Numerical Study of Interface Behaviour in Composite SHCC-Concrete Beams

Abstract

A parking garage of Eindhoven airport partially collapsed in 2017. It was caused by failure of the longitudinal joint (long-side) of the composite plank floor or breedplaatvloer on the roof level. Experimental and numerical research showed that the joint suffered high positive bending moment in its transverse direction. However, from the investigation results, there was not enough resistance at the concrete-to-concrete interface around the joint to transfer the tensile force from the precast to the cast-in-situ section, which led to delamination of the two layers of concrete and resulting in failure when the delamination crack reaches the end of the coupling reinforcements. As happened in that case, the concrete-to-concrete interface usually is the weakest link and has a critical role on behaviour of a composite concrete system, especially the one which has unreinforced interface (no stirrup near joint). Further studies have been conducted to understand the influence of various details around the joint on the interface behaviour. In this thesis, details in spacing between lap splices (coupling reinforcements in cast-in-situ layer and bottom reinforcements in precast layer), spacing between connecting reinforcements (stirrups crossing the interface near the joint), the role of connecting reinforcement, and the sensitivity of interface parameters were studied numerically using DIANA finite element analysis software. Since the spacings are in direction of the specimen’s width, interface behaviour was analysed in both longitudinal and transverse directions. An additional study about compressive membrane action or arching was also conducted to understand the influence of lateral restraint, which usually occurs in composite plank floor systems used in buildings, including the one used in Eindhoven parking garage, to the capacity of the structure. This action was suspected of providing additional strength to the existing composite plank floor

Two composite SHCC-concrete beam specimens from the experimental research by Harrass [1] were used in this numerical study since the experiment had both unreinforced and reinforced interface specimens which were important for this study. The unreinforced interface beam (Sample 1) was suitable for the study of lap splice spacing and lateral restraint without any influence from reinforcement crossing the interface, while the reinforced interface beam (Sample 7) was suitable for the study of stirrups spacing. The specimens are solid beams (without weight-saving element) consisting of a SHCC (Strain Hardening Cementitious Concrete) precast layer with a joint in the mid-span, and a regular concrete cast-in-situ layer. By using DIANA 10.4 finite element analysis software, this study is able to simulate both specimens in 2D and 3D numerical models. The models represented Sample 1 failed with a horizontal crack along the interface and a flexural crack at the end of coupling reinforcement reaches the top of the cast-in-situ layer, while the models represented Sample 7 failed with a horizontal crack along the interface, a flexural crack at the end of coupling reinforcement, and a crack at the stirrup location in precast layer. From the verification study, the reinforcement bond-slip function (CEB-FIB 2010) was chosen not to be used for the rest of the study since it did not affect much the load capacity and the failure mechanism of the specimens. Consequently, pull-out failure of the stirrup is excluded for the rest of this study.

Prior to the main study, the influence of each interface parameter was studied in both unreinforced and reinforced interface models by varying the interface parameters. It was observed that interface tensile strength and stiffness are governing in the unreinforced interface model. Within the range of those parameters, the load capacity was increased and decreased by more than 50% in compared to the reference model verified by Sample 1. By adding a rectangular stirrup near the joint of unreinforced interface model, the model with reinforced interface has a different governing parameter, the cohesion, and the variability of the results decrease. Within the range of the cohesion, the load capacity was increased and decreased up to 30% in compared to the reference model verified by Sample 7. Since the interface parameters influence the capacity of both unreinforced and reinforced interface beams, two different interface types are used for the whole of the study. They are known as “smooth interface” which uses the parameters obtained from the verification with the experimental specimen, and “perfect bonded interface” which use rigid connection between the elements of the two concrete layers.

To study the influence of lap splices spacing, models with three lap splices setup from Harrass’ experiment (three coupling reinforcements and three bottom reinforcements) were compared to models with a single lap splice (one coupling reinforcement and one bottom reinforcement) with the same total reinforcement area. As a result, with perfect bonded interface, model with single lap splice has higher load capacity by more than 10% in compared to model with three lap splices though both models failed with the same failure mechanism which was the horizontal crack along coupling reinforcement and a flexural crack at the end of coupling reinforcement reaches the top of the cast-in-situ layer. Stress concentration around coupling reinforcements were observed in all models, especially in model with single coupling reinforcement. However, different horizontal crack propagation occurred in each models. Uniform horizontal crack propagation along the interface were observed in models with smooth interface, while more concentrated crack propagation around the coupling reinforcements were observed in models with perfect bonded interface, especially in model with single lap splice. This different crack propagation in models with perfect bonded interface could be the cause of different load capacity since more uncracked elements could provide more tensile force transfer from precast layer to cast-in-situ layer.

In the study of the influence of stirrups spacing, models with rectangular stirrup (two legs) setup were compared to models with a single leg vertical stirrup setup with the same total reinforcement area. In both cases the presence of the stirrup near the joint stopped the propagation of the horizontal crack. As a result, all models could reach yielding of the coupling reinforcement for both interface types although different structural stiffness is observed. The plausible cause for this difference in structural stiffness was the higher tensile stress in stirrups of model with two legs stirrups compared to model with single leg stirrup. This higher tensile stress might be resulted by the more distributed stirrups across the width of the beam.

In the additional study, models with full height lateral restraint at the support were compared with models with simple support. As a result, with perfect bonded interface, model with lateral restraint has higher load and displacement capacity by more than 14 and 2.5 times consecutively compared to the model without lateral restraint. In compared to the collapse load, the numerical result of model with lateral restraint has higher load capacity by almost 4 times. With smooth interface, model with lateral restraint has higher load and displacement capacity by more than 10 and 5.5 times consecutively compared to the model without lateral restraint. In compared to the collapse load, the numerical result of model with lateral restraint has higher load capacity by more than 2 times. These increases are in accordance with a research by Ockleston [2] which found a considerable increase of load capacity on concrete slab with lateral restraint compared to the yield line theory. Part of the increase of capacity was resulted by the fix boundary action due to bending moment at support. However, it was observed that compressive membrane action started to develop after the first flexural crack as the horizontal force at support rapidly increased after that crack. Although the models with lateral restraint of both interface types had a different failure mechanism compared to the models without lateral restraint, they have a similar final stage of the failure mechanism, which is the flexural crack at the end of coupling reinforcement. This flexural crack propagation can be prevented from occurring earlier due to the high compressive stress at the top part of the cast-in-situ layer. When the concrete crushed at the support due to limited rotation capacity of the concrete, the compressive stress dropped causing the flexural crack at the end of coupling reinforcement to develop.

In conclusion, there was an influence of the interface behaviour to the failure and the capacity of composite SHCC-concrete beam. However, the influence was varied, depending on the coupling reinforcement spacing, presence of stirrup crossing the interface near the joint, spacing of stirrup, and interface type. It is also concluded that compressive membrane action in addition to fix bending action, which occurred due to the lateral restraint, increased the capacity of the structure. This increase depended on the interface type and rotation capacity of the concrete. A wider range of the influencing parameters are needed in future studies to get a more robust results which are beneficial for a more general conclusions. However, the series of experimental research based on this study are essential to provide a verification on the results of this numerical study.