Edge-focused wire arc additive manufacturing: Method development with ANN-based stress-strain and mass-efficiency

This study evaluates Edge-Focused Wire Arc Additive Manufacturing (EF-WAAM) for CT38 steel using an ER70S-6 filler. EF-WAAM employs an edge-guided toolpath with a prescribed travel angle to localize heat input per unit length, narrow the heat-affected zone, and mitigate residual stress relative to conventional WAAM. Under standardized 3-point bending with identical specimen geometry and span, the maximum flexural stress of EF-WAAM builds ranges from 2,414.21 to 3,338.11 MPa (n = 5 per condition). The best case improves 171% over the CT38 substrate (1,231.5 MPa) and exceeds typical values for conventional WAAM (< 2,000 MPa). Mass efficiency, reported as strength-to-density (σ/ρ), reaches 420.4 MPa·cm³·g ⁻ ¹, representing gains of 40.4% versus the substrate and 83.3% versus conventional WAAM. An artificial neural network (ANN) maps process variables-current, step-over distance, travel angle, travel speed, layer thickness, and strain-to stress and reconstructs full stress-strain curves with high agreement on training, validation, and held-out test sets. ANOVA/S-N and sensitivity analyses indicate layer thickness is the dominant factor within the explored window, with beneficial interactions from travel speed and current that moderate thermal gradients. The study also demonstrates the feasibility of internal features (e.g., 3D spiral channels) while maintaining controlled thermal fields. Overall, EF-WAAM delivers higher flexural strength and improved mass-specific performance within a standardized, reusable evaluation pipeline, offering a transferable workflow for other WAAM variants. Beyond a single case, we provide a reusable evaluation pipeline, actionable parameter windows, and cross-variant metrics (σ/ρ, AER) together with an ANN routine that reconstructs full stress-strain curves from process vectors. These assets enable practitioners to transfer the method to related WAAM variants without additional sensing or bespoke hardware.