Abstract: Digital twin technology provides a theoretical structural framework for monitoring and autonomous decision-making and forecasting throughout the production lifecycle of the workshop, promoting the digital and intelligent transformation of the workshop. The digital twin architecture system is built from five aspects: physical entity fusion, data interaction, virtual entity construction, autonomous model update, decision, and prediction in a big data environment. The architecture solves the problems of data interaction delays and dynamic perturbations in physical space that are common in real workshops. The architecture of a digital twin workshop model based on state transfer in a big data environment is built. State-based twin model adaptive update strategy approach and self-decision scheduling algorithms considering artefact priority, rework, and insertion orders were proposed. The data structure and operating rules of the digital twin model are updated by combining the stage data generated by the process with historical big data. Synchronize the actual production status to provide a rational re-scheduling strategy and obtain credible production forecasts. Taking the processing of an aerospace non-standard product as an example, the proposed digital twin model is built, and the data generated by the state interval is used to update the model, generate and validate a re-scheduling solution for reworked parts, and prove the effectiveness of the digital twin model.