You are on page 1of 12

This article was downloaded by: [Amir Kabir University] On: 11 March 2012, At: 09:27 Publisher: Taylor

& Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Pavement Engineering


Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gpav20

ABAQUS model for PCC slab cracking


Anastasios M. Ioannides , Jun Peng & James R. Swindler Jr.
a a a a

Department of Civil and Environmental Engineering, University of Cincinnati (ML-0071), PO Box 210071, Cincinnati, OH, 45221 0071, USA Available online: 24 Nov 2006

To cite this article: Anastasios M. Ioannides, Jun Peng & James R. Swindler Jr. (2006): ABAQUS model for PCC slab cracking, International Journal of Pavement Engineering, 7:4, 311-321 To link to this article: http://dx.doi.org/10.1080/10298430600798994

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

International Journal of Pavement Engineering, Vol. 7, No. 4, December 2006, 311321

ABAQUS model for PCC slab cracking


ANASTASIOS M. IOANNIDES*, JUN PENG and JAMES R. SWINDLER Jr.
Department of Civil and Environmental Engineering, University of Cincinnati (ML-0071), PO Box 210071, Cincinnati, OH 45221 0071, USA (Received 13 September 2005; revised 5 May 2006)

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

To contribute towards the development of improved failure criteria for pavement systems that could potentially replace Miners hypothesis in future pavement design guides, Hillerborgs Fictitious Crack Model can be used to simulate crack propagation in concrete pavement slabs, thereby dispensing with the need to conduct time consuming and expensive physical experiments in the laboratory and the eld. Commercial nite element program ABAQUS is used for slabs assumed to rest on a dense liquid foundation, and to be loaded by an edge load. Both notched and unnotched slabs are considered, and the effects of various loading parameters, notch size, size of the loaded area, slab thickness and slab size are examined. A comparison is made between displacement and loading-controlled testing of the slabs. Keywords: Concrete pavement fracture; Fictitious crack model; ABAQUS; Finite element analysis

1. Introduction The majority of current pavement analysis and design procedures have two primary features: (a) With respect to the prediction of behavior from initial loading until shortly before failure, current methods are based on the theory of linear elasticity; (b) With respect to the prediction of performance, distress, and failure, current methods resort to rather simple, mostly empirical and phenomenological concepts, such as Miners cumulative linear fatigue hypothesis (Miner 1945). The conventional approach to pavement design is commonly a two-stage one: rst, a critical primary response is calculated, which is subsequently passed into a statistical/empirical algorithm that converts it into a measure of performance. A cursory review of existing analytical and design procedures for pavements might lead to the impression that these two aspects are decoupled, when in fact they are closely interrelated. It should be appreciated that expediency is the only justication for such practices, pending the development of more reliable and rational (mechanistic) alternatives. It is often the case, meanwhile, that the choice of a particular empirical and phenomenological performance criterion is by far the most overriding consideration in any design exercise. Consequently, derivation of improved performance relationships, preferably ones that recover their interrelationship with the primary response calculation process, is an on-going objective of pavement
*Corresponding author. Email: anastasios.ioannides@uc.edu Email: jun.peng@uc.edu Email: swindljr@email.uc.edu

researchers. The study reported herein is intended as a contribution to this effort, which seeks to develop models that are implementable in sophisticated nite element codes and allow parametric studies and predictions of structural behavior. More specically, this paper focuses on the application of the ABAQUS/STANDARD nite element software (Hibbitt et al. 1994) in tracking crack propagation in Portland Cement Concrete (PCC) pavement slabs, subject to the usual restrictive assumptions of Westergaard (1926), namely: (a) full contact (no temperature differential); (b) single slab (no load transfer); (c) single placed layer (no subbase); (d) semi-innite foundation (no rigid bottom); and (e) one tire-print. Fracture mechanics, particularly the Fictitious Crack Model (FCM) proposed by Hillerborg et al. (1976) to simulate crack propagation, can be an important tool toward a better understanding of crack formation and propagation in pavements, which in turn can provide us with models capable of capturing these phenomena.

2. Context of investigations at the University of Cincinnati The context for the present study is provided by a longterm, step-by-step effort that started in the late 1990s, with a historical review of the major research activities concerning the development of fatigue cracking in both

International Journal of Pavement Engineering ISSN 1029-8436 print/ISSN 1477-268X online q 2006 Taylor & Francis http://www.tandf.co.uk/journals DOI: 10.1080/10298430600798994

312

A. M. Ioannides et al.

PCC and bituminous pavements (Ioannides 1997b). Efforts spanning almost a century were examined, with the objective of identifying the sequence of events that have led to the formulation of current approaches used to account for fatigue in pavement design codes. In particular, the roots of Miners cumulative linear fatigue hypothesis were traced, and its advantages and limitations were discussed. It was shown that the foundations for this crucial aspect of current design procedures were surprisingly feeble, and the desirability of enhanced mechanistic approaches to fatigue cracking prediction was established. The scarcity of suitable candidates to replace Miner was highlighted, and fracture mechanics was proposed as a promising realm to explore. An exhaustive examination of the advantages and limitations of a variety of fracture mechanics options for pavement engineering was then embarked on (Ioannides 1997a). Early pavement fracture mechanics efforts were primarily associated with Paris Law (Paris and Erdogan 1963), a phenomenological construct that was eventually shown to offer few breakthroughs compared to Miners hypothesis. Of primary historical importance are the studies by Majidzadeh et al. (1971) at Ohio State University (OSU), and of Prof. Robert L. Lytton (Lytton and Shanmugham 1982) at Texas A&M University (TX A&M). The major shortcomings of these efforts, which probably account for the lack of progress achieved, were identied as their acceptance of the validity of Linear Elastic Fracture Mechanics (LEFM) (Broek 1986) as applied to (bituminous) concrete mixtures, and of Paris law for explaining pavement fatigue. Having established the inability of Paris Law to address the fundamental weakness of current pavement design procedures satisfactorily, a number of alternative approaches were then examined. Noteworthy among these were investigations conducted by Prof. Heshmat A. Aglan (Aglan and Figueroa 1993) at Tuskegee University, whose more mechanistic avor was clearly discernible. Advanced concepts of thermodynamics and viscoelasticity were employed for the development of the Modied Crack Layer Model for the characterization of the nearfailure behavior of bituminous mixes. Another very promising investigation was conducted by Prof. YeouShang. Jenq (Jenq and Perng 1991) of OSU, who applied to asphalt pavements the FCM introduced by Swedish investigator Arne Hillerborg. Such research establishes the fact that intercontinental collaboration is invaluable in the development of effective procedures to account for the fatigue cracking and fracture phenomena in pavements. Its limitations notwithstanding (Bazant 2002), the FCM was also identied as a most promising tool in this effort, and additional pavement applications thereof were sought. Of particular interest were found to be the contributions of two Danish investigators, Hans H. Bache and Ib Vinding, who applied Hillerborgs FCM to concrete pavement engineering (Bache and Vinding 1990). They also suggested a number of similitude considerations that ow naturally from the application of this model, and that

