Task-driven data fusion for additive manufacturing: Framework, approaches, and case studies

Abstract

Additive manufacturing (AM) has been envisioned as a critical technology for the next industrial revolution. Due to the advances in data sensing and collection technologies, a large amount of data, generated from multiple sources in AM production, becomes available for relevant analytics to improve process reliability, repeatability, and part quality. However, AM processes occur over a wide range of spatial and temporal scales where the data generally involves different types, dimensions and structures, leading to difficulties when integrating and then analysing. Hence, in what way and how to integrate the heterogeneous data or merge the spatial and temporal information lead to significant challenges in data analytics for AM systems. This paper proposed a task-driven data fusion framework that enables the integration of heterogeneous data from different sources and modalities based on tasks to support decision-making activities. In this framework, the data analytics activities involved in the task are identified in the first place. Then, the data required for the task is identified, collected, and characterised. Finally, data fusion techniques are employed and applied based on the characteristics of the data for integration to support data analytics. The fusion techniques that best fit the task requirements are selected as the final fusion approach. Case studies on different research directions of AM, including AM energy consumption prediction, mechanical properties prediction of additively manufactured lattice structures, and investigation of remelting process on part density, were carried out to demonstrate the feasibility and effectiveness of the proposed framework and approaches.