Section 1.3   Issues for Practice

 

Could another disaster occur due to faults in modelling?

 

 

In Section 1.3 the potential risk in using structural analysis is discussed and in Sections 3.8.1, 3.8.2 and 3.8.3 three real case studies of collapses caused by fundamental errors in the analysis models used.  These are not recent events and the question arises could it happen again with significant loss of life.  A main aim in promoting the use of the modelling process described in Chapter 3 is to seek to prevent such an event occurring.

 

This event described here is not real by has been synthesised to demonstrate what could happen if the modelling process is neglected

 

 

 

Failure Case

 

 

Context:  Long span roof over a conference centre.

 

Structure:  Steel beams with cable stayed supports.  The beams were pinned at the centre of the span.

 

Event:  The roof collapsed during a severe gale while a conference was in session with significant loss of life.

 

Modelling errors:  The cables were modelled as bar elements but one wind load case caused uplift such that the cables went into compression.  That these cables should have been made ineffective for this situation was not identified by those doing the modelling.  Because of the uplift and the lack of stay action, the end connections of the main beams went into reverse bending for which they were not designed and failed. When the cables came back into tension the failed end connections caused the back cables to become ineffective and the roof as a whole collapsed.  The centre pin in the main beams was a negative feature because when the end connections failed, the beam system became a mechanism.  Also there was a lack of secondary action to spread the load across the structure after the initial connection failure.

 

Management errors:  The modelling was checked by a third party but there was pressure to get the contract going and the checkers were not allowed enough time to do the work properly.

 

Summary:  The fundamental cause of the collapse was due to a modelling error.  The connection details contributed to a lack of robustness which also contributed to the failure.

 

 

 

  Note:  This case was developed by Alan Mann