See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/236171502 Stiffness-oriented numerical model for non-linear reinforced concrete beam systems Thesis · February 2011 CITATIONS READS 0 430 1 author: Agnieszka Jędrzejewska Silesian University of Technology 87 PUBLICATIONS 525 CITATIONS SEE PROFILE All content following this page was uploaded by Agnieszka Jędrzejewska on 27 May 2014. The user has requested enhancement of the downloaded file. SILESIAN UNIVERSITY OF TECHNOLOGY Faculty of Civil Engineering master thesis Stiffness-oriented numerical model for non-linear reinforced concrete beam systems Author: SEng Agnieszka KNOPPIK–WRÓBEL KBI-A CIS, year 2010/2011 February 22, 2011 Supervisor: PhD SEng Grzegorz WANDZIK Contents 1 Introduction 1.1 Objective of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Range of problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Scope of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5 5 6 2 Models for analysis of reinforced concrete beam elements 2.1 Material non-linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1 Models based on theory of elasticity . . . . . . . . . . . . . . . . . . 2.1.2 Models based on theory of plasticity . . . . . . . . . . . . . . . . . 2.1.3 Rheological models . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Material models in standards . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 Concrete models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Steel models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Geometrical non-linearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 9 10 10 11 15 15 17 18 3 Refinement of linear-elastic analysis results 3.1 Analysis of reinforced concrete members beyond elastic phase . . . . . . . . 3.1.1 Linear-elastic vs. non-linear analysis . . . . . . . . . . . . . . . . . 3.1.2 Plastic properties of reinforced concrete . . . . . . . . . . . . . . . . 3.2 Linear-elastic analysis with moment redistribution . . . . . . . . . . . . . . 3.3 Plastic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Limit equilibrium method . . . . . . . . . . . . . . . . . . . . . . . 3.4 Non-linear analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 19 19 20 23 25 27 28 4 Stiffness degradation in R/C flexural members 4.1 Stiffness of reinforced concrete flexural members . . . . . . . . . . . . . . . 4.2 Bending stiffness of cross-section . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Material behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.2 Cross-section behaviour . . . . . . . . . . . . . . . . . . . . . . . . 4.2.3 Creep and shrinkage effects . . . . . . . . . . . . . . . . . . . . . . 4.2.4 Tension stiffening effect . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.5 Mean moment–curvature relationship . . . . . . . . . . . . . . . . . 4.3 Bending stiffness of member . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Influence of cracks . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Influence of reinforcement . . . . . . . . . . . . . . . . . . . . . . . 31 31 32 32 36 39 39 40 40 40 42 3 4 CONTENTS 5 Numerical model 5.1 Static scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1.2 Discretisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Material model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Cross-section model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.1 Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.2 Bending moment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3.3 Moment–curvature relationship . . . . . . . . . . . . . . . . . . . . 5.4 Stiffness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4.1 Stiffness of cross-section . . . . . . . . . . . . . . . . . . . . . . . . 5.4.2 Stiffness of segment . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Static analysis with FEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.1 Static calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2 Bending moment distribution . . . . . . . . . . . . . . . . . . . . . 43 44 44 45 45 46 46 47 50 51 51 53 54 55 57 6 Summary 6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Abstract (in Polish) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 65 67 67 Chapter 1 Introduction 1.1 Objective of thesis The objective of this thesis is to derive a numerical model for designing of flexural reinforced concrete beams taking into consideration a non-linear behaviour of reinforced concrete and redistribution of internal forces as a result of stiffness degradation of the elements due to crack formation at flexure, thus providing a unified algorithm for static calculations and dimensioning. 1.2 Range of problems The use of the linear-elastic analysis in determination of internal forces distributions in reinforced concrete structures has a virtue of simplicity and allows of results from a series of analyses to be combined using the principle of superposition. Popularity and long-lasting tradition of application of the approach is undeniable. The linear-elastic approach eliminates problems in numerical analysis, such as incremental calculations, in which the stiffness of the loaded element is locally modified along with the changing distribution of internal forces, iterative calculations connected with mutual correlation between amount and distribution of reinforcement and distribution of bending moments and, as a result, problems with convergence of computational process. The assumption of the linear-elastic behaviour is reasonable at low levels of loading but becomes invalid at higher loads due to cracking and development of plastic deformations. Once an element cracks, the behaviour becomes non-linear, so application of the elastic theory provides unreliable values of internal forces and has a series of disadvantages. In the limit states analysis complex material models are used while the static analysis is based on a very primitive model. The assumptions are inconsistent [20]: the load-bearing capacity of the section MRd is determined according to the ultimate limit state of flexure (section failure criterion) but it is compared with the values of internal forces (MSd ) determined for a completely different state of work, even before the level of cracking moment is reached. The results of the static analysis do not consider the influence of the amount and distribution of reinforcement or the capacity use of section. Concrete cracking is considered to be a disadvantage requiring reinforcement and diminishing structural tightness and durability. The positive effects of cracking – reduction and redistribution of cross-sectional forces – are either neglected or poorly simulated [30]. In static calculations of elements subjected to non-mechanical loads (thermal loads, fire) in the design process “corrected”, reduced values of temperatures, safety factors or moduli of elasticity are 5 6 CHAPTER 1. INTRODUCTION introduced. This frequently results in uneconomic and unsafe structures, and should be improved by a broader application of stiffness-oriented design procedures. The notion of the moment redistribution in cracked elements is useful for practical design as it allows of some flexibility in the arrangement of reinforcement. Bending moments may be transferred into the less congested areas or standard reinforcement layouts may be applied where small differences occur in the bending moment distributions for a series of beams, thus avoiding the need to detail each beam separately. In addition, economy can be achieved when moment redistribution is applied to different load combinations, resulting in a smaller bending moment envelope which still satisfies equilibrium. Moment redistribution in beams has traditionally been considered as an ultimate limit state (ULS) phenomenon, but experiments [31] prove that a significant portion of this redistribution will almost always occur at the serviceability limit state (SLS) because of the mismatch between the flexural stiffness assumed when calculating moments for the ULS and those actually occurring at the SLS due to variations in the reinforcement layout along the member and the influence of cracking. That is why the ULS and SLS assumptions and procedures should be conjugated and permissible level of redistribution should be defined. The analysis of the plastic properties of reinforced concrete structures has been performed throughout the years in stages [21]. In the first stage changes of the internal forces under increasing loads were determined experimentally, especially under loads close to the load-bearing capacity of the structure. In the second stage the limit equilibrium method was adopted, primarily to the steel structures. The notion of the fact that the plastic hinge in a reinforced concrete structure differs from the plastic hinge in a steel one was a motive for the next, third stage, covering development of general methods for calculation of structures according to the theory of plasticity. The fourth stage, which has not yet been finished, is characterised by development of simple design methods, useful in everyday practice, and finally the last stage, still under development, is connected with incorporation of numerical calculations thanks to the possibility of the use of computers. The linear-elastic FE analysis is already world-wide accepted but for the reasons mentioned before there arises an urgent need for implementation of the non-linear analysis into design procedures. This results from the use of more and more complex material models, close-to-reality modelling of multi-phenomena processes or advanced analysis of whole complex models instead of their subdivisions. Nevertheless, the nature of the nonlinear FE analysis makes it much more difficult to provide the same level of automatisation as for the linear one. The factors which prevent wide acceptability of the non-linear FE analysis procedures are connected with [24]: no full identification of constitutive properties of concrete, no generally accepter material law available to model concrete behaviour, high costs and required experience. 1.3 Scope of thesis Given the range of problems stated in Sec. 1.2, the thesis challenges a range of issues aimed at deriving of a close-to-reality numerical model of reinforced concrete structures on the example of continuous flexural beams. The theoretical background is a concise overview of the Polish and foreign literature connected with the topic of non-linear concrete mechanics as well as international standards, primarily the European Standard EN 1992 – Eurocode 2 with detailed useful information from the CEB-FIP Model Code 1990 and final draft of Model Code 2010, the Polish 1.3. SCOPE OF THESIS 7 Standard PN-B-03264 (the author’s previous national standard), the British Standard BS 8110 and the American Standard ACI 318 1 . The practical part covers a creative analysis of the solutions proposed in the literature and description of the numerical model derived in cooperation with the supervisor of this thesis, PhD Grzegorz Wandzik, proposed on the basis of the referred theoretical background and personal experience in the topic. Chapter 2 presents an overview of the reasons of non-linearity in the analysis of concrete structures. In this chapter the two general aspects of non-linear mechanics are described: non-linear material models and behaviour of structures considering deformations and their influence on the distribution of internal forces. Chapter 3 presents the methods of refinement of the linear-elastic analysis result proposed in the literature, taking into consideration plastic properties of reinforced concrete and the consequent mechanism of stiffness degradation and redistribution of internal forces. Chapter 4 introduces the theory of stiffness degradation in a flexural reinforced concrete member. The idea of a cross-section model is presented and stiffness a segment is defined. The influences of rheological phenomena occurring in steel and concrete are considered. Chapter 5 is a description of a numerical model for calculations of continuous reinforced concrete beams derived in co-operation with the supervisor of this thesis. This chapter provides a description of the algorithm along with a concise overview and evaluation of computer methods which can be applied in the approach aimed in this thesis. Chapter 6 summarises the work done in this thesis and evaluates the applicability of the presented model together with the future prospects. 1 Model Code is a generalised theoretical basis for codes while the Eurocode 2 is the currently valid standard. Each time a reference to the standard is made, the reader should understand the EC2, unless stated otherwise. Chapter 2 Models for analysis of reinforced concrete beam elements Two main reasons of non-linearity in engineering structures can be distinguished: 1. material (physical) if the material has a non-linear σ– characteristic, 2. geometrical (kinematic), when initial and final configuration of the system is considered or as a result of initial deformations. Two characteristics are important for good understanding of these phenomena [21]: a material behaviour and a cross-section behaviour. The first one defines the relationship between stress and strain (σ–) for a given material while the other represents the relationship between the internal force in the section and its deformation (moment and curvature of the deformation line M –κ). The structure can be analysed in a number of ways, depending on the initial assumption of a designer. One must remember that no model can perfectly represent the real behaviour of the structure and that the assumptions of the model are made to provide optimal simplicity and accuracy of results for satisfaction of the design requirements. The choice of the method depends on the assumed behaviour of the material and the possible consideration of the effects of deformation on the action effects (second order effects). This thesis presents the issues connected with implementation of some variants of the non-linear analysis to reinforced concrete beam systems (considering 1-dimensional stress state). 2.1 Material non-linearity There exists a great number of models representing the stress–strain behaviour of engineering materials. These models can be divided into groups according to the exhibited properties they have in common. This might be the phenomena they represent, level and time of loading, properties of occurring strain or whether the behaviour of the material at compression and tension is the same or different. This section presents basic models for structural analysis and dimensioning of reinforced concrete elements. Concrete exhibits a complex structural response with various non-linearities: a nonlinear stress–strain behaviour, tensile cracking and compression crushing material failures, and creep strains. This non-linearities together with non-linearities introduced by reinforcing steel should be taken into account [24]. 9 10 CHAPTER 2. MODELS FOR ANALYSIS OF REINFORCED CONCRETE BEAM ELEMENTS 2.1.1 Models based on theory of elasticity This group constitutes of a number of elastic models, from simple linear elasticity, through various types of non-linear elasticities, to non-linear incremental elasticity. Such models are characterised with the strain being dependent only on the actual state of stress, not on the history of loading. Strains are totally reversible and the material behaves in the same manner at loading, unloading and reloading. Assumption of the linear elasticity signifies that the relationships between the components of stress and strain are linear (Fig. 2.1a1 , Fig. 2.1b). It is valid for infinitesimal strains or small deformations, and for stress states that do not produce yielding. (a) linear-elastic model (b) perfectly-rigid model Figure 2.1: Models based on theory of elasticity The linear elasticity is the simplest example of the more general non-linear theory of elasticity, convenient for the materials in elements which undergo large deformations, and includes such material models as the Cauchy model, hyperelastic or Green model and hypoelastic model. Nevertheless, these models are good representation for materials such as polymers, and are rather useless with respect to reinforced concrete. 2.1.2 Models based on theory of plasticity (a) elastic–perfectly-plastic model (b) rigid–perfectly-plastic model Figure 2.2: Models based on theory of plasticity 1 cu and fc signify the ultimate compressive strain and strength while tu and ft the ultimate tensile strain and strength, respectively. 11 2.1. MATERIAL NON-LINEARITY (a) plastic phase with hardening (b) plastic phase with softening Figure 2.3: Elasto-plastic models with hardening and softening Operation of that group of models can be divided into initial elastic phase (usually linear or rigid) and non-elastic phase, where relationship between stress and strain can be presented in an incremental manner (plastic flow). Here the strain depends not only on the actual stress state but also on the loading history. Though, such models exhibit both reversible and irreversible strains and unloading goes on the path other than initial loading. The idealised stress–strain diagram in Fig. 2.2a2 is referred to as the elastic-perfectly plastic behaviour. Plastic deformations are often significantly larger than in the elastic phase so the model can be further simplified to the rigid-perfectly plastic (Fig. 2.2b). Plastic phase can be treated in more sophisticated way as plastic with hardening (Fig. 2.3a), if the progressive increase of yield stress after yielding needs to be taken into account (such as in steel), or softening (Fig. 2.3b) in which stress decreases with increasing strain (i.e. in materials experiencing brittle damage due to cracks such as concrete). Linear hardening and linear softening in plastic phase is a convenient approximation. However, a more realistic description of hardening in metals can be achieved with non-linear hardening. On the other hand, softening in quasi-brittle materials such as concrete can be modelled in a bilinear or exponential manner [6]. 2.1.3 Rheological models In some materials like concrete, steel, soil or polymers a phenomenon of flow can be observed even under very small stresses. The phenomenon is observed to demonstrate the features of viscous flow and the material models can be identified as rheological or rate-dependent. For these materials the elasto–plastic, let alone linear-elastic models are highly inaccurate, so the viscous effects must be incorporated. The viscous effects can be considered in the material model in different manners, by assuming that they [8]: • are connected only with elastic strains (viscoelastic and viscoelasto-plastic models), • are connected only with plastic strains (elasto-viscoplastic models), • occur in the whole range of material work (viscoelasto-viscoplastic models). 