THERMAL-PHASE TRANSFORMATION MODELLING AND NEURAL NETWORK ANALYSIS OF FRICTION WELDING R44 Welding in the World, Vol. 53, n° 3/4, 2009 – Peer-reviewed Section 1 INTRODUCTION Friction welding is now well established as a means of joining many different types of materials, because it has proved itself to be a reliable and economical way of producing high quality welds. The subject has been reviewed thoroughly [1]. The main advantages of the process may be summarized as follows: i In FW due to the highly concentrated heat at the weld interface, the heat-affected zone (HAZ) is very narrow. As a result of the very small HAZ, the wel- ding distortion is kept minimal, and also the variati- ons in mechanical properties of the base metal are limited to a small region. ii The process is very energy efficient resulting in fast joining time on the order of seconds. The power required for the process is as low as 10-25 % of that used in butt welding. iii Because it is a solid-state process, defects asso- ciated with melting-solidification phenomenon are not present. iv An airtight weld is made across 100 % of the joint interface, eliminating the risk of porosity, voids, leaks or cracks. Friction welds have superior strength wit- hout sacrificing product integrity. v Joint preparation in FW is minimal and filler metal, flux and shielding gas are not required, therefore, the reduction in cost in FW is one of the most signi- ficant benefits of the process. vi The process is environmentally friendly, producing minimum smoke, fume and slag. THERMAL-PHASE TRANSFORMATION MODELLING AND NEURAL NETWORK ANALYSIS OF FRICTION WELDING OF NON-CIRCULAR EUTECTOID STEEL COMPONENTS M. Maalekian H. Cerjak ABSTRACT A thermal phase transformation finite element model is presented to study the orbital friction welding of eutectoid steel components. The experimental microstructure observations and the hardness profile in the weld and heat- affected zone verify the predictive capability of the proposed model. Moreover, based on the experimental friction welding data, a neural network analysis is presented which relates the weld integrity in terms of the size of the flash (upset) to the friction welding parameters. IIW-Thesaurus keywords: Artificial intelligence; Friction welding; Martensite; Reference lists; Simulating; Steels. Dr. Mehran MAALEKIAN (mehran.maalekian@ubc.ca) was with Christian Doppler Laboratory for Early Stages of Precipitation, Vienna University of Technology, Vienna (Austria) and with Institute for Materials Science and Welding, Graz University of Technology, Graz (Austria), and now is with Department of Materials Engineering, The University of British Columbia, Vancouver (Canada). Em. Univ.-Prof. Dipl.-Ing. Dr. Horst CERJAK (cerjak@ tugraz.at) is with Institute for Materials Science and Welding, Graz University of Technology, Graz (Austria). Doc. IIW-1945-08 (ex-doc. IX-2275r1-08) recommended for publication by Commission IX “Behaviour of metals subjected to welding”.