REAL-TIME DEFECT MONITORING AND ADAPTIVE PROCESS CONTROL IN ADDITIVE MANUFACTURING SYSTEMS

Abstract

Additive manufacturing has transformed modern production by enabling layer-wise fabrication of geometrically complex components, rapid design modification, lightweight structures, customized products, and material-efficient manufacturing. However, industrial adoption remains constrained by process instability, porosity, lack of fusion, excessive melting, balling, cracking, dimensional distortion, surface irregularity, residual stress, and inconsistent mechanical properties. Conventional quality assurance frequently depends on post-process inspection, destructive testing, computed tomography, dimensional measurement, and offline analysis, which identify defects only after substantial material, machine time, and energy have already been consumed. This paper proposes a real-time defect monitoring and adaptive process control framework for additive manufacturing systems. The proposed framework integrates heterogeneous in-situ sensors, high-speed imaging, thermal monitoring, acoustic sensing, photodiode signals, machine operating data, edge computing, multi-sensor data fusion, machine learning, temporal prediction, digital process representations, secure application programming interfaces, and closed-loop control. Real-time observations involving melt-pool temperature, melt-pool geometry, layer images, acoustic characteristics, laser power, scan speed, energy behavior, powder-bed condition, environmental variables, and machine states are continuously acquired and transformed through validation, preprocessing, synchronization, feature extraction, anomaly detection, defect classification, and control decision stages. Multiple analytical models, including Random Forest, Support Vector Machine, Gradient Boosting, Convolutional Neural Network, Artificial Neural Network, and Long Short-Term Memory networks, are incorporated to identify porosity risk, lack-of-fusion conditions, overheating, unstable melt behavior, surface anomalies, and progressive process drift. The adaptive controller modifies approved process variables such as laser power, scan speed, feed behavior, cooling intensity, and selected machine parameters within predefined safety and quality boundaries. A representative prototype-oriented evaluation indicates that the proposed framework can improve defect-detection accuracy, reduce monitoring latency, increase warning lead time, lower scrap and rework, and stabilize process behavior compared with offline inspection and fixed-parameter manufacturing. The framework also integrates model lifecycle management, secure service interfaces, communication integrity, enterprise data modernization, and energy-aware computation. The study demonstrates that the convergence of in-situ sensing, machine learning, edge-cloud analytics, digital twins, and adaptive control provides a scalable foundation for intelligent and selfcorrecting additive manufacturing systems.