Adaptive time-domain acoustic emission framework for post-deposition cooling transition detection in directed energy deposition

During the metal additive manufacturing (AM) process, each deposited layer undergoes rapid solidification driven by extremely high cooling rates. Directed Energy Deposition (DED) is a method within metal AM that allows the high deposition rates necessary for large-scale metallic manufacturing. During and after deposition, the resulting cooling behavior governs thermal and structural transitions that influence material evolution and defect formation. Acoustic emission (AE) provides a passive, high-frequency sensing modality for monitoring these evolving dynamics through stress-wave activity generated during and after deposition. Delineating the corresponding cooling transition zones is important for separating active process behavior from post-deposition cooling and for enabling physically interpretable analysis of AE signal evolution. In this study, we present a data-driven framework for delineating transient zones between the operational and cooling phases of powder-based metal DED using only AE data. An adaptive unsupervised time-domain detection method was developed to automatically identify the onset and termination of the transitional zone () using only AE data, leveraging a combination of root mean square (RMS) peak tracking, Hilbert envelope smoothing, and curvature-based stabilization via unsupervised clustering. This approach enabled the segmentation of the AE signal into three physically interpretable periods: the operational zone (), the transition zone (), and the cooling zone (), without manual intervention. Quantitative statistical analysis revealed distinct behaviors across these zones. In particular, maintained compact, stable features across all metrics, consistent with steady-state deposition. Meanwhile, exhibited elevated standard deviation and entropy, reflecting dynamic thermal and structural changes. Finally, demonstrated high kurtosis values (up to 20.31), indicating heavy-tailed behavior and isolated bursts of energy near the end of the cooling process. These results show that AE-based adaptive segmentation can support automated detection of cooling transitions in DED and provide a foundation for process-aware monitoring of post-deposition dynamics.