Ontwikkeling van Procesbewaking en Real-time Defectdetectie voor Wire-Arc Additive Manufacturing

Wire Arc Additive Manufacturing (WAAM) is a promising technology for fabricating large and complex metal components with reduced material waste and improved efficiency. To fully realize its industrial potential, ensuring consistent build quality through robust monitoring is essential. However, real-time process monitoring remains a challenge due to the complexity of the underlying welding physics and the occurrence of defects such as porosity, cracks, and geometrical deviations. This thesis advances WAAM monitoring through four key contributions. First, a multi-modal, high-resolution sensing system was developed, integrating acoustic, electrical, and optical modalities. The change of the data acquisition system from 500 kHz to 2 MHz enabled high-fidelity capture of transient events such as spatter and bubble bursts, with synchronized high-speed video providing robust physical validation. Second, a systematic comparison of acquisition rates revealed that while 2 MHz sampling allowed the use of high-frequency acoustic emission sensors, these did not improve defect classification. Instead, microphones operating in the 500 Hz-10 kHz range consistently provided the most informative signals, demonstrating that effective defect detection does not require extreme sampling rates. Third, an interpretable modeling framework—Sparse Wavelet Networks (SWaN) was introduced, combining wavelet scattering transforms with sparse principal component analysis to retain classification accuracy while highlighting physically meaningful features linked to melt pool dynamics and process stability. An optimized lightweight version (SWaN-Lite) further reduced computational overhead, making real-time monitoring feasible. Finally, the monitoring system and models were validated in laboratory and industrial settings, including case studies in laser cladding, demonstrating both robustness and transferability. Together, these contributions provide a sensor-integrated, interpretable, and real-time capable framework for WAAM defect detection, bridging the gap between high-performance data-driven modeling and the transparency required for industrial adoption.