2 In the diagram the symbols el and pl signify the elastic and plastic component of strain. Their meaning is only qualitative. 12 CHAPTER 2. MODELS FOR ANALYSIS OF REINFORCED CONCRETE BEAM ELEMENTS (a) Hooke’s element (b) Saint-Venant’s element (c) Newton’s element Figure 2.4: Basic elements for decription of rate-dependent materials The physical characteristics of the rheological materials can be described as a sequence of basic elements: 1. Perfectly elastic – represented by the Hooke’s spring element (Fig. 2.4a) with stiffness E. The response of the spring to stress σ is instantaneous and the recovery after release of the stress is instantaneous and complete. The stress σ applied is proportional to the deformation , and the proportionality constant is the modulus of elasticity E: σ = E · . 2. Perfectly plastic – represented by the Saint-Venant’s sliding frictional element (Fig. 2.4b) in which the friction force acts against the exerted tensile force P . When P reaches its limit value P = Plim , the element undergoes permanent and irreversible elongation. The behaviour of the model can be expressed as: σ = σy where σy is the yield limit. 3. Perfectly viscous – represented by the Newton’s dash-pot element (Fig. 2.4c) characterised by a viscosity η. There is no instantaneous response and no recovery takes place. Stress rate is proportional to the rate of deformation, and the proportionality constant is the viscosity η: d . σ=η· dt The viscoelastic or viscoplastic behaviour is comprised of elastic and/or plastic and viscous components modelled as a linear combination of springs, dash-pots and sliding frictional elements. Viscotic behaviour in concrete is connected with creep (increase of deformation under constant load) while in steel – with relaxation (decrease of stress under constant deformation). The following rheological models can be distinguished [8]. The Maxwell model is a combination of perfectly elastic and perfectly viscous elements combined in series (Fig. 2.5a). For a description of creep in concrete the model assumes that the response under constant load is the sum of elastic and viscous response (Fig. 2.5b): ! σ σ 1 1 = 1 + 2 = + = σ · + . E η E η 13 2.1. MATERIAL NON-LINEARITY (a) scheme (b) creep in concrete (c) relaxation in steel Figure 2.5: Viscoelastic model (Maxwell model) To represent the relaxation in steel, the stress change in time under constant deformation is analysed (Fig. 2.5c). An exponential decay can be observed. A relaxation time τ is defined as a time after which the initial stress is reduced by 1/e. In the generalised Maxwell model3 a set of spring–dash-pot Maxwell elements are used to represent that the relaxation occurs at a distribution of times. (a) scheme (b) creep in concrete Figure 2.6: Viscoelastic model (Kelvin–Voight model) The Kelvin–Voight model is used for description of creep. In the model, the elastic and viscous elements are combined in parallel (Fig. 2.6a). Subjected to constant stress (σ1 + σ2 = const.), it demonstrates an exponential increase of strain with relaxation time σ τ asymptotically approaching the steady-state value max = : E σ(t) = E · (t) + η · d . dt After removal of the load also an exponential decay of strain (reversible) can be observed with the same relaxation time (Fig. 2.6b). The Standard Linear Solid model (Fig. 2.7a) is a combination of the Maxwell model and the Kelvin–Voight model. Under constant load total deformation tot is a sum of three types of deformations (Fig. 2.7b): spontaneous elastic deformation 1 , delayed elastic deformation (reversible creep) 2 and irreversible creep (flow) 3 . The elasto-viscoplastic model in a general form (Fig. 2.8a) consists of a spring, a dash-pot and a sliding frictional element connected in parallel, i.e. the spring element of 3 also known as the Maxwell–Wiechert model 14 CHAPTER 2. MODELS FOR ANALYSIS OF REINFORCED CONCRETE BEAM ELEMENTS (a) scheme (b) creep in concrete Figure 2.7: Viscoelastic model (Standard Linear Solid model) (a) general elasto-viscoplastic (b) rigid-viscoplastic Figure 2.8: Elasto-viscoplastic models the Kelvin–Voight model replaced by the sliding frictional element acting as a rigid body. This model is used for representation of creep in concrete. A specific type of that model is the Bingham model (Fig. 2.8b) without a spring element (a rigid-viscoplastic model). In that model the strain is irreversible afrer reaching the yield stress (yield limit) σy . Figure 2.9: Viscoelasto-plastic model Figure 2.10: Viscoelasto-viscoplastic model The viscoelasto-plastic model (Fig. 2.9) is a modification of viscoelastic model in which a plastic phase is considered. The model might be regarded as development of the Kelvin–Voight model in which a frictional sliding element is added to simulate the stress–strain relationship after the yield limit is reached. Further improvement into the elasto-viscoplastic and viscoelasco-plastic models is 2.2. MATERIAL MODELS IN STANDARDS 15 consideration of viscous effects both in elastic and plastic phase. This leads to viscoelastoviscoplastic models. Despite their best representation of rheological properties of concrete, such models are not widely presented because of their complexity. Nevertheless, development of computer methods should increase their applicability. Two basic viscoelastoviscoplastic formulations are given by Perzyna and Duvaut–Lions. In Fig. 2.10 the viscoelasto-viscoplastic model according to Perzyna formulation is presented. 2.2 Material models in standards 2.2.1 Concrete models Concrete in compression Concrete is a brittle material as it has good properties in compression but fractures abruptly when its tensile strength is reached as a result of decohesion. The linear elastic model is used extensively in structural analysis and engineering design of concrete. Major design codes such as the Model Code, BS 8110, EC2 and ACI 318, as well as PN-B-03264 endorse this model. However, in the linear-elastic analysis the plastic properties of concrete in compression, which may in reality increase the load-bearing capacity of concrete member, are neglected. Therefore, the standards enable application of non-linear material models, taking these favourable properties into account. The models are based on Model Code and incorporated into other European standards. In this section they are presented acc. to EC2. For the structural analysis of elements under short-term loading the following stress– strain relationship is proposed, presented in Fig. 2.11a and described by the formula: kη − η 2 σc = fcm 1 + (k − 2)η (2.1) where: σc – compressive stress, fcm – mean compressive strength, c η = , c1 c – compressive strain, c1 – compressive strain at reaching maximum compressive strength fcm , cu1 – ultimate compressive strain4 , c1 . k = 1.05Ecm fcm This model is an example of the (elasto-)plastic model with softening. The linear-elastic behaviour can be assumed for stresses not greater than 0.4fcm , above which the behaviour becomes non-linear. When the material (compressive) strength fcm is reached, the element can still transfer load until the ultimate strain ecu1 is reached which signifies failure. Dimensioning of sections can be based on one of two proposed models: parabola– rectangle model, presented in Fig. 2.11b, and the bi-linear model, presented in Fig. 2.11c. 4 For normal concretes cu1 = cu2 = cu3 = 3.5h, c2 = 2.0h, c3 = 1.75h, n = 2.0 [36]. 16 CHAPTER 2. MODELS FOR ANALYSIS OF REINFORCED CONCRETE BEAM ELEMENTS (a) model for static analysis (b) parabola–rectangle model (c) bi-linear model Figure 2.11: Standard models for concrete in compression [36] The parabola–rectangle model is given by the formula: n fcd 1 − 1 − c for 0 < c < c2 c2 σc = f for c2 < c < cu2 cd and the bi-linear model by the formula: f c cd c3 σc = fcd where: fck , fcd c2 , c3 cu2 , cu3 n for 0 < c < c3 (2.2) (2.3) for c3 < c < cu3 – characteristic and design compressive strength, – compressive strain at reaching maximum compressive strength fcd , – ultimate compressive strain, – exponent. Concrete in tension Tensile failure of concrete is always a discrete phenomenon. Therefore, to describe the tensile behaviour a stress–strain diagram (Fig. 2.12a) should be used for uncracked concrete and a stress–crack opening diagram (Fig. 2.12b) should be used for cracked section [35]. (a) σ– diagram for uncracked concrete (b) σ–w diagram for cracked concrete Figure 2.12: Standard models for concrete in tension [35] For uncracked concrete a bilinear stress-strain relationship is given in equations: Eci ct σct = ctu − ct f 1 − 0.1 ctm 0.9f ctm − ctu Eci for σct ≤ 0.9fctm for 0.9fctm < σct ≤ fctm (2.4) 17 2.2. MATERIAL MODELS IN STANDARDS where: Eci fctm σct ct ctu – tangent modulus of elasticity, – mean tensile strength, – tensile stress, – tensile strain, – ultimate tensile strain5 . For cracked section a bilinear stress-crack opening relationship is given in equations: w fctm 1 − 0.8 w1 σct = w fctm 0.25 − 0.05 w1 for w ≤ w1 (2.5) for w1 < w ≤ wc where: w w1 wc GF – crack opening, = GF /fctm , crack opening for σct = 0.2fctm , = 5GF /fctm , crack opening for σct = 0, – fracture energy. 2.2.2 Steel models Plastic material models are also ideal for representation of the properties of steel, namely ductility, i.e. the ability to deform plastically without rupture, specifically under tensile stresses. This is endorsed by the standards where two models are proposed: for hot-rolled steel (Fig. 2.13a) and cold-worked steel (Fig. 2.13b). (a) hot-rolled steel (b) cold-worked steel (c) idealised and design model Figure 2.13: Standard models for steel [36] In the hot-rolled–steel model the yield stress is explicit as the yielding of that type of steel can be easily defined (visible yield plateau). The value of the yield stress is referred to as the yield strength fyk of steel. The behaviour of the cold-worked steel is much more difficult to describe. As the boundary between elastic and plastic behaviour is not that visible, the assumptions for the model are more implicit. The yield limit is determined as the f0.2k proof stress, i.e. the stress after reaching 0.2% strain. The tensile strain uk 6 is related to reaching the tensile strength ft 7 of steel, after which steel begins to flow until rupture. For design, simplified models presented in Fig. 2.13c are proposed. 5 ctu = 0.15h [35] uk ≥ 2.5% for steel A, 5.0% for steel B and 7.5% for steel C 7 ft = kfyk or ft = kf0.2k , where k is coefficient dependent on the class of steel [36] 6 18 CHAPTER 2. MODELS FOR ANALYSIS OF REINFORCED CONCRETE BEAM ELEMENTS 2.3 Geometrical non-linearity Long elements subjected to a large compressive force are on the verge of buckling, as under such a load the lateral stiffness of the element reduces significantly and a small lateral load may cause the element to buckle. Consideration of the influence of the secondary effects caused by reduction of that “geometrical” stiffness is referred to as P –∆ analysis, since additional moments arising in the element are the product of compressive force P and displacement ∆. The problem is marginal for the issues analysed in this thesis and is introduced juts for the notion of the reader. Assuming the displacements in the structure to be small, the linear buckling theory can be applied to determine the value of critical force Pcr (which leads to buckling of compressed element). Buckling of the beam elements is assumed to be caused only by bending moment exerted on that element; influence of transverse forces and shortening of the element’s axis is neglected. The value of the critical force is derived from a differential equation of deformation line. It occurs that the solution of that equation is not exact, i.e. there is an infinite number of critical forces and buckling modes that satisfy that equation. Buckling occurs for the smallest value of the critical force for the first mode of buckling (described by sine/cosine functions). The value of critical force is therefore: Pcr = where: EI lw µ π 2 EI lw2 (2.6) – bending stiffness of section, – buckling length of compressed bar; lw = µl, – buckling coefficient dependent on the support conditions of the element. If the displacements in the structure are large, the curvature of buckled elements should be described by a precise non-linear differential equation. Hence, the P (∆) relationship becomes exact and signifies that after reaching the critical force any increase of compressive force leads to extensive deflections and catastrophic increase of normal stresses. The assumption that the “axial” force is subjected to the element along its axis and that the element is straight is just an approximation. In reality, the elements have an initial geometrical imperfection, so as a result the subjected force acts eccentrically which leads to increase of bending moments exerted on this element. Nevertheless, that phenomenon has no qualitative influence of the structure, merely defines the direction in which the element will buckle and provides the exact P (∆) relationship. In the presence of transverse forces the critical force decreases, so the element is more prone to buckling under the axial compression. The shortening of the element’s axis may have additional influence on the reduction of the critical force, especially in short elements with a high elasticity limit. Chapter 3 Refinement of linear-elastic analysis results 3.1 Analysis of reinforced concrete members beyond elastic phase 3.1.1 Linear-elastic vs. non-linear analysis Under a given load bending moments as well as normal and shear forces take given values in every point of the structure, depending on the static scheme and material properties of the structure. As calculation of the structure (or cross-section) determination of distribution and values of the internal forces in the structure, usually presented in a form of graphs, with the assumed properties under given internal or external load should be understood. Thus, distribution of forces and moments is a known state which can be determined according to different theories of mechanics. Computation according to the theory of elasticity is based on the assumption that the relationship between stresses and strains in every cross-section of the structure remains linear until the moment of failure while the theory of plasticity assumes that the stress–strain relationship is non-linear in a whole range. The linear-elastic analysis is the most frequently used for analysis of structures. However, distribution of internal forces determined in such a way differs from the real distribution because it poorly resembles the real behaviour and random characteristics of the materials. Hence, the non-linear analysis is the only consistent way to verify the safety of the structure satisfying both equilibrium and compatibility. The Model Code specifies that “non-linear analysis is a realistic description of the physical behaviour and therefore a method completely consistent with the assumptions used for the local verification and member design” and “it should be used as a reference for other more simplified approaches”. The following idealisations of structural behaviour are proposed: 1. linear-elastic behaviour, 2. linear-elastic behaviour with limited redistribution (Sec. 3.2), 3. plastic behaviour (Sec. 3.3), 4. non-linear behaviour (Sec. 3.4). 19 20 CHAPTER 3. REFINEMENT OF LINEAR-ELASTIC ANALYSIS RESULTS Eurocode 2 specifies that linear analysis of elements based on the theory of elasticity may be used for both the serviceability and ultimate limit states, having assumed that: • cross-sections are uncracked, • stress–strain relationships are linear, • mean value of elastic modulus is considered. For thermal deformation, settlement and shrinkage effects at the ultimate limit state, a reduced stiffness corresponding to the cracked sections, neglecting tension stiffening but including the effects of creep, may be assumed. For the serviceability limit state a gradual evolution of cracking should be considered. As it will be presented in the following sections, these assumptions are very simplified and allow for structural analysis of only a limited range of cases. Therefore, the more complex idealisations should be introduced in the design. 3.1.2 Plastic properties of reinforced concrete In a rationally reinforced concrete beam reinforcement is located in the tensile zone while concrete works in compression in all loading stages. When the tensile strength of concrete is exceeded at the tensiled side, concrete is cracked and the work of reinforcing steel becomes more intensive taking the overall tensile stress. As a result, particular parts of the beam are in different stress stages, called phases, and the stiffness of the element is reduced and no longer constant. In general, three phases of reinforced concrete member work can be distinguished [13]: 1. phase I before the formation of the first crack, 2. phase II when section is cracked in tensile zone and steel takes all tensile forces, 3. phase III when equilibrium of forces is reached at the moment of failure. Figure 3.1: Real work of flexural reinforced concrete beam [13] Under the change of load in the most-loaded parts of the structure plastification of concrete or steel occurs. Plastified parts of the structure work in a different manner than before plastification. Incremental applied load, greater than the load producing first yielding, is assumed to produce inelastic rotation at the yielded section, but no change to the applied moment, so incremental moments are developed at sections other than the initially yielded section. Static scheme and initial distribution of internal forces is changed. 3.1. ANALYSIS OF REINFORCED CONCRETE MEMBERS BEYOND ELASTIC PHASE 21 Under increasing load in the structure with a new static scheme a new distribution of internal forces is formed in the parts which are still in elastic or partially-plastic phase. Redistribution of forces and moments can be observed. Implicit in the current use of the moment redistribution is the assumption that sections possess sufficient ductility for the requisite plastic deformations to occur. The parts in which plastification of sections is observed (i.e. where plastic deformations occurred in concrete or steel) are called plastification zones. Plastic deformations from the whole plastification zone are concentrated in a plastic hinge. Three types of hinges can be distinguished [21]: 1. real hinge, not capable of transferring any moments (Fig. 3.2a), 2. partial plastic hinge in which rotation occurs and which can transfer the increment of bending moment up to failure (Fig. 3.2b), 3. full plastic hinge which can undergo limited or unlimited plastic deformations but cannot transfer the increment of bending moment over a given value of moment Mu (Fig. 3.2c). (a) real hinge (b) partial plastic hinge (c) full plastic hinge Figure 3.2: Types of hinges [21] A place in the structure where the plastic hinge can be formed or failure can occur is called a critical section. In a statically-indeterminate structure there are usually several critical sections – their number is generally greater than the level of statical indetermination of the structure. Location of the critical sections can be estimated according to the elastic theory – they usually occur in the places of the extreme bending moments. The classical theory of reinforced concrete focuses on phase I and II. Dimensioning of flexural members is based on the comparison of stresses in steel and concrete with their allowable values. Deflections are calculated on the basis of Hooke’s law, both in phase I and II, but for different values of modulus of elasticity. According to that theory, transition between these two phases is instantaneous, which is not true – formation of a single crack determines treatment of the member as in phase II while in the reality most parts which are not yet cracks still work in phase I. The theory of ultimate limit state focuses only on the critical sections neglecting neighbouring sections in which the load-bearing capacity is not fully used, in such a way that limit state (phase III) occurs almost independently of the previous stage, as if a rigid, undeformable member would instantaneously undergo local failure. It is not true in reality. In an arbitrarily supported beam under loading all of the phases can be observed simultaneously, as presented in Fig. 3.1. Formation of a plastic hinge (phase III) is demonstrated by widening and deepening of cracks, and excess deformations of the member, and there is a transition between phase I and II and then between phase II and III. The influence of stiffness degradation in a flexural element is crucial for adequate representation of the behaviour of that element under incremental load. There is a mutual relationship between the stiffness, the load and the deformation, which can be defined by the equality on the basis of Bernoulli hypothesis: 22 CHAPTER 3. REFINEMENT OF LINEAR-ELASTIC ANALYSIS RESULTS κ= where: v(x) φ(x) κ M B dφ d2 v M = = dx d x2 B (3.1) – function of deformation line, – function of rotation of the cross-section, – function of curvature of deformation line, – bending moment in cross-section, – stiffness of the cross-section. Thus, the deformation function can be found by the solution of such a second-order differential equation. The stiffness law has to be known, which is not that difficult in case of statically determinate elements but becomes highly analytically complicated in continuous ones, as there is a mutual dependency between bending moment and stiffness distributions. (a) real [21], [6] (b) approximated [20], [6] Figure 3.3: M –κ relationship for flexural reinforced concrete elements Figure 3.3a shows the relationship between moment Mx and curvature κ as a result of stiffness degradation, which in statically-indeterminate elements decides about the actual distribution of internal forces. Analysing that diagram it is visible that for small values of load (section in phase I) the value of stiffness is constant and so the curvature is proportional to the value of bending moment: κel = M . B0 (3.2) Nevertheless, along with the increase of bending moment the value of the moment M0 is reached for which this relationship is no longer linear. The increment of bending moment is connected with displacements greater than for M < M0 , as the tensiled concrete undergoes plastic deformations until the first crack is formed and the stiffness decreases. After cracking, the compressed concrete continues to act linearly-elastic but it finally starts to deform plastically, too. For that reason, the increase of curvature becomes faster and is no longer linear when concrete is in plastic phase (κtot = κel + κpl ). Further increase of load leads to the bending moment reaching its ultimate value Mu . Depending on the proposed model, beyond this load limit the section undergoing further deformations can be assumed to either sustain moments to transfer them into other sections, until the ultimate curvature κu is reached. 3.2. LINEAR-ELASTIC ANALYSIS WITH MOMENT REDISTRIBUTION 23 The failure of an engineering structure falls into one of two simple categories: material failure or structural instability. If the section fails due to exceeding its resistance this may happen either as a result of tension (excess deformation or rupture of reinforcement) or compression (crushing of concrete). When calculations are based on the plastic hinges method, global (local) failure occurs if the structure with nth level of statical indetermination looses its global (local) stability as a result of formation of n + 1 plastic hinges (turning into a mechanism). The load at which a mechanism forms in any span is called the limit load in that span. Precise modelling of the behaviour of a structure is significant in a process of development of new structural solutions, with a tendency to minimise the weight of structure and economically use the materials. It is undeniable that, although we have precise methods for determination of deformation properties of steel and concrete, distribution of internal forces is determined based on the assumption of ideally-elastic cooperation of steel and concrete, neglecting the phenomenon that plastification of both materials influences vast areas beyond critical sections of the member. The method based on plastic deformations seems to be rational, especially in case of economical dimensioning, as it takes into account: • plastic properties of concrete, • magnitude of bending moment leading to failure of the section (phase III), against which the structure is protected with safety factors > 1, • assumption of simultaneous destruction of steel and concrete (condition of allowable percentage of reinforcement). Algorithms for dimensioning according to the theory of plasticity are based on experimental data, considering deformability of concrete and steel, changes of neutral axis location, magnitude of fracture moment and deflection as well as general properties of concrete such as fire resistance, corrosion of steel and concrete, resistance to cracking, water permeability, resistance to high temperatures, etc. The approaches to refine the linear-elastic analysis results presented in the following sections are presented in a form proposed by Eurocode 2. 3.2 Linear-elastic analysis with moment redistribution The plastic behaviour of reinforced concrete at the ultimate limit state affects the distribution of moments in a structure. To allow of this, the moments derived from the elastic analysis my be redistributed based on the assumption that plastic hinges have formed at the sections with the largest moments. EC2 allows of such an adjustment of bending moment with the ratio of the redistributed and initial bending moment diagrams being an indicator of the amount of redistribution which has occurred. The percentage of moment redistribution δ[%] at a section along a beam is calculated as follows: Mred · 100% (3.3) Mel where Mred signifies the moment after redistribution and Mel the moment before redistribution, calculated for linear state. δ= 24 CHAPTER 3. REFINEMENT OF LINEAR-ELASTIC ANALYSIS RESULTS The initial elastic bending moment diagram thus forms the baseline for the redistribution calculation and any assumptions or approximations made in its determination will directly affect the level of redistribution calculated using the above expression. It must be noted that in the static analysis based on the linear-elastic analysis with limited redistribution the elastic–plastic σ– relationships are used, so the name “linear” becomes a bit ambiguous as the method is non-linear in a mathematical sense. The linear analysis with limited redistribution may be applied to the analysis of structural members only for the verification of ULS and only for continuous beams and slabs predominantly subjected to flexure and with the ratio of the lengths of adjacent spans in the range of 0.5 to 2. Formation of plastic hinges requires relatively large rotations with yielding of the tension reinforcement. Design codes achieve this by specifying rules which ensure that the tension steel must have yielded, explicitly in the case of ACI 318 (which specifies a minimum reinforcement strain of 7500 microstrains) and implicitly in the case of PN-B, BS 8110 and EC2 (which link percentage redistribution to neutral axis depth) [31]. To ensure large strains in the tension steel, EC2 restricts the depth of the neutral axis: xu for fck ≤ 50MPa, δ ≥ k1 + k2 d xu δ ≥ k3 + k4 for fck > 50MPa, d δ ≥ k5 if reinforcement type B or C is used, δ ≥ k6 if reinforcement type A is used1 . where: xu d – the depth of the neutral axis at the ultimate limit state after redistribution – the effective depth of the section. 0.0014 The recommended values of ki are : k1 = 0.44, k2 = 1.25 0.6 + , k3 = 0.54, cu2 0.0014 k4 = 1.25 0.6 + , k5 = 0.7 and k6 = 0.8. Therefore, the limitation in the method cu2 due to the limitation of the rotation in the plastic hinge results in the requirement that the maximum allowable moment redistribution cannot exceed ±30% [20]. Redistribution should not be carried out in circumstances where the rotation capacity cannot be defined with confidence. The BS 8110 imposes a minimum neutral axis depth of 0.11d (where d – effective depth of beam) which, acc. to [31], has the effect of restricting reinforcement strains to a maximum of 28000 microstrains when making the usual assumption of linear strain distribution across the section. In reality, this value is largely meaningless since gross yield of the reinforcement will have occurred by the time this neutral axis depth has been reached, leading to strains greatly in excess of this nominal value. Therefore, the reinforcement will be able to develop the required strain and the failure of section will be caused by crushing of the compressed concrete. The ACI Code provides two methods for determination of allowable redistribution [16]. In the first method the negative moments at the supports can be changed by the value: 2 1 2 A stands for low ductility steels while B and C for high ductility steels In PN-B k1 = 0.44, k2 = 1.25, k5 = 0.7, k6 = 0.8. High strength concretes are not considered. 25 3.3. PLASTIC ANALYSIS ρ − ρ0 δ ≤ 20 1 − ρb ! (3.4) for ρ − ρ0 ≤ 0.5ρb , where: ρ ρ0 ρb – ratio of tension reinforcement, – ratio of compression reinforcement, – reinforcement ratio producing balanced strain condition3 . Alternatively, the negative moments can be changed by no more than 1000t %4 , with a maximum of 20%, provided that t ≥ 7.5h at the section where moments are being reduced. The method of the moment redistribution presented in standards is a great simplification of the real process of redistribution. Only a fact of transferring bending moments from the most loaded to the less loaded cross-sections of the flexural element is considered without a detailed analysis of the process itself. For that reason it is obvious that the proposal is provided in a very implicit and uncontrollable way. 3.3 Plastic analysis Methods based on plastic analysis shall only be used for check at the ULS. The effects of previous applications of loading may generally be ignored and a monotonic increase of the intensity of actions may be assumed. The plastic analysis without any direct check of the rotation capacity may be used if the ductility of the critical sections is sufficient for the envisaged mechanism to be formed. The ductility condition is assumed to be satisfied if [36]: xu ≤ 0.25 for d concrete strength classes at most C50/60, and ≤ 0.15 for concrete strength classes at least C55/67, 1. the area of tension reinforcement is limited such that at any section 2. reinforcing steel is either class B or C5 , 3. the ratio of the moments at intermediate supports to the moments in the span shall be between 0.5 and 2. If the rotation capacity has to be controlled, the simplified procedure is proposed based on the rotation capacity of beam zones over a length of approximately 1.2 times the depth of the section (Fig. 3.4). It is assumed that these zones undergo plastic deformation (formation of yield hinges is observed) under the relevant combination of actions. Thus, the plastic hinge in reinforced concrete is not a single-section phenomena, but occurs as a result of large curvatures in the section and in its neighbourhood due to concentration of deformations [9]. The verification of the plastic rotation in the ultimate limit state is considered to be fulfilled if it is shown that under the relevant action the calculated rotation θs is less than or equal to the allowable plastic rotation θpl,d . The rotation θs should be determined on the basis of the design values for actions and materials. In the simplified procedure, the allowable plastic rotation may be determined 3 i.e. when tension reinforcement reaches the strain corresponding to its specified yield strength fy as 26 CHAPTER 3. REFINEMENT OF LINEAR-ELASTIC ANALYSIS RESULTS Figure 3.4: Plastic rotation θs of reinforced concrete sections in continuous beams [36] by multiplying the basic value of the allowable rotation by a correction factor kλ that depends on the shear slenderness, according to the formula: s kλ = λ 3 (3.5) where λ is the ratio of the distance between point of zero and maximum moment after redistribution and effective depth d. As a simplification λ may be calculated for the concordant design values of the bending moment and shear, as: λ= MSd . VSd · d (3.6) Figure 3.5: Allowable plastic rotation θpl,d of reinforced concrete sections [36] The recommended values of θpl,d for steel classes B and C (the use of class A steel is not recommended for plastic analysis) and concrete strength classes less than or equal to concrete in compression reaches its assumed ultimate strain (3h). 4 t signifies the tensile strain in the reinforcement 5 In PN-B requirements for steel are expressed as “high ductility steel”, which actually coincides with EC2 steel class B and C 27 3.3. PLASTIC ANALYSIS C50/60 and C90/105 are given in Fig. 3.5. The values for concrete strength classes C55/67 to C90/105 may be interpolated accordingly. The values apply for the shear slenderness λ = 3.0. For different values of the shear slenderness the θpl,d should be multiplied by kλ . It must be noted that the concept of the rotational capacity of a reinforced concrete section is not well-examined [9]. There is a number of factors to be controlled with respect to their influence on the plastic properties of reinforced concrete, such as: scale (dimensions of the element), length of segment in which the concentration of deformations occurs, compression reinforcement, cyclic loads as well as normal and transverse forces. 3.3.1 Limit equilibrium method The plastic analysis should be based either on the lower bound (static) method or on the upper bound (kinematic) method. This can be presented on the simple example of the limit equilibrium method. The idea of the limit equilibrium method, contrary to the redistribution method in which the history of loading is important, is to find the limit load. The distribution of the internal forces is determined based on the assumption that the cross-section of the structure behaves elastically up to some given value of stresses (which generates the ultimate moment Mu ), above which the full plastic hinge is formed (refer to Fig. 3.2c) which is no longer able to carry any increments of stress but can undergo unlimited deformations (Fig. 3.3b). The remaining cross-sections in which the value of acting moment M < Mu behave in a linear-elastic manner. The limit equilibrium method is acc. to [26] a simplification of the linear-elastic analysis with redistribution. As it does not satisfy requirements of EC2, it should be – as a consequence – forbidden. Nevertheless, under certain precautions, the method has a number of practically approved advantages, such as: 1. economy is reinforcement design, especially visible with the increasing q ratio; g+q 2. unification of reinforcement layout in mid-span and over supports (with a little additional reinforcement of the edge span); 3. simplification of calculations. Kinematic method The kinematic method is referred to as the upper bound method as is provides the upper boundary for the searched limit load. In the kinematic approach the limit load is determined by comparison of a virtual work of load and a virtual work in plastic hinges under an exerted bending moment (a detailed scheme of the procedure is presented in Fig. 3.6 acc. to [20]). Figure 3.6: Determination of limit load Qu with kinematic method [20] 28 CHAPTER 3. REFINEMENT OF LINEAR-ELASTIC ANALYSIS RESULTS Under the incremental force Q two hinges are formed along with the displacement of the point 1. From the equality of the virtual works of the external forces and bending moments in the hinges, the value of the limit force can be derived as: 1 1 + MBu . (3.7) bc c This solution is very easy, but the location of plastic hinges has to be known prior to calculations of the limit load, which is not obvious in all cases. Qu = M1u Static method The static method is referred to as the lower bound method as is provides the lower boundary for the searched limit load. In the static approach the analysis of transferring of the incremental load q to the continuous beam is performed (a detailed scheme of the procedure is presented in Fig. 3.7 acc. to [20]). The proportionally increased load q eventually reaches the value q1 which leads to the internal support bending moment MBq1 reaching the ultimate value MBu . This results in the plastic hinge formation in that cross-section. Hence, the static scheme of the structure is modified and the additional load transferred to that model would be analysed for the new scheme. Along with the further increase of the load from q1 to q2 , the ultimate moment MCu = MCq1 + MCq2 is reached over the next support, another plastic hinge if formed and the static scheme is again modified. The process of the load increase resulting in the formation of the subsequent plastic hinges can be repeated as long as the number of the hinges does not exceed the level of static indetermination of the structure. In that case for a 3-span beam with 4 supports and consequently 4 unknown support reactions, 5 hinges can occur under the load q3 in the structure until it turns into a mechanism. qu = q1 + q2 + q3 is the limit load for that structure. 