| description abstract | Abstract. The rising demand for high-value electronics necessitates advanced manufacturing techniques capable of meeting stringent specifications for precise, complex, and compact devices, driving the shift toward innovative additive manufacturing (AM) solutions. Aerosol jet printing (AJP) is a versatile AM technique that utilizes aerosolized functional materials to accurately print intricate patterns onto diverse substrates. Due to inherent process uncertainties and complex spatiotemporal dynamics, effective characterization of AJP outcomes often requires information from multiple sensing modalities. While machine learning has been widely applied to analyze AJP processes, existing approaches largely rely on single-modality data, which limits their ability to comprehensively capture the structural characteristics of printed features. To address this limitation, this study proposes a diffusion-based generative data fusion framework for integrating multimodal AJP sensing data. The proposed method first performs spatial and temporal registration of heterogeneous inputs and then fuses optical microscopy (OM) images, which provide high spatial resolution, and confocal profilometry (CP) data, which offer height measurements using a denoising diffusion implicit model. Through a case study on AJP-printed lines, the proposed approach demonstrates effective integration of complementary information, producing fused representations that preserve spatial and height-related features. Quantitative evaluations using multiple fusion metrics show that the proposed method achieves improved information preservation compared to conventional convolutional neural network (CNN)-based fusion approaches. The resulting fused representations provide a data-driven foundation for enhanced process monitoring and offer potential for future digital twin–oriented analysis in AJP manufacturing. | |
