In industrial manufacturing, oil contamination on the surface of the workpiece is difficult to completely avoid during laser beam welding (LBW). In this study, LBW experiments were conducted on stainless steel sheets with controlled oil contamination located on the top surface of the upper plate and at the interface between the upper and lower plates in an overlap configuration. The experimental results demonstrate that oil evaporation affects the pressure balance within the keyhole, inducing violent keyhole oscillations and unstable melt pool dynamics. This leads to the formation of various weld defects, including lack of fusion, oil-induced porosity, keyhole-induced porosity, and blowout defects. Therefore, in industrial production, contamination of workpieces by oil should be detected. For automatic detection of oil contamination, multiple machine learning-based algorithms were deployed using acoustic signals. Airborne and structure-borne Acoustic Emission (AE) signals were recorded simultaneously during the welding process. Feature extraction from the acoustic signal channels is carried out by the Wavelet Scattering Transform (WST). The results show that the identification F1-macro score for oil contaminated regions exceeds 99% with 10 ms temporal resolution. This indicates that contaminated and clean welding regions can be reliably distinguished. The study suggests that even under industrial acoustic interference from cross-jet with 5 bar air pressure, AE provides a promising approach for in-situ detection of oil contamination and process monitoring in LBW.