ABSTRACT
This paper tends to discuss the relation between welding parameters (Voltage, Arc
Current and Travel Speed) and the deposited lines shape (width, height and
penetration) in metal deposition process. Due to the high nonlinearity of the process
ANN (artificial neural network) was found to be the best chose for representing it.
ANN is trained off-line under different operating conditions then used for prediction of
the system model for further optimization of the process. The results show the
capability of the developed ANNs to represent the process properly. Supervised
learning with back propagation technique was used to make the networks. Best
network for width consisted of tansig functions with an output layer of linear function.
Traingd function gave the best result for width. Best network for penetration consisted
of tansig functions with an output layer of linear function. Traingd function gave the
best result for penetration. Best network for height consisted of mixture of radial
basis, tansig functions with an output layer of linear function. Traingd function gave
the best result for height.