You are on page 1of 12
‘Struetura) Concrete 2001, 2, No. 4 December, 201-242 Non-linear finite element analysis of reinforced concrete: at the crossroads? F. 4, Vecchio University of Toronto, Canada This a concrete structures. Their potential usefulness as a practe examining 1 ‘when applying non-linear analysis methods are identified and discussed. Fina ‘takes a critical look atthe relevance and value of non-tinear finite element procedures for analysis and design of reinforced ‘design office tool is illustrated through a sample application. Then, in results of recent prediction competitions, the accuracy of such analysis procedures Is examined. Reasons for caution of a test programme involving shear ertical beams are presented In support of the contention that the behaviour of reinforced concrete Is still not always well understood: the test results provide a ch lenge for validating current procedures Notation Agm maximum aggregate size A, cross section area of rebar rebar diameter Fc compressive strength of concrete cylinder at 28 days f', tensile strength of concrete f, yield strength of reinforcement My. section moment capacity of beam (nand calculated) ultimate load capacity of beam (finite element analysis) Yur sectional shear capacity of beam (Simple Method of CSA 123.3) Yoo. sectional shear capacity of beam (General Method of CSA 423.3) migspan deflection at ultimate load (finite element analysis) concrete strain at geak compressive stress shear strain feinforcement ratio + shear stress: Introduction Non-linear finite element analysis (NLFEA) of reinforced concrete has taken tremendous strides forward since intial applications about 40 years ago. Much research activity has occurred in the realm of constitutive modelling of reinforced concrete behaviour ‘and in the development of sophisticated analysis algorithms. These advancements ate well documented in various state-of 41464-4177 © 2001 Thomas Taford Ltd ane > the-art reports, and still remain the subject of many specialty symposia and workshops. ‘Occurring at the same time, and no less significant, has been the prodigious advancement in computing technology and hard ‘ware. Data compiled by Bentz, shown in Figure 1, provide a clear measure of the exponential growth in computing power in recent yyears.* Shown isthe time required to conduct a non-linear shear ‘analysis of a prestressed T-beam using a layered beam element algorithm. Its seen from the grah that, in 25 years, computing ‘speed has increased by five orders of magnitude. Analyses that required several days of CPU time on supercomputers two dec- fades ago run in minutes on personal desktop computers today. The advent of powerful and relatively inexpensive computers has areatly expanded the size and complexity of problems that can be analysed, and has greatly reduced the computer time required for their solution ‘The state-of-the-art in NLFEA has thus progressed to the point where such procedures are close to being practical, every-day tools for design office engineers. No longer solely within the pur View of researchers, they ate finding use in various applications: many relating to our aging infrastructure, NLFEA procedures can be used to provide reliable assessments of the strength and integrity of damages or deteriorates structures, or of structures built to superseded codes, standards or practices deemed to be deficient today. They can serve as valuable tools in assessing the expected behaviour from retrofitted structures, or in investigating ‘and rationally selecting amongst various repair alternatives. In situations that have not turnes out wall, NLFEA procedures are 201 veeehio 100 on SPECiNeS oot 001 0.0001 "970 1975 780 7065 18s Penn amin 20m 2eaye 2h days 790 7985 2009 2008 ‘Year of CPU invedcton Fig. 4 Increase in computing power in ecent yeors* finding applications to forensic analyses and tigations. that follow. In the near future, they will likely form the main engine in computer-based automated design software, although in a form likely invisible to the user A sample application [As an example of the usefulness of NLFEA methods, the studies Undertaken subsequent to the collapse of the Sleipner A offshore platform will be briefly reviewed. The gravity base structure of the platform consisted of a cluster of 24 cell, four of which extended upwards to form shafts (see Figure 2). While the exterior of the walls of the cells were circular, the interior walls separating the cells were straight. At the intersection of these interior walls, a ‘small triangular vold called a tricell was formed (32 in total. These tricells had openings at the top, and hence had to resist ‘substantial hydrostatic pressure when submerged. ‘On 23 August 1994, the gravity base structure was slowly being submerged as part of the deck-mating operation. The intent was to lower the structure until its base was 104 m below the ocean surface. However, when a depth of 99m was reached, a loud rumbling noise emanated from one of the drill shafts and water could be heard pouring in. Within minutes, the structure began to ‘sink in an unconttolleg manner. Moments after disappearing below the surface, a series of implosions were felt as the buoy. ancy cells collapsed. Evidence showed that the loss of the struc: ture was attributable to the shear failure of one of the tricell walls, To develop a better understanding ofthe factors influencing the failuce of the tricell wall, a series of nor-linear finite element ‘analyses were undertaken. The finite element model used, repre- sented in Figure 3(@), is fuly described by Collins et al.” Initial analyses indicated that the as-built structure would fail in shear when the apalied water pressure on the inner faces ofthe tricells, 202 reached 625 kN/m?. This corresponded to a head of seawater of 62m, a value that agreed well with the estimated 65 m head active at the cell wall location where collapse occurred, ‘The designers of the structure wete interested in learning how the strength of the tricell would have changed if the stirrups, which were terminated just below the failure location, had been Continued further up the wall They also wanted to know how the length of the T-headed bars used in the throat of the tricell walls influenced the behaviour. Additional finite element analyses were conducted, and the results are summarised in Figure 3(b). The results of these studies indicated that the tricells could have resisted an additional 20 m of water head if either the stirrups had been continued up the wall or if the T-headed bars had been ‘about 500 mm longer.