Page 126 - Welding Robots Technology, System Issues, and Applications

Robotic Welding: System Issues 113

4.2.1 Knowledge Base
As already explained in Chapter 2, the welding process is very difficult to model
which suggests that a rule base approach [8],[22],[23] would be advisable instead
of an explicit mathematical model approach. That means exporting to a knowledge
base the relevant information for a specific welding process and setup, so that the
adaptive behavior could be still obtained. That can be done in several ways: using
neural networks or fuzzy rule base approaches, simple lookup tables for the
relevant parameters, etc. As always, simplicity is desirable and a solution that
could grow with experience, just by adding new rules or more training, would be
ideal for a physics process as complex as welding.

4.2.2 Sensors and Interfaces

The existence of smart sensors is fundamental to achieve a good solution for a
robotic welding system that is necessarily distributed, i.e., it is based in the
distribution of functions through the components of the system as a policy for
efficiency and organization. Since the most promising sensors and sensor
techniques have already been presented in Chapter 3, the focus here is given to the
interfaces and system architecture.

The basic components of a single cell robotic welding system include a robot
manipulator, the robot controller, the welding torch, the welding power source, and
the sensors adopted to monitor and sense the process parameters (Figure 4.6).

Figure 4.6. Single cell robotic welding system

Nevertheless, an industrial application may include several robots, and sensors, and
welding power sources, which means that data networks must be present to