3.4 Non-linear analysis The EC2 allows of application of non-linear methods of analysis for both ULS and SLS, provided that equilibrium and compatibility are satisfied and an adequate non-linear behaviour for materials is assumed. The analysis may be first or second order. Nevertheless, no consistent design procedure is provided except for some design requirements which must be satisfied: • At the ultimate limit state, the ability of local critical sections to withstand any inelastic deformations implied by the analysis shall be checked, taking appropriate account of uncertainties. • For structures predominantly subjected to static loads the effects of previous applications of loading may generally be ignored and a monotonic increase of the intensity of the actions may be assumed. • The use of material characteristics which represent the stiffness in a realistic way but take account of the uncertainties of failure shall be used when using non-linear analysis. Only these design formats which are valid within the restricted fields of application shall be used. 3.4. NON-LINEAR ANALYSIS 29 Figure 3.7: Determination of limit load qu with static method [20] • For slender structures, in which second order effects cannot be ignored: – equilibrium and resistance shall be verified in the deformed state; – deformations shall be calculated taking into account the relevant effects of cracking, non-linear material properties and creep; – where relevant, analysis shall include the effect of flexibility of adjacent members and foundations (soil-structure interaction); – the structural behaviour shall be considered in the direction in which deforma- 30 CHAPTER 3. REFINEMENT OF LINEAR-ELASTIC ANALYSIS RESULTS tions can occur and biaxial bending shall be taken into account when necessary; – uncertainties in geometry and position of axial loads shall be taken into account as additional first order effects based on geometric imperfections. Having understood the real behaviour of the flexural reinforced concrete members and based on the methods proposed by the standards along with the requirements for non-linear analysis, the model can be derived. Chapter 4 Stiffness degradation in R/C flexural members 4.1 Stiffness of reinforced concrete flexural members The stiffness of a reinforced concrete beam decreases as the magnitude of the load increases. Analysis and mathematical interpretation of this phenomenon is complicated for such a complex material as reinforced concrete. Classic plastic properties have influence only on the compressive side of the section. On the other side, cracks in concrete, cooperation of tensiled concrete with reinforcing steel and non-uniform stress state in steel in cracks and between them can be observed. Because of the complexity of these processes considering all of the factors in a form of a multi-parameter cause analysis becomes difficult. It is very important for an engineering practice to formulate a theory describing the real work of statically-indeterminate elements in which there is a significant influence of stiffness degradation in the process of flexure. In the analysis of a reinforced concrete beam considering the changing stiffness at the length of the beam, classic approach is insufficient. As it was stated in Sec. 3.1.2, in the beam under heavy loading all of the phases of the work can be distinguished simultaneously. Therefore, the stiffness of the beam is different at the supports and in the mid-span. Moreover, the stiffness, apart from differing at the length of the beam, degrades in a given cross-section along with the loading being increased. Hence, it is basically a function of two parameters [13]: location x ξ = and load P . However, there is a number of phenomena influencing the stiffness, l such as cooperation of steel and concrete at crack due to the bond or the effects of creep and shrinkage regardless of the applied load. The notion of stiffness degradation allows of solution of two groups of problems: calculation of displacements and analysis of the work of statically-indeterminate beams. In determination of deflection (as a consequence of action of bending moment) three main approaches can be distinguished [20]: 1. consideration of the local degradation of stiffness in the cracked cross-section and the neighbouring cross-sections, 2. averaging of the stiffnesses of cracked and uncracked sections, 3. averaging of the stiffness for the whole element. It was experimentally proved that the process of deflection of the beam under the load is continuous [13], and so the (1) approach is used very rarely [20]. The effects of the 31 32 CHAPTER 4. STIFFNESS DEGRADATION IN R/C FLEXURAL MEMBERS cracks in the tensile zone, even if instantaneous, are compensated on one side by elastic properties of steel and on the other – in the compressive zone – by elasto–plastic properties of concrete. A drop in stiffness caused by formation of the crack occurs in a very small part of the beam and the process is performed together with the increment of the moment, which results in almost unnoticeable discontinuity. That is why the cross-section behaviour M –κ can be represented by a smooth curve (as it was shown in Fig. 3.3a). Of course, for such a non-uniform material as concrete, this theory to be true must be based on the statistical data (mean values). Only in that sense concrete can be treated as continuum and, consequently, the material and cross-sectional characteristics can be described by analytical functions (for which the differential and integral calculus can be applied)[13]. To derive the model for a reinforced concrete flexural member based on the theory of stiffness degradation, the following assumptions have to be made [13]: • a specified reinforced concrete material and cross-section model, • stiffness change both along the length of the beam and with the increase of load, • validity of the theory in statistical sense, • validity of the Hooke’s law in the initial phase of the beam’s work, • limit state of failure occurring for stresses in steel reaching yield strength and in compressed concrete compressive strength at flexure, equivalent to formation of plastic hinge1 , • formation of cracks connected with cracking moment being equivalent to the second ultimate limit state of the beam, • curvature of the beam being related to the central axis of the beam (line along the centres of gravity of the sections) and in physical meaning not coinciding with the neutral axis (line along the zero-stresses in the sections), with neutral axis being a theoretical (not material) axis. 4.2 Bending stiffness of cross-section 4.2.1 Material behaviour Figure 4.1 presents the behaviour of the cross-section of the flexural reinforced concrete member in different phases of work. Phase I - uncracked section For a reinforced concrete section in phase I a perfect bond between steel and concrete can be assumed. The section, called a transformed section, is composed of concrete section and n-time reinforcement section. The area of the transformed section Atrans for a doubly-reinforced section: Atrans = bh + n(As1 + As2 ) (4.1) 1 Note that in all proposed elasto–plastic reinforced concrete models failure is not a direct result of exceeding material strength but reaching the ultimate compressive strain leading to crushing of compressed concrete fibers. 4.2. BENDING STIFFNESS OF CROSS-SECTION 33 Figure 4.1: Phases of work of reinforced concrete member. Stresses [13] where: b, h – width and height of section, As1 – reinforcement in tensile zone, As2 – reinforcement in compressive zone. Phase Ia (Fig. 4.1a). Under small loads stresses are linear, which means they are still within the range of relative proportionality. Elastic modulus of concrete is constant so it does not influence the change of stresses. The use of reinforcing steel is very small, however, the neutral axis is located a bit lower than the centre of gravity of the section. The depth of the compressive zone x for a rectangular cross-section: bh2 + nAs1 d + nAs2 a2 xI = 2 Atrans where: d a1 a2 (4.2) – effective height of beam, d = h − a1 , – distance of centroid of reinforcement in tensile zone from tensiled edge, – distance of centroid of reinforcement in compressive zone from compressed edge. The moment of inertia II : II = b(h − xI )3 bx3I + + nAs2 (xI − a2 )2 + nAs1 (d − xI )2 . 3 3 (4.3) The static moment of reinforcement about the centroid of the section SI : SI = As1 z1,I + As2 z2,I (4.4) where z1,I and z2,I denote location of the reinforcement As1 and As2 with respect to the central line of the section2 . For a singly-reinforced section the component As2 is omitted. Phase Ib (Fig. 4.1b). Increase of load, thus increase of bending moment leads to non-linear stress distribution in the tensile zone. In the compressive zone, where stresses are still 2 Central line of the section (line along the center of gravity) does not necessarily coincide with a neutral axis of the section for a given geometry and level of loading. 34 CHAPTER 4. STIFFNESS DEGRADATION IN R/C FLEXURAL MEMBERS within proportionality range, their distribution remains linear. The use of reinforcement increases but is still insubstantial. The neutral axis is raised. Determination of the equilibrium equations for phase Ib is much more difficult than in phase Ia as distribution of stresses is no longer linear. Phase I last until the tensile strength of concrete is reached. The respective bending moment reaches the value of the cracking moment Mcr . The cracking moment can be determined under the assumption that a crack is formed when the tensile stress in concrete reaches the mean tensile strength fctm and can be calculated as [9]: Mcr = fctm Wc (4.5) where the bending index Wc is usually calculated for the concrete section (i.e. without reinforcement), although it is allowed to use the transformed section characteristics. There used to be a practice to account for plastic properties of concrete in tension, which led to increase of the value of the cracking moment, but it no longer functions. Phase II - cracked section Phase IIa (Fig. 4.1c). Fracture of the reinforced concrete section (formation of cracks) signifies transition between phase I and II. Initially, cracks are short and small but along with increase of the bending moment they get wider and lengthen towards the compressed side. Formation of the first crack is a random phenomenon, thus its location is also random and unpredictable. Only after the first crack occurs, a pattern in formation of subsequent cracks can be observed. In the compressive zone stresses remain linear. Only a small part of concrete section works in tension – the gross tensile stresses are transferred to reinforcing steel. The range of phase IIa depends on the class of concrete and for very strong ones lasts almost up to the moment of failure. For singly-reinforced section: • the depth of the compressive zone (from equilibrium of static moments): s xII = • the moment of inertia: nAs1 2bd 1+ − 1 , b nAs1 bx3II III = + nAs1 (d − xII )2 , 3 (4.6) (4.7) • the static moment of reinforcement about center of gravity of the section SII = As1 z1,II . (4.8) v u n(As1 + As2 ) u 2b(A d + A a ) s1 s2 2 t1 + xII = − 1 , (4.9) For doubly-reinforced section: • the depth of the compressive zone: b n(As1 + As2 )2 35 4.2. BENDING STIFFNESS OF CROSS-SECTION • the moment of inertia: III = bx3II + nAs1 (d − xII )2 + nAs2 (xII − a2 )2 , 3 (4.10) • the static moment of reinforcement about center of gravity of the section SII = As1 z1,II + As2 z2,II . (4.11) Phase IIb (Fig. 4.1d). Along with further increase of the load, the tensiled concrete undergoes heavy cracking while in the compressive zone significant plastification of concrete can be observed. Stress distribution becomes totally non-linear and the Hooke’s law no longer applies. Therefore, similarly as in phase Ib, determination of that distribution is very difficult. Phase III - ultimate limit state Figure 4.1e presents the cross-section in the ultimate limit state of flexure (when bending moment reaches its maximum value). Distribution of stresses is non-linear. In compressive zone its shape is close to the 2nd or 3rd order curve. In practice a non-linear diagram of stress in concrete – as a result of chosen diagram for dimensioning (refer to Fig. 2.11b) – can be substituted with [18]: 1. the rectangular–parabolic stress block (Fig. 4.2b), 2. the equivalent rectangular stress block (Fig. 4.2c). (a) cross-section (b) rectangular–parabolic (c) equivalent rectangular Figure 4.2: Simplified diagrams for distribution of stresses in ULS [18] In the classical reinforced concrete design it is assumed that the structural failure occurs when the ultimate limit state is reached in one section (i.e. when the bending moment induced exceeds the maximum allowable moment). Actually, the load-bearing capacity of the section is not yet depleted when the material strength is reached – the section can still undergo some limited deformations (strains). When yielding of steel occurs (i.e. when σs = fyd ), a plastic hinge is formed and the section posseses some rotational capacity to sustain loads and transfer them into the less congested areas. Not until the number of plastic hinges exceeds the degree of statical indetermination, does the failure occur as a result of the global stability loss. Hence, the load-bearing capacity is not a property of a single cross-section but of a structure as a whole. 36 4.2.2 CHAPTER 4. STIFFNESS DEGRADATION IN R/C FLEXURAL MEMBERS Cross-section behaviour Models for a cross-section behaviour (M –κ relationship) are collectively presented in Fig. 4.3 acc. to [20]. (a) model 1 (b) model 2 (c) model 3 (d) model 4 (e) model 5 (f) model 6 (g) model 7 (h) model 8 (i) model 9 Figure 4.3: Models for cross-section behaviour [20] Figure 4.3a presents the linear-elastic model for the flexural element. The constant stiffness is a product of the modulus of elasticity Ec and the moment of inertia of the section Ic , i.e. B = Ec Ic . Depending on the standard, the value of the modulus of elasticity is assumed to be Ec = Ec0 (line 1), Ec = 0.85Ecm (line 2) or Ec = 0.625Ecm (line 3), while the moment of inertia is derived for the uncracked transformed section Ic = II . The model presented in Fig. 4.3b is a modification of the previous model taking into account occurrence of the cracks. Phase I is neglected in that model. The models in Fig. 4.3c and Fig. 4.3d are the first models considering the stiffness degradation in a flexural reinforced concrete element in different phases of work. The first one (Fig. 4.3c), given by Muraszow and incorporated by old Russian standards, assumes a stiffness jump at crack formation. This model was applied initially in Polish Standard as 37 4.2. BENDING STIFFNESS OF CROSS-SECTION well, but in the revision published in 1976 it was replaced by the one in Fig. 4.3d, where a transition zone between the cracked and uncracked section under the load in the vicinity of the cracking moment is more smooth. The influence of stiffness degradation is well approximated by the model in Fig. 4.3e (line 1), especially for the loads M ≤ 0.9Mu . The value of strain derived for the uncracked section is increased by the value of strain caused by crack formation. The behaviour of the section in phase II can be treated as either linear or non-linear. The ultimate limit state is determined for the modified version of that model in which the failure is assumed to occur for the value of (Mu , κu ) (line 2). However, that far-from-reality assumption leads to serious design errors, which in contrary is not a problem for a solution developed by Levi presented in Fig. 4.3f. The best representation in a form of a broken lines was derived by Macchi for CEB and is shown in Fig. 4.3g. The last group of models, presented in Fig. 4.3h, assumes a continuous stiffness degradation. The most popular are the models proposed by Kuczyński in [13]. These models satisfy the assumptions stated in Sec. 4.1 and coincide with the experimental results (Fig. 4.3i shows the experimental data in which discrete values are approximated with a parabolic function). The bending stiffness B with the initial value of B0 is degraded in the whole range of work of the flexural element proportionally to the ratio of the active bending moment Mi and ultimate bending moment Mu in the section under the load qi . This proportionality can be either direct (linear) or power. In the first case the stiffness degradation is referred to as mutation φ and defined by the formula: Mi 1−φ , Mu B = B0 (4.12) while in the other it is referred to as mutation ψ and defined by the formula: " Mi B = B0 1 − Mu ψ # . (4.13) The values φ and ψ are empirical coefficients so in order to achieve a satisfactory approximation, these values must be indicated empirically. The approximation accuracy is limited to the value Mi /Mu < 0.85, but the analyses prove that the mutation φ can be used up to Mi /Mu = 1 [20]. For a given state of stress in the section, the curvature of deflection line at this section is defined as [21]: s − c κ= (4.14) d where: c – maximum compressive strain (of outermost concrete fibre), s – maximum tensile strain (of tensile reinforcement), d – effective depth of cross-section. Phase I - uncracked section The cross-section works within an elastic range and concrete cooperates in transferring tensile stresses (M < Mcr ). The maximum compressive and tensile strains are equal to (Fig. 4.4a): 38 CHAPTER 4. STIFFNESS DEGRADATION IN R/C FLEXURAL MEMBERS Figure 4.4: Phases of work of reinforced concrete member. Strains [18] c = − σc M xI =− , Ec Ec II (4.15) s = − σs M (d − xI ) =− , Ec Ec II (4.16) M Ec II (4.17) and the curvature: κI = where: Ec II – elastic modulus of concrete, – moment of inertia for uncracked section, acc. to Eq. 4.3. Phase II - cracked section The curvature of the cracked section (Fig. 4.4b) can be derived analogically to the curvature of uncracked section, neglecting the cracked tensile part of concrete section, for the moment of inertia of the cracked section III acc. to Eq. 4.7 or Eq. 4.10 [18]: κII = M . Ec III (4.18) Phase III - ultimate limit state The ultimate value of the curvature κu can be easily determined assuming the ultimate value of c = cu and x = xlim (Fig. 4.4c), for which: cu κu = − . (4.19) xlim For normal concretes the ultimate value of compressive strain should be taken as 3.5h. The ultimate value of the location of the neutral axis should be limited according to the balanced design recommendation, i.e. in a way that provides simultaneous depletion of concrete and reinforcement resistance (concrete and tension steel reach their ultimate strains at the same time). When no distribution is considered this value is xlim = xbal = 0.45d, and is smaller if moments are redistributed [18]. 