Modelling and prediction of the mechanical properties of TIG welded joint for AISI 4130 low carbon steel plates using Artificial Neural Network (ANN) approach

Abstract

The mechanical properties (Ultimate Tensile Strength (UTS), modulus of elasticity (E), elongation and strain (e)) for twenty samples of AISI 4130 Low carbon steel plate were studied in this paper. Statistical design of experiment (DOE) using the central composite design method (CCD) was employed in Design Expert 7.01 software to generate DOE for twenty (20) experimental runs as input variables (current, voltage and gas flowrate) which were used in predicting and optimizing the output parameters (maximum UTS and maximum modulus of elasticity with corresponding elongation and strain). One out of the 20 welding runs was found to be optimum using the Artificial Neural Network (ANN) optimization approach. The same twenty (20) predicted variables were subjected to TIG welding experimentation which showed close proximity between the predicted and experimental values. Optimized ANN predicted output parameters were UTS of 421 MPa, modulus of elasticity of 793 MPa, strain of 0.61 and elongation of 61% while experimental values using the optimized input variables produced output parameters of 427 MPa for UTS of 421 MPa, 806 MPa for modulus of elasticity, strain of 0.62 and 62% elongation. Visuals of the weldment obtained from Scanning Electron Microscopy with Energy Dispersive Spectroscopy (SEM/EDS) revealed a uniformly distributed grain sizes in the weldment primarily composing of iron (Fe), chromium (Cr), molybdenum (Mo), and nickel (Ni). To save time, energy and resources required for welding experimentation processes, conventional software such as ANN can be used to obtain accurate results.

Keywords: Modelling, Prediction, Low carbon steel, TIG welding, Welding variables