Chapter 5 Theoretical basis of LimitState:RING 3.

0
5.1 Background
Masonry arch bridges are statically indeterminate compression structures which resist external applied loads primarily as a result of the thickness of the masonry and their inherent self weight. They tend to be resilient to small support movements, with these typically transforming a structure into a statically determinate form. Cracks which might accompany support movements are therefore not normally of great concern, making the notion of crack widths or other conventional serviceability criteria not applicable. Consequently engineers are generally primarily interested in guarding against the ultimate limit state (i.e. structural collapse) condition. This typically occurs when a sufficient number of hinges or sliding planes are present between blocks to create a collapse mechanism.

5.2 Analysis methods used by LimitState:RING 3.0

LimitState:RING 3.0 idealises a masonry arch bridge structure as an assemblage of rigid blocks and uses computational limit analysis methods (also known as plastic or mechanism methods) to: 1. analyse the ultimate limit state, determining the amount of live load that can be applied before structural collapse; 2. permit investigation of the mode of response when supports undergo small movements. Limit analysis techniques were originally developed for steel components and structures, but it has since been shown that these can be applied to masonry gravity structures, such as piers and arches (Heyman 1982). To help understand why limit analysis theory is applicable, compare and contrast the response of a steel column with uniform plastic cross-section and a weakly mortared masonry pier, both subject to a horizontal load F, as shown in Figure 5.1.

Figure 5.1: Laterally loaded (a) steel column, (b) masonry pier, and (c) idealised response curves It can be deduced that: whilst the tensile and compressive strength of the steel column endow it with a finite plastic moment of resistance, M p , the absence of tensile strength means that the masonry pier does not possess a comparable moment capacity derived from material strength; however, the thickness and self weight of the pier mean that there is some resistance against overturning and the masonry pier could conceptually be considered as possessing a moment capacity, albeit one that varies with height (moment capacity equal in magnitude to the normal force at a given cross-section multiplied by half the pier thickness); furthermore, provided pier displacements do not become large, the resistance of the masonry pier against overturning at a given cross-section will remain broadly constant; hence, the response of the pier can be considered ductile, an important requirement in order for limit analysis theory to be applicable. As already indicated, in masonry structures the moment of resistance effectively varies continuously and consequentially this can make conventional bending moment diagrams difficult to interpret. It is normally more useful to plot the eccentricity of the compressive force, or thrust, at each crosssection (where eccentricity = moment / thrust). Figure 5.2 shows the resulting lines of thrust at collapse for two different configurations of masonry blocks and loading types:

Figure 5.2: Thrust line at collapse in (a) masonry pier, and (b) masonry arch In Figure 5.2 it can be seen that the lines of thrust lie entirely within the masonry and also that hinges form at the locations where the lines of thrust touch the exterior faces of the blocks. Formation of a sufficient number of hinges (and / or planes of sliding) leads to collapse. In the case of the masonry pier shown in Figure 5.2(a), the structure is statically determinate and statics alone may be used to uniquely determine the position of the thrust line both prior to and at ultimate failure. In contrast, the masonry arch shown in Figure 5.2(b) is statically indeterminate and there are many possible positions of the thrust line prior to failure. Therefore, the actual position can only be uniquely determined at the point of ultimate failure. In addition to basic equilibrium considerations, in the context of masonry gravity structures, the following conditions may be used to test for ultimate collapse (assuming both hinging1 and sliding2 failures at masonry joints are considered possible): 1. The yield condition, which may be deemed to be satisfied providing the line of thrust both lies entirely within the masonry and does not cross any joint at a subtended angle () of less than tan-1 (), where is the coefficient of friction. 2. The mechanism condition, which may be deemed to be satisfied providing the line of thrust either touches exterior faces of the masonry blocks and/or crosses sufficient joints at an angle () of tan-1 () to create the releases required to transform the structure into a mechanism. Thus, if a line of thrust satisfies the equilibrium and yield conditions, then the true plastic collapse load cannot be less than the applied load i.e. it is a lower bound. Similarly, if a line of thrust satisfies the equilibrium and mechanism conditions, then the plastic collapse load cannot be higher than the applied load i.e. it is an upper bound. It is possible to perform limit analysis by hand. For example, an upper bound hand analysis could be carried out by: 1. choosing a likely mechanism of collapse; 2. using equilibrium (or a work method) to calculate the collapse load; 3. trying other likely collapse mechanisms until the critical one is found. However, even in the case of a single span, single ring, arch the curved geometry makes such a hand based procedure extremely laborious and liable to human error. LimitState:RING 3.0 effectively automates this process. However, rather than adopting a trial and error procedure to find the mechanism associated with the absolute minimum collapse load, LimitState:RING 3.0 uses rigorous mathematical optimisation techniques to quickly and accurately determine the correct solution (refer to Appendix A.1) for details of the mathematical problem formulation). Furthermore, whilst the above discussion has implied that masonry possesses infinite compressive strength, this is clearly not the case in reality. In fact the presence of finite strength masonry means that the line of thrust mentioned previously is better thought of as a zone of thrust, which should have sufficient thickness at any point to carry the compressive force, with the required thickness depending on the crushing strength of the masonry. Clearly, this consideration would add extra complexity to an already tedious hand calculation. However, if a finite masonry crushing strength is specified by the user then this thickness is automatically computed by LimitState:RING 3.0. It is assumed that the thrust is carried by an area of material under a uniform level of stress (i.e. assuming a rectangular stress-block, in accordance with a rigid-plastic idealisation of the masonry crushing response).