4.2. BENDING STIFFNESS OF CROSS-SECTION 4.2.3 39 Creep and shrinkage effects The works focused on derivation of a good model for the behaviour of the flexural reinforced concrete elements result in creation of newer and better models. Nevertheless, so far there is no model which, with sufficient accuracy, resembles all of cases of the element’s work, let alone under the long-term loading. It is essential to prepare a model taking into consideration the influence of rheological effects, which at the moment are modelled with: • change of moduli of elasticity of steel and concrete, • application of multiplication factors increasing the final value of deflection calculated for short-term loading, • application of additional multiplication factors increasing the initial value of deflection to account for creep and independently addition of the value of deflection due to shrinkage. Creep The effect of creep results in increasing deflections with time and should be included in calculations by using an effective modulus of elasticity Ec,ef f [18]: Ec,ef f = Ecm 1 + φ(∞, t0 ) (4.20) where: φ(∞, t0 ) – creep coefficient equal to the ratio of creep strain to initial elastic strain3 , Ecm – mean elastic modulus of concrete. Shrinkage The effect of shrinkage of the concrete is a change of curvature by κcs and consequently the deflection of the member; it can be calculated according to the formula [18]: κcs = Es cs S Ec I (4.21) where: cs Es S I – free shrinkage strain4 , – modulus of elasticity of steel, – static moment of reinforcement about the centroid of the section (S = SI or SII ), – moment of inertia of the section (I = II or III ). 4.2.4 Tension stiffening effect In a cracked cross-section all tensile forces are balanced by the steel only. However, between adjacent cracks, tensile forces are transmitted from the steel to the surrounding concrete by bond forces. The contribution of the concrete may be considered to increase 3 4 For detailed procedure of determination of φ(∞, t0 ) refer to [36]. For detailed procedure of determination of cs refer to [36]. 40 CHAPTER 4. STIFFNESS DEGRADATION IN R/C FLEXURAL MEMBERS the stiffness of the tensile reinforcement. This effect is called the tension stiffening effect. If the tension stiffening effect is neglected, the stiffness of a reinforced concrete bar or a structural member is underestimated. The influence of the tension stiffening effect can be introduced with a coefficient [18]: ξ(M ) = 0 if M < Mcr Mcr 1 − β M if M ≥ Mcr (4.22) where: β Mcr M – load duration factor; β = 1 for short-term load, β = 0.5 for sustained loads, – cracking moment, – design bending moment for calculation of curvature and deflection. 4.2.5 Mean moment–curvature relationship The mean curvature κmean should be based on both cracked and uncracked sections, taking into consideration the effects of creep, shrinkage and cooperation of steel and concrete at crack in a form of tension stiffening effect. The modified values of curvatures κI and κII allowing for creep and shrinkage effects are given by formulas: κI = M − Es cs SI , Ec,ef f II (4.23) κII = M − Es cs SII . Ec,ef f III (4.24) An average value of curvature κm can be obtained using a formula [18]: κmean = (1 − ξ)κI + ξκII (4.25) where: κI κII ξ – curvature for uncracked case, considering creep and shrinkage, acc. to Eq. 4.23, – curvature for cracked case, considering creep and shrinkage, acc. to Eq. 4.24, – coefficient allowing for tension stiffening effect. 4.3 Bending stiffness of member 4.3.1 Influence of cracks Distribution of stiffness The most popular is an assumption that the stiffness of a member is constant along its length (Fig. 4.5b). Providing the considerations of this thesis it becomes obvious that such an assumption is far from reality and more sophisticated models should be introduced. The influence of cracks may be treated as a local phenomenon, thus locally influencing the stiffness of the member (Fig. 4.5c). It can be assumed that beyond the cracked area sections behave as in phase I, while the cracked parts are considered as either parts of limited length, working in phase II or as points where additional elastic rotations occur. 41 4.3. BENDING STIFFNESS OF MEMBER The continuous stiffness degradation acc. to [13] is presented in Fig. 4.5d. This is quite a good approximation, except that in the zone of minimum moments there should be a segment of constant stiffness (which results from the analysis of the experimentally obtained M –κ relationship presented in Fig. 4.3i). This problem is solved by modification proposed in [20] presented in Fig. 4.5e. (a) cracked flexural beam (b) constant stiffness (c) local stiffness degradation (d) continuous stiffness degradation (e) modified continuous stiffness degradation Figure 4.5: Stiffness of flexural beam along the element [20] Cracking pattern During the state of crack formation one crack after another occurs decreasing the stiffness of the member. When cracks appear, single cracks play an important role. In this state some parts of the area between cracks remain in phase I. After the crack formation is finished, the mean spacing between cracks where bonded reinforcement is fixed at reasonably close centres within the tension zone (spacing ≤ 5(c + φ/2)) can be taken as [36]: sr,max = k3 c + k1 k2 k4 φ ρp,ef f (4.26) 42 where: c φ k1 k2 k3 k4 ρp,ef f CHAPTER 4. STIFFNESS DEGRADATION IN R/C FLEXURAL MEMBERS – concrete cover, – diamater of reinforcement, – coefficient dependent on bonding properties; k1 = 0.8 for ribbed bars, – coefficient dependent on strain distribution; k2 = 0.5 for bending, – coefficient; recommended value k3 = 3.4, – coefficient; recommended value k3 = 0.425, – effective reinforcement ratio, given by formula: As ρp,ef f = Ac,ef f where: As Ac,ef f (4.27) – area of reinforcement, – effective area, for rectangular section Ac,ef f = hef f b, where: ! h−x hef f ≤ min 2.5(h − d); . (4.28) 3 Where spacing of the bonded reinforcement exceeds 5(c + φ/2) or where there is no bonded reinforcement within the tension zone, the maximum crack spacing may be assumed as: sr,max = 1.3(h − x) (4.29) where: h – height of section, x – depth of neutral axis. In the analysis with numerical methods it can be assumed that cracks are either a fuzzy or concentrated phenomenon [9]. 4.3.2 Influence of reinforcement In the analysis, the flexural stiffness EI is used to determine the values of the internal forces in statically indeterminate systems. Since the reinforcement is not know until the end of the design process, in static calculations the concrete section is most frequently used. The use provides simplicity but also has implications for the moment redistribution. It is intuitive that distribution of bending moments in a statically indeterminate system is different if initially a constant flexural stiffness is assumed and when a local degradation of stiffness is considered where reinforcement is applied, even though all sections along the member are behaving in a linearly-elastic fashion. Therefore, it can be said that the so-called relative not the absolute values of EI determine the moment distribution and that the stiffness changing along the member influences not only the ULS but also the SLS (when reinforcement is behaving elastically) [31]. The redistribution resulting from the mismatch between the assumed and actual stiffness values can be termed the elastic redistribution while the post-yield redistribution (after the reinforcement yields) can be referred to as the plastic redistribution [31]. The plastic redistribution is widely recognised and used. Nevertheless, the total redistribution in ULS design is actually a sum of elastic and plastic redistribution, as the elastic redistribution will occur even if it is not anticipated. In order to take into account the actual stiffness distribution of the reinforcement, the reinforcement layout has to be known prior to undertaking the analysis. Such an approach is possible only with iterative numerical methods. Chapter 5 Numerical model There are two main criteria that should be satisfied by the model of a physical phenomenon [24]: the model should be as simple as possible, but reproduce the important characteristics consistent with experimental results and should be theoretically sound and computationally stable, so that reliable analysis results are obtained. A distinction has to be made between a continuous and a discrete approach [22]. A discrete approach is such a situation in which a finite number of components can be extracted while a continuous approach is defined using the mathematical functions of an infinitesimal which leads to differential equations or equivalent statements implying an infinite number of elements. Discrete problems can be solved easily, even for situations with a great number of elements, which is not so obvious in continuous problems. That is why a discretisation of continuous problems can be applied where it is inefficient or impossible to solve those problems exactly by mathematical manipulation. A design performed for a non-linear constitutive material model assumed in this thesis requires discretisation and numerical approach. Derivation of a numerical simulation of a physical phenomenon is an interdisciplinary task. In case of this thesis a structural engineer and a software developer must be engaged in the process, which is schematically preseted in Fig. 5.1. Figure 5.1: Process of numerical simulation of physical phenomenon This thesis is oriented towards the first and the second phase of the model derivation process. If a problem in the model can be solved in a number of ways, their advantages and disadvantages are analysed and the applicability evaluated. No choice is made as the optimisation of the computational performance is not the aim of this thesis. 43 44 CHAPTER 5. NUMERICAL MODEL 5.1 Static scheme 5.1.1 Geometry As an input data an arbitrary continuous beam is chosen with known geometry and reinforcement layout. An exemplary beam is shown in Fig. 5.2. Figure 5.2: Exemplary continuous beam Such a beam can be represented as a bar structure in which each span is a bar of a given effective length Lef f and defined geometrical and material properties. It is assumed that each bar spans between two nodes and that the bar is supported in the nodal points only. The static scheme of the beam is shown in Fig. 5.3. Figure 5.3: Static scheme of the beam According to Sec. 4.3.2, location and amount of reinforcement has the influence on the stiffness of cross-sections and consequently on the distribution of internal forces. Therefore, reinforcement is assumed from the beginning of the computational process and as a result four types of sections are distinguished. Definition of sections is provided in Fig. 5.4. For a beam with a greater number of spans and different reinforcement layouts the amount of types of section may be greater. Figure 5.4: Material and geometrical characteristics 5.2. MATERIAL MODEL 5.1.2 45 Discretisation To perform the finite element analysis, the beam has to be discretised, i.e. divided into a finite number of sub-elements. In this model the beam is divided into the sub-elements, referred to as segments, of the same length l and in such a way that each ith segment includes cross-sections of one type only; the type of the ith segment is denoted j = t(i). Thus, one can conclude that the geometry of the beam and the reinforcement layout should be defined with the precision not greater than the precision of the definition of the segments, i.e. if the length of the segment is assumed to be l, then the length of the beam and the length and location of reinforcement has to be a multiplication of l (Lef f = n · l). Figure 5.5 shows the static scheme of the beam after discretisation. Figure 5.5: Static scheme of the beam after discretisation The segment has the same properties (in a statical sense) as a bar: it has a given length, a start and an end node, and defined geometrical and material characteristics. At the beginning of the computational process each ith segment has the stiffness of the j th type of cross-section it includes, determined for the the phase I (Bt(i) = Bj = EII,j ). 5.2 Material model The material model for concrete is chosen to be composed of the MC’s models: the model presented in Fig. 2.11a and described in detail in Sec. 2.2.1, to represent the reinforced concrete σ– relationship in compression and the model presented in Fig. 2.12a and described in detail in Sec. 2.2.1 for tension, for stresses ≤ fctm , beyond which cracking is assumed and cooperation of concrete in section is neglected. The input concrete material model is shown in Fig. 5.6a. The material model for steel in chosen to be an EC2 model presented in Fig. 2.13c and described in detail in Sec. 2.2.2. The steel is considered to be an isotropic material, i.e. its characteristic in tension and compression is the same. The input steel material model is shown in Fig. 5.6b. The strengths of concrete can be taken as characteristic values fck , fctk , design values fcd , fctd or mean values fcm , fctm , depending on the nature of the problem being solved. The ultimate strains for characteristic, designs and mean values are cu = 3.5h and ctu = 0.15h for compression and tension, respectively. The strengths of steel can be also taken as either characteristic (ftk = fck and fyk ) or design (ftd = fcd and fyk ) values. The ultimate strain cu = ctu = u depends on the class of steel, and for characteristic values is equal to 25h, 50h and 75h for class A, B and C, respectively. Note that an arbitrarily complex material model can be used without depriving the model of its computational performance. 46 CHAPTER 5. NUMERICAL MODEL (a) for concrete (b) for steel Figure 5.6: Chosen σ– relationship 5.3 Cross-section model The cross-section model (M –κ relationship) is derived separately for each type of crosssection. To find one point of the M –κ diagram a numerical procedure of finding the strain distribution in the section is performed. It must be noted that there is a possibility that each section can be subjected to either a positive or negative bending moment, i.e. compression can occur in either top or bottom fibers1 . 5.3.1 Curvature The procedure is carried out in increasing steps for a constant increment of curvature ∆κ. A geometrical interpretation of the curvature is a tangent of the inclination angle α of the strain line (a) representing the strain distribution as a function of a vertical location a ∈ [0, h]. The strain line inclination angle must be assumed as α ∈ [−90◦ , 90◦ ] to determine M –κ values within both positive and negative range of M –κ set (Fig. 5.7). (a) κ > 0 (b) κ < 0 Figure 5.7: Geometrical interpretation of curvature For the assumed κk the location of a strain line is determined in iterations in such a way that there is pure bending in the section (N = 0). The bisection method is directly applicable, because the strain distribution in a section is a monotonic function with a root within the range of interest. It is also the simplest and the most stable iterative method. Other iterative methods include linear interpolation methods (the regula falsi method and the secant method) or the Newton–Raphson method 2 . 1 Conventionally, a positive bending moment denotes tension of bottom fibres while negative their compression. 2 also known as Newton method or tangent method 47 5.3. CROSS-SECTION MODEL The group of linear interpolation methods is similar to bisection method, but although they all assure convergence of the solution, the former reach it faster. i.e. in a smaller number of iterations, thanks to introduction of the further knowledge of the function. Choosing between the linear interpolation and the Newton method it is advised to think what is easier to calculate for a given function – its value or the value of its derivative. Because the function to analyse is the function of axial force N (a) (as formulated in Eq. 5.3) and taking into account the problems of convergence in the Newton method, the linear interpolation methods, or at least the bisection method, seem to be better solution in this case. The procedure of finding the strain line is presented on the example of the bisection method. Two parameters are required to determine the strain distribution in the section. For a known curvature κ the values of strains along the height of the section can be determined assuming either the value of strain at one edge or the location of the neutral axis x ((x) = 0). In the first case, when the value of strain is assumed – let it be the maximum strain on top edge tmax – the strain distribution function has a form: (a) = κa + tmax . (5.1) In the other case, this function has a form: (a) = κ(a + x). (5.2) In each approach addition and multiplication is needed to find one value of strain, so for a defined finite number of strains m to be calculated both approaches require the same amount of operations, i.e. 2m operations. As the strain distribution is needed to determine the stress distribution, the choice of the approach is based on the convenience it will provide in the next steps of the process. However, both the values of the outermost strains and the location of the neutral axis are necessary for further computations. The choice can be arbitrary, so let us choose that the procedure will be presented for the second approach. In the first step, the curvature κ1 = ∆κ is assumed. The boundaries for bisection are the cases when x = 0 and x = h, for which the resultant forces have the opposite signs. These two boundary forces can be denoted N |x=0 = N01 and N |x=h = Nh1 . N0k and Nhk always have the opposite signs. Thus, in the first iteration the neutral axis is assumed to lie in the middle of the section, i.e. x11 = h2 . The top and the bottom strains are equal (t,11 = b,11 ). The resultant axial force N11 is calculated. If N11 6= 0 then it is either a compressive or tensile force. The bisection procedure must continue within the range of (0, h ) if N01 · N11 < 0 or ( h2 , h) if Nh1 · N11 < 0, and then again and again, until in the F th 2 iteration there is an x1F found for which N1F = 0. The x1F is the location of the neutral axis for the first step. A graphical representation of that procedure is presented in Fig. 5.8. The process is preformed for the next steps: κ2 = κ1 + ∆κ, κ3 = κ2 + ∆κ, ..., κk = κk−1 +∆κ = k∆κ, until after K steps, for κK the ultimate value of strain (compressive or tensile) is reached on one edge, which refers to reaching the ultimate value of curvature κu . The following steps of the process are schematically depicted in Fig. 5.9. 