This study aims to analyze key performance metrics such as surface roughness and material removal rate in the WEDM of Hardox 450 steel using artificial intelligence-based (ANN) and statistical (RSM) modeling approaches. The study also aims to establish a multidisciplinary relationship between modeling and experimental results by examining the effects of process parameters on surface integrity using microstructural characterization methods (SEM, EDX and XRD).
In the experimental studies, input parameters were Ton, Toff, wire feed and voltage. Output parameters were surface roughness and material removal rate. The data were analyzed and compared using the RSM and ANN models. Furthermore, microstructural and chemical changes of the cut surfaces were investigated using SEM, EDX and XRD analyses.
Analysis showed that Ton and voltage parameters are particularly decisive for surface roughness. A dominant effect of Ton was observed on MRR. The ANN model demonstrated superior performance to RSM, with higher accuracy (R2 > 0.98) and lower error rates. Microstructural analyses revealed surface resolidification, microcracks, and element transfer from the wire material to the workpiece, particularly at higher energy levels.
This study is one of the few that comprehensively evaluated the WEDM method in terms of performance outcomes (SR and MRR), microstructural changes, and surface integrity. The study developed and validated high-accuracy predictive models by combining statistical (RSM) and artificial intelligence (ANN) methods with experimental data. In this respect, the study demonstrates the effectiveness of AI-based approaches in modeling the complex processes of WEDM. It significantly contributes to the precision manufacturing processes of difficult-to-machine materials such as Hardox 450.
