SENSOR-DRIVEN PROCESS INTELLIGENCE FOR REALTIME DEFECT DETECTION IN METAL ADDITIVE MANUFACTURING

Abstract

Metal Additive Manufacturing (MAM) has revolutionized advanced manufacturing by enabling the fabrication of highly complex metallic components with exceptional design flexibility, reduced material waste, and shortened production cycles. However, ensuring consistent product quality remains a significant challenge because the manufacturing process involves complex thermal, mechanical, and metallurgical interactions that are highly sensitive to process parameter variations. Defects such as porosity, lack of fusion, keyhole instability, cracking, balling, residual stresses, and dimensional inaccuracies frequently occur due to unstable process conditions, reducing component reliability and increasing production costs. Conventional quality assurance methods primarily rely on post-process inspection techniques, which are incapable of detecting defects during fabrication and often result in expensive rework or component rejection. Recent advancements in Industrial Internet of Things (IIoT), intelligent sensor networks, artificial intelligence, and real-time manufacturing analytics have enabled continuous monitoring of manufacturing processes through sensor-driven process intelligence. This research proposes a comprehensive sensor-driven process intelligence framework for real-time defect detection in metal additive manufacturing by integrating multi-sensor data acquisition, edge computing, machine learning, Digital Twin technology, and adaptive process optimization within a unified intelligent manufacturing architecture. The proposed framework continuously collects thermal images, melt pool characteristics, acoustic emissions, vibration signals, laser power measurements, environmental conditions, and machine operating parameters using IIoT-enabled sensors. The acquired data are processed through feature engineering and analyzed using advanced machine learning algorithms, including Random Forest, Support Vector Machine, Extreme Gradient Boosting, Artificial Neural Networks, Convolutional Neural Networks, and Long ShortTerm Memory networks, to predict manufacturing defects in real time. Intelligent optimization algorithms subsequently recommend corrective process adjustments through a closed-loop feedback mechanism to maintain stable manufacturing conditions. Experimental evaluation demonstrates that the proposed framework achieves defect detection accuracy exceeding 98.5%, reduces manufacturing defects by approximately 44%, improves process stability by 37%, decreases inspection time by 58%, and enhances overall production productivity by nearly 40% compared with conventional manufacturing quality monitoring approaches. The proposed sensordriven process intelligence framework provides an effective, scalable, and intelligent solution for autonomous defect detection and quality assurance in Industry 4.0-enabled metal additive manufacturing systems.