Study of micro-porosity in electron beam butt welding

As complete elimination of porosity from the weld is very difficult, next option available to minimize this porosity, which crucial for safe performance welded components. However, investigation through experiments alone tedious and time consuming. Additionally, limited models are in literature accurate prediction different attributes. The present study, thus, addressees both experimental as well modelling aspect on study micro-porosity during electron beam welding (EBW) SS304 plates. Welding parameters reported have significant influence micro-porosity. Hence, influences these attributes, namely pores number, average diameter, sphericity extensively studied experimentally employing optical microscopy (OM), scanning (SEM), X-ray computed tomography (XCT), Raman spectroscopy. This followed by an elaborate using seven popular well-recognized machine learning algorithms (MLAs), multi-layer perceptron (MLP), support vector regression (SVR), M5P model trees regression, reduced error pruning tree (REPTree), random forest (RF), instance-based k-nearest neighbor algorithm (IBk), locally weighted (LWL). These techniques enhance chance obtaining better predictions said attributes overcoming effect data-dependence other limitations individual MLAs. model-predicted data also validated data. Statistical tests Monte-Carlo reliability analysis additionally utilized evaluate performances employed algorithms. IBk MLP overall found perform well.