Abstract
Apollo-Clad Laser Cladding, an Edmonton-based division of Apollo Machine & Welding Ltd., applies wear and corrosion resistant coatings to downhole tool components using high-power laser cladding processes. Building on the successful completion of a multi-partner Industrial Research Assistance Program (IRAP) project involving Canadian and German institutions, Apollo is advancing the integration of machine-learning (ML) based coating defect detection models into active production environments.
The project focuses on deploying trained ML models as an operator-assist recommendation engine, providing near real-time feedback during laser cladding operations. The intent is to identify process instabilities and defect precursors that are traditionally only observable during post-process inspection or via destructive metallurgical evaluation, which is not feasible for production components. The ML framework utilizes spatiotemporal thermal history data acquired in situ and correlates these signals with metallurgically validated outcomes, including defect classification and coating quality metrics. This approach has demonstrated strong performance for nickel–tungsten carbide abrasion-resistant coating systems.
This work examines the practical challenges associated with integrating ML decision-support tools into high-mix, low-volume production laser cladding operations, including data latency, operator interaction, and process variability. The implications of scaling the methodology to approximately 35 active production alloy systems and 11 production laser cladding cells are also discussed, with emphasis on model generalization, data requirements, and deployment robustness. The successful implementation of ML-enabled laser cladding operations establishes a foundational capability for the adoption of more complex metal additive manufacturing processes, supporting a key strategic growth area for both Apollo’s business and the broader Canadian advanced manufacturing sector.