highlight the signicance of the application of the principles of dimensional analysis. Their work validated some earlier observations made by pavement engineers concerning the relative size of beam specimens compared to that of in situ pavement slabs. The specimen size effect (Bazant and Planas 1998) was thus found to be at the heart of the concrete fracture problem, and its resolution to be essential before unraveling the complex phenomenon of pavement fatigue cracking. An opportunity to overcome the specimen size limitation was identied in the work of Russian investigator Vyacheslav D. Kharlab (Kharlab 1995, personal communication) of St. Petersburg State University of Architecture and Civil Engineering (SPSUACE). Kharlabs approach was shown to have similarities to Hillerborgs FCM, but also to be in contradiction to it in some respects. These two proposals were selected in this study as the most promising tools available for PCC pavement fracture mechanics applications at this time. To begin with, nite element analysis was used in simulating crack propagation in PCC beams (Ioannides and Sengupta 2003). Experimental data (Liu 1994) pertaining to the load vs. deection (P 2 d) and the load vs. crack mouth opening displacement (P-CMOD) behavior of simply supported PCC beams subjected to a point load at mid-span were successfully reproduced in this way, past the elastic limit to failure. Finite element package GTSTRUDL (1993) was used to generate the exibility matrix pertaining to the linear elastic aspects of structural response, whereas fracture behavior in accordance with the FCM was examined using CRACKIT, a FORTRAN computer program coded during the course of the study. The GTSTRUDL/ CRACKIT combination was then used to generate numerical analysis data for different beam sizes, and these data were interpreted in dimensionless format. These beam test results were subsequently conrmed by implementing the FCM in the commercial package ABAQUS (Ioannides and Peng 2004). The validity of the ABAQUS simulation for beams was checked through comparison with results obtained using CRACKIT, and from other independent laboratory and analytical investigations. The methods adopted for the analysis of pavement slabs in the present paper are an extension of those applied to beams, thereby afrming the suitability of the step-by-step approach adopted in this project.

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

3. Fracture analysis of slabs using JOINTC elements A slab is assumed to be resting on a dense liquid foundation loaded by a single, square (or rectangular) edge load. The dimensions of the slab are selected to correspond roughly to those in an actual concrete pavement, and to lend themselves for a series of nite element runs without undue demands on memory and other computer resources, as follows: length, L 6.10 m (240 in.), width, W 3.05 m (120 in.) and thickness,

ABAQUS model for PCC slab cracking Table 1. Baseline pavement system considered Slab geometry Length 240 in. Width 120 in. Thickness 6 in. Number of slab layers 3 (or 2) Foundation characteristics WINKLER Dense liquid Subgrade modulus 200 psi/in. Metric conversions 1 in. 25.4 mm 1 lb 4.44822 N 1 psi 6.89476 kPa 1 psi/in. 0.27145 MN/m3 Slab properties Modulus 4 Mpsi Poissons ratio 0.15 Tensile strength 463 psi Fracture energy 4.31 1024 kips/in. Applied load Edge loading 12 12 in. Pressure 100 psi Westergaard responses Maximum bending stress 768 psi Maximum deection 40 mils

313

h 0.152 m (6 in.) (see also table 1). Additional advantages of the slab dimensions selected include the elimination of slab-size effects by retaining a value above 5 for the ratio of slab size or width to the radius of relative stiffness of the slab-subgrade system, and the adoption of a rectangular rather than of a square slab. Both notched and unnotched slabs are considered, with the notch (when present) specied to extend both through the slab thickness (vertically), and along the slab symmetry line (horizontally). Consequently, the crack is assumed to follow the symmetry line, as it propagates from the bottom up. The notch is described by two ratios, of notch-to-slab thickness, (az/h), and of notch-to-slab width, (ay/W), in the vertical and horizontal directions, respectively. The values of these ratios are limited by the number of elements used in the two directions, since only notches spanning an entire element are considered. Figure 1 illustrates the geometry of the slab and the denition of notches in the z- and y-directions. In cross section, the slab is subdivided into three layers, a value that allows meaningful consideration of crack propagation through the thickness without undue penalty in terms of execution time. In discussing the use of three dimensional nite element analysis to slabs on grade,