5.3 Output from a LimitState:RING 3.0 analysis

5.3.1 Identification of the adequacy factor Although for simplicity the previous section considered a case where a collapse load (e.g. Fp ) was to be computed, it is generally more useful to compute the factor which would, when applied to some specified pattern of live loads, lead to collapse. This factor (or multiplier) is commonly termed the adequacy factor and its determination for a given bridge is the principal goal of a normal LimitState:RING 3.0 analysis. When appropriate partial factors are included in the model, this adequacy factor must be greater than 1.0 for a safe structure3. For example, if a 1kN single axle load is specified and LimitState:RING indicates a computed adequacy factor of for example 154, this means that

the load which would cause collapse is 154kN. Alternatively if a 100kN single axle load was specified, the adequacy factor computed would be 1.54 (1.54100 = 154kN). When the applied load comprises a series of axle loads, the adequacy factor is the multiplier which, when applied to all axle loads simultaneously leads to collapse. For example, if a 1400kN rail vehicle comprises four 200kN axles and four 150kN axles and LimitState:RING indicates a computed adequacy factor of 3, this means that the loading at failure comprises four 600kN axles and four 450kN axles (3200 = 600kN; 3150 = 450kN). Full details of the mathematical formulation are provided in Appendix A.1. 5.3.2 Bridge behaviour when subjected to support movements Although the usual (standard analysis mode) goal of a LimitState:RING 3.0 analysis is to identify the adequacy factor, LimitState:RING 3.0 can also be used to model the effects of support movements. Movement of a support will lead to formation of hinges and/or sliding failures in the same way as when excessive applied loading is applied by a highway or railway vehicle. When in support movement analysis mode, vehicle loads (if present) can be pre-factored by the user and the critical arrangement of hinges and/or sliding failure planes are identified by finding the energy associated by moving the supports. The line / zone of thrust is also identified. Figure 5.3 shows the result of a support movement analysis. Full details of the mathematical formulation are provided in Appendix A.2.

