Civil Engineering Reference
In-Depth Information
Torch
Standoff
Desired
Crown
Width
Forward
Current
Reverse
Current
Desired
Root
Width
Travel
Speed
FIGURE 7.11
A neural network used for VPPA welding equipment parameter selection.
Torch
Standoff
Desired
Resulting
Crown
Width
Crown
Width
Forward
Current
Reverse
Current
Desired
Resulting
Root
Width
Root
Width
Travel
Speed
Equipment
Parameter
Selecting
Neural Net
VPPA Weld
Modeling
Neural Net
FIGURE 7.12
A parameter selection network determines the welding equipment parameters necessary to achieve
desired bead geometry.
widths are specified and applied to the inputs of the equipment parameter selection network which, in
turn, determines the suitable torch standoff, forward and reverse current, and travel speed. As a substitute
for the actual welding process, the neural network VPPA weld model, discussed earlier, is fed with these
equipment parameters and its crown and root widths are compared with the desired ones.
The deviations in the ultimate crown and root widths from the desired ones are mostly small. The
worst case is again for the welding run in which the root width was recorded as larger than the crown
width. Generally the welding data which was left out of the training of the parameter selector and the
model does not exhibit noticeably worse performance than the training data. The conclusions from this
experiment suggest that neural networks may be reliably used in selecting welding equipment parameters
for the VPPAW process.
Weld Bead Profile Analysis and Control
A welding seam tracker, based on a laser scanner, was used at the NASA Marshall Space Flight Center for
