Civil Engineering Reference
In-Depth Information
parameters are ramped to the setpoint values specified by the neural network. When the setpoint values
have been reached, at time
t
T
, the closed-loop control is enacted. The bead width from the process can
be monitored in real-time. A real-time penetration sensor is not readily available, therefore, a second neural
network is run in parallel with the process to yield estimates of the penetration. The measured bead
width and the estimated penetration are subtracted from the respective reference values, processed
through proportional-plus-integral controllers, and added to the final values obtained from the setpoint
sequencer. When a workpiece thickness variation is encountered in the process, the system adjusts the
current and the travel speed accordingly to maintain constant bead geometry.
The control loop experiment was run using mild steel for the workpiece material. Plates of two
thicknesses, 0.125 in. and 0.250 in., were joined together and an arc using the nominal arc parameters
(
I
6 ipm) was run across the boundary between the plates, from the thicker
section to the thinner one. The bead width and penetration were 0.142 in. and 0.035 in., respectively,
on the thicker plate. With the controller disabled (and thus the equipment parameters remaining
unchanged), the bead width increased to 0.157 in. and the penetration increased to 0.047 in. when the
weld pool entered the thinner plate. With the controller enabled, the width and the penetration remained
the same on the thin plate as they were on the thick plate, with only a slightly discernible transient.
100 A,
L
0.100 in.,
v
VPPA Weld Modeling and Control
Artificial neural networks were evaluated for monitoring and control of the variable polarity plasma arc
welding process by Cook, et al. in [5]. Three areas of welding application were investigated: weld process
modeling, weld process control, and weld bead profile analysis for quality control.
Plasma arc welding (PAW) of aluminum has been extensively used and developed at NASA's Marshall
Space Flight Center using square wave ac with variable polarity (VPPA) [67]. With the variable polarity
process, oxide removal before welding is not required for most aluminum alloys. NASA and its subcon-
tractors use the VPPAW process in the space shuttle external tank fabrication.
The plasma arc differs from the ordinary unconstricted welding arc in that, while ordinary weld arc
jet velocities range from 80 to 150 m/s, plasma arc jet velocities range from 300 to 2000 m/s in welding
usage [68]. The high plasma arc jet velocities are produced by heating the plasma gas as it passes through
a constricting orifice. In VPPA welding of certain ranges of metal thickness, appropriate combinations
of plasma gas flow, arc current, and weld travel speed will produce a relatively small weld pool with a
hole (called a keyhole) penetrating completely through the base metal.
VPPAW process modeling was investigated to determine if attributes of the weld bead, such as crown
width, root width, or penetration could be predicted based on the corresponding equipment param-
eters (arc current, torch standoff, travel speed, etc.). Emphasis was placed on direct control of the weld
bead geometries, where the user could specify the desired crown width and root widths, and the
proposed control system would select and maintain the equipment parameters necessary to achieve
the desired results. Various neural network methodologies were investigated for the purpose of arc
weld modeling and control.
Additionally, artificial neural networks were used to analyze digitized weld bead profiles, obtained
from a laser-based bead surface scanner presently used at NASA. The networks provided an improved
method for automatically detecting the location of significant bead attributes, such as crown, undercuts,
and edges.
Process Modeling
Data used for the VPPAW process modeling was obtained from the NASA Marshall Space Flight Center
[8], which consisted of the following parameters: torch standoff, forward current, reverse current, travel
speed, crown width, root width. Of the 13 VPPA welding runs, 10 were used for network training while
the other three were used exclusively for testing.
Several network configurations were tested by varying the number of nodes between 5 and 20 in the
