Vision-based control of contact tip to work distance in dot-by-dot WAAM of Al4043

The dot-by-dot approach in Wire Arc Additive Manufacturing (WAAM) represents a fundamental shift from conventional continuous deposition, enabling the fabrication of complex, lightweight three-dimensional aluminum lattice structures. While this strategy offers superior morphological control, it is highly susceptible to heat accumulation, which triggers a progressive and uncontrolled increase in the Contact Tip to Work Distance (CTWD). In aluminum alloys, the lack of a reliable correlation between electrical arc signals and CTWD complicates real-time height regulation, often leading to process instability and compromised shielding gas effectiveness. To address this challenge, this study proposes a vision-based framework leveraging a Convolutional Neural Network (CNN) for real-time CTWD sensing, by detecting the wire tip during the discrete cooling intervals between consecutive droplets. The output of the supervised machine learning architecture is subsequently mapped to the CTWD through a linear regression model. Two different control strategies are assessed within this research to regulate the CTWD during the dot-by-dot deposition of Al4043 single struts: a rule-based scheme without inter-droplet waiting times and an adaptive robot motion strategy that dynamically adjusts the vertical step-up based on real-time height estimation. The experimental results demonstrate that the proposed approaches enable stable deposition of aluminum single struts up to the target height while preserving process stability. In addition, a comprehensive geometrical and mechanical characterization was performed for the different control strategies, allowing their respective advantages and limitations to be identified. Finally, the approach provides a solid basis for future extensions toward more complex, multi-strut assemblies.