5.3.2 Bending moment The values of the internal forces acting in the section are computed with integration of the function of stress distribution in the section σ((a)) corresponding to this strain distribution acc. to the chosen concrete and steel material models. Although the chosen 48 CHAPTER 5. NUMERICAL MODEL (a) curvature κ1 (c) first iteration (b) boundary forces N01 and Nh1 (d) second iteration (e) last iteration Figure 5.8: Determination of strain distribution in cross-section with bisection method (a) first step (b) intermediate step (c) last step Figure 5.9: Distribution of strains in cross-section for different levels of loading material models are defined with the analytically differentiable functions, application of numerical integration is chosen because of its flexibility – a material model of any complexity can be used in further modifications of the program, including models obtained in laboratory tests (presented in a form of set of measurements). For a known strain distribution function in the f th iteration of the k th step, kf (a), a stress distribution σkf (kf (a)) can be derived based on the assumed material model. If the stress in concrete is represented in a form of a density of load as a function of location, i.e. b · σc,kf (kf (a)) and the stress in ith of Ns groups of reinforcing steel of the area Asi and location asi as a resultant force σsi,kf (kf (asi )) · Asi , then the axial force acting in the section can be calculated from the formula: Nkf = b Z h 0 σc,kf (kf (a)) d a + Ns X σsi,kf (kf (asi ))Asi (5.3) i=1 and the corresponding bending moment from the formula: Mkf = b Z h 0 (xkf − a)σc,kf (kf (a)) d a + Ns X (xkf − a)σsi,kf (kf (asi ))Asi . (5.4) i=1 Analytic integration can be efficiently replaced with numerical integration in which instead of summation of the infinitesimal increments, an approximate method of the 49 5.3. CROSS-SECTION MODEL finite summation is performed. The values of stresses are derived for a finite number m of strains at the height of the section at constant distances ∆a. Precision of numerical integration depends on the number of sub-intervals (value of m/length of ∆a) and method of σ((a)) function approximation in subintervals. The most popular methods of function approximation are the Rectangular Rule (interpolation with a 0th order polynomial) and the Trapezium Rule (interpolation with a 1st order polynomial). Let us divide the height of the section into m intervals of width ∆a and denote each ith interval as [ai , ai+1 ]. In the Rectangular Rule one value of the stress function is required in each interval, lat us say σ((ac,i )), such that ai ≤ ac,i ≤ ai+1 . It is equivalent to the assumption that ∀a ∈ (ai , ai+1 ) σ((a)) ≃ σ(ac,i ). The function can be now easily integrated, as: Z h σ((a)) ≃ ∆a 0 m X σ((aci )). (5.5) i=1 The ac,i can be any value from [ai , ai+1 ] interval. If ac,i = ai , then it is a left or lower Riemman sum, if ac,i = ai+1 it is a right or upper Riemman sum and if ac,i = 12 (ai + ai+1 ) it is a middle Riemman sum. The value of ac,i should be chosen in such a way that provides optimum performance (maximum possible precision at minimum possible computational effort). In the Trapezium Rule two values of stress function are needed: σ((ai )) and σ((ai+1 )). Then the stress function is interpolated with a linear function and the integral can be calculated as: Z h σ((a)) ≃ ∆a 0 m X σ((ai )) + σ((ai+1 )) 2 i=1 . (5.6) Of course, the function can be approximated with a polynomial of any desired order or any other easily integrable function, however, their use in numerical integration becomes more and more questionable along with the increasing complexity of the approximating/interpolating function. Moreover, linearisation of the function for very small intervals provides the greatest flexibility in the choice of material model as an input data without the necessity of adjustment of the approximation method to best suite the material model function. The process of numerical integration of the stress distribution function is presented for the middle Riemman sum. The axial force is calculated in each f th iteration of each k th step according to the formula: Nkf = b m X σc,kf kf i=1 1 i− ∆a 2 ∆a − Ns X σsi,kf (kf (asi ))Asi . (5.7) i=1 The corresponding bending moment, calculated after the last iteration of the k th step and denoted as Mk , is calculated from the formula: Mk = b m X i=1 xk − i − 1 ∆a σc,k k 2 i− 1 ∆a 2 ∆a + Ns X (xk − asi )σsi,k (k (asi ))Asi . i=1 (5.8) The operation of finding of the internal forces with numerical integration is presented in Fig. 5.10 for an exemplary stress distribution. 50 CHAPTER 5. NUMERICAL MODEL (a) cross-section (b) strains (a) (c) stresses σ((a)) (d) forces M and N Figure 5.10: Determination of internal forces in section with numerical integration 5.3.3 Moment–curvature relationship A set of pairs of Mj,k –κj,k values create a M –κ relationship – the cross-section model for the j th type of section. The M –κ diagrams are prepared in both positive and negative range of bending moments (curvatures), but it must be remembered that the M –κ diagram is symmetrical (with respect to the origin of coordinate system) only for symmetrical cross-sections. An exemplary M –κ diagram obtained in that method is presented in Fig. 5.11. Figure 5.11: Exemplary numerically-determined M –κ diagram Mu denotes the ultimate bending moment – the maximum moment which the section can transfer. There are two independent Mu values for each section: Mu,pos which signifies the moment resistance of the section against positive bending moment and Mu,neg the moment resistance against negative bending moment. Mu values for a given type of sections can be determined from the Mj,k –κj,k set according to the conditions that Mu,pos,j,k = max{Mj,k } and Mu,neg,j,k = min{Mj,k }. It should be noted that such a solution has one drawback – a constant increment of curvature does not necessarily result in a constant increment of bending moment, so the M –κ pairs cannot be prepared for all desired values of bending moments (at least 5.4. STIFFNESS 51 not that easily). This will result is a certain approximation error in static calculations in which the values of bending moments, not curvatures, are needed. Nevertheless, the simplicity of determination of the bending moment value with a known curvature instead of doing otherwise was decided to be more computationally convenient, having in mind considerations from Sec. 5.4.1. 5.4 Stiffness 5.4.1 Stiffness of cross-section The basis of this thesis is the notion that the stiffness of the section is a function of bending moment acting in this section, i.e. B = B(M ). In a mathematical sense, the value of the stiffness of the section Br after the rth loading step with the bending moment Mr is the first derivative of M (κ) function, i.e. Br = M 0 (κr ). In a geometrical sense it is tangent of the inclination angle β of the line tangent to the M –κ function in a given (κr ,Mr ) point, as presented graphically in Fig. 5.12. Figure 5.12: Geometrical interpretation of section’s stiffness For the purpose of this model, one should be able to determine the value of the stiffness for every stress state occurring in every segment of the analysed member. However, the M –κ function derived in the procedure described in Sec. 5.3 is only a finite set of M –κ pairs. Therefore, some approximation has to be made in the stiffness determination process. Determination of the cross-section stiffness is a two-step process. First, for a given bending moment Mr acting in the cross-section the corresponding curvature κr must be found, so that (κr , Mr ) point is known, and then a derivative of M (κ) in that point, i.e. M 0 (κr ) must be determined as the stiffness B(Mr ) = Br . The easiest solution is preparation of a very “dense” diagram, i.e. M and κ values are determined for a very “small” increment of curvature ∆κ. The terms “dense” and “small” should be understood in the context of the precision of results: if the maximum difference between two adjacent values of bending moments in the M –κ set is smaller than the precision of the result values of bending moments obtained in a static analysis, then such a solution is satisfactory. Then the value of active bending moment Mr is rounded to the nearest value Mk ∈ {Mk }, for which κk is known. The rounding is proceeded in such a way that the values of Mk1 , Mk2 adjacent to Mr are found, i.e. Mk1 = max Mk : Mk < Mr and Mk2 = min Mk : Mk > Mr , and Mr is taken as Mr = Mk1 if |Mr − Mk1 | < |Mr − Mk2 | or Mr = Mk2 otherwise. 52 CHAPTER 5. NUMERICAL MODEL The value of derivative can be then determined with one of the finite difference methods, which enable to replace derivative expressions with approximately equivalent difference quotients. The basis for this group of methods is the assumption that the domain of the function is uniformly discretised – it this case that the spacing between the subsequent κk is equal (∆κ = const.). The finite difference may be of any order – the order determines the number of points (or (κk , Mk ) pairs) to be used to calculate the value of the derivative. The most commonly considered are the forward, backward and central differences. For a (κk , Mk ) pair the forward difference requires the knowledge of the (κk+1 , Mk+1 ) pair such that κk+1 > κk and κk+1 − κk = ∆κ, the backward difference the knowledge of the (κk−1 , Mk−1 ) pair such that κk−1 < κk and κk − κk−1 = ∆κ, while the central difference existence of two pairs, (κk− 1 , Mk− 1 ) and (κk+ 1 , Mk+ 1 ), such that κk− 1 < κk < κk+ 1 and 2 2 2 2 2 2 κk+ 1 − κk− 1 = ∆κ. The central difference to be used for {Mk , κk } obtained in this model 2 2 must be calculated for (κk−1 , Mk−1 ) and (κk+1 , Mk+1 ), such that κk−1 < κk < κk+1 and κk+1 − κk−1 = 2∆κ. Provided that Mk is the value of bending moment from {Mk } nearest to Mr , the stiffness Br is calculated with a forward, backward and central finite difference method as: Br = M 0 (κk ) ≃ Mk+1 − Mk , ∆κ (5.9) Br = M 0 (κk ) ≃ Mk − Mk−1 , ∆κ (5.10) Mk+1 − Mk−1 . (5.11) 2∆κ In a geometrical sense the stiffness Br determined with this operation is a tangent of the inclination angle of the line crossing points {(κk+1 , Mk+1 ), (κk , Mk )}, {(κk−1 , Mk−1 ), (κk , Mk )} and {(κk+1 , Mk+1 ), (κk−1 , Mk−1 )} for the forward (Fig. 5.13a), backward (Fig. 5.13b) and central (Fig. 5.13c) finite difference method, respectively. Br = M 0 (κk ) ≃ (a) forward difference (b) backward difference (c) central difference Figure 5.13: Stiffness of section after rth step with first-order finite difference methods A different idea is an operation of curve fitting – an approximation of the set with an analytically differentiable function, preferably a polynomial. The operation of approximation with a differentiable function enables to proceed with analytic computations, because differentiation of polynomials is not only analytically possible but also very simple. The approximation can be performed for the whole set at once or segmentally. 5.4. STIFFNESS 53 The first method of curve fitting is interpolation with a polynomial. The interpolation methods assume that the interpolating function – let us denote it as W (κ) – for a chosen set of values of the interpolated function – in this case M (κ) – has the same values, i.e. as the the M (κ) function is a set of points, {Mk , κk }, interpolation with a polynomial W (κ) is an operation of finding such a polynomial that ∀k W (κk ) = Mk . If the {Mk , κk } has K elements, then the W (κ) will be a polynomial of the K − 1 order. It must be noted, that the higher the order of the polynomial (the greater the number of the points), the less stable becomes the interpolating function3 . Moreover, as the value of κr for a given Mr must be known, the procedure of root finding of the W (κ) polynomial has to be performed. This is analytically possible only for polynomials of at most 4th order. For higher order polynomials the iterative methods for non-linear equations root finding must be used. Hence, a choice of interpolation is a compromise between precision and computational convenience. Another solution can be derivation of an approximating function. In that sense, approximation is a standard optimisation problem of minimalisation of the error function, or bluntly speaking, minimalisation of the differences between the corresponding values of the approximated (M (κ)) and the approximating (A(κ)) function. The approximation method is deprived of the Runge’s phenomenon problem. The process of interpolation/approximation can be also performed locally, for a defined neighbourhood of the desired Mr value. The method which can be considered a half-way solution between the interpolation and approximation of the whole set is spline interpolation, a form of interpolation where the interpolant is a special type of piecewise polynomial called a spline. The spline is composed of a set of polynomials, preferably of a low degree, defined over the subintervals of the interpolated function, in such a way that the polynomials have the same values in the joining points, called knots. The proper use of splines minimises the risk of polynomial oscillation and preserves the characteristic of the interpolated function. For the purpose of this model it is demanded that the spline was a smooth curve of the smoothness at least C 1 , i.e. the spline function and its derivative are continuous. It is possible to achieve using at least a quadratic spline (the spline of the 2nd order). However, to assure stability of the spline, the special type of spline called a natural cubic spline would be more convenient. The operation of differentiation of the spline is very similar to that used in typical interpolation and approximation. The only additional condition, which may be regarded as a drawback of this method, is a need of checking between which knots is the desired value of Mr to indicate which segment of the spline to consider. 5.4.2 Stiffness of segment The stiffness determined in the process described in Sec. 5.4.1 is the stiffness of a single cross-section. It was stated in Sec. 5.1 that initially each segment is given the stiffness of the cross-section type j it includes as a mean stiffness, i.e. Bmean,i = Bt(i),0 = Bj,0 , and this stiffness is changing directly according to the proper cross-section model. As long as there is no crack in the segment, it is valid. However, once the crack is formed, i.e. the value of the active bending moment Mr = Mcr , the stiffnesses of particular cross-sections in a segment differ. The stiffness of the cracked cross-section is degraded from BI towards BII . Thus, the mean stiffness of the segment must be some average of the stiffnesses of 3 The problem of oscillation of the interpolating polynomial is known as the Runge’s phenomenon. 54 CHAPTER 5. NUMERICAL MODEL the cracked and uncracked sections. The tension stiffening effect coefficient ξ, given by Eq. 4.22, can be used, and the mean stiffness after the rth loading step Bmean,t(i),r can be determined as: Bmean,t(i),r = (1 − ξi,r )Bt(i),r (5.12) where Bt(i),r is the stiffness determined from the Mj –κj diagram and ξi,r is the tension stiffening effect coefficient for the ith segment for the bending moment Mi,r obtained after the rth loading step. With an increasing load, the stiffness will be decreasing till, for the ultimate moment Mu , it will degrade to 0 (M 0 (κMu ) = const.) It is when the plastic hinge is formed. It must be remembered that once the stiffness of a segment Bt(i),r is reduced after the rth loading step, it cannot be “regained” if the value of the bending moment is reduced in that segment in the following steps, nor can be the (1 − ξi,r ) coefficient. This fact can be mathematically defined as: ξi,r = max (ξ(Mi,r ); ξi,r−1 ) Bt(i),r = max 0; min Bt(i),r ; Bt(i),r−1 5.5 (5.13) . (5.14) Static analysis with FEM In is not possible to determine the values of internal forces in a statically-indeterminate structure directly with simple equilibrium conditions. A method must be used which takes into consideration stiffnesses of all structural components. A displacement method is a very good choice. In a matrix formulation, the method can be easily implemented to be used in computer calculations. The numerical version of this method is referred to as the displacement finite element method. Displacement finite element method is a method of finding a displacement function of the structure with defined stiffness of the components and external loading pattern. Computational procedure is a two-phase process. In the first phase the structure is descretised, i.e. divided into the finite number of bars, each bar having two nodes and the bars joining in the nodes. In case of the exemplary beam analysed in this chapter, as shown in Fig. 5.5, these bars are referred to as segments. In the second phase a global stiffness matrix K is formed as a result of definition of relationship between generalised forces (in a form of a matrix of nodal loads P ) and generalised displacements of nodes (in a form of a nodal displacements matrix u), which in a linear-elastic approach leads to a matrix formulation of the system of linear equations: Ku = P. (5.15) In the non-linear analysis the elements of the stiffness matrix K are no longer constant as they are functions of displacements u (stiffness is a function of bending moment, while bending moment is a function of displacement), and the resulting system is a system of non-linear equations of the form: K(u)u = P. (5.16) 5.5. STATIC ANALYSIS WITH FEM 55 Solution of such a system of non-linear equations must be found in steps and with iterative search in each step. Thus, the methods of numerical solution of non-linear problems are ofter referred to as incremental–iterative methods. The approximated solution can be also found in a process of linearisation of the problem: the global stiffness matrix can be calculated in steps, for very small increments. In each step it can be then assumed that the problem is no longer non-linear and the stiffness is independent on the deflection. Precision of such a solution increases (in comparison to iterative method) as the value of increment decreases. The incremental approach is presented in this section. 