Edinburgh Napier University

  The results of a series of full-scale torsion tests on rectangular hollow sections (RHS) are presented and discussed. The observed torque–twist behaviour is compared to that predicted by an extended version of Marshall’s simplified thick wall torsion theory and finite element (FE) analysis, and significant differences are highlighted and examined. The behaviour predicted by the FE models is shown to be identical to that predicted by Marshall’s thick wall theory, which forms the basis of the British and European design procedures. However, even though the experimental measurements agree with the FE and theoretical predictions in the elastic range, the measurements of torsional capacity are significantly lower than those calculated, and this has important implications for design that may be wider than just torsion of RHS. A number of potential causes for this behaviour are examined, but it is yet to be fully explained. Evidence of similar behaviour in previous large-scale testing is highlighted and discussed.