Ioannides and Donnelly (1988) had found that even two layers are adequate for linear elastic analysis. A more detailed examination of through-the-slab thickness cracking will probably require a ner subdivision. In plan view, the subdivision is into 40 elements in the x-direction and 10 elements in the y-direction; each element is, therefore, 152 305 mm (6 12 in.). The element adopted in modeling the slab is the C3D27R, which is described as an isoparametric, 3D, 27-node, reduced integration element. For the purpose of simulating crack propagation, a series of JOINTC elements is used to connect each pair of nodes on either side of the symmetry plane, which thereby serves as the potential fracture plane. There are three main classes of spring-type elements available in ABAQUS/STANDARD that allow the user to dene explicitly the desired fracture process: SPRING, ITS (tube support elements), and JOINTC (exible joint element). The stiffness (force per relative displacement) for all these is dened using the *SPRING option. SPRING elements include three distinct types: SPRING1, SPRING2, and SPRINGA, and all three can be linear or nonlinear. SPRING1 is a spring between a node and the subgrade, acting in a xed direction and is not useful at this time. SPRING2 and SPRINGA connect two nodes, but whereas the rst acts in a xed direction, the line of action of the second can rotate, as might be necessary in large displacement analyses. Moreover, SPRINGA requires that the two nodes it connects be separated by a nite distance, whereas SPRING2 can connect two nodes that occupy the same geometrical location. ITS elements are not relevant to this work. JOINTC elements consist of translational and rotational springs and connect two nodes that are essentially at the same geometric location, thereby dispensing with the need to dene a spring length. Ioannides et al. (2005) report that when ABAQUS/STANDARD is used, SPRINGA appears to be sensitive to the choice of spring length, for which no rational method of determination could be devised. Analyses were conducted to compare the behavior of SPRING2 and JOINTC elements, and to justify the choice of the latter in this study. It was observed that these two element types give

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

Figure 1. Slab geometry and Notch denition

314

A. M. Ioannides et al.

Figure 2. Bilinear s 2 w curve for PCC, per FCM

very similar results and reinforce the credibility of the model using the JOINTC element. The latter was chosen because the plot of the deected shape produced by ABAQUS/POST is more attractive. There is no other compelling reason to pick one over the other. In this study, a bilinear closing pressure vs. crack opening displacement (s 2 w) curve, required by the FCM, is assumed, for the denition of which three points are needed, depending primarily on the tensile strength of the concrete, ft. The rst point is, of course, ( ft, 0), while the two other dene the location of the intersection point ( fI, wI), and the value of the critical crack opening at (0, wc). Following the proposals by Petersson (1981) and Gustafsson (1985), wc is set to 3.6 Gf/ft, in which Gf is the concrete fracture energy, while the knee of the bilinear curve is located at 1/3 ft and 2/9 wc. The experimental beam FCM bilinear curve employed by Liu (1994) is retained here, as shown in gure 2. For this curve, critical CMOD, wc 0.085154 mm (0.0033525 in.); intersection point, (wI, f I) (0.0189 mm, 0.1064988 MPa) or (0.000745 in., 0.1544233 ksi); for comparison purposes, a second, linear FCM curve with critical CMOD, wc 0.047262 mm (0.0018607 in.) was also considered; both these curves correspond to Gf 75.4 N/m (0.431 lb/in.). The nonlinear response of the JOINTC elements is dened in accordance with this curve, converted into a cohesive force vs. crack mouth opening displacement relationship, on the basis of energy equivalence and moment balance considerations (Shah et al. 1995). Because of the inability of ABAQUS to accommodate truly unnotched slabs, a ctitious notch extending half an element in each of the two directions of interest is introduced in such cases. This operation is compensated by doubling the stiffness of the two neighboring JOINTC elements. The load is applied at the middle of the long edge of the slab, and the size of the loaded area considered is 305 305 mm (12 12 in.) in most cases. The maximum bending stress predicted by Westergaard (1948) under a pressure of 689 kPa (0.100 ksi) is 5292 kPa (0.768 ksi) and the corresponding maximum vertical

deection is 1.01 mm (39.6 mils). This guarantees that the slab will crack under a reasonable pressure, producing simultaneously a measurable deection. For the sake of additional comparisons, loaded areas of 305 610 mm (12 24 in.) and 610 305 mm (24 12 in.) are also examined. Fracture in slabs is simulated the only two available options with regard to load application, i.e., (vertical) displacement control (maintaining a maximum displacement) and loading control (maintaining a maximum load). By default, under displacement control, the displacement is applied as a RAMP function, starting at 0 and reaching a relative maximum value of 1.0 at the end of the load step. The absolute maximum displacement is set to a value of 25.4 mm (1.0 in.) downward at each of the nodes dening the loaded elements. In most cases, two fully loaded elements were considered, and the displacement was xed at 18 nodes. Under loading control conditions, the load may be applied as a uniformly distributed load or as a series of concentrated nodal loads. The latter is particularly useful in cases involving partially loaded elements. For the distributed load, the applied pressure will increase linearly from 0 at time 0 to a relative maximum value of 1.0 at time 1.0. The absolute value of the maximum pressure is typically xed between 1380 kPa (0.2 ksi) to 34,450 kPa (5 ksi). These choices correspond to approximately 0.5 to 10 times the tensile strength of the material ( ft 3192 kPa or 0.4633 ksi), depending on the particular stage of the fracture process one is interested in. It is freely admitted that several of these choices are rather unrealistic when in situ pavements are considered, and that the mesh idealization described is not ideal. They were considered, however, expedient and quite adequate for the purposes of this preliminary analytical investigation, which aims primarily at verifying the most signicant aspects of the formulation, and at delineating the most prominent trends to be expected. Occasional tests with much ner meshes verify this assertion. Figure 3 displays the stress contours for a typical case examined. It is noted that the analyses presented are not aimed at calculating the stress or displacement levels per se, but only to create a robust numerical model. Increasing the mesh neness would improve the accuracy of the solution, but would not serve identify any weaknesses of the numerical model, while at the same time it would inhibit the efciency of the study, increasing execution times and resource expenditures. Consistent with this approach, results obtained are presented primarily in the form of Tables, which lend themselves better to verifying their veracity by those wishing to adopt the approach proposed. A practical interpretation for the purposes of modifying existing pavement design procedures is pursued only to the limited extent possible at this time. It is anticipated that future research will explore the agreement between the nite element results and in situ measurements, pending the performance of pertinent experiments.

