A Fault Tolerant Neural Network for Space-based 3D Printing Quality Assessment

Space manufacturing technology has garnered considerable interest from institutes and commercial groups due to its potential to reduce launch costs and logistic pressures associated with future space missions. The on-demand manufacturing capabilities offer increased flexibility in tool and equipment supply, thereby reducing Earth's dependence on space missions. In particular, 3D printing, known for its ability to produce complex geometries in various materials, is a prime candidate for space-based manufacturing. Past studies have demonstrated several concepts and designs for space-based 3D printing; however, the quality of 3D printed pieces is critical to the success of space missions. To this end, we propose a failure detection method with fault tolerance ability that allows the manufacturing system to autonomously detect multiple failure modes on 3D printed pieces without human intervention. We evaluate the performance of the proposed failure detection and fault tolerance methods using a group of test samples.