In-situ process evaluation for continuous fiber composite additive manufacturing using multisensing and correlation analysis

Multi-sensing and correlation analyses are essential for online process evaluation and optimization to improve the quality of as-fabricated components. Defect-free process control is important for additively manufactured (AM) continuous fiber-reinforced composites (CFRP) because the number of defects and poor-quality control in AM-fabricated CFRP restrict their mechanical performance and product service life. In this study, a framework of multi-sensor fusion for CFRP additive manufacturing is proposed for in-situ process evaluation and to establish correlations between process parameters/pattern features with layer wise defects and surface quality. Infrared (IR), visual cameras, force, and laser-displacement sensors were integrated with the printing head to obtain online datasets. Multiple signal denoising, feature extraction, and classification were performed to incorporate deep-learning neural networks and correlation analyses using feature-level fusion approaches. The critical features of these signals were extracted for a quantitative analysis of the layer wise surface roughness, level of fiber misalignment (LoM), and number of defects. Multi-sensor fusion is an effective approach to online monitoring and process evaluation. The established knowledge base is helpful for predicting and adjusting the localized process parameters during the fabrication process.