LoRe-GRNN: A Hybrid Deep Learning Framework for Real-Time Anomaly Detection and Stress Distribution Prediction in 3D Printing Processes

Abstract

Advanced 3D Printing (A3P) revolutionizes manufacturing with precision, speed, and innovation, unlocking limitless design possibilities and superior material performance for next-generation industrial and creative applications. A3P epitomizes a paradigm shift in manufacturing, seamlessly merging additive fabrication with advanced 3D printing to construct intricate geometries unattainable through conventional methods. However, inherent challenges persist, including structural deformations in Stereolithography (SLA) and nozzle occlusions in Fused Deposition Modeling (FDM), necessitating intelligent intervention. This study introduces LoRe-GRNN, a groundbreaking Deep Learning (DL) framework for real-time anomaly detection and stress distribution prediction. Leveraging a novel fusion of Longformer-Reformer (LoRe) architectures with Gated Recurrent Neural Networks (GRNN), the system optimizes feature extraction and predictive accuracy. A meticulously curated 3D model repository, synergized with Finite Element (FE) simulations, enhances SLA stress predictions, while an integrated multisensory module ensures FDM process monitoring. The hybrid approach demonstrates unparalleled precision, achieving 99.23% anomaly detection accuracy, significantly mitigating computational overhead compared to traditional FE simulations. This transformative framework enhances the resilience of additive manufacturing, heralding an era of intelligent, high-fidelity, and resource-efficient 3D printing systems.