Wire Arc Additive Manufacturing (WAAM) enables cost-effective fabrication of complex metallic components but faces challenges in achieving consistent tensile strength for cylindrical parts with intricate internal features (e.g., cooling channels, helical grooves), where conventional machining is often infeasible or prohibitively expensive. This study introduces a novel stacked ring substrate strategy with pre-formed low-carbon steel rings defining complex internal geometries, followed by external WAAM deposition using ER70S-6 wire to overcome these limitations. Five process parameters (welding current: 110–130 A; offset distance: 2.5–3.0 mm; Step Length: rotary to straight; torch speed: 400–500 mm/min; weld thickness: 2.0–3.0 mm) were optimized using a Taguchi L25 orthogonal array (25 runs in triplicate). ANOVA identified Step Length as the dominant factor, with straight paths significantly reducing thermal cycling and improving interlayer bonding, alongside a notable current × speed interaction. Optimal settings achieved tensile strengths of 280–290 MPa, significantly below wrought ER70S-6 benchmarks (400–550 MPa) due to interfacial weaknesses at ring fusion zones and thermal accumulation from stacked cylindrical geometry, a limitation acknowledged in the absence of microstructural or thermal history data. A Random Forest Regressor predicted strength with R² = 0.9312, outperforming conventional models. This hybrid approach significantly enhances design freedom and mechanical reliability for high-value cylindrical components in aerospace and tooling, establishing a scalable, data-driven framework for geometry-constrained WAAM optimization.

This study investigates the fatigue bending strength, measured as the Mean of Fatigue Cycles (N), of cylindrical components produced by external wire arc additive manufacturing (WAAM) through a Taguchi L25 orthogonal array and linear regression analysis. Five welding parameters—welding current (Ampe), offset distance (mm), step length (mm), welding speed (mm/min), and specimen gauge diameter (mm)—were evaluated to maximize N using signal-to-noise (S/N) ratios. Nominal bending stresses (σ) ranged from 45 to 54 MPa, ANOVA on raw replicate data (25 runs, 3 replicates) confirmed specimen gauge diameter = 17 mm and weld current = 125 A as dominant, with F = 171.62 (p < 0.001, eta² = 0.62 [95% confidence intervals (CI) 0.55–0.68]) for specimen gauge diameter and F = 6.13 (p < 0.001, eta² = 0.13 [95% CI 0.08–0.18]) for weld current, accounting for ~75% of the variance. Optimal settings (offset distance = 3.0 mm, step length = 1000 mm, welding speed = 550 mm/min, specimen gauge diameter = 17 mm) achieved S/N = 111.35 dB, predicting N ≈ 350,000–380,000 cycles, a 22–33% improvement. Interactions between specimen gauge diameter and speed, and between weld current and offset distance, suggested enhanced strength at speed = 400–450 mm/min for specimen gauge diameter = 17 mm. Basquin’s law (b ≈ 0.72, R² = 0.992) confirmed weld current as key. The linear regression model (adjusted R² = 0.9506) had coefficients for specimen gauge diameter (+70,120 cycles/mm, p < 0.001) and weld current (+1088 cycles/Ampe, p = 0.02), but lower test R² = 0.7212 via cross-validation (60/20/20 split) indicates overfitting due to small dataset size (25 runs), suggesting larger datasets or nonlinear models (e.g., polynomial regression, RSM). Confirmation runs (N = 317,082, 95% CI [287,000–347,000]) validated the results within ~13% error. WAAM reaches 80–90% of traditional manufacturing (TM) fatigue performance, with a 10–20% gap due to the microstructure; recommendations include post-treatments and safety factors (~1.2).

The implementation of artificial intelligence models on edge devices is increasingly popular, bringing many values in reducing latency, effectively utilizing bandwidth, improving data security, enhancing privacy and reducing costs for users. However, this work poses many challenges in terms of accuracy, processing speed, hardware resources and model size for devices constrained by limited hardware. Binary Neural Network (BNN) is proposed as a potential solution to reduce resource requirements by using only 1 bit for quantizing. In this study, BNN network is optimized by binary quantizing both weights and activation functions with XNOR-popcout multiplication to optimize BNN network. The results show that BNN network model is lighter in memory footprint when deployed on hardware with limited computational resources, less computational time than conventional BNN network which helps the model execute faster as the network architecture becomes less complex, with acceptable accuracy on two datasets MNIST and Fashion MNIST. The proposed BNN model resul can be deployed on edge devices for IoT applications.

This study investigates the fatigue performance of 3D-printed metal shafts fabricated via Wire Arc Additive Manufacturing (WAAM) with stacked steel ring substrates under rotating bending (ISO 1143:2021). A Taguchi L25 orthogonal array was used to analyze five process parameters: ring diameter, current intensity, torch speed, ring thickness, and contact tip to workpiece distance (CTWD). Analysis of Variance (ANOVA) identified ring diameter as the dominant factor, significantly enhancing fatigue life at 14.0 mm by reducing stress concentrations. Current intensity (125 A) and torch speed (550 mm/min) further improve weld quality and microstructure, while ring thickness (1.0 mm) and CTWD (1.5 mm) have minor effects. A linear regression model (R² = 0.9603) accurately predicts fatigue life, with optimal settings yielding 299,730 cycles. The stacked-ring configuration enables intricate structures like cooling channels, ideal for aerospace and automotive applications. The 3.5% unexplained variance suggests parameter interactions, warranting further investigation into shielding gas effects and multiaxial loading to broaden material and loading applicability.