Figure 5.3: Modelling the effects of support movement at the base of a pier

5.4 Range of applicability of LimitState:RING 3.0

5.4.1 Spans When the adequacy factor is sought, LimitState:RING 3.0 is most suited to the analysis of single and multi-span masonry arch bridges with short to medium span lengths, where foreseeable live loadings are typically non-negligible in comparison to structural self-weight. For spans longer than 20 to 30m, foreseeable live loadings are often essentially negligible and other considerations become more important, such as long term masonry creep effects (due to persistent moderately high stresses). Additionally, in the case of very long span bridges, the presence of high compressive stresses may give rise to non-negligible second order deformations, making the adequacy factors computed by LimitState:RING 3.0 potentially non-conservative. 5.4.2 Block shape LimitState:RING 3.0 provides a realistic model of bridges comprising single or multi-ring arch barrels constructed using regular stone blockwork or brickwork. Since in LimitState:RING 3.0 the constituent blocks are assumed to be rectangular, the software provides a less realistic model of random rubble stone masonry arches. Nevertheless, given that it is unfeasible to model the actual layout of stones in a random rubble arch, the software may be applied to such arches provided suitably conservative smeared properties for the masonry are adopted. 5.4.3 Stress-related failures In common with other limit analysis (or mechanism) programs, LimitState:RING 3.0 may not accurately predict the ultimate strength of a bridge if either of the following apply: 1. The bridge comprises a long (e.g. > 20 - 30m) and/or flat arch (e.g. span/rise > 6, or the arch contains very flat sections, for example in the case of an elliptical arch) and it can be expected that elastic deformations prior to collapse will significantly change the arch geometry; 2. A brittle response of some part of the structure may be expected to prevent the formation of a ductile collapse mechanism (e.g. abrupt failure of the bond between rings; brittle hinge formation). However, even in these cases LimitState:RING 3.0 can provide an invaluable upper bound estimate of the likely strength of the bridge, which can be used as a benchmark for alternative analysis methods. Such alternative methods may comprise: in the case of (i) a geometrically non-linear elastic analysis; in the case of (ii) a non-linear elastic analysis incorporating a masonry material model respecting fracture mechanics principles (e.g. using

Hillerborgs cohesive crack theory, Hillerbourg 1976). Alternatively if in the case of (ii) the brittle response stems from shear failure at masonry joints, it may alternatively be assumed that the initial bond strength at these joints is zero, and that only compressive and frictional forces may be transmitted. In this case LimitState:RING 3.0 may be used and can normally be expected to provide lower bound (conservative) results. Unfortunately the decision as to when cases (i) and (ii) might apply is complicated by the fact that both are dependent on stress levels. If stresses are very low in comparison to the elastic modulus and/or bond strength of the masonry, then brittle fracture of the bond between rings for example is unlikely to be an important issue. Since it can be shown that stresses increase with the size of structure being considered 4, this implies that brittle fracture of the bond between rings may not be of great concern in the case of very short span bridges. Readers are referred to Section 7.8.2 for further discussion on modelling multi-ring arch bridges. 5.4.4 Fill depth Additionally, since LimitState:RING 3.0 has been calibrated in situations when fill depths are relatively small in comparison to the arch span, for bridges with relatively small fill depths at the crown (or with no fill) LimitState:RING 3.0 can be expected to provide reasonable results. Conversely, when the fill depth at the crown is large (e.g. > span/2) results from the program should be treated as being very approximate. In such cases it may often be found that the predicted load carrying capacity is in excess of the bearing capacity of the arch fill material (i.e. is unattainable in practice). 5.4.5 3D effects In general spandrel walls at the edges of a bridge can stiffen the arch prior to failure and, depending on their end restraint conditions, may also enhance the ultimate limit strength. Studies of the influence of spandrel walls on the carrying capacity of full-scale single and multi-span laboratory bridges are detailed elsewhere (Melbourne & Gilbert 1995 and Melbourne et al. 1997). However, if a bridge is wide in comparison to its span then the effects of the spandrel walls on the central section of the bridge may be quite minimal. Furthermore, a common defect observed in masonry bridges is detachment of the spandrel walls (this is evident by the presence of longitudinal cracks running close to the edges of the bridge). For these reasons spandrel walls are not modelled in LimitState:RING 3.0. Since the software idealises the arch in two-dimensions, it is most suited for assessing masonry arch bridges which span squarely between abutments; readers are referred to Section 7.3 for further advice on modelling skew bridges. By default, LimitState:RING 3.0 utilises a user specified fixed bridge width in the analysis. However, an effective width which varies according to simple user-defined transverse distribution rules can alternatively be used; readers are referred to Section 8.1.4 and Section 8.2.2 for further guidance on choosing an effective bridge width. 5.4.6 Range of collapse modes identifiable The general problem formulation and rigorous mathematical solvers employed mean that LimitState:RING 3.0 can identify numerous potential failure mechanisms. Figure 5.4 shows a selection of those mechanisms that have been observed whilst using the program to assess real bridges. The ability of LimitState:RING 3.0 to identify hitherto unknown failure modes has led to some interesting findings. For example, it has previously been suggested that a multi-span bridge can safely be analysed as a series of separate single spans if the piers are stocky (i.e. thick in comparison to their height). However, this is not in general the case. For example, the adequacy factor associated with the mechanism shown on Figure 5.4(f) is actually much lower than that computed if the outer spans are omitted from the analysis - indicating that the presence of stocky piers (and backing in this case) is no guarantee that the structure can be safely idealized as a series of separate single spans. In general, users should avoid using such rules of thumb and should where possible model as much of the structure as is practicable (refer to Section 7.4 for further advice on modelling multispan bridges).