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

ABAQUS model for PCC slab cracking

315

Figure 3. Stress contours for typical case considered

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

4. Discussion of results 4.1 Effect of loading parameters A series of runs with the smallest notch considered in this study was performed. The notch has a depth of one layer (50.8 mm or 2 in.) and a width of one row of elements (305 mm or 12 in.); it is formed by removing two JOINTC elements in each of the vertical and horizontal directions. Loading control is employed, but the comments and observations below apply to displacement control, as well. Load application requires at least two parameters: the initial time increment (ITI), and the time period of the step (TPS). Two additional parameters may also be specied: the minimum time increment (MnTI), and the maximum time increment (MxTI), for which there is no default value, i.e. it is left unbounded. The TPS is set at 1.0, for convenience, ensuring that all time values are also prescribed in relative terms. In earlier linear elastic analyses (Ioannides et al. 2005), it was found that MxTI has no inuence on the results, as might have been expected. Changing the maximum displacement or velocity leads to proportional changes in the maximum reactive force as predicted by linear elasticity. The

following procedure for selecting these parameters was formulated: it is desired to have about 30 time increments during the entire loading process, each increment lasting 3 1022 relative units of time or 3.33% of the total time; this denes the MxTI. The ITI is set to about 1/5th of this value (6 1023); the MnTI is chosen as approximately 1/5th of the ITI (1 1023). In particular, the following responses are monitored in table 2: the rst peak stress occurring at the horizontal crack tip, stip; the applied pressure when stip is achieved, pc; the load line displacement (LLD), i.e. the vertical displacement at the center edge node at the top surface of the slab; and the crack mouth opening displacement (CMOD), which is twice the value of the horizontal (xdirection) displacement along the loaded edge of the center edge node at the bottom surface of the slab. Table 2 also gives the number of layers (NL) into which the 152.4 mm (6 in.) slab was subdivided, the specied maximum pressure to be applied, pmax (set to a value many times greater than ft, to ensure cracking occurs), the number of loading increments (NINC) sustained, and the execution time (CPU) consumed. It is observed that the choice of loading parameters can inuence the maximum responses calculated by up to about 25%. This reects

Table 2. Effect of loading parameters. NL 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 5.00 2.50 2.00 1.50 1.50 1.50 5.00 2.50 1.50 1.50 1.50 1.50 5.00 2.00 5.00 ITI 10 1025 1025 1025 1025 1025 1025 1025 1025 1025 1025 1025 1023 1023 1023
25

MxTI N/S N/S N/S N/S N/S N/S N/S N/S N/S N/S N/S N/S 2.00 1022 2.00 1022 2.00 1022

pmax (ksi) 50 50 50 50 25 10 50 50 50 25 10 5 5 5 2

NINC 37 44 46 34 49 41 44 43 48 45 39 38 54 59 53

CPU (min) 56 66 66 46 76 61 152 144 151 156 132 91 112 120 121

LLD (mils) 37.1 42.7 33.9 33.8 43.7 39.1 36.7 42.0 37.9 42.9 38.7 43.6 38.7 40.4 40.4

CMOD/2 (mils) 1.0 13.1 0.9 1.1 1.4 1.2 0.7 0.8 0.8 0.9 0.8 1.0 0.8 0.8 0.8

stip (ksi)
0.609 0.768 0.557 0.630 0.787 0.642 0.579 0.667 0.600 0.683 0.612 0.694 0.613 0.641 0.641

pc (psi) 82.9 94.9 75.9 85.7 96.9 87.4 82.9 94.9 85.7 96.9 87.4 98.5 87.5 91.3 91.3

Note: Runs employ loading control; slabs notched: (ay/W) 10%; (az/h) 33.3%; N/S: not specied; MnTI 1.00 1025.

316

A. M. Ioannides et al.

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

Figure 4. Effect on loading parameters on stress distribution

the sensitivity of the maximum responses to the particular instant in time at which they are sampled, which in effect means to the particular applied pressure under which they occur. This follows from the fact that time elapsed is linearly related to the applied pressure. Additional investigations are necessary to establish the procedures needed to produce accurate as well as precise predictions of the maximum responses desired. In contrast, the distribution of the responses does not appear to be nearly as sensitive to the input loading parameters, as illustrated in gure 4, in which the various curves cluster over one another. It is noted that the peak stress, stip, in table 2 assumes values well in excess of ft. A possible explanation for this phenomenon is provided by the stress gradient criteria of strength (SGCS) for quasi-brittle materials, rst proposed by Kharlab and Minin (1989) and later elaborated by Kharlab (1989, 1990). The formulation of the SGCS is based on the experimental observation that local strength of the material is higher where elastic stress distribution is more non-uniform. Kharlabs SGCS are based on the hypothesis that a material may not fail under theoretically predicted high (or even innite) stresses, if these are sufciently localized in nature. Stated in another way, this hypothesis recognizes the signicance of the stress gradient, a factor recognized in the West, as well (Siemes 1982). Additional discussion of the incorporation of SGCS in fracture mechanics analyses has been presented by Khazanovich and Ioannides (1993). 4.2 Inuence of Notch size To investigate the inuence of the two aforementioned notch ratios on slab response, a series of runs were performed using loading control, with the input par-

ameters specied above. In each case, the JOINTC elements were removed from an additional row of elements or an additional layer, simulating increased notch sizes. The unnotched case is modeled using a ctitious notch of half an element row or (ay/W) 5%, and half an element layer or (az/h) 16.7%, and compensating for these changes by doubling the stiffness of the rst JOINTC elements at every point of the forcedisplacement curve dened by the FCM. Four different responses are tracked, namely, the maximum stress at the node located at the notch tip, stip, occurring when the crack begins to propagate; the applied pressure, pc, corresponding to the development of stip; the horizontal displacement at the bottom of the slab at the middle of the loaded edge, directly below the center of the applied load, corresponding to the development of stip, and being equal to one-half the CMOD; and the vertical displacement at the top of the slab at the middle of the loaded edge, at the center of the applied load, corresponding to the development of stip, and being equal to the LLD. In each case, a pair of these parameters is recorded, pertaining to the two directions of crack propagation: at the notch front in the vertical, z, or at the notch back in the horizontal, y, directions. In general, the crack propagates in each direction at a different time and load level, but in each instance when the crack opening exceeds wc, as indicated by the results in tables 3 and 4. It is observed that the crack propagates rst in the vertical direction and then horizontally, indicating that the vertical notch size ratio, (az/h), is more signicant. This is not unexpected, since the primary contributor to the stiffness of the slab is its thickness; consequently even a small reduction in thickness leads to pronounced deterioration in performance. The maximum stip is much smaller for the notched cases than for the unnotched case.