5.5.1 Static calculations Computations are proceeded with an incremental method. A uniformly distributed load q is applied to the whole member in constant increments ∆q. Initially, the beam has a static scheme as shown in Fig. 5.5 and is loaded with a load q1 = ∆q (Fig. 5.14). The value of the bending moment is determined in each node, and distribution of bending moments is interpolated. As the decisive bending moment for each segment, Mi , the maximum bending moment in this segment is understood. Figure 5.14: Static calculations. Step 1 After this step the stiffness of each segment Bt(i),0 is modified according to the relevant Mj –κj relationship and the procedure is being repeated for the next increments q2 = ∆q, q3 = ∆q, ..., qr = ∆q, each time for corrected stiffness value Bt(i),r−1 = B(Mi,r−1 ). The total value of bending moment for each segment i after the rth loading step is the sum of P all bending moments obtained in all the previous steps, i.e. Mtot,r = rm=1 Mm . Finally, in the Rth step, the total bending moment Mtot,R in one node reaches the ultimate value Mu , which is equivalent to depletion of the load-bearing capacity of this section and formation of a plastic hinge (Fig. 5.15). These R steps are the “first cycle”. As it is known, formation of the plastic hinge does not signify failure of the member, because although the section can no longer transfer additional bending moments, the moments can be redistributed and the section can undergo further deformations as long as they do not exceed their ultimate values. In the model, a static scheme of the beam is changed (a structural hinge is introduced in the place of the plastic hinge), and the beam can undergo another loading “cycle”. The “correcting” bending moments Mdif f are introduced in the section where plastic hinge is formed to substitute the process of moment capacity decrease in section acc. to the material model assumed (Fig. 5.16). The value of the additional moments is equal to the difference between the value of the bending moment capacity of the section before and after the given loading step. 56 CHAPTER 5. NUMERICAL MODEL Figure 5.15: Static calculations. Formation of first plastic hinge Figure 5.16: Static calculations. Step R1 + 1 57 5.5. STATIC ANALYSIS WITH FEM The process continues until the number of structural hinges exceeds the degree of statical indetermination of the beam. Each “cycle” c terminates in the step Rc up to the P c final step RC of the final cycle C. The final value of the load qtot,RC = R m=1 qm = Rc ∆q, leading to the formation of the last plastic hinge, is the ultimate load qu for the beam (Fig. 5.17). Figure 5.17: Static calculations. Formation of the last plastic hinge Formation of the second plastic hinge is equivalent to the global loss of stability and failure of the member. Hence, the total number of cycles C = 2, R2 is the last step of the whole procedure and the load qtot,R2 = R2 ∆q is the ultimate load for the exemplary beam. 5.5.2 Bending moment distribution Solution of the exemplary beam with a finite element method is based on the Euler– Bernoulli partial differential equation for bending with application of Galerkin’s method of weighted residuals to develop the finite element formulation and the corresponding matrix equations [1]. A deformation function (deformation line) can be derived with a finite element method from a displacement function which assigns displacements to the nodes. In the displacement method a displacement function is a (1xDOF ) vector, where DOF is a global number of degrees of freedom (number of nodes x degrees of freedom/node). It can be calculated after transformation of Eq. 5.15 as: u = K −1 P. (5.17) The deformation of a beam must have continuous slope and continuous deflection at any two neighbouring elements. In case of the analysed beam any two neighbouring beam elements (segments) have common deflection vi and slope φi at the shared nodal point i. Based on the Bernoulli hypothesis of plane sections it is concluded that slope is the first 58 CHAPTER 5. NUMERICAL MODEL derivative of deflection in terms of location along the beam x, φ = dd xv . As there are four nodal variables in each beam element (4 degrees of freedom), the deformation line can be described with a 3rd order polynomial function: v(x) = c0 + c1 x + c2 x2 + c3 x3 (5.18) The notion of this fact is crucial for derivation of bending moment function. Firstly, the stiffness matrices Ki are prepared separately for each segment. A segment with stiffness Bj which spans over two nodes – the first node being denoted as P and the second as K – and with 2 degrees of freedom in each node (vertical displacement v[m] and rotation φ[rad]) has a total number of 4 degrees of freedom. 4 values: forces VP and VK and moments MP and MK are found for 4 cases: unit vertical displacement of node P and K (vP = 1 and vK = 1) and unit rotation of node P and K (φP = 1 and φK = 1) with all the remaining displacements taken as 0. This is possible with application of the Hermitian shape functions, the functions which are derived from the deformation function given by Eq. 5.18 for the appropriate boundary conditions, i.e. vP = 1, φP = 1, vK = 1 and φK = 1, respectively, and have the form: 3x2 2x3 + 3 l2 l 2 2x x3 H2 (x) = x − + 2 l l 3x2 2x3 H3 (x) = 2 − 3 l l 2 x x3 H4 (x) = − + 2 l l H1 (x) = 1 − (5.19) (5.20) (5.21) (5.22) The analysed sub-cases with corresponding shape functions and resultant V and M forces are shown in Fig. 5.18. The resulting deformation function within one element is a linear combination of shape functions, decribed by the formula: v(x) = H1 (x)vP + H2 (x)φP + H3 (x)vK + H4 (x)φK (5.23) In such a way the deformation function can be derived for the whole length of the element when exact values of displacements are introduced. The calculated values of the forces are the second derivatives of the shape functions for the assumed stiffness and create the stiffness matrix of the segment Ki , of the form: " Ki = kP,P,i kP,K,i kK,P,i kK,K,i # (5.24) where kP,P,i , kP,K,i , kK,P,i and kK,K,i are the stiffness sub-matrices which represent the reaction of the unit-displacement force applied in the node P on node P , in the node P on node K, in the node K on node P and in the node K on node K, respectively, of the forms [19]: 59 5.5. STATIC ANALYSIS WITH FEM (a) unit vertical displacement of P node for H1 (x) (b) unit vertical displacement of K node for H3 (x) (c) unit rotation of P node for H2 (x) (d) unit rotation of K node for H4 (x) Figure 5.18: Determination of elements of stiffness matrix of segment Ki acc. to [19] 60 CHAPTER 5. NUMERICAL MODEL 12Bt(i),r l3 kP,P,i = kP,K,i = VP,i |vP =1 VP,i |φP =1 MP,i |vP =1 MP,i |φP =1 VP,i |vK =1 VP,i |φK =1 kK,P,i = MP,i |vK =1 MP,i |φK =1 VK,i |vP =1 VK,i |φP =1 MK,i |vP =1 MK,i |φP =1 = kK,K,i = VK,i |vK =1 VK,i |φK =1 MK,i |vK =1 MK,i |φK =1 6Bt(i),r l2 (5.25) 6Bt(i),r 4Bt(i),r − 2 l l 12Bt(i),r 6Bt(i),r − 2 − l3 l = 6Bt(i),r l2 2Bt(i),r l (5.26) T = kP,K,i (5.27) 12Bt(i),r l3 6Bt(i),r l2 6Bt(i),r l2 l − = (5.28) . 4Bt(i),r The complete stiffness matrix of the element is a (4x4) matrix. Formation of a global stiffness matrix K is carried out in a process of aggregation. The aggregation is a process of addition of the matrices of segments, Ki , into the appropriate places of the global matrix K. Aggregation ensures equal displacements of the nodes belonging to more than one segment. For a continuous beam, being an example in this model, a Ki is a stiffness matrix for a segment i with the first node P = i and the second node K = i + 1, and not more than two elements joining in one node. The sub-matrices kP,P,i , kP,K,i , kK,P,i and kK,K,i have the same form for all segments, with only the value of stiffness Bt(i),r changing between segments and in loading steps. The beam has 11 segments and 12 nodes, with 2 degrees of freedom in each node, so the total number of degrees of freedom is DOF = 24. The resultant global stiffness matrix K can be presented in a (24x24) matrix form, with all values of forces corresponding to all the degrees of freedom, or in a simplified way in a (12x12) matrix form with sub-matrices corresponding to the subsequent nodes, in a form: k1,2,1 k2,2,1 + k2,2,2 k3,2,2 .. . 0 k2,3,2 k3,3,2 + k3,3,3 .. . 0 0 k3,4,3 .. . 0 0 0 0 0 0 0 k1,1 k2,1 0 = ... k1,2 k2,2 k3,2 .. . 0 k2,3 k3,3 .. . 0 0 k3,4 .. . ... ... ... .. . 0 0 0 0 0 0 0 k11,10 0 k1,1,1 k2,1,1 0 K= ... 0 0 0 0 0 0 0 0 ... ... ... .. . 0 0 0 = 0 k11,12,11 k10,11,10 k11,10,10 k11,11,10 + k11,11,11 0 k12,11,11 k12,12,11 0 k11,12 k10,11 k11,11 k12,11 k12,12 0 0 0 (5.29) 61 5.5. STATIC ANALYSIS WITH FEM where ka,b,c is a sub-matrix representing the reaction of the forces acting in the node a on the node b, calculated for segment c, or in a graphical form in Fig. 5.19a. (a) global stiffness matrix K (b) vector of nodal loads P Figure 5.19: Graphical presentation of matrices formation in aggregation process It must be emphasised that for a continuous beam only the entries of the matrix laying on its diagonal (hatched entries) are being modified in the aggregation process and that the resultant matrix is a band matrix. Figure 5.20: Determination of elements of matrix of nodal loads Pi The matrix of loads represents the loads to which the beam is subjected in a form of generalised nodal loads, which are statically equivalent to the external load acting on the beam. It is a (1xDOF ) vector. For each segment a pair of transverse forces and bending moments is obtained. For a single segment subjected to the uniformly distributed load qr the resultant nodal loads are presented in Fig. 5.20. The resultant vector of nodal loads has a form: 62 CHAPTER 5. NUMERICAL MODEL VP,i |qr MP,i | " # qr pP,i Pi = = pK,i VK,i | qr MK,i |qr qr l = 2 qr l 2 − 12 qr l 2 qr l 2 12 T . (5.30) For a general arbitrary loading pattern q(x) such a segment would produce a load vector of a form: H1 (x) H (x) Pi = q(x) 2 d x. H3 (x) 0 H4 (x) Z l (5.31) The aggregation process can be carried out to form a load vector for the whole beam and boundary conditions can be introduced for the nodes where supports are defined. The known shear force and or bending moment can be included in the system force vector; otherwise, they remain unknown. The resultant load vector has a form: V1,1 M 1,1 V2,1 + V2,2 M +M 2,1 2,2 . . P = . V11,10 + V11,11 M 11,10 + M11,11 V12,11 (5.32) M12,11 where Va,c is an entry representing the equivalent nodal force in the node a calculated for segment c subjected to the load qr , or in a graphical form in Fig. 5.19b. It must be noted that for the range of work where κr > κ|Mu additional bending moments Mdif f are introduced in the node where plastic hinge was formed, to reflect the softening nature of concrete. The last step is introduction of boundary conditions, i.e. conditions of support. Introduction of boundary conditions assures zero displacements in the points where supports are defined. There is a number of ways to achieve this goal. One of them is limiting or depriving the node of the possibility of displacement – depending on the support a vertical displacement or rotation – by introducing a very big force responsible for causing this displacement, e.g. multiplying the value by 1020 . In case of the exemplary beam, the hinge supports are introduced in nodes 1, 7 and 12. This is equivalent to the condition that v1 = 0, v7 = 0 and v12 = 0. Therefore the vertical forces responsible for causing this displacements must be increased to the “computer’s infinity”. In the global stiffness matrix only the ki,i entries are modified to account for the direction of dicplacement being blocked. In that case the vertical forces causing unit displacements of the supported nodes in the k1,1 , k7,7 and k12,12 matrices are modified as follows: 63 5.5. STATIC ANALYSIS WITH FEM VP,1 |v1 =1 · 1020 k1,1 = kP,P,1 = VP,1 |φ1 =1 MP,1 |v1 =1 k7,7 = kK,K,6 + kP,P,7 = MP,1 |φ1 =1 (5.33) VK,6 |v7 =1 + VP,7 |v7 =1 · 1020 VK,6 |φ7 =1 + VP,7 |φ7 =1 MK,6 |v7 =1 + MP,7 |v7 =1 MK,6 |φ7 =0 + MP,7 |φ7 =1 (5.34) k12,12 = kK,K,12 = VK,12 |v1 =1 · 1020 VK,12 |φ1 =1 MK,12 |v1 =1 MK,12 |φ1 =1 . (5.35) In case of the nodal loads vector these boundary conditions must signify that there is no possibility of bending moment transfer in the external nodes where hinge supports are introduced, i.e. M1 = 0 and M12 = 0. Another solution, coming directly from the classic displacement method, is to neglect the displacements which values are known. In the FEM it means crossing-out columns and rows which corresond to zero displacement. In the analysed case the rows and columns in K, u and P matrices corresponding to vertical displacements in nodes 1, 7 and 12 should be crossed out – in 24x24 presentation these are columns no. 1, 13 and 23 in K matrix and rows no. 1, 13 and 23 in all matrices. Finally, the u vector can be calculated. The u vector returns a finite sequence of displacements of a form: v1 φ1 v2 = φ2 . .. u1 u2 u= .. . u12 v12 (5.36) φ12 where ua is a vector of displacements for a node a which includes a vertical displacement va and a rotation φa in the node, when the complete matrix is analysed (then it should be v1 = v7 = v12 = 0), or in a condensed form, with only unknown displacements provided. In computer calculations the K matrix is not actually inversed (it is computationally uneconomiacal). Solution of the matrix form of the system of equations is proceeded with one of the methods of solutiton of linear equations systems, e.g. Gaussian elimination, LU decomposition, Hausholder or Gram-Schmidt orthonormalisation process, Cholesky decomposition (as K is a symmetrical, positive-definite matrix) or with simple-type iterative methods (Chebyshev method). The values of internal forces along the whole memeber can be obtained by summation of forces resulting from the displacements of the nodes and forces resulting from external loads to which the memeber is subjected, in any cross-section. This must be performed for a local case, so for each segment. The values if internal forces in an every node i must 64 CHAPTER 5. NUMERICAL MODEL be equal regardless of the choice of segment i (for which node i = P ) or i − 1 (for which node i = K). For a single element the forces resulting from the displacements of the nodes are the sums of the products of multiplication of the forces resulting from the unit displacements of the nodes and displacement of the corresponding nodes obtained from Eq. 5.17 in a form of the vector u, given by Eq. 5.36, and the forces resulting from the action of the external load. In case of a segment, as analysed in this thesis, these forces are obtained by addition of five cases: four representing the unit-displacement forces, presented in Fig. 5.18 and the fifth being a solution of the substitute static scheme of the segment under the uniformly distributed load qr , which is presented in Fig. 5.20. For segments in which a plastic hinge was formed, the additional bending moments Mdif f must be considered in calculations. The border values of bending moments in the segment i with nodes i and i + 1 can be calculated acc. to the superposition rule, for the nodes i and i + 1, respectively: Mi = MP |vP =1 vi + MP |vK =1 vi+1 + MP |φP =1 φi + MP |φK =1 φi+1 + MP |qr (5.37) Mi+1 = MK |vP =1 vi + MK |vK =1 vi+1 + MK |φP =1 φi + + MK |φK =1 φi+1 + MK |qr (5.38) The same can be done with shear forces: Vi = VP |vP =1 vi + VP |vK =1 vi+1 + VP |φP =1 φi + VP |φK =1 φi+1 + VP |qr (5.39) Vi+1 = VK |vP =1 vi + VK |vK =1 vi+1 + VK |φP =1 φi + + VK |φK =1 φi+1 + VK |qr . (5.40) It must be noted that since the deformation function is given by a differentiable function and since internal forces are derivatives of the deformation line, the resultant distribution of internal forces can be obtained as a continuous, not a discrete solution. Chapter 6 Summary 6.1 Conclusions The process of design of a structure requires a series of idealisations: idealisation of geometry, idealisation of materials and idealisation of loads. Hence, this process is actually performed on a model. The model should be as simple as possible, but reproduce the important characteristics consistent with experimental results and should be theoretically sound and numerically stable, so that reliable analysis results are obtained. In the structural analysis of reinforced concrete members four main approaches are proposed by the standards: linear-elastic analysis, linear-elastic analysis with limited redistribution, plastic analysis and non-linear analysis. In the linear-elastic analysis the values of internal forces are derived based the assumption that the material has a linear-elastic characteristic. This simplification is not much far from reality, because the values of bending moments MSd obtained with such an approach are usually beyond the level of cracking moment Mcr , so in the range where the material is in phase I and its behaviour is very close to linear. However, this values are compared to the cross-sectional moment resistance MRd determined from the ultimate limit state of flexure, as for the cross-section being in the phase III, so for a completely different stage of work. This inconsistence is the reason for uneconomic design. The notion of the plastic properties of reinforced concrete, so the ability of concrete to undergo plastic deformations without rupture, provides additional load-bearing capacity, the so-called plastic reserve, which – if applied reasonably – allows of designing of lighter, more economical elements. It also extends the understanding of the behaviour of reinforced concrete in the plastic phase, when a cracked element is subjected to irreversible strains. This notion is the basis for the non-linear methods of analysis. The linear-elastic analysis with limited redistribution is the simplest method of refinement of linear-elastic analysis results. Bending moments obtained in the linear-elastic static analysis are redistributed within a limited range from the most loaded cross-sections to other, less congested areas. The method is simple, but diifcult to control and thus questionable. The standards provide a designer with information about the plastic properties of concrete. The notions of plastic hinge formation and rotational capacity of the plastified regions are introduced. Moreover, it is emphasised that the reinforced concrete behaviour is in reality totally non-linear due to stiffness degradation as a result of material characteristics, crack formation and damage of internal structure, which finally leads to depletion of loadbearing capacity and fracture. 