Process monitoring and control for laser-aided additive manufacturing

• Main Author: Chen, Lequn
• Other Authors: Moon Seung Ki
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Manufacturing::Flexible manufacturing systems; Engineering::Mechanical engineering
• Online Access: https://hdl.handle.net/10356/149504
• Institution: Nanyang Technological University

In-situ monitoring and closed-loop control are two critical methodologies for quality assurance in laser-aided additive manufacturing (LAAM). In particular, geometric conditions of fabricated parts need to be monitored in real-time so that surface defects can be detected and corrected early to avoid further deterioration. Closed-loop control of laser power based on melt pool feedback can enhance the mechanical integrity and reduce defect occurrences. However, the state-of-the-art surface quality inspection techniques require data post-processing with undesirable process intermittence, and conventional closed-loop control requires pre-build parameter optimization, which is cumbersome and time-consuming. In this research, an AI-assisted rapid surface defect detection method with in-situ point cloud data processing and semi-supervised machine learning is proposed, and a novel data-driven adaptive controller with automatic parameter tuning algorithm is presented. The surface monitoring and adaptive control algorithms are implemented in a Robot Operating System (ROS) based software platform. The integrated multi-nodal software architecture enables on-the-fly surface defect detection and real-time laser power control without process intermittence. A defect correction algorithm is implemented to automatically generate repairing tool path if geometric distortions were identified during the deposition process. The main advantage of the proposed monitoring and control system is its efficiency and adaptivity for industrial adoption. Multiple subprocesses can run simultaneously, and surface defects are detected and corrected without human intervention. The proposed control technique is robust to various deposition conditions. Pre-build parameter optimization is not needed even when the deposition material or the parts’ geometries are changed. Experimental results have shown 93.15% defect identification accuracy based on the proposed semi-supervised model and significant improvement in dimensional accuracy of the printed parts attributed to the proposed data-driven adaptive control method.