The influence of semi-rigid connections on the performance of steel framed structures in fire
Observation of fire damaged structures and recent fire tests at the Cardington LBTF have suggested that even nominally `simple' connections are capable of providing significant restraint at elevated-temperatures. As most frames are designed assuming pinned response at ambient-temperature, with no account being taken of the reduction in mid-span moments, this is an aspect of connectivity which may be utilised in the assessment of the fire resistance of steel framed buildings, without necessitating changes in the approach adopted in ambient-temperature design or construction. To date the assessment of the influence of connection response on frame behaviour has been limited by the quantity of available test data, although initial studies based on postulated moment-rotation- temperature characteristics concluded that the failure temperatures for beams are increased due to the rigidity of `simple' connections. Moment-rotation relationships have been measured for a flush end-plate connection, both as bare-steel and as composite with a concrete slab across a range of temperatures. To define accurately the full moment-rotation-temperature response a series of tests have been conducted for each arrangement, where specimens were subject to varying constant levels of load and increasing temperatures. Observed failure mechanisms have been compared with those for a nominally identical specimen tested at ambient-temperature, and initial recommendations presented for the degradation of ambient-temperature connection characteristics. A mathematical expression is proposed in order to represent the test data at a number of temperatures. It is clearly unrealistic to expect that many such tests can be anticipated in the future, and as such a spring-stiffness model has been presented for both bare-steel and composite flush end-plate connections. The use of a spring-stiffness model compares favourably with other forms of modelling due to the combination of efficient solution and the ability to follow accurately the full non-linear range of connection response, based on an understanding of the response of the component parts. A multi-linear representation of response has been adopted, where the stiffness of the connection is revised as elements enter the plastic range of response. Comparison has been made between the response predicted and that recorded experimentally. Experimentally derived connection characteristics have been incorporated within analysis of typical sub-frames, with parameters including connection stiffness, capacity and temperature being varied. Further studies are presented considering the sensitivity of overall frame behaviour to inaccuracies in the representation of connection response and the use of simplified models to generate elevated-temperature connection characteristics. Based on postulated elevated-temperature moment-rotation characteristics for the connections contained within the Cardington test frame, predictions have been presented for the response of the structure subject to a series of full scale fire tests, with semi-rigid behaviour being compared with the common assumptions of pinned and rigid characteristics.