Optimization of geometric parameters in a welded joint through response surface methodology Eusebio J. Martinez-Conesa a,⇑ , Jose A. Egea b , Valentin Miguel c , Carlos Toledo a , Jose L. Meseguer-Valdenebro a a Department of Building Technology, Universidad Politécnica de Cartagena, Paseo Alfonso XIII, 48, 30203 Cartagena, Spain b Department of Applied Mathematics & Statistics, Universidad Politécnica de Cartagena, Av/ Dr. Fleming s/n, 30202 Cartagena, Spain c Department of Applied Mechanics and Projects Engineering, Universidad de Castilla La Mancha, Campus Universitario s/n, 02071 Albacete, Spain highlights  Experimental desing for a GMAW welding process to maximize the amount of information.  Response surface-based modelling (RSM) to quantify response variables of interest.  Statistical model selection to obtain the most informative models.  Statistical model checking for definitive models to ensure inference capabilities.  Multiobjective optimization to identify the Pareto front of optimal solutions. article info Article history: Received 7 April 2016 Received in revised form 19 July 2017 Accepted 22 July 2017 Keywords: Response surface methodology Welding process Multiobjective optimization Geometric parameters abstract This work makes use of experimental design and response surface methodology to model Gas Metal Arc Welding processes. The correlations among three key geometric parameters, ie., penetration, bead width and overthickness, and four technological variables that define the welding process are quantified. Based on experimental data and using model selection techniques, a mathematical model has been deduced for each of the response variables herein presented. Using these models, a multiobjective optimization is car- ried out to find the space of optimal solutions (i.e., the Pareto front). After a preliminary study of the rela- tionships between independent and response variables, regression models are built. These models capture the data variability reasonably well (e.g., around 70% of the variability). These models are the basis to perform the multiobjective optimization using the e-constraint approach. Results reveal that the conditions which favour a good balance between maximum penetration and minimum bead width and overthickness, involve a high value for gas flow rate, low values for electrode feed rate and voltage, and an intermediate value for the electrode position. This permits the authors to define the welding con- ditions that lead to an optimum joint geometry and then to guarantee its properties. Ó 2017 Elsevier Ltd. All rights reserved. 1. Introduction The mechanical properties of a welded joint depend on the geometry of the bead together with other factors. The study of the geometric factors of the weld bead has an important consider- ation for the design and manufacturing of welded constructions. The geometry of the bead directly affects the quality of the welding in the building of structures [1]. In order to obtain a correct weld, it is essential that the fusion between the base metal and the mate- rial deposited is appropriate. The surface of the base metal which is part of the joint must be completely melted until it forms a suffi- ciently deep bead. If the drops of metal from the electrode and the heat of the arc are not able to melt the base metal, then the bead will have little penetration. The dimensions which best define the geometry of the bead are its width, its penetration, and its overthickness, Fig. 1 (overthickness is referred to ‘‘height” in the figure). This geometry depend on the technological parameters of the specific welding process, in relation with the heat contributed in the process and with the thermal conditions in which it occurs [2]. It is therefore important to establish appropriate welding parameters in order to produce a stable weld bead. In general, the optimization criteria are aimed at maximizing the weld bead’s penetration while maintaining the values of bead width and over- http://dx.doi.org/10.1016/j.conbuildmat.2017.07.163 0950-0618/Ó 2017 Elsevier Ltd. All rights reserved. ⇑ Corresponding author. E-mail address: eusebio.martinez@upct.es (E.J. Martinez-Conesa). Construction and Building Materials 154 (2017) 105–114 Contents lists available at ScienceDirect Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat