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


3.7. EXPERT SYSTEMS FOR WELDING

Up to the present, many expert systems have been developed and implemented in the welding industry, especially for welding procedure selection and weld defect diagnosis. This section briefly discusses the features of expert systems for welding. These include online and offline expert system techniques to help the user to analyze weld tasks in the preweld, weld, and postweld phases.

In the preweld phase, expert systems emulate weld tasks such as joint design, including edge preparation, welding procedure selection, and material selection. It is also recommended to perform welding procedure selection based on preweld inspection. Expert systems produce procedures that achieve acceptable weld bead geometries and avoid hydrogen cracking, solidification cracking, and fracture toughness impairment.

During the weld phase, sensors to detect process features are used, and based on this information along with the preweld information, conditions varying from the present welding procedure are identified, analyzed, and modified accordingly.


FIGURE 2 Example of software organization of expert system for welding.

Postweld expert systems conduct fault diagnosis and weld inspection interpretation. For weld defect diagnosis, the user is asked for information about welding process variables and the welding methods used, and then the system lists its conclusions about the most probable causes of the defect.

Figure 2 shows the software structure of an integrated expert system for welding procedure selection and process control (Yasuda and Tachibana, 1996). To clarify the practices of expert systems in welding, the expert system is described briefly in the following:

For welding automation, the expert system and a welding process controller are combined with a welding robot, various sensors, and/or a CAD system. The procedure selection software is part of the weld task planner that contains all the necessary knowledge to supervise the weld task planning. The user defines the preset variables such as base material composition, plate thickness, joint type, and welding position. The user also defines acceptable limits on parameters such as weld beat geometry, penetration, and cooling rate. Data such as electrode material and size, and desired type of metal transfer are optionally required. The system then can generate values for voltage, current, wire feed rate, and welding speed. Conventional reasoning methods, such as forward reasoning and backward reasoning are contained in the expert system to support different and bidirectional reasoning according to the user's needs at arbitrary stages.

The secondary manipulated variables are selected using a set of production rules that control the access to the weld joint and welding equipment databases. The hierarchical class structure indicating inheritance relations between objects in the weld joint database is shown in Figure 3.


FIGURE 3 Hierarchical class structure of weld joint data.


Kb::
[Thickness, `mm', positive,[ ], [1, 100]],

[Welding position, `', atom, [ ], [`flat', `horizontal', `vertical', `overhead']],

[Groove, `', atom, [Thickness, Welding position], [`V', `J', `X', `K', `L', `U']]

Rule1:: Welding position == `flat',
        Thickness >= 6,
        Thickness <= 30

        ==> Groove = `V'


FIGURE 4 Example of knowledge and rule representation for welding procedure selection.

For example, the type of weld groove is selected based on the plate thickness and the welding position using a production rule as shown in Figure 4, where the associated knowledge and the rule are written in a conventional expert system shell language.

Forward reasoning by data-driven or bottom-up inference is used for adjustment of parameters to satisfy specified constraints in the welding condition setting as shown in Figure 5. The system can schedule preheating temperature, and postweld heat treatment temperature and time. All feasible solutions to meet specified constraints are graphically visualized for the user to select the most allowable candidate on the screen.

Object-oriented approaches have been used for building databases in welding engineering. In the developed example, data of weld joint, weld equipment, and weld task are represented as frames into the respective class hierarchy. The static data of an object is expressed by "attribute" slots of the frame. Numerical calculating procedures and heuristic logic concerning the object are expressed by "method" slots and "rule" slots, respectively.

Figure 6 shows a neural expert system for the determination of submerged arc welding conditions. The input nodes of the neural network represent the preset, fixed variables or the secondary manupulated variables which include base material type, plate thickness, joint type, welding position, electrode material type, groove type, etc. The output nodes represent primary manipulated variables, which include welding current, welding speed, preheating temperature, and postheating temperature and time. The prototype can learn a provisional set of samples from industrial standards by the back propagation algorithm with inertia terms, and can be embedded in the developed knowledge-based expert system (Yasuda and Tachibana, 1996).


FIGURE 5 Data and control flow diagram in welding condition adjustment.


FIGURE 6 Neural network for the determination of welding conditions.

In the neural expert system, the user can change the number of the hidden layer and the steepness of the sigmoid function of the neuron model so that the performance of learning is satisfactory to determine the welding conditions in the allowable range of process variables.

The realized hardware architecture is composed of an engineering workstation and a network of transputers (IMS T805). The former, including the neural expert system, deals with the expert welding procedure selection, while the latter implements the real-time expert welding control. Owing to the distributed properties of the expert system, some wide area network such as Internet and ISDN systems can be utilized for communication and cooperative reasoning between the global expert system on the host workstation and local expert control systems at remote sites. At remote sites, only knowledge concerned with the specified items such as plate thickness, joint type, welding method, and equipment, can be loaded into the knowledge base when the expert system is executed.


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.