While the concept of the Digital Twin is gaining traction in Cultural Heritage (CH), its application is often limited to static "digital replicas," failing to capture the rich semantic and symbolic dimensions of heritage objects. A significant gap exists in formally modeling the complex, dynamic narratives inherent in CH entities in a machine-actionable way. This paper addresses this gap by introducing "Digital Cultural Heritage Twins"—autonomous, richly structured digital entities built using the Narrative Knowledge Representation Language (NKRL). Our approach employs an "augmented n-ary" framework to formally represent complex events, relationships, and abstract notions with a high degree of semantic precision. We demonstrate the methodology through a detailed formal analysis of the "pentimenti" (artistic revisