PSI - Issue 25

Romanin Luca et al. / Procedia Structural Integrity 25 (2020) 149–158 Author name / Structural Integrity Procedia 00 (2019) 000–000

150

2

1. Introduction Large structures bring with themselves complications in computational procedures even for structural calculations because of limitation in computational capacity and subsequent post calculations such as fatigue. Shell elements represent a well-known solution in industry to simplify the numerical model and keep the computational cost low. The computational problem is a limit for the prediction of welding deformations using Computational Welding Mechanics. To obtain the maximum level of reliability, one method is modelling only the welded joint and the near region with solid elements while connecting them to the structure modelled by shell elements. However, it is often not feasible both for the time involved in prepossessing and for the computational resources needed for solving large models. Moreover, a design change in the plate thickness or weld throat involves expensive changes in terms of pre processing and re-solution times. In fact, computational welding mechanics has high computational requirements even for small components, for this reason it can be hardly applied for large assemblies such the case as civil structures or ships. Moreover, in a manual production context, the scatter on input parameters does not justify the use of such sophisticated procedures. In a one-of-a-kind production system, the testing of different welding sequences is not economical. The problem is often overcome by the skills of welders aquired over years of experience; in case the assemblies do not respect tolerances, they are corrected using flame straightening or press bending. As a matter of fact, also line heating, which is used to straighten plates or impose a curvature, is a field of research in the shipbuilding industry, as reported by Ha and Shipyard (2007); Vega et al. (2010); Wang et al. (2013). A multitude of approaches has already been developed in literature, but few of them have reached maturity and found an industrial application. The inherent deformation method by Deng et Murakawa (2007) has been implemented in Virfac Mega simulation code while the Direct Strain as Boundary Conditions, Kim et al. (2015), is probably used by Samsung Heavy Industries. The purpose of the present work is to develop another simplified finite element-oriented approach, named Virtual Bead method, to predict residual deformation. The accuracy of the proposed method is based on an inverse analysis procedure. The inherent deformation method takes advantage of a dedicated interface element which does not permit to use this method on a general-purpose FE code. The concept of Strain as Direct Boundary has been extended to have more flexibility using orthotropic shell elements to control the direction of loads and apply them on regions of elements (rather than nodes), making the method mesh-insensitive. The advantages of lowering the cost of simulation are self-evident, welding of large assemblies could become a more common analysis in a similar way to foundry, additive manufacturing or forging simulation. The purpose of the simplified approach is based on reutilizing the structural shell element model which is already available in numerical codes. Computational times are reduced compared to solid elements models and the method can be easily integrated with the normal design workflow. The only additional step is to apply some divisions in the original geometry near the welded joint; loads have to be applied on these regions. Multi-run welds are included using a lumped approach neglecting the transient phase of the weld built up. Compared to methods such inherent strain, the stress-strain filed is skipped in favour of residual deformations only. A specific inverse analysis has been developed to calculate equivalent loads to be applied. Since the inverse analysis is an integral part of the procedure, a section is dedicated to describe it in detail. The main steps of the proposed method are (Fig. 1):  Obtaining deformation data from an experimental test or numerical simulation, as reference  Performing an Inverse Analysis to calibrate the FE Equivalent Loads to be applied to the model  Storing the calculated equivalent load in a database Nomenclature CWM Computational Welding Mechanics PSO Particle Swarm Optimizer GMAW Gas Metal Arc Welding