The thesis deals with the development of numerical models for composite connection and composite frame behaviour. The final models have been fully checked by means of verification against test results. The verified model has provided an opportunity to examine several detailed aspects of behaviour that have then formed the basis for the development of design procedures for composite connections. This chapter summarises the main findings of the study and identifies the future scope for research in this area.
A numerical model has been developed which can represent the full structural response of several different types of composite connections. Before developing the composite connection model, important components were modelled separately and the modelling verified separately. Test measurements of response for both local and overall behaviour of composite connections were used to verify the model. It was observed that among the various different types of connections the endplate connection is the easiest to model as it does not exhibit bolt slip. The development of composite connection models are presented in chapter 3 and the preliminary modelling of various components in chapter 2. Results of parametric studies using the model described in chapter 3 are presented in chapter 4.
The results that are available from tests designed to investigate the effect of the shear to moment ratio were found to exhibit somewhat contradictory trends. Both theoretical and numerical analyses have been conducted to predict the effect of changes in the shear to moment ratio on composite connection moment capacity. Equations have been developed that incorporate these variations. A design method for cruciform flush endplate connections has been developed that relates the shear span directly with the connection moment capacity. The detailed results are presented in chapter 5.
At present no tests are reported that investigate the effect of column axial loading on composite connection moment capacity. The design equations for column web compression strength of EC3 are based on a few (bare steel) tests. It is doubted that these tests actually reflected the effect of column axial loading on the buckling resistance of the column web - not on the compression resistance of the column web. Also the column web shear strength is assumed to be constant irrespective of the level of column axial load. Both theoretical and numerical analyses have been conducted to study the effect of varying levels of column axial loading on composite connection moment capacity. It has been found that the present equation for column web compression resistance is too conservative, the equation for column web shear strength is unsafe, and that final bolt forces should be calculated from the equilibrium of the joint at the interface of the beam and the column - instead of directly using the values from equations to compute the moment capacity. New equations have been developed to predict the effect of the column loading on column web compression strength, column web shear strength and bolt force; these can be used to replace the existing EC3 equations. A design method has been developed to consider the effect of column loading on the composite flush endplate connection moment capacity. The detailed results are presented in chapter 6.
A unified design method has been developed which can consider the effect of shear to moment ratio and column axial loading for both symmetrical and non-symmetrically loaded connection. Care has been taken to include all possible modes of failure that can occur in a composite connection. The method utilises plastic theory and a simplified stress block approach, based on evidence from the numerical studies. The method is capable of determining the failure modes very accurately. It was observed that out of the 32 major axis flush endplate connections tested the proposed method predicted the correct failure mode for 27 cases. Predictions from the proposed method have been compared with a total of 53 test and finite element results. These comparisons gave an overall prediction to test ratio of 0.99 with a standard deviation of 0.14, thereby demonstrating that the proposed method can accurately predict the resistance of composite flush endplate connections under a variety of different connection arrangements and loading conditions. Detailed results are presented in chapter 7. The method is suitable as a design procedure for flush endplate connections in EC4, where, at present, there is no design procedure for connections.
On the basis of the theoretical studies a method has been developed to estimate the initial stiffness of composite endplate connections. Comparison with 28 major axis flush endplate test results gave an average of 0.99 with a standard deviation of 0.21 for the proposed method. At the same time a method to predict the available connection rotation capacity has been developed that is fully compatible with the prediction method for moment capacity. Care has been taken to include the effect of the extension of the rebar and the bolts and the shear stud slip. Also, the method allows for the effect of the depth of beam web in compression during the calculation of available rotation capacity. Both methods are described in chapter 7. These two methods have been combined with the moment capacity calculation method of flush endplate composite connections to predict the overall behaviour of the connections that has also been described in chapter 7. They are suitable for inclusion in EC4 for the computation of initial stiffness and available rotation capacity of composite connections.
Following an approach similar to that used for flush endplate connections, design procedures have been developed for composite finplate and angle cleated connections. Results are compared for 6 finplate connection tests which gave an average of 1.06 with a standard deviation of 0.18. The comparison for 16 tests on angle cleated connections gave an average of 0.98 with a standard deviation of 0.13. These have been described in detail in chapter 8 and are also suitable for inclusion in structural design codes e.g., EC4.
A numerical model has been developed for composite non-sway frames that can accurately represent behaviour observed in actual frame tests. It has been found that it is possible to model the response in terms of the frame moment distribution, connection moment rotation and the beam load displacement history very accurately. This provides an economic tool to study different aspects of the behaviour of composite non-sway frames. The frame model has been described in chapter 9.
A numerical model has been developed for unbraced bare steel frames. This model was verified against numerical results obtained by other researchers. Using this model some studies have been performed that provide basic guidance for developing equations for estimating sway. This is also reported in chapter 9.
There is a need for further theoretical and numerical work on the available rotation capacity of connections. In the proposed model some aspects of behaviour have been approximated. For example, the stiffness of the bolts and the length of reinforcement that is to be taken into consideration. Also, the exact way in which the extension of the rebar is to be calculated i.e., on the basis of an effective length only or by taking account of its stiffness if the rebar is not found to yield when predicting moment capacity. For medium and low levels of reinforcement the rebar is usually found to yield. Test results show the range of reinforcement strain to be 3000 to 15,000micro-strain, these will vary according to type of reinforcement (e.g., mesh or bar) and also on their diameter and spacing. This indicates that the selection of yield strain can change the rotation capacity by more than 100%. For connections with high levels of reinforcement ratio the rebar usually does not yield; for this situation it may be more appropriate to calculate the extension from force and rebar stiffness.
To experimentally investigate the moment-rotation behaviour of different types of connections when the connection is loaded in an upward direction. This will provide the data required to verify or modify the connection model with upward beam loading. These are needed to construct a complete moment rotation curve that can than be used for the composite sway frame numerical model development. Once this has been done the model can be utilised to study frame moment distribution, connection required and available rotation capacities, serviceability deflection with changes to load and member sizes. The findings can be utilised to develop a comprehensive design procedure for composite frames.
There is a further need for theoretical and numerical work on sway of steel frames which can then be extended for composite frames. It is essential to explore the sway behaviour of the composite frames experimentally. The results can be used to verify the developed numerical model for composite frames for the sway mode as described above.
The traditional approach of frame analysis assumes that the contraflexure point is located at a certain distance from the beam to column connection. Once a validated numerical model for the composite frame in the sway mode is developed the model can be used to numerically investigate the effect of horizontal load on the location of contraflexure points. Results from the numerical study can form the basis of design procedures for composite sway frames.
