Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


2. HISTORICAL BACKGROUND

An extensive survey on most of the aspects of manufacturing automation, taking into account different levels of automation, reveals the hierarchical or multilayered architecture of expert systems applying various types of information and intelligent components at different levels. Some information is about the structure of the system and other parts are more connected to the lower level of signals and their interpretation. Each type of information has its own appropriate knowledge representation.

Figure 1 shows the hierarchical feedback control diagram for submerged arc welding as an example. Generally, welding is considered to be one of the most difficult fields for automation. Welding processes involve many physical factors, such as geometrical, thermohydrodynamical, mechanical, and metallurgical. Furthermore, very diverse welding methods and equipment types must be taken into account. On the other hand, many pieces of knowledge concerning the determination of welding conditions and welding process control are very difficult to quantify.

Since welding is situated downstream in the whole flow of manufacturing, more restrictions due to accumulated variability of the factors are imposed. The existing standards cannot correspond to these changes (Fukuda, 1988). Furthermore, different kinds of welding for different kinds of metal or nonmetal, and welding for rehabilitation that differs from case to case, are also considered. Thus, the great complexity of the weld planning task requires computer-aided or automated welding procedure selection and online process control techniques for reducing the burden of welding engineers.

In real-time control applications, an intelligent method can be used to generate a control action directly or adapt a conventional control scheme by tuning its parameters according to the performance of the system. At the lowest control level, a variety of intelligent control schemes have been used in place of conventional control algorithms. For batch weighing control of powdery or granular materials, for example, time parameters of feeding rate are searched and learned from experience of previous batches to improve the weighing performance. The initial feeding pattern is determined based on the approximately optimum control strategy using an empirical process model (Yasuda and Tachibana, 1988). Such intelligent control schemes are called learning control or optimizing control.


FIGURE 1 Control diagram of process variables.

However, control using only a single input variable may not yield optimum conditions for most of manufacturing applications. For example, in robotic welding, arc length and weld shape should be kept constant during welding to obtain a good quality weld. The mass balance and the heat balance are simultaneously controlled. Since the state equations of the wire feed system and welding phenomena are complicated, feedback control algorithms based on modern control theory are nearly inapplicable. Therefore, an intelligent control method such as fuzzy logic control, which is suitable for cooperative control of several process variables, should be developed.

In machining, conventional computer control schemes suffer from the drawback that their operating parameters, such as feed rate, depth of cut, and cutting velocity, are programmed offline. If perfect and fixed mathematical models of the manufacturing process were available, it would be possible to select optimum operating conditions prior to the process. However, no reliable models are available due to the complexity of the process, inherent variations in tool and workpiece properties, and other varying environmental conditions. User-chosen constraints also are variable according to different manufacturing conditions.

Formerly, theoretical relationships expressing each of output variables in terms of the input variables have been modeled empirically, but are not very useful in practical situations because of various differences in workpiece and tool properties and machine conditions. A large number of empirical data have to be generated and processed offline to develop the suitable model. So, there is a strong motivation for developing schemes that are able to learn appropriate input-output mappings based solely on online sensor measurements. The success of unattended manufacturing depends, to a great extent, on the development of appropriate sensors and computer-based learning strategies.

Furthermore, adaptive machining control must be achieved in the presence of more crucial constraints, such as tool wear, tool deflection, and chatter vibration. These phenomena are difficult to measure directly and must be inferred from process variables that can be monitored. The knowledge-based system could incorporate these constraints easily as compared to conventional control systems.

Up to the present, many expert systems have been developed and implemented in the welding and machining industries. However, many problems exist in practical use to utilize the outcome provided by such expert systems, because they are mainly confined to the selection of plausible and conservative procedures and their conditions from those available in a database based on industrial handbooks or textbooks. From the experience of expert system development for the determination of welding conditions, it was considered that pieces of knowledge represented as production rules in a knowledge base would serve nothing to welding engineers, so it should be necessary to allow the user to process the pieces of knowledge in the knowledge base according to his own way, which is based on his experience (Fukuda, 1988). Ultimately, domain experts should be able to develop, modify, and integrate their rule bases and resolution strategies easily without intervention by computer engineers.

From these facts, hierarchical integrated expert systems are required which support knowledge representation, including inference mechanism. At the higher levels, explicitly described knowledge with advanced reasoning strategies for problem-solving plays an important role. At the lower levels, normal signal processing for in-process monitoring and control plays an important role, as used in conventional mathematical methods. To sum up, generally, effectively integrated expert systems are not found, or under development, corresponding to multilayered control structures from problem-solving or task planning to real-time adaptive control.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.