ABAQUS model for PCC slab cracking Table 3. Inuence of notch Size (at notch Front). Unnotched in both directions (ay/W) (%) 0 (az/h) 33.3% 10 20 30 50 70 100 (az/h) 66.7% 10 20 30 50 70 100

317

stip (ksi) 0.796


0.693 0.708 0.689 0.694 0.694 0.694 0.412 0.353 0.350 0.395 0.393 0.393

% 100.00 87.06 88.94 86.56 87.19 87.19 87.19 51.76 44.35 43.97 49.62 49.37 49.37

CMOD/2 (mils) 1.00 1.01 1.19 2.11 1.18 1.15 1.18 2.23 2.94 4.47 9.22 9.39 9.49

% 100 101 119 211 118 115 118 223 294 447 922 939 949

LLD (mils) 66.2 50.1 47.5 45.9 45.9 45.9 46.0 41.0 42.7 56.6 101.6 100.0 102.4

% 100.00 75.68 71.75 69.34 69.34 69.34 69.49 61.93 64.50 85.50 153.47 151.06 154.68

pc (psi) 153.1 113.1 105.1 101.1 101.1 101.1 101.1 85.1 85.1 113.1 189.1 185.1 189.1

% 100.00 73.87 68.65 66.04 66.04 66.04 66.04 55.58 55.58 73.87 123.51 120.90 123.51

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

Note: Runs employ loading control: pmax 0.2 ksi; slab dimensions 120 240 6 in.; ITI 5 1023; MnTI 1 1025; MxTI 2 1022; Columns marked % provide the results of the preceding columns as a ratio of the corresponding unnotched case response.

In general, as the notch size increases, stip, LLD, and pc decrease, whereas CMOD increases; trends contrary to this may occur at large notch sizes, presumably because the notch in such cases is beyond the limits of the loaded area, which is 305 305 mm (12 12 in.), i.e. extends only one element-row width on either side of the slab center-line. In some cases, the crack does not even propagate at all in the y-direction by the end of the specied load step (maximum applied pressure of 1378 kPa or 0.2 ksi). The latter was dened by considering the response of the unnotched beam, in a manner that would apply to all cases analyzed and result in efcient utilization of available computer resources. The maximum bending stress predicted by Westergaard (1948) noted earlier is already almost twice the specied tensile strength of the material.

A similar series of runs was also conducted using displacement control, with maximum applied displacement arbitrarily set at 25.4 mm (1 in.). The variation of the CMOD is tracked as a function of the applied load, RF3, calculated as the sum of vertical reaction forces under loaded area. The pressure to be applied in not specied in this case, but for the cases considered it was as high as 6000 kPa (0.900 ksi). It is observed in gure 5 that, in general, the fracture process consists of three stages: an initial quasi-linear elastic region, extending to about 178 kN (40 kips) or pressure of 3824 kPa (0.555 ksi); a main fracture region, extending to about 267 kN (60 kips) or pressure of 5739 kPa (0.833 ksi); and a collapse region, which is also quasi-linear at loads above about 267 kN (60 kips). The linearity of the latter region can be ascribed to the more signicant role that the Winkler foundation, itself linearly elastic, plays in this region. The

Table 4. Inuence of notch Size (at notch back) Unnotched in both directions (ay/W) (%) 0 (az/h) 33.3% 10 20 30 50 70 100 (az/h) 66.7% 10 20 30 50 70 100

stip (ksi) 0.874


0.833 0.605 0.553 0.348 0.065

% 100.00 95.31 69.22 63.27 39.82 7.44

CMOD/2 (mils) 1.00 1.14 2.41 6.40 9.08 9.10

% 100

LLD (mils) 66.2

% 100.00 78.40 90.48 109.37 159.06 159.52

pc (psi) 153.1 117.1 129.1 145.1 200.0 200.0

% 100.00 76.49 84.32 94.77 130.63 130.63

114 51.9 241 59.9 640 72.4 908 105.3 910 105.6 No crack tip 157 32.6 274 40.5 426 54.4 986 104.9 872 93.4 No crack tip

0.313 0.316 0.307 0.306 0.199

35.81 36.16 35.13 35.01 22.77

1.57 2.74 4.26 9.86 8.72

49.24 61.18 82.18 158.46 141.09

69.1 81.1 105.1 193.1 173.1

45.13 52.97 68.65 126.13 113.06

Note: Runs employ loading control; pmax 0.2 ksi; slab dimensions 120 240 6 in.; ITI 5 1023; MnTI 1 1025; MxTI 2 1022; Columns marked % provide the results of the preceding columns as a ratio of the corresponding unnotched case response. The value given is the stress observed at the last step increment; no maximum was observed.

318

A. M. Ioannides et al.

4.3 Effect of loaded area size A few cases were selected for an investigation of the effect of the size of loaded area, and additional runs were performed using loaded areas of 305 610 mm (12 24 in.) and 610 305 mm (24 12 in.). The ITI was 1.5 1025, the MnTI was the default (1 1025) and the MxTI was not specied. The response tracked in table 6 is the applied pressure, pc, corresponding to the development of the maximum stress at the notch tip, which occurs when the crack rst begins to propagate. It is observed that pc rst decreases as the notch size increases, as expected, but then increases indicating that the notch tip is now beyond the limits of the loaded area. These observations also explain the fact that, in general, pcvalues for a 610 305 mm (24 12 in.) area are higher than those obtained using the 305 610 mm (12 24 in.) area, since the horizontal notch extends into the slab away from the edge, i.e. in the y-direction.

