Remainders control is a major challenge facing spacecraft products, but spacecraft remainders online detection is affected by insufficient marker data. This paper studies the physical characteristics of remainders generation, establishes a remainders model in spacecraft digital twin systems, and proposes a cross-domain adaptive spacecraft remainders detection method with digital twin enhancement. This method, through digital twin augmentation, which first acquires spacecraft real-time data and then combines similar structural historical marker data to apply a cross-domain adaptive approach to assist current online reasoning. This paper also proposes a new type of cross-domain adaptive approach model to better mine prior knowledge from complex tasks through a shared network structure and gating mechanism. By combining cross-domain adaptive techniques and digital twins to simulate real structures on a computer, this approach achieves more efficient, accurate and real-time predictions for comprehensive detection of remainders states of different components of spacecraft.