the CCB, prestressing load, and an internal pressure were
the CCB, especially for the ring beam, to enable the use of also considered in the BOSOR5 model [15]. Equivalent load
shell segments for the modelling. Different simplifications approaches were used to simulate the prestress in all P-T
made in the BOSOR6 model and the ANSYS and LS-DYNA cables. The equivalent pressures/forces used in the
models might cause quite different local stresses and strains BOSOR5 model are in good agreement with those calculated
in those regions; however, the simplifications should not and used for the ANSYS and LS-DYNA models. For example,
have a significant influence on the predicted overall behav- a constant equivalent pressure of 0.271 MPa was calculated
iour in the upper dome and perimeter wall. for the cables in the upper dome, and equivalent pressures
of 0.253 MPa, 0.258 MPa, and 0.264 MPa were used in
In the BOSOR5 model, the foundation was modelled as an the BOSOR5 model for 3 portions of the upper dome,
axisymmetric Winkler foundation, in which the restraint to respectively [15].
the base slab is concentrated at a discrete number of points.
However, the foundation was not modelled in the ANSYS Material properties used in the BOSOR5 model are in good
and LS-DYNA models. The difference in the modelling of the agreement with those assumed in the ANSYS and LS-DYNA
base slab and foundation should not have a significant impact models.
on the predicted stress–strain data in the perimeter wall and
upper dome, which are the major interest of this paper. The BOSOR5 model was considered to be suitable for the val-
idation based on the above discussions. Figure 8 presents the
Loading conditions considered in the BOSOR5 model are predicted stresses at the inside and outside surfaces from the
consistent with those studied in the ANSYS and LS-DYNA BOSOR5 and the ANSYS and LS-DYNA models with an
internal pressure of 20.7 psig. As seen in Figure 8, there is stresses and strains, which are induced by temperature gra-
good agreement in the circumferential surface stresses and dients through the CCB components. The ANSYS model has
meridional stresses obtained from these 3 models. The cur- been modified and used successfully for the study of the in-
rent version of LS-DYNA can only provide stresses in the fluence of temperature loads on strain measurements for a
global Cartesian coordinate system; therefore, Figure 8 does CANDU 6 CCB as reported in Ding et al. [3]. To simplify the
not give the meridional surface stresses for the upper dome modelling work, a linear temperature distribution profile
and inner dome from the LS-DYNA model. Instead, normal along the thickness of the reactor building concrete was
stress components in the global Cartesian coordinate system assumed. A steady-state thermal analysis was performed
predicted from the ANSYS and LS-DYNA models were com- first to predict the temperature distribution in the CCB using
pared. Good agreement was observed in the predicted nor- the inside and outside surface temperature data as boundary
mal stress components at the outside and inside surfaces of conditions in the model. Then, a nonlinear structural analysis
the upper dome and inner dome. was performed to examine the structural behaviour of the
CCB under the thermal and other existing loads, such as the
7. Applications of the FEMs dead weight of the concrete structures and prestressing
loads. Temperature data obtained from the results of the
The ANSYS and LS-DYNA models have been modified for pre-
thermal analysis was applied as nodal thermal loads in the
liminary studies of the effect of temperature loads on strain
structural analysis. The analysis results demonstrated that
and stress measurements. Both models can be used to study
the temperature distribution within the containment concrete
the ageing effects in CCBs. A brief introduction to the meth-
has a significant impact on the strain, which was seen in the
odology used for these studies is given here.
numerical simulation results and the strain measurements [3].
7.1. Understanding of temperature effects
CCBs of nuclear reactors exposed to external and internal However, the linear temperature distribution assumed in the
temperature changes are likely to be subjected to significant preliminary analysis only provides a rough estimation of the
46
CNL NUCLEAR REVIEW DEVELOPMENT OF FINITE ELEMENT MODELS FOR THE STUDY OF AGEING EFFECTS IN CANDU 6
VOL 5, NUMBER 1, June 2016 CONCRETE CONTAINMENT BUILDINGS – YUQING DING AND SHAHZMA JAFFER
temperature effects. CANDU 6 CCBs are made of concrete of prestressing losses can be created from the ANSYS base-
with relatively large dimensions. The temperature variation line model by directly updating the equivalent loads used to
at greater depths from the outside surface, in particular at simulate the prestressing effect in the model.
the inside surface of the CCB wall, is much less than the tem-
perature variation in the ambient atmosphere. The tempera- 8. Summary and Conclusions
ture profile in CANDU 6 CCBs might be nonlinear [3].
Nonlinear transient thermal and structural analyses could FEMs have been developed for CANDU 6 CCBs using 2 com-
be considered in the future work to better predict the tem- mercial finite element programs, ANSY and LS-DYNA. The
perature distribution in the CCBs and the response caused modelling assumptions/simplifications, approach and valida-
by temperature loads. tion have been introduced.
creep and shrinkage of concrete, corrosion of reinforcing pendent finite element programs have been validated
bars and P-T cables, postulated breakage of local P-T cables. using multi-source data, including simulation results
and experimental data from a scaled test structure for
Design codes (e.g., ACI 209R-92 [16]) can be used to deter- a standard CANDU 6 CCB, and have been proven to pro-
mine the creep strains or coefficients and shrinkage strains vide reasonable results.
for different portions of the CCBs considering the inside and • The validated FEMs can be used for the study of the
outside environmental conditions. An effective modulus overall behaviour of a CANDU 6 CCB during long-term
approach can be used to roughly estimate the bound values operations with ageing degradation issues to better
of strains caused by creep using ANSYS models. The effective understand the instrumentation data collected from
modulus approach is practical with assumed constant stress CANDU 6 CCBs and provide guidance for the develop-
For personal use only.
conditions; hence, the approach is suitable for a rough esti- ment of novel inspection and monitoring approaches
mation of the band values of strains induced by creep. For for CANDU 6 reactors.
more accurate results, equivalent temperature loads can be
applied to each portion of the CCB to simulate the creep The modelling methodologies used in this study could be fur-
effect based on the creep coefficients and stresses of specific ther improved. Complex temperature distribution in CANDU
elements using APDL (ANSYS Parametric Design Language) 6 CCBs should be included in the FEMs for a better under-
[5]. The shrinkage strains estimated from design codes can standing of temperature and ageing degradation effects.
be converted to equivalent temperature loads applied to the
inside and outside surfaces of the CCB to predict the strain
changes due to concrete shrinkage. Creep and shrinkage are
significantly affected by environmental variables, such as
temperature and humidity [17]. The temperature distribu- References
[1] CSA, 2008, In-service Examination and Testing Requirements for Concrete
tion in the CCB is a complex 3-D issue that may have a signifi-
Containment Structures for CANDU Nuclear Power Plants, Canadian
cant impact on creep and shrinkage predictions. Complex Standards Association, Standard N287.7-08 (Reaffirmed 2013).
temperature distribution in the CCB should be included in [2] J.P. Radulescu, L. Pradolin and J.C. Mamet, 1997, “Ultimate Pressure
the numerical models for a better prediction of degradation Capacity of CANDU 6 Containment Structures,” Proceedings, 1997 CNA/
effects. CNS Annual Conference, Volume 2, Session 4A, June 8–11, 1997, Canadian
Nuclear Society, Toronto, ON.
[3] Y. Ding, S. Jaffer, P. Angell, 2014, “Influence of Temperature on Strain
Analysis cases can be created based on the baseline cases by Monitoring of Degradation in Concrete Containment Buildings,” Fontevraud
updating the input parameters for the study of the effect of 8, Avignon, France, September 15–18, 2014, Paper O-T10–069.
corrosion of reinforcing bars and corrosion or fracture of [4] Fib, 2001, “Nuclear Containments,” State-of-art Report, Bulletin 13,
P-T cables. International Federation for Structural Concrete, Federation Internationale
de Beton (fib), Lausanne, Switzerland.
[5] ANSYS, Inc., 2009, Theory Reference for the Mechanical APDL and
It would be difficult to model the effect of each contributing Mechanical Applications, Release 12.0, Canonsburg, PA, USA.
factor of loss of prestressing forces because limited data are [6] LSTC, 2007, “LS-DYNA® Keyword User’s Manual,” Vol. I and II. Version
available on some time-dependent degradation factors, par- 971, Livermore Software Technology Center (LSTC), Livermore, CA.
ticularly the relaxation of prestressing steel. Sensitivity [7] L. Schwer, 2011, “The Winfrith Concrete Model: Beauty or Beast?
Insights into the Winfrith Concrete Model,” 8th European LS-DYNA® Users
analyses can be performed to directly study the relationship
Conference, May 23–24, 2011, Strasbourg, France.
between the prestressing loss and the resulting concrete [8] L. Schwer, 2014, “Modeling Rebar: The Forgotten Sister in Reinforced
strain changes assuming that the prestressing loss is uni- Concrete Modeling (v2),” 13th LS-DYNA® International Conference, June 8–
formly distributed in CCBs. Cases with assumed percentage 10, 2014, Dearborn, MI.
[9] P. Desayi, and S. Krishnan, 1964, “Equation for the Stress-Strain Curve of [14] J.G. McGregor, S.H. Simmonds and S.H. Rizkalla, 1980, “Test of a
Concrete,” Journal of American Concrete Institute, 61(3), pp. 345–350. Prestressed Concrete Secondary Containment Structure,” A Technical
[10] D. Kachlakev, T. Miller, S. Yim, K. Chansawat and T. Potisuk, 2001. Report to the Atomic Energy Control Board, Structural Engineering
“Finite Element Modeling of Reinforced Concrete Structures Strengthened Report No. 85, University of Alberta, Department of Civil Engineering,
with FRP Laminates,” Final Report SPR-316, Oregon Department of Edmonton, AB.
Transportation, Salem, OR. [15] D.W. Murray, C. Wong, S.H. Simmonds and J.G. MacGregor, 1980, “An
[11] A.M. Ibrahim, and M. Mahmood, 2009, “Finite Element Modeling of Inelastic Analysis of The Gentilly-2 Secondary Containment Structure,” A
Reinforced Concrete Beams Strengthened with FRP Laminates,” European Technical Report to the Atomic Energy Control Board, Structural
Journal of Scientific Research, 30(4), pp. 526–541. Engineering Report No. 86, University of Alberta, Department of Civil
[12] Comite Euro-International du Beton, 1993, CEB-FIP Model Code 1990: Engineering, Edmonton, AB.
Design Code, Thomas Telford, London, United Kingdom. [16] ACI, 1992, Prediction of Creep, Shrinkage, and Temperature Effects in
[13] S.-J. Jeon, and C.-H. Chung, 2005,“Axisymmetric Modeling of Concrete Structures, American Concrete Institute, Standard ACI 209R-92
Prestressing Tendons in Nuclear Containment Dome ,” Nuclear Engineering (Reapproved 2008).
and Design, 235(23), pp. 2463–2476. doi: 10.1016/j.nucengdes.2005. [17] ACI, 2008, Guide for Modeling and Calculating Shrinkage and Creep in
06.007. Hardened Concrete, American Concrete Institute, Standard ACI 209.2R–08.
CNL Nuclear Review Downloaded from pubs.cnl.ca by 210.6.150.96 on 01/26/22
For personal use only.
48