A digital twin can be seen as "the digital representation of a specific physical product, in the form of computational models, identified by their serial numbers, at a specific time and with sufficient fidelity to simulate the corresponding physical product", as described by Singh and Willcox (2018). In this way, the digital twin can have different levels of fidelity that represent the physical model in its life cycle and specific moment. In this article, a new method for synchronizing digital twin counterparts is developed, attributing sufficient levels of fidelity to the dynamic flexibility demands required by Industry 4.0. Experiments with the proposed approach were carried out on a virtual simulation platform that supports the new SyncLMKD method. The results obtained showed that the SyncLMKD method is capable of both tracking and measuring synchronization efficiency and adjusting synchronization for virtual and physical counterparts, for digital twin applications.