Defect monitoring in wire arc additive manufacturing using frequency domain analysis

Abstract

Efforts to integrate Wire Arc Additive Manufacturing (WAAM) into industrial settings drive a focus on refining in-process defect detection. WAAM commonly employs waveformcontrolled welding techniques, like pulsed or controlled dip transfer processes, to enhance material properties and reduce heat input. The cyclic nature of voltage and current waveforms in these processes suggests that valuable information exists in their frequency content for assessing the process state and potential defects. This study introduces the employment of frequency domain analyses, utilizing Fast Fourier transform (FFT) and discrete wavelet transform (DWT) methodologies, to identify anomalies in welding signal data. Statistical assessments reveal the efficacy of online frequency domain analysis in extracting valuable insights across various WAAM processes. The research showcases the utility of this information in developing unsupervised learning techniques for online anomaly detection systems tailored to WAAM, proficient in identifying issues like arc instability, porosity, and geometrical defects caused by arc blow and humping.