Figure 5. The three stages of the loading-and-fracture process. (RF3/2: sum of vertical reaction forces under loaded area)

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

notch size appears to have a signicant impact on the responses obtained during the rst two stages of loading. Thus, the initial slope of the curves decreases as the notch size increases. Similarly, the CMOD corresponding to any given load level increases as the notch size increases, while the maximum load sustained before entering the second stage decreases as notch size increases. Responses are much more sensitive to increases in notch size in the vertical than in the horizontal direction. The pattern exhibited by the unnotched curve suggests that the modeling of this case may call for some additional renement. On the basis of these observations, the value of half the CMOD is tracked as the notch size increases, at two different levels for half the total applied force, between 169 kN (38 kips) and 178 kN (40 kips) and between 267 kN (60 kips) and 285 kN (64 kips), respectively. The LLD is not tracked in table 5 since this is controlled directly. At the lower level of applied force, the CMOD increases as the notch size increases, whereas at the higher level, the CMOD does not change much as notch size increases. This indicates that whereas at the low load level the slab is primarily responsible for carrying the load, at the higher load level the subgrade is carrying a bigger portion of the load.
Table 5a. Inuence of notch size (at Notch front): displacement control.

4.4 Effect of slab thickness To investigate the effect of slab thickness on the fracture process, a run was conducted using a thickness of 229 mm (9 in.) and the results are compared to those from an identical run using the standard thickness of 152 mm (6 in.). Displacement control was used in both cases, and the notch size was one slab layer by one element row. The impact of the notch is more pronounced on the thicker slab, which exhibits a lower initial slope, as well as a lower sustained load prior to the main fracture region (gure 6). At any given applied load, the thicker slab shows a higher CMOD. This observation can be explained by recalling that the FCM model is dened in terms of cohesive stress vs. crack opening near the crack tip; consequently, a larger CMOD is required to produce the same crack opening near the tip if the thickness of each slab layer increases. It is also observed in table 7 that the thicker slab requires fewer load increments to complete the fracture process, and that the load steps for it are larger than those applied to the thinner slab. This is probably due to the internal manner in which ABAQUS determines the appropriate load step at every stage. In the initial few stages of loading, the thicker slab allows the load step to increase faster, and once large steps begin to be taken, they are continued to the end of the fracture process, thereby resulting in fewer load steps. In contrast, the thinner slab dictates a slower increase in the load step magnitude, and leads to a greater total number of steps required to complete the entire process. Thus, at any particular load level, the thicker slab exhibits a larger CMOD than the thinner slab. Finally, the thicker slab requires a total load of 1620 kN (364 kips) to complete the fracture process, compared to only 1190 kN (267 kips) for the thinner slab. It is apparent that these observations are inuenced to a great extent by the use of displacement control and the selection of the three loading parameters, ITI, MnTI, and MxTI.

Unnotched in both directions (ay/W) (%) 0 (az/h) 33.3% 10 30 70 (az/h) 66.7% 10 30 70 CMOD/2 (mils) 1.1132 1.6983 1.7938 1.9409 2.8498 4.7591 5.2540 % 100 153 161 174 256 428 472 STEP 7 8 7 12 7 7 13

Note: RF3/2: sum of vertical reaction forces under loaded area; RF3/2 3840 kip; slab dimensions 120 240 6 in.; ITI 1 1022 ; MnTI 2 1023; MxTI 5 1022. The Column marked % provides the results of the preceding column as a ratio of the corresponding unnotched case response.

ABAQUS model for PCC slab cracking Table 5b. Inuence of notch size (at notch front): displacement control Unnotched in both directions (ay/W) (%) 0 (az/h) 33.3% 10 30 70 (az/h) 66.7% 10 30 70 CMOD/2 (mils) 11.433 11.107 11.926 12.479 12.566 12.607 13.072 % 100 97 104 109 110 110 114 STEP 9 13 8 8 8 8 8

319

Note: RF3/2: sum of vertical reaction forces under loaded area; RF3/2 60 kip; slab dimensions 120 240 6 in.; ITI 1 1022; MnTI 2 1023; MxTI 5 1022. The Column marked % provides the results of the preceding column as a ratio of the corresponding unnotched case response.

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

Figure 6. Slab thickness effect on loading response. (RF3/2: sum of vertical reaction forces under loaded area)

5. Relevance to concrete pavement design The bilinear curve used in this study is based on experimental results on PCC of different qualities obtained during stable tensile tests, and seems to be a reasonable approximation for this type of material (Gustafsson 1985). The experimental data can also be represented by a best-t bilinear curve in dimensionless form, as s/ft vs. wft/Gf, but this is by no means an arbitrary choice. Commenting on this issue, Gustafsson noted that Gf is often a suitable and convenient parameter (for this purpose because) the value of Gf can often be determined rather easily by means of ordinary testing equipment. Moreover, it is difcult to obtain and dene any accurate value of w that corresponds to s 0, wc as, at least in the case of concrete, it seems that ds/dw is close to zero when s approaches zero. Instead, Gustafsson suggested setting wc 3.6 Gf/ft, following an earlier recommendation by Petersson (1981). The practical implication of this is that the s-w curve is exclusively determined by the values of Gf and ft, its bilinear shape, as well as the location of its knee being established by curve tting. Gustafsson presented his results in dimensionless form as well, as the ratio between the predicted ultimate bending moment capacity, ff, and the tensile strength, ft, vs. the ratio of the beam depth, d, to the characteristic length, lch, of the material, i.e. ff/ft vs. d/lch. It is recalled