® Changes were made to the design of replacement structure accordingly The question of accuracy Despite the increasing sophistication of NLFEA tools, users must be ever minctul of the question of how accurately and reliably they represent the behaviour of reinforced concrete structures. In this regard, its useful to examine the results of three ‘prediction competitions’ ‘At the 1981 IABSE Symposium in Delft, “blind competition was organised involving four panels tested in a comprehensive research programme then underway at the University of Toronto. The test panels were orthogonally reinforced, and subjected to Uniform, proportional and menotonically increasing stress condi- tions: seemingly a very simple problem to model and analyse. The results of these panel tests were not disclosed prior to analysts submitting their predictions of strength and load-deformation response, Approximately 30 entries were received, many from the leading researchers in the field at the time. Snown in Figure 4 Is the range of responses for Panel C, one of the better pre: Structural Conerete, 2001, 2, No, & Sea lve uring ck mating fore e25m Moan se lave during operation at st 26 cals 52 cats ein | smn 2.25 mm bars ceneead at 170mm eT 70 70 16mm ‘Stirups eae mies te 2 Retina i ‘neon bean rena Fig. 2 Details ofthe Sleiner A concrete base structure dicted of the four panels. The analysis results submitted showed 2 wide variation in predictions of the panel's shear strength, and ‘an even wider divergence in computed load-deformation responses.* Clearly the collective ability to mode! non-linear behaviour of reinforced concrete, particularly in shear-critical conditions, was not well advanced. Structural Concrete, 2001, 2, No. NNomtinear fite element analysts of reinforced concrete More recently, in 1995, the Nuclear Power Engineering Cor poration of Japan (NUPEC) staged a preciction competition invol- ving 2 large-scale 3.0 shear wall subjected to dynamic eyclle loading * The flanged shear wall exhibited highly non-linear beha- Viour before sustaining a sliding shear failure along the base of the web. One facet of the competition called for estimates of the Ltimate strength, and corresponding displacement, ofthe wall as, etermined from static push-over analyses. Again, over 30 sets of predictions were received; the results are summarised in Figure 5. The predictions of strength, as a group, showed better corcela- tion than was seen with the Toronto panels; however, the defor: mation estimates still showed large scatter. Nevertheless, it ‘could be concluded that the ability of NLFEA to accurately capture the behaviour of reinforced concrete had measurably advanced. It should be noted, however, that this was not a completely bling Competition since some of the test results had been disclosed to analysts prior to competition, More recently, ASCE-ACI Committee 447 organised an informal Competition centred on results from a series of large-scale col umns tested at the University of California at San Diego. Many of the analyses undertaken are documented in papers contained within an ACI Special Pubiication.® From these, it can be noted that: (I) a number of quite different analysis approaches were taken; (I) predictions of strength and pre-peak response gener ally correlated well with the experimental results; and (il) predic: tions of post peak response were generally not es accurate and Still require further attention. (Bear in mind, once again, that this was not a blind competition. Analysts had the opportunity to calibrate parameters, optimise material models, and refine an. Iyses. One should also bear in mind that experimental results themselves are subject to scatter and error. Repeating a test, particulary f conducted at different laboratories, may yield difer: ing results) Nevertheless, itis an inevitable conclusion that our ability to accurately model the behaviour of reinforced concrete structures has seen significant improvement over the past 20 years. It has ‘approached a stage of development where we may be inclined to proceed with a certain degree of confidence. Reasons for caution Despite these significant advancements in our ablity to accu: Fately mode! the response of reinforced concrete, the users of NLFEA procedures need to be mindful of several issues and potential dangers: Diversity of theoretical approaches [A number of rather diverse approaches exist for NLFEA modelling of reinforced concrete structures. Among those available are ‘models built on non-linear elasticity, plasticity, fracture ‘mechanics, damage continuum mechanics, endechronic theory ‘or other hybrid formulations. Cracking can be modelled discretely, fr using smeared crack approaches; the latter can range trom fully rotating crack models, to fixed crack models, to multiple non: orthogonal crack models, to hybrid crack models. Some ‘approaches place heavy emphasis on classical mechanics formu: lations, others draw more hesvily on empirical data and phenom: tenological models. it can generally be said of any approach that it will be more suited to certain structure/loading situations and less s0 to others. No one approach performs well over the entire 203

You might also like