Open Access
Development of an intelligent system for cooling rate and fill control in GMAW
C. J. Einerson,H. B. Smartt,J. A. Johnson,P. L. Taylor,K. L. Moore +4 more
- pp 1-5
Reads0
Chats0
TLDR
In this article, a control strategy for gas metal arc welding (GMAW) is developed in which the welding system detects certain existing conditions and adjusts the process in accordance to pre-specified rules.Abstract:
A control strategy for gas metal arc welding (GMAW) is developed in which the welding system detects certain existing conditions and adjusts the process in accordance to pre-specified rules. This strategy is used to control the reinforcement and weld bead centerline cooling rate during welding. Relationships between heat and mass transfer rates to the base metal and the required electrode speed and welding speed for specific open circuit voltages are taught to a artificial neural network. Control rules are programmed into a fuzzy logic system. TRADITOINAL CONTROL OF THE GMAW PROCESS is based on the use of explicit welding procedures detailing allowable parameter ranges on a pass by pass basis for a given weld. The present work is an exploration of a completely different approach to welding control. In this work the objectives are to produce welds having desired weld bead reinforcements while maintaining the weld bead centerline cooling rate at preselected values. The need for this specific control is related to fabrication requirements for specific types of pressure vessels. The control strategy involves measuring weld joint transverse cross-sectional area ahead of the welding torch and the weld bead centerline cooling rate behind the weld pool, both by means ofmore » video (2), calculating the required process parameters necessary to obtain the needed heat and mass transfer rates (in appropriate dimensions) by means of an artificial neural network, and controlling the heat transfer rate by means of a fuzzy logic controller (3). The result is a welding machine that senses the welding conditions and responds to those conditions on the basis of logical rules, as opposed to producing a weld based on a specific procedure.« lessread more
Citations
More filters
Journal ArticleDOI
Multivariable adaptive control of the bead profile geometry in gas metal arc welding with thermal scanning
TL;DR: In this paper, the dynamics of the weld profile geometry (i.e., the bead width and reinforcement height) are modeled experimentally with respect to the process conditions (weld speed and wire feed).
Journal ArticleDOI
Geometry Regulation of Material Deposition in Near-Net Shape Manufacturing by Thermally Scanned Welding
TL;DR: In this paper, a thermally scanned material deposition control method for near-net shape manufacturing of metal parts by welding is introduced, where the material is simultaneously deposited by a gas metal arc welding torch, with monitoring of the weld profile by two laser stripe profilometers.
Journal ArticleDOI
Multiplexed and distributed control of automated welding
TL;DR: In this paper, a continuous heat distribution and temperature monitoring on the entire weld surface is adopted to maximize the range of achievable weld features, and the necessary vector-scanning trajectories of the torch are regulated in real time by a distributed-parameter control strategy, integrated to the weld design software.
Journal ArticleDOI
Temperature distribution control in scanned thermal processing of thin circular parts
C.C. Doumanidis,N. Fourligkas +1 more
TL;DR: A new analytical description of the thermal field in thin disk-shaped parts, based on superposition of Green's functions, was developed for off-line analysis and embedded to a thermal distribution control scheme, driving the scanned torch motion and power by a new real-time simulated annealing optimization strategy.
Journal ArticleDOI
Advantages of an alternative form of fuzzy logic
J.A. Johnson,H.B. Smartt +1 more
TL;DR: A specific implementation of fuzzy logic that uses multiplication rather than finding the minimum to determine the conjunction between antecedents and is equivalent to a certain class of neural networks and can be trained to optimum values of the output actions of the system.