Mathematical models have been devised expressing the levels of
controllable welding factors as a function of the joint geometry, such that
acceptable weld beads are produced. Weld beads were required to be both
geometrically acceptable and mechanically sound despite changes in the root
face thickness (0.5 mm - 2.5 mm) and root gap (0 - 1.5 mm). The equations
are intended to form part of an adaptively controlled robotic arc welding
system. Simulation was used to develop the adaptive expressions.
The study was applied to the root weld bead of the closing seam of
railway bogie side frames fabricated from structural steel. The self shielding
flux cored electrode arc welding process was used to weld single J
preparations orientated in the horizontal-vertical position. Single sided full
penetration welds with underbeads were required. The weld bead geometry was defined in terms of ten responses.
Mathematical models derived from factorially designed experiments were used
to relate the weld bead geometry, incidence of porosity and the occurrence of
electrode stubbing to a function of upto seven factors. A data base of almost
1000 test welds was generated, in which each test was characterised by 76
pieces of information. Analysis of variance was used to determine which
factors most influenced each of the responses. Multiple regression enabled an
expression for each response to be derived as a function of the welding factor
levels. The weld bead geometry model consisted of ten equations, each a
function of upto six factors, whilst the soundness model related the optimum
welding voltage to a function of three factors so that porosity and electrode
stubbing would not be encountered.
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough University.