Table 6. Inuence of the size of the loaded area on pc (ksi). Notch Unnotched (%) (az/h) 33.3 (ay/W) 10 (az/h) 33.3 (ay/W) 20 (az/h) 33.3 (ay/W) 30 (az/h) 33.3 (ay/W) 50 (az/h) 33.3 (ay/W) 70 (az/h) 33.3 (ay/W) 100 (az/h) 66.7 (ay/W) 10 (az/h) 66.7 (ay/W) 20 (az/h) 66.7 (ay/W) 30 (az/h) 66.7 (ay/W) 50 (az/h) 66.7 (ay/W) 70 (az/h) 66.7 (ay/W) 100 View B F B F B F B F B F B F B F B F B F B F B F B F B F 12 12 in. 0.1683 0.1683 0.1122 0.1122 0.1122 0.0748 0.1683 0.0748 0.2525 0.1122 0.3788 0.1122 0.1122 0.0499 0.0748 0.0748 0.1683 0.0748 0.0748 0.1683 0.0499 0.2525 0.0499 0.0499 12 24 in. 0.1683 0.1683 0.1122 0.1122 0.1683 0.1122 0.1683 0.1122 0.3788 0.1122 Did Not Peak 0.1122 0.1122 0.0748 0.2525 0.0748 0.0748 0.1122 0.2525 0.2525 0.2525 0.3788 0.0748 0.2525 % 100.00 100.00 100.00 100.00 150.00 150.00 100.00 150.00 150.02 100.00 100.00 No crack tip 100.00 149.90 337.57 100.00 44.44 150.00 337.52 150.03 506.01 150.02 149.9 No crack tip 506.01 24 12 in. 0.0982 0.0982 0.0748 0.0748 0.0748 0.0748 0.0748 0.0499 0.1683 0.0499 0.2244 0.0499 0.0499 0.0332 0.0499 0.0499 0.0332 0.0499 0.0332 0.1122 0.0332 0.1683 0.0332 0.0332 % 58.35 58.35 66.67 66.67 66.67 100.00 44.44 66.71 66.65 44.47 59.24 44.47 44.47 66.53 66.71 66.71 19.73 66.71 44.39 66.67 66.53 66.65 66.53 66.53

Note: pmax 0.5 ksi; B: at notch back; F: at notch front.

320

A. M. Ioannides et al. Table 7. Inuence of slab thickness

Slab Thickness (in.) 6 9 6 6 5.00 5.00 1.00 1.00

ITI 10 1022 1022 1022


22

MnTI 1.00 1.00 2.00 2.00 10 1025 1023 1023


25

MxTI N/S N/S 5.00 1022 5.00 1022

NINC 18 12 30 25

CPU (s) 3662 2425 5001 3816

CMOD/2 (mils) 55.5 99.1 55.1 58.1

RF3/2 (kips) 2266.7 2364.2 2266.7 2266.7

Note: Displacement control; maximum displacement 1 in. Slab dimensions 120 240 in. Unnotched. All other cases are notched: (az/h) 33.3%; (ay/W) 10% RF3/2: sum of vertical reaction forces under loaded area; N/S: not specied.

that d/lch is the brittleness number, B. All contributors to these ratios are input parameters to the numerical simulation procedure, with the exception of ff. The latter is obtained from the value of the ultimate load, Pu, which is the load at the peak of the load, P, vs. CMOD or of the P vs. LLD, curves. Once Pu is determined, ff is calculated using:

6. Conclusion This paper focuses on the application of the commercial software nite element package ABAQUS to tracking crack propagation in PCC concrete slabs. This is achieved through the implementation of Hilleborgs FCM using a series of JOINTC elements in a slab-on-grade of typical PCC pavement dimensions. Analyses examine the effects of loading parameters, notch size, size of loaded area and slab thickness. Results are corroborated to the extent possible by Westergaards analysis, thereby ensuring the smooth and condent transition from the current state-ofthe-art and engineering intuition into a scantily explored area of research. This paper demonstrates that through a painstaking, step-by-step approach, a nonlinear nite element approach for concrete pavement slabs is feasible using a commercially available code. It is hoped that the use of fracture mechanics concepts in such a procedure will eventually lead to the denition of a more reliable and realistic mechanistic failure criterion, that will address the weaknesses of transfer functions commonly employed in current pavement design guides. References
Aglan, H.A. and Figueroa, J.L., Damage-evolution approach to fatigue cracking in pavements. J. Eng. Mech., ASCE, 1993, 119(6), Paper No. 1024, June, 12431259 New York, NY. Bache, H.H., Principles of similitude in design of reinforced brittle matrix composites, Presented at the International Workshop on High Performance Fiber Reinforced Cement Composites, Mainz, Germany, June 1991. Bache, H.H. and Vinding, I., Fracture mechanics in design of concrete pavements. Proceedings, Second International Workshop on the Design and Evaluation of Concrete Pavements, 45 October, pp. 139164, 1990 (CROW/PIARC: Siguenza, Spain). Bazant, Z.P., Concrete fracture models: testing and practice. Engineering Fracture Mechanics, Vol. 69, pp. 165205, 2002 (Elsevier Science Ltd. Amsterdam, The Netherlands). Bazant, Z.P. and Planas, J., Fracture and size effect in concrete and other quasibrittle materials, p. 616, 1998 (CRC Press: Boca Raton, FL). Broek, D., Elementary Engineering Fracture Mechanics, Fourth Revised Edition, p. 516, 1986 (Martinus Nijhoff Publishers: Dordrecht, The Netherlands). GTSTRUDL, Finite element computer software system for structural analysis and design. Users Manual, 1993 (Georgia Tech Research Corporation, Georgia Institute of Technology: Atlanta, GA). Gustafsson, P.J., Fracture mechanics studies of non-yielding materials like concrete: modeling of tensile fracture and applied strength analysis, Report TVBM-1007, Division of Building Materials, University of Lund, Sweden, 1985. Hibbitt, Karlsson and Sorenson, Inc., ABAQUS/standard users manual. General Purpose Finite Element Analysis Program, Version 5.6, Pawtucket, RI, 1994.

