play. They are a line tracking system and part memory.
The cap welding phase has similar requirements to the fill
welding phase, but the placement of the individual weld runs
relative to the original underlying edges of the joint is less critical
than during fill welding.
Arc spatter, sunlight or welding environment influence
against image processing results because these have the same
wavelength of semiconductor laser used for laser vision sensor.
To eliminate various kinds of noise images and detect accurate
weld profiles by filtering the noise images, a new type of image
processing algorithm was developed. A new pattern matching
algorithm for detection of the weld center position was de-veloped based on [4]. The developed pattern matching algorithm
focuses on minimization of image processing time and error
caused by arc noise. The pattern matching algorithm contains the
followings:
• Save the original image without noise and extract line char-
acteristics as shown in Fig. 3: line angle, line type, and line
magnitude.
• Eliminate noise images (Fig. 5) caused by open arc and sur-
roundings from captured image (Fig. 4).
• Reconstruct the image data reflected from the weld groove
exactly to match the original weld profile (Fig. 6).• Examine the reconstructed image data with the line char-
acteristics saved before welding (the new pattern matching
algorithm).
• If there exists any matching point between the reconstructed
image and the line characteristics, determine the weld seam
position as shown in Fig. 7 (linear regression).
Figure 8 shows a typical example of the proposed image process-
ing algorithm. In this case, various types of noise are involved,
however, the weld seam center is exactly detected.
3 System architecture and application
To maximize system reliability and to prevent the occurrence
of defects, it is preferable to use more than one source of data
for weld joint tracking and torch positioning. In the systems de-
scribed here, three main data sources have been used:• Laser sensor and welding system,
• Line tracking system,
• Path memorization.
Figure 9 shows the general welding phases for multi-pass and
multi-layer welding.
3.1 Laser sensor and welding system
The laser sensor used was specially designed for the large joint
types, including narrow gap and semi-narrow gap grooves, found
in thick wall SAW. It uses the scanning spot principle in prefer-
ence to the more common laser stripe principle. While scanning
spot sensors are more complex and hence more expensive than
the laser stripe type, the benefits significantly outweigh the addi-
tional cost. These include:
• Immunity to surface type,
• Immunity to reflection problems,
• Independence of horizontal and vertical measuring fields,• Programmable width of field, scan rate, and number of
samples.
The joint profiles, especially for narrow gap and semi-narrow
gap joints, are often machined just before being welded. This
presents significant challenges to the sensor which must be able
to get an adequate return signal from the steep and shiny side-
wall of the joint. Use of a scanning spot concentrates the laser
on a single very small spot on the joint and produces a good sig-
nal showing the complete joint profile to be obtained (Figs. 10
and 11).
The welding system consists of the laser vision sensor, anti-
drift system, and sensor for detection of one cycle of pipe rota-
tion. Figure 12 shows the developed welding system for a pres-
sure vessel.
3.2 Line tracking system
Although the scanning spot laser sensor is ideal for the root pass 多传感器全自动焊接系统英文参考文献和翻译(3):http://www.751com.cn/fanyi/lunwen_370.html