Figure 5.4: Selection of potential failure modes identifiable using LimitState:RING 3.0

5.5 Use of reinforcement

The key assumptions made when modelling bridges containing reinforcement using LimitState:RING 3.0 are as follows: The masonry (or concrete) arch and the reinforcement are assumed to behave in a rigid-plastic manner. This means that at a failing contact the reinforcement is assumed to yield at the specified force if it is located away from the neutral axis. This idealization means that the software is likely to be most suitable when sections are lightly reinforced using ductile steel bars. Shear failures in reinforced sections can be modelled by specifying a limiting shear force. Anchorage failure of the reinforcement is not considered. For details on how to specify reinforcement in a bridge model, see Chapter 17.

5.6 Finite masonry strength

In LimitState:RING 3.0 it is assumed that in the vicinity of hinges the thrust in the arch is transmitted across joints either: 1. through an infinitely thin strip of material lying on an exterior face, or, if a finite material strength is specified; 2. through a rectangular stress block located adjacent to an exterior face. For both cases, moment vs. normal force failure (yield) envelopes for a contact can be plotted, as shown on Figure 5.5:

Figure 5.5: Contact surface moment vs. normal force failure envelopes for: (i) infinite; (ii) finite masonry crushing strengths It is evident from Figure 5.5 that the envelope for case (i) is defined by linear constraints whereas in the case of (ii) these are non-linear. In LimitState:RING 3.0 the envelope for case (ii) is approximated using sufficient linear constraints to ensure that any deviation from the true non-linear yield envelope is negligible. To achieve this, linear constraints are adaptively added, using an iterative procedure which is described in Appendix A.3. (The procedure ensures that converged solutions are obtained more rapidly and reliably than was the case with RING version 1.x.)

5.7 Sliding failures

Unlike many other masonry arch bridge analysis programs, LimitState:RING 3.0 does not rule out the possibility that sliding failures might occur. A saw tooth model for friction is used (also referred to as associative friction). This means that separation is assumed to accompany sliding. The main advantage of using a saw tooth model is that the linear character of the problem is preserved. Whilst it can be shown that use of a saw tooth model for friction can lead to non-conservative adequacy factors being obtained if sliding is involved in the critical failure mode (Drucker 1954), when previously applied to multi-ring brickwork arch bridges reasonably good agreement between experimental and numerical results were obtained (in fact it was found that the numerical multi-ring model always under-estimated the experimentally observed carrying capacity).

5.8 Backfill
5.8.1 General The vertical dead weight of backfill material effectively pre-stresses the masonry in an arch, thereby increasing its load carrying capacity (provided the constituent masonry has sufficient compressive strength). The backfill also has two other beneficial effects as shown in Figure 5.6: (i) it disperses live loads; (ii) it can restrain movement of the arch when the latter sways into the fill. This is often termed passive restraint.

Figure 5.6: Masonry bridge soil-structure interaction Each of the above effects has the potential to significantly enhance the carrying capacity of a masonry arch bridge. However, whereas the constituent masonry blocks of a bridge are modelled explicitly in LimitState:RING 3.0, the backfill is presently modelled indirectly according to the simplified model described in the following sections. 5.8.2 Dispersion of live loads The vertical live load pressures on the back of an arch are assumed by LimitState:RING 3.0 to be either: 1. uniformly distributed, the intensity being governed by the depth of fill under the centre of a given axle and the specified limiting dispersal angle; or 2. dispersed according to a Boussinesq type distribution, with a limiting distribution angle specified by the user. The Boussinesq distribution model generates a suitable bell shaped distribution of load, which laboratory tests have indicated better approximates reality than uniform pressure distributions, and which also models the effects of overlapping dispersed loads more appropriately. This is therefore the default distribution model in LimitState:RING 3.0 (except for the surface layer, if present, where a uniform distribution is adopted). A default cutoff angle of 30 is used to prevent excessive distribution. Further details of the Boussinesq model are provided in Appendix B.1. 5.8.3 Passive restraint One-dimensional bar elements, for convenience hereafter termed backfill elements (Figure 5.7(a)) are used to model the passive restraint experienced by sections of the arch moving into the fill material. Backfill elements compress at constant force (e.g. h vertical projected area) where h is the horizontal stress, but have no tensile capacity (Figure 5.7(b)). Use of these elements ensures that pressures are mobilised in the correct sense. For example, in Figure 5.7(c) the backfill elements only apply a force to the non-loaded side of the bridge. Note that in LimitState:RING 3.0, active pressures on the loaded side of the bridge, which are usually relatively small, are for simplicity ignored.

Figure 5.7: (a) Arch restrained with uniaxial backfill elements; (b) backfill element response; (c) LimitState:RING 3.0 representation of backfill elements Classical lateral earth pressure theory developed originally for vertical retaining walls is often used in masonry arch analysis to estimate the amount of horizontal passive restraint which can be realised. This approach is also used in LimitState:RING 3.0. According to vertical retaining wall theory the horizontal passive restraining stress h applied to the back of a smooth wall is: (5.1) Where Kp = cohesion of the fill material. For a frictional backfill, drained strength properties In LimitState:RING 3.0, equation (5.2) where m p and m pc are modification factors designed to account for additional effects not represented by simple vertical retaining wall theory: Determination of m p Consider a wall which retains a frictional soil. In the case of a vertical wall which rotates at failure, full passive pressures will not be mobilised until rotations are very large (Figure 5.8(a)). In practice, a reduced earth pressure coefficient is often used to limit rotations to acceptable levels (Figure 5.8(b)). If the wall is actually curved (i.e. part of an arch) then the coefficient can be assumed to be further reduced (Figure 5.8(c)). is the effective angle of friction of the fill material, and c is the c would be used.

Figure 5.8: Commonly used horizontal earth pressure coefficients: (a) large wall rotation, (b) small wall rotation (0.5%), (c) arch segment rotation into surrounding fill It has been found that taking m p = 13 gives rise to a restraint force which is approximately equal to that measured in laboratory tests (at least for backfills with relatively high angles of friction). Hence this value is used as the default in LimitState:RING 3.0. i.e. when the user inputs the angle of friction for a granular backfill this is used to compute a Kp value which in turn is multiplied by the vertical stress, v, and the m p modification factor (default = 0.33) to compute the horizontal restraining stress h . Determination of m pc The cohesive strength of clayey backfill materials may enhance the load carrying capacity of short span bridges. However, limited experimental evidence is available at present and a conservative default modification factor on Kpc of m pc= 0.05 is therefore currently used in LimitState:RING 3.0. (This has been increased from the default value of 0.01 utilised in LimitState:RING 2.0 in the light of additional experimental and numerical data.) It should be noted that this value may be inappropriate for very soft, low stiffness clay backfills. However these are unlikely to be encountered in practice. Once equation 5.2 has been used to determine a value of the horizontal restraining stress h applied to a given block, this is checked to see whether it exceeds the magnitude of horizontal stress that can be applied without causing the strip of backfill acting on the block to slide. If so, then the reduced horizontal stress associated with the sliding failure is used (this can be overridden if a user-defined pressure is specified). Further details are given in Appendix B.2. Finally it should be appreciated that there are several important simplifications inherent in the way LimitState:RING 3.0 treats passive (and active) restraining horizontal pressures. For example: (i) backfill pressures are assumed to be mobilized by infinitesimal structural movements; (ii) the failure mode is assumed not to influence the magnitude of peak pressure mobilized. Further background information on these assumptions is provided in Appendix B.3 and Appendix B.4.
1Initially assuming that the masonry possesses infinite compressive strength, so the line of thrust can be transmitted through a hinge point lying on an exterior face of the arch. 2Initially assuming that sliding failures follow a sawtooth friction model (i.e. obey an associative flow rule, where sliding is accompanied by dilatancy). 3The adequacy factor may need to be commensurately greater than 1.0 for a safe structure if dynamic and/or other factors have still to be applied following completion of a LimitState:RING analysis. 4Consider for example the stresses at the base of two geometrically similar solid masonry piers. If the second pier is n times as large, in all dimensions, as the first then its volume and hence self weight will be n3 larger. However the area at the base of the pier will only be n2 times larger, so it follows that the gravity stresses at the base (and elsewhere) in the second pier will be n3n2 = n times larger.