In the dynamic environment of manufacturing, the gradual evolution of data patterns presents a challenge to artificial intelligence (AI) mechanisms and so the consistent integration of digital twins. Phenomenon called concept drift refers to the statistical properties of input data or the target variable in a predictive modeling task that changes over time, necessitating adjustments in the model to maintain accuracy and reliability. With the consolidation of industry 4.0 technologies, factories shop floor provide thousands of data each day creating a great opportunity for AI applications, but also defining some challenges for data-drift handling in data-based digital twins. In this context, drift detection tools empower manufacturers to rapidly address potential defects and maintain the desired level of product quality, production line performance and accuracy of predictive maintenance AI models. The integration of technologies mentioned above not only improves real-time quality assurance but also establishes a proactive initiative for continuous improvement. This article aims explore the impact of concept drift in smart manufacturing through a systematic mapping approach, focusing on digital twin systems. By consolidating current research, this mapping study contributes to the understanding of drift challenges in smart manufacturing and identifies avenues for future research to enhance the robustness of AI systems embedded in digital twin solutions. \(\dots \)