Page 446 - Sensors and Control Systems in Manufacturing

400
Ei g h t
Cha p te r

8.7.1 Runtime Phase
The runtime phase of the LFF acquires images of partially visible
objects and determines their identities, positions, and orientations.
This processing occurs in four steps:
1. Reading task information
2. Locating local features
3. Hypothesizing objects
4. Verifying hypotheses
The procedure (Fig. 8.15) is to input the object model together
with the list of focus features and their nearby cofeatures. Then, for
each image, the system locates all potentially useful local features,
forms clusters of them to hypothesize object occurrences, and finally
performs template matching to verify these hypotheses.

8.8 Ultrasonic End Effector
An end effector on a welding robot (Fig. 8.16) contains an ultrasonic
sensor for inspection of the weld. An ultrasonic sensor detects such
flaws as tungsten inclusions and lack of penetration of weld. The end
effector determines the quality of a weld immediately after the weld
contact has been made, while the workpiece is still mounted on the
weld apparatus; a weld can be reworked in place, if necessary. The delay
caused by the paperwork and setup involved in returning the work-
piece for rework is thereby avoided.
The ultrasonic end effector can be mounted on any standard gas
tungsten arc welding torch. It may also be equipped with a through-
the-torch vision system. The size of the ultrasonic end effector is the
same as that of a gas cup with a cathode.
A set of extension springs stabilizes the sensor and ensures that
its elastomeric dry-couplant pad fits squarely in the weldment sur-
face. The sensor can be rotated 360° and locked into alignment with
the weld lead. A small force-actuated switch halts downward travel
of the robot arm toward the workpiece and sets the force of contact
between the sensor and the workpiece.

8.9 End Effector Sound-Vision Recognition Sensors
The sound recognition sensor consists of a source that emits sound
waves to an object and a sound receiver that receives the reflected
sound waves from the same object (Fig. 8.17). The sound recognition
sensor array consists of one sound source and one to as many as 16
receivers fitted intricately on an end effector of a robot.
The sound-vision recognition sensor array measures reflections
from some surface of interest on the object, called the measured surface,