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

ff

3 Pu L 2 td 2

This suggests that ff can be thought of as the linear elastic bending stress arising at the bottom ber of the beam under the action of a point load, Pu. In addition, the ratio ff/ft represents the reserve strength available beyond rst cracking, i.e. beyond the onset of ft at the bottom ber of the beam. A value of ff/ft 1 denotes the linear elastic condition, for which LEFM is applicable. The correspondence of this quantity to the conventional ratio of (bending stress/modulus of rupture) used in applications of Miners hypothesis is immediately apparent, especially in light of a reinterpretation of the latter by Bache and Vinding (1990), who criticize the conventional application of Miners hypothesis, involving testing of small, simply supported beam specimens and applying the results to full-scale pavement sections. They stress that such an approach is pertinent only to crack initiation, and point out that it tells us nothing about the consequences of local fracture for example whether this leads to total failure or only to the formation of harmless, small cracks. To address this fundamental limitation in conventional design practices, Bache (1991) advocates the use of similitude principles. Such application rests on the postulate that geometrically similar objects exhibit similar behavior if the same ratio exists between the signicant forces or energies. It should be noted that Bache is concerned primarily with the determination of load capacity, or strength, i.e., with the determination of the maximum sustainable load before failure. At rst sight, this might be construed to refer only to one load repetition and to be inapplicable to fatigue loading considerations. Bache and Vinding (1990), however, reinterpret the conventional application of fatigue laws as involving the substitution of the static exural strength, MR, by a (lower) fatigue strength, Mf, which is a function of the number of repetitions, N. In conventional design, the allowable stress in the slab is set equal to Mf.

ABAQUS model for PCC slab cracking Hillerborg, A., Modeer, M. and Petersson, P.E., Analysis of crack formation and crack growth in concrete by means of fracture mechanics and nite elements. Cem. Concr. Res., 1976, 6(6), 773 792. Ioannides, A.M., Fracture mechanics in pavement engineering. Transportation Research Record 1568, pp. 1016, 1997a (Transportation Research Board, National Research Council: Washington, DC). Ioannides, A.M., Pavement fatigue concepts: a historical review. Proceedings of the Sixth International Purdue Conference on Concrete Pavement Design and Materials for High Performance, Indianapolis, Indiana, Vol. 3, pp. 147159, 1997b. Ioannides, A.M. and Donnelly, J., Three-dimensional analysis of slab on stress-.dependent foundation. Transportation Research Record 1196, pp. 72 84, 1988 (Transportation Research Board, National Research Council: Washington, DC). Ioannides, A.M. and Peng, J., Finite element simulation of crack growth in concrete slabs: implications for pavement design. Proceedings, Fifth International Workshop on the Design and Rehabilitation of Concrete Pavements, Istanbul, Turkey (April 1 2) pp. 56 68, 2004. Ioannides, A.M. and Sengupta, S., Crack propagation in Portland cement concrete beams: implications for pavement design. Transportation Research Record 1853, pp. 110117, 2003 (Transportation Research Board, National Research Council: Washington, DC). Ioannides, A.M., Peng, J. and Swindler, J.R. Jr, Analysis of concrete fracture using ABAQUS. Proceedings, 8th International Conference on Concrete Pavements, Colorado Springs, CO, August, Vol. 3, pp. 11381154, 2005. Jenq, Y.-S. and Perng, J.-D., Analysis of crack propagation in asphalt concrete using cohesive crack model. Transportation Research Record 1317, pp. 9099, 1991 (Transportation Research Board, National Research Council: Washington, DC). Kharlab, V.D., Singular strength criteria. Issledovaniya po Mekhanike Stroilnykh Konstruktsii i Materialov, pp. 58 63, 1989 (LISI: Leningrad), (in Russian). Kharlab, V.D., On singular strength criteria. Issledovaniya po Mekhanike Stroilnykh Konstruktsii i Materialov, pp. 82 85, 1990 (LISI: Leningrad), (in Russian).

321

Downloaded by [Amir Kabir University] at 09:27 11 March 2012

Kharlab, V.D. and Minin, V.A., Strength criteria accounting for inuence of stress state gradient. Issledovaniya po Mekhanike Stroilnykh Konstruktsii i Materialov, pp. 5357, 1989 (LISI: Leningrad) (in Russian). Khazanovich, L. and Ioannides, A.M., Glasnost and concrete pavement engineering: insights concerning slabs-on-grade from the former Soviet Union. Proceedings, Fifth International Conference on Concrete Pavement Design and Rehabilitation, 1, pp. 318, 1993 (Purdue University: West Lafayette, IN), (Apr. 20-22). Liu, P., Time-dependent fracture of concrete, PhD dissertation. Ohio State University, Columbus, Ohio, 1994. Lytton, R.L. and Shanmugham, U., Analysis and design of pavements to resist thermal cracking using fracture mechanics. Proceedings, 5th International Conference on Structural Design of Asphalt Pavements, Vol. 1, pp. 818 830, 1982 (The University of Michigan: Delft, The Netherlands). Majidzadeh, K., Kauffmann, E.M. and Ramsamooj, D.V., Application of fracture mechanics in the analysis of pavement fatigue. Proceedings, AAPT, Vol. 40, pp. 227 244, 1971, Oklahoma City, OK, February, disc.: pp. 245-246. Miner, M.A., Cumulative damage in fatigue. Transactions, American Society of Mechanical Engineers, New York, NY, September, Vol. 67, pp. A-159 A-164, 1945. Paris, P.C. and Erdogan, F., A critical analyis of crack propagation laws. J. Basic Eng., Trans. ASME, Ser. D, 1963, 85(4), 528553. Petersson, P.E., Crack growth and development of fracture zones in plain concrete and similar materials. Report TVBM-1006, 1981 (Division of Building Materials, University of Lund: Sweden). Shah, S.P., Swartz, S.E. and Ouyang, C., Fracture Mechanics of Concrete, p. 552, 1995 (John Wiley and Sons, Inc.: New York, NY). Siemes, A.J.M., Miners rule with respect to plain concrete variable amplitude tests. Special Publication SP-75, pp. 343372, 1982 (American Concrete Institute: Ann Arbor, MI). Westergaard, H.M., Stresses in concrete pavements computed by theoretical analysis. Public Roads, 1926, 7(2), 25 35. Westergaard, H.M., New formulas for stresses in concrete pavements of airelds. ASCE Trans., 1948, 113, 425 439.

You might also like