A Novel Spatter Detection Algorithm for Real-Time Quality Control in Laser-Directed Energy Deposition-Based Additive Manufacturing.

Laser-Directed Energy Deposition (LDED) has recently been widely used for 3D-printing metal components and repairing high-value parts. One key performance indicator of the LDED process is represented by melt pool stability and spatter behavior. In this research study, an off-axis vision monitoring system is employed to characterize spatter formation based on different anomalies in the process. This study utilizes a 1 kW fiber laser-based LDED system equipped with a monochrome high-dynamic-range (HDR) vision camera and an SP700 Near-IR/UV Block visible bandpass filter positioned at various locations. To extract meaningful features from the original images, a novel image processing algorithm is developed to quantify spatter counts, orientation, area, and distance from the melt pool under harsh conditions. Additionally, this study analyzes the average number of spatters for different laser power settings, revealing a strong positive correlation. Validation experiments confirm over 93% detection accuracy, underscoring the robustness of the image processing pipeline. Furthermore, spatter detection is employed to assess the impact of spatter formation on deposition continuity. This research study provides a method for detecting spatters, correlating them with LDED process parameters, and predicting deposit quality.