In-Situ Porosity Detection in LPBF Using Machine Learning - Augmented Ultrasonic Emissions

Abstract

Laser powder bed fusion (LPBF) is a metal additive manufacturing process that selectively fuses regions of a distributed metal powder bed using a high-energy laser beam. While LPBF enables the manufacture of complex geometries and processing of challenging materials like titanium alloys, producing defect-free components remains a significant challenge due to the process's highly dynamic nature. Among the various defects in LPBF, porosities are particularly concerning as they directly compromise the component's mechanical properties and structural integrity. Their subsurface nature makes detection especially challenging, as conventional optical monitoring methods prove ineffective. While X-ray computed tomography (XRCT) offers capable post-production inspection, its technical complexity and high costs render it impractical for in-situ monitoring applications.This work investigates ultrasonic emissions as a promising method for in-situ porosity detection, building upon its established use as a non-destructive testing (NDT) technique for defect detection in metal components. However, conventional pulse-echo ultrasonic NDT cannot be directly implemented in LPBF processes, as it fails to provide sufficient information about the size and distribution of clustered porosities which is crucial for robust process control. Additionally, the necessity of adapting the mounting configuration of the transducer to avoid interference with the LPBF process increases wave attenuation and signal interference. To address these challenges, machine learning (ML) is employed to decode and characterize the reflected signals, enabling rapid, real-time detection of porosity formation. This method could also provide a reference for mitigation of porosity formation through in-situ process control.