65 66 CHAPTER 6. SUMMARY Nevertheless, modelling of the non-linear behaviour of reinforced concrete is not an easy task. The phenomena occurring in reinforced concrete members, although repeatably observed in laboratory tests and in full-scale objects, are very difficult in mathematical description. Thus, analytic computations are impossible and, as a consequence, numerical analysis is required. Derivation of a numerical simulation of the physical phenomena is an interdisciplinary task. The process requires co-operation between the scientist/engineer and the software developer. A numerical model, derived on the basis of theoretical background of the physical phenomenon and creative application of mathematics and numerical methods, can be implemented and optimised for performance. The results obtained with such a numerical analysis can be then validated with experimental results and the applicability of the model can be evaluated by the comparison to the currently used methods. The objective of this thesis, which was to derive a stiffness-oriented numerical model for non-linear analysis of flexural reinforced concrete beams, was achieved. The literature review was carried out to prepare a theoretical background of the phenomenon of stiffness degradation in flexural reinforced concrete elements as a result of plastic behaviour of concrete and crack formation at flexure. The currently used methods of static analysis and dimensioning of reinforced concrete structures were investigated. The basic requirements given by standards, mainly Model Code and Eurocode 2, were cited. The numerical model was derived in cooperation with the supervisor of this thesis on the basis of the referred theoretical background and personal experience in the topic. The procedure of calculations was presented in a very detail, mathematical or geometrical interpretations of physical phenomena were introduced, supported with schematic graphical representations. Wherever noticed possible, simplifications and patterns resulting from the specification of the design task were emphasised. The aim of these actions was to generalise the physical problem in a way that enables the software developer, who is not necessary familiar with the analysed physical problem, to implement the model. Each time a step in the procedure could be solved in more than one way, the advantages and disadvantages of the possible methods were analysed and the applicability evaluated. Although it is believed that the software developer would have great experience in this field, such a pre-analysis was performed to show the possible problems or – in contrary – convenient solutions resulting from the nature of the design task. The whole procedure of numerical analysis is presented on a chosen example. A very simple – from a mechanical point of view – case was chosen because the objective of this model is qualitative and quantitative assesment of redistribution process. It is believed that a more complicated case could make validation of the model more difficult and bias the results of the analysis. The objective of this thesis was hence two-fold. The main aim was connected with the necessity of derivation of a unified design algorithm for non-linear analysis of reinforced concrete. The other, minor aim, was to take the floor in the topic of the future prospects of engineering. The accessibility of fast computing machines and variety of computer programming environments and numerical methods provides excellent possibilities for solution of engineering task that used to be impossible to solve. Nevertheless, beside the necessity of understanding the engineering problem there arises an urgent need of programming skills. Therefore, there are two choices for the engineer on these crossroads: either to learn programming or to establish co-operation with software developer. The second choice has always been a privilege of big commercial companies, leaving the first option the only 6.2. ACKNOWLEDGMENTS 67 choice for a common engineer or scientist. Have the fact been acknowledged that each expert is the best in their domain, both parties would gain, let alone the benefits for engineering itself. Thus, the actions should be undertaken which aim at triggering such co-operation at all levels of engineering/scientific work. The author and the supervisor of this thesis hope that it will be possible to successfully implement the derived numerical model. It would be then possible to evaluate the process of redistribution of internal forces in reinforced concrete elements and define its real magnitude and nature. The results obtained in the numerical analysis could be compared to the currently used simplified methods. Probably a more efficient and more general algorithm for concrete members design and reinforcement could be derived. 6.2 Acknowledgments To the supervisor of this thesis, PhD SEng Grzegorz Wandzik, for co-operation, guidance, knowledge and time spent helping me complete this thesis. To my husband, Michał Wróbel, for engagement, creativity, encouragement and support. 6.3 Abstract (in Polish) Proces projektowania konstrukcji, w tym konstrukcji żelbetowych, wymaga szeregu idealizacji, stąd też proces ten przeprowadza się właściwie na modelu. Model ten powinien być możliwe prosty, ale w sposób powtarzalny odwzorowywać najważniejsze charakterystyki zgodnie z wynikami badań oraz powinien być teoretycznie poprawny i numerycznie stabilny, aby uzyskane dzięki niemu wyniki pokrywały się z rzeczywistymi obserwacjami. Normy proponują cztery grupy metod analizy elementów żelbetowych: analizę liniowo-sprężystą, analizę liniowo-sprężystą z ograniczoną redystrybucją, analizę plastyczną oraz analizę nieliniową. W analizie liniowo-sprężystej wartości sił wewnętrznych otrzymuje się przy założeniu liniowo-sprężystej charakterystyki materiału. Uproszczenie to nie jest bardzo dalekie od rzeczywistości, ponieważ n.p. wartości momentów zginających MSd otrzymanych przy takim podejściu są zwykle poniżej poziomu momentu rysującego Mcr , a więc w zakresie, gdzie element znajduje się w fazie I, a jego zachowanie jest bliskie liniowemu. Jednakże wartości te w metodzie stanów granicznych porównywane są z nośnością przekroju na zginanie MRd , wyznaczoną dla stanu granicznego nośności na złamanie, jak dla przekroju w fazie III, więc w zupełnie innej fazie pracy. Ta niespójność jest powodem nieekonomicznego projektowania. Wiedza na temat plastycznych właściwości betonu zapewnia dodatkową nośność, zwaną rezerwą plastyczną, która – wykorzystywana rozsądnie – pozwala na projektowanie lżejszych, bardziej oszczędnych elementów. Wiedza ta jest podstawą nieliniowych metod analizy konstrukcji. Analiza liniowo-sprężysta z ograniczoną redystrybucją jest najprostszą z metod urzeczywistniania wyników analizy liniowo-sprężystej z uwzględnieniem plastycznych właściwości betonu. Momenty zginające uzyskane w liniowej analizie statycznej są redystrybułowane w ograniczonym zakresie z najbardziej obciążonych przekrojów w mniej wytężone obszary. Metoda ta jest prosta, ale trudna do kontrolowanie, przez co kontrowersyjna. Normy dostarczają projektantowi informacje dotyczące plastycznych właściwości betonu. Wprowadzone są informacje na temat powstawania przegubów plastycznych oraz 68 CHAPTER 6. SUMMARY zdolności do obrotu obszarów uplastycznionych. Co więcej, podkreśla się, iż zachowanie żelbetu w rzeczywistości jest całkowicie nieliniowe z powodu spadku sztywności jako efektu właściwości materiałowych, zarysowania oraz zniszczenia wewnętrznej struktury, co w konsekwencji prowadzi do wyczerpania nośności i zniszczenia. Niemniej jednak modelowanie nieliniowego zachowania żelbetu nie jest prostym zadaniem. Zjawiska zachodzące w elementach żelbetowych, choć w sposób powtarzalny obserwowane w badaniach laboratoryjnych jak i w pełno wymiarowych obiektach, są trudne w opisie matematycznym. Obliczenia analityczne są niemożliwe, stąd konieczne jest wykorzystanie numerycznych metod analizy. Stworzenie numerycznej symulacji zjawisk fizycznych jest zadaniem interdyscyplinarnym, wymagającym współpracy pomiędzy naukowcem/inżynierem a programistą. Model numeryczny, stworzony na postawie teoretycznego podłoża oraz kreatywnego zastosowania matematyki i metod numerycznych, może zostać zaimplementowany. Wyniki uzyskane dzięki takiej analizie numerycznej mogą następnie zostać porównane z obserwacjami oraz wynikami badań, a przydatność modelu oceniona w odniesieniu do obecnie istniejących, uproszczonych metod analizy. Celem tej pracy było stworzenie modelu numerycznego do nieliniowej analizy żelbetowych belek ciągłych przy uwzględnieniu degradacji ich sztywności. Cel ten został osiągnięty. Przeprowadzono przegląd literaturowy w celu przygotowania teoretycznego podłoża dla zagadnienia degradacji sztywności w zginanych elementach żelbetowych zakładając sprężysto–plastyczne zachowanie żelbetu. Zbadano obecnie stosowane metody analizy konstrukcji oraz wymiarowania. Zacytowano również najważniejsze wymagania normowe, głównie CEB-FIP Model Code oraz Eurokod 2, dotyczące analizy i wymiarowania elementów żelbetowych. Model numeryczny został opracowany przy współpracy z kierującym tę pracę, dr inż. Grzegorzem Wandzikiem, na podstawie przytoczonego przeglądu literaturowego oraz własnego doświadczenia w tym zakresie. Szczegółowo opisano algorytm obliczeń, wprowadzono matematyczne i geometryczne interpretacje poszczególnych zjawisk fizycznych oraz przedstawiono ich graficzne prezentacje. Gdziekolwiek było to możliwe, podkreślono uproszczenia oraz zależności wynikające z charakteru analizowanego zagadnienia. Celem takiego działania była generalizacja analizowanych zjawisk, aby umożliwić programiści, który niekoniecznie jest biegły w omawianym zagadnieniu, implementację modelu. Ilekroć dany krok w procedurze można byłoby wykonać na kilka sposobów, przydatność każdego rozwiązania została oceniona na podstawie przytoczonych zalet i wad. Taka wstępna analiza pozwoliła na zwrócenie uwagi na możliwe problemu lub, przeciwnie, korzyści wynikające z zastosowanie danego rozwiązania do tego konkretnego zagadnienia. Procedura analizy numerycznej została przedstawiona na prostym przykładzie, ponieważ celem modelu była ilościowa i jakościowa ocena procesu redystrybucji. Wprowadzenie bardziej skomplikowanego przypadku mogłoby w znaczny sposób utrudnić ocenę wyników analizy. Niemniej jednak za pomocą tego modelu w łatwy sposób można analizować dowolne układy prętowe, o dowolnych charakterystykach geometrycznych i materiałowych oraz pod dowolnym obciążeniem. Autor i kierujący tą pracą wierzą, iż będzie możliwe zaimplementowanie tego modelu numerycznego. Umożliwiłoby to jego weryfikację, a następnie ocenę badanego zjawiska. Docelowo rozwój tego modelu mogłoby zaowocować powstaniem zaawansowanego oprogramowania do bezpieczniejszego i bardziej ekonomicznego projektowania oraz zbrojenia elementów żelbetowych. List of Figures 2.1 Models based on theory of elasticity . . . . . . . . . . . . . . . . . . . . . . 2.2 Models based on theory of plasticity . . . . . . . . . . . . . . . . . . . . . . 2.3 Elasto-plastic models with hardening and softening . . . . . . . . . . . . . 2.4 Basic elements for decription of rate-dependent materials . . . . . . . . . . 2.5 Viscoelastic model (Maxwell model) . . . . . . . . . . . . . . . . . . . . . . 2.6 Viscoelastic model (Kelvin–Voight model) . . . . . . . . . . . . . . . . . . 2.7 Viscoelastic model (Standard Linear Solid model) . . . . . . . . . . . . . . 2.8 Elasto-viscoplastic models . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.9 Viscoelasto-plastic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.10 Viscoelasto-viscoplastic model . . . . . . . . . . . . . . . . . . . . . . . . . 2.11 Standard models for concrete in compression [36] . . . . . . . . . . . . . . 2.12 Standard models for concrete in tension [35] . . . . . . . . . . . . . . . . . 2.13 Standard models for steel [36] . . . . . . . . . . . . . . . . . . . . . . . . . 10 10 11 12 13 13 14 14 14 14 16 16 17 3.1 3.2 3.3 3.4 3.5 3.6 3.7 Real work of flexural reinforced concrete beam [13] . . . . . . . . . . . . . Types of hinges [21] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . M –κ relationship for flexural reinforced concrete elements . . . . . . . . . Plastic rotation θs of reinforced concrete sections in continuous beams [36] Allowable plastic rotation θpl,d of reinforced concrete sections [36] . . . . . Determination of limit load Qu with kinematic method [20] . . . . . . . . . Determination of limit load qu with static method [20] . . . . . . . . . . . . 20 21 22 26 26 27 29 4.1 4.2 4.3 4.4 4.5 Phases of work of reinforced concrete member. Stresses [13] . . . . . . . . . Simplified diagrams for distribution of stresses in ULS [18] . . . . . . . . . Models for cross-section behaviour [20] . . . . . . . . . . . . . . . . . . . . Phases of work of reinforced concrete member. Strains [18] . . . . . . . . . Stiffness of flexural beam along the element [20] . . . . . . . . . . . . . . . 33 35 36 38 41 5.1 Process of numerical simulation of physical phenomenon . . . . . . . . . . 5.2 Exemplary continuous beam . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Static scheme of the beam . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Material and geometrical characteristics . . . . . . . . . . . . . . . . . . . 5.5 Static scheme of the beam after discretisation . . . . . . . . . . . . . . . . 5.6 Chosen σ– relationship . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7 Geometrical interpretation of curvature . . . . . . . . . . . . . . . . . . . . 5.8 Determination of strain distribution in cross-section with bisection method 5.9 Distribution of strains in cross-section for different levels of loading . . . . 5.10 Determination of internal forces in section with numerical integration . . . 5.11 Exemplary numerically-determined M –κ diagram . . . . . . . . . . . . . . 43 44 44 44 45 46 46 48 48 50 50 69 70 LIST OF FIGURES 5.12 Geometrical interpretation of section’s stiffness . . . . . . . . . . . . . . . . 5.13 Stiffness of section after rth step with first-order finite difference methods . 5.14 Static calculations. Step 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.15 Static calculations. Formation of first plastic hinge . . . . . . . . . . . . . 5.16 Static calculations. Step R1 + 1 . . . . . . . . . . . . . . . . . . . . . . . . 5.17 Static calculations. Formation of the last plastic hinge . . . . . . . . . . . 5.18 Determination of elements of stiffness matrix of segment Ki acc. to [19] . . 5.19 Graphical presentation of matrices formation in aggregation process . . . . 5.20 Determination of elements of matrix of nodal loads Pi . . . . . . . . . . . . 51 52 55 56 56 57 59 61 61 Bibliography Books [1] Kwan Y. W., Bang H.: The Finite Element Method using MATLAB. CRC Press LLC, New York (1997) [2] Garstecki A.: Wytrzymałość materiałów. Archiwum AlmaMater Politechniki Poznańskiej (2004/2005) [3] Gawęcki A.: Mechanika materiałów i konstrukcji prętowych. Archiwum AlmaMater (2003) [4] Ghali A., Favre R., Eldbadry M.: Concrete streuctures - stresses and deformations, 3rd ed. Spon Press, London (2002) [5] Godycki–Ćwirko T.: Mechanika betonu. Wydawnictwo Arkady, Warszawa (1982) [6] Jirásek M., Bažant Z.: Inelastic analysis of structures. John Wiley & Sons, Chichester (2002) [7] Knoll F., Vogel T.: Design for robustness. IABSE, Zurich (2009) [8] Klemczak B.: Modelowanie efektów termiczno-wilgotnościowych i mechanicznych w betonowych konstrukcjach masywnych. Wydawnictwo Politechniki Śląskiej, Gliwice (2008) [9] Kanuff M., ed.: Podstawy projektowania konstrukcji żelbetowych i sprężonych według Eurokodu 2. Dolnośląskie Wydawnictwo Edukacyjne, Wrocław (2006) [10] Klamka J., Pawełczyk M., Wyrwał J.: Numerical methods. Wydawnictwo Politechniki Śląskiej, Gliwice (2001) [11] Konderla P.: Metoda Elementów Skończonych. Teoria i zastosowania. Konspekt wykładu - Studia doktoranckie Wydziału Budownictwa Lądowego i Wodnego Politechniki Wrocławskiej (2007) [12] Kreja I.: Mechanika ośrodków ciągłych. Materiały pomocnicze do wykładów dla słuchaczy Środowiskowego Studium Doktoranckiego Inżynierii Lądowej i Architektury Politechniki Gdańskiej, Gdańsk (2003) [13] Kuczyński W.: Konstrukcje betonowe. Kontynualna teoria zginania żelbetu. Państwowe Wydawnictwo Naukowe, Łódź (1971) 71 72 BIBLIOGRAPHY [14] Łodygowski T.: Metody komputerowe. Archiwum AlmaMater Politechniki Poznańskiej (2004/2005) [15] Łodygowski T., Kąkol W.: Metoda elementów skończonych w wybranych zagadnieniach mechaniki konstrukcji inżynierskich. Archiwum AlmaMater Politechniki Poznańskiej (2003) [16] MacGregor J.: Reinforced concrete. Prentice Hall, Englewood Cliffs (1997) [17] Majewski S.: Mechanika betonu konstrukcyjnego w ujęciu sprężysto-plastycznym. Wydawnictwo Politechniki Śląskiej, Gliwice (2003) [18] Mosley B., Bungey J., Hulse R.: Reinforced concrete design to Eurocode 2, 6th ed. Palgrave MacMillan. New York (2007) [19] Pietrzak J., Rakowski G., Wrześniowski K.: Macierzowa analiza konstrukcji, Wyd. 2 zm. Państwowe Wydawnictwo Naukowe, Warszawa (1986) [20] Starosolski W.: Konstrukcje żelbetowe. Tom I, wyd. 8 zm. Wydawnictwo Naukowe PWN, Warszawa (2003) [21] Tichý M., Rákosník J.: Obliczanie ramowych konstrukcji żelbetowych z uwzględnieniem odkształceń plastycznych. Wydawnictwo Arkady, Warszawa (1971) [22] Zienkiewicz O.C., Taylor R.L.: The Finite Element Method. Vol. 1: The Basis, 5th ed. Butterworth–Heinemann, Oxford (2000) [23] Zienkiewicz O.C., Taylor R.L.: The Finite Element Method. Vol. 2: Solid Mechanics, 5th ed. Butterworth–Heinemann, Oxford (2000) Articles [24] Bathe K.J., Walczak J., Welch A., Mistry N.: Nonlinear analysis of concrete structures. Computers & Structures vol. 32, no. 3/4, p. 563–590 (1989) [25] Bondy K.B.: Moment Redistribution: Principles and Practice Using ACI 318-02. PTI Journal, p. 3–21 (2003) [26] Dąbrowski K.: Kilka uwag na temat metody plastycznego wyrównania momentów stosowanej w konstrukcjach żelbetowych. Inżynieria i Budownictwo nr 8/2002, str. 431–432 (2002) [27] Jędrzejczak M., Knauff M.: Redystrybucja momentów zginających w żelbetowych belkach ciągłych – zasady polskiej normy na tle Eurokodu. Inżynieria i Budownictwo nr 8/2002, str. 428–430 (2002) [28] Noakowski P.: Advanced Structural Design. Basis of non-linear concrete mechanics. Materials for the course of Advanced Structural Design for Master degree course at the faculty of Civil Engineering at the Silesian University of Technololgy, Gliwice (2009/2010) BIBLIOGRAPHY 73 [29] Noakowski P.: Continuous Theory for the Determination of Crack Width under Consideration of Bond. Beton- und Stahlbetonbau, 80, Nos. 7 and 8, p. 19 (1985) [30] Noakowski P., Moncarz P.: Stiffness-Oriented Design of R.C. Structures. Close-toReality and Practicable Computations. Procedures and Their Applications. Księga jubileuszowa z okazji 70-lecia profesora Tadeusza Godyckiego-Ćwirko, Politechnika Gdańska, p. 145–155 (1998) [31] Scott R.H., Whittle R.T.: Moment redistribution effects in beams, Magazine of Concrete Research, 57, No. 1, p. 9–20 (2005) Standards [32] ACI 318:2002 Building Code Requirements for Structural Concrete [33] BS 8110:1997 Structural Use of Concrete Part 1: Code of Practice for Design and Construction [34] CEB-FIP Model Code 1990 [35] CEB-FIP Model Code 2010 (final draft) [36] EN 1992-1-1:2008 Eurocode 2: Design of Concrete Structures. Part 1-1: General Rules and Rules for Buildings [37] PN-B-03264:2002: Konstrukcje betonowe, żelbetowe i sprężone. Obliczenia statyczne i projektowanie View publication stats