Steel-concrete-steel (SCS) sandwich panels are an efficient means of achieving a strong and stable composite wall. Development in the 70's and 80's focussed on tunnelling, with other applications, particularly in the defence and offshore sectors, appearing later.

Renewed focus has been placed on the system in recent years due to a proliferation of proposals for new nuclear power stations in Europe. Many new nuclear projects that have been completed in recent years have been significantly delayed by problems with reinforcement congestion. SCS construction offers a potential solution to this, since reinforcement is either significantly reduced or eliminated entirely in most designs. As a result of this renewed interest, industry has sought to develop improved design rules, both for economy and easier regulatory approval.

As with any composite system, the strength of the system is derived from the ability of the materials to interface efficiently with each other where they are connected. Review of existing design guides and research showed a gap in understanding of the effects of shear connection on the overall behaviour of the system, particularly when resisting out-of-plane loads. This thesis aims to improve this understanding, leading to improved design provision and a wider range of applications for SCS panels in industry.

An extensive literature search found a large body of test results. However, the majority of these tests are for designs where shear connection is over-provisioned, meaning shear connection is not critical. The tests that were conducted with lower degrees of shear connection were found to be insufficient to draw definitive conclusions about changes in behaviour. For this reason, numerical modelling using finite element analysis was used to supplement the test data. A validation and verification exercise was performed, which showed that the model accurately predicted the behaviour seen in testing, for all of the relevant failure modes.

This thesis focusses on the three design checks that are required for panels subject to out-of-plane loads; bending resistance, shear resistance and deflection. The effect of reduced shear connection on each of these design checks is explored in turn.

For bending resistance, design rules based on first principles cross-section equilibrium are found to accurately predict the point of failure for the majority of cases. However, the existing assumption of a smooth profile of shear connector force is found to be incorrect on the tension plate, with tensile cracking leading to discontinuities in the stud force profile. Further interpretation of this result shows that this can lead to an unconservative prediction of the failure load when a panel with a low degree of shear connection is subject to a uniformly-distributed load (UDL). A new design rule is presented for this situation.

Design equations for shear resistance are found to vary considerably between design codes and countries. As with the bending check, the test database is found to be lacking in tests with low enough degrees of shear connection to draw definitive conclusions about any changes in behaviour. A parametric FE study is presented to investigate these effects. The study focusses on varying the degree of shear connection for groups of beams loaded at different shear-span to depth ratios. Different behaviour is observed in each group, with the influence of shear connection varying, depending on which shear transfer action is dominant. The study shows that unconservative predictions are made for a number of the design models, particularly for slender beams with low degrees of shear connection. A new adjustment is presented for the Eurocode shear resistance model that removes the unconservative predictions. The models from the fib Model Code are suggested as a better alternative, again with some adjustment to account for reduced degree of shear connection.

Deflection of SCS panels is usually predicted using linear-elastic models. Debate has occurred about whether to base the stiffness used on the contribution of the steel plates only, or whether the concrete stiffness should be included. This work finds that a partial concrete contribution should be assumed. It is also found that simple bending prediction models, based on Euler-Bernoulli principles, tend to overestimate stiffness for beams with low shear-span to depth ratios. In these cases, models that include shear deformation (such as the model by Timoshenko) are found to produce more accurate predictions. Reduced shear connection is found to lead to non-linear load deflection response curves, which cannot be easily approximated with linear-elastic models. A new load-stiffness curve is proposed for simplified non-linear modelling, which could be easily implemented in most current software packages with non-linear solvers.

Finally, partial resistance factors for the bending and shear design checks are calculated, using the procedure presented in Annex D of Eurocode 0. This method takes into account the precision and conservativeness of a particular design equation through a systematic comparison with available test data, and penalises studies that are based on limited test data. The procedure is found to be deficient when the design model includes contributions from multiple materials and large numbers of parameters. To overcome this, a novel extension to the existing procedure is proposed, termed the 'matrix method'.

In general, it is concluded that lower degrees of shear connection are not immediately detrimental to the performance of the system. This thesis highlights the changes in behaviour that can occur, which designers should account for when calculating the resistance of panels. This thesis also presents new adjustments and design rules to allow resistance to be accurately calculated in such cases.