Session: 03-09-01: Data-Driven Innovation in Smart Product Design and Manufacturing

Paper Number: 172558

Real-Time Melt Pool Height Control in Ded Using Thermal Feedback 

Maintaining geometry accuracy in metal additive manufacturing is particularly challenging in Directed Energy Deposition (DED), where melt pool behavior is highly sensitive to thermal fluctuations and geometric variations. Inconsistencies in layer height during the build process can accumulate over time, resulting in shape distortion, surface irregularities, and degradation of mechanical performance. To mitigate these issues, we propose a real-time control framework that continuously monitors the melt pool using infrared thermal imaging. By analyzing thermal feedback in real time, the system dynamically adjusts laser power to stabilize melt pool height throughout the DED process.

The proposed control strategy was implemented on a custom-built metal 3D printer specifically designed for DED. The system comprises a high-power fiber laser, a powder delivery system, a gas delivery system, and a 3-axis motion stage, all managed through own software tailored for synchronized printing and monitoring. A long-wave infrared (LWIR) camera is integrated into the optical path of the laser system using a beam splitter and bandpass filter, enabling coaxial alignment with the laser beam. This configuration allows thermal imaging of the melt pool in real time, aligned coaxially with the laser beam. Designed to support multi-bead, multi-layer printing and adaptable to complex geometries, the printer serves as a flexible platform for validating advanced process control algorithms.

The control method relies on continuous, non-contact thermal monitoring of the melt pool to regulate melt height during deposition. Real-time thermal images acquired from the coaxially aligned LWIR camera are analyzed to extract the peak-to-boundary temperature difference (TD), a quantitative indicator of melt pool energy distribution. Calibration experiments confirmed a strong correlation between TD and melt pool height under varying laser power conditions. Leveraging this relationship, a closed-loop control scheme was developed in which TD serves as a feedback parameter to dynamically adjust laser power. This approach enables height stabilization throughout the build process, regardless of geometry or deposition location, with thermal data processed at 25 Hz to ensure timely actuation.

Experimental validation was conducted using two geometrically distinct models: a cube and a stepped pyramid structure, each printed under both uncontrolled and laser power-controlled conditions. Without active control, significant variations in melt pool height were observed, particularly at contour regions, resulting in dimensional errors exceeding 10%. In contrast, the implementation of thermal-feedback-based control led to substantial improvements in deposition uniformity and geometric accuracy. Quantitative assessment using 3D scanning and cross-sectional analysis revealed a reduction in geometric error to below 4%, and a root mean square error (RMSE) in layer height of less than 0.4 mm across all measured sections.

This study demonstrates that real-time thermal feedback can effectively control melt pool height in DED, significantly improving dimensional accuracy. The proposed method provides a practical and reliable approach for enhancing build quality, with potential for future expansion to control additional process parameters.

Presenting Author: Subin Shin Korea Advanced Inst. of Science & Technology (KAIST)

Presenting Author Biography: Subin is currently a Ph.D. candidate in the Department of Civil and Environmental Engineering at Korea Advanced Institute of Science and Technology (KAIST), under the supervision of Professor Hoon Sohn. His research focuses on infrared thermography-based non-destructive evaluation (NDE), with particular emphasis on weld defect detection and process monitoring and control in metal additive manufacturing. His work aims to improve the reliability and precision of inspection and manufacturing technologies through real-time sensing and data-driven analysis.

Authors:

Subin Shin Korea Advanced Inst. of Science & Technology (KAIST)
Ikgeun Jeon Korea Institute of Machinery & Materials (KIMM)
Hoon Sohn Korea Advanced Institute of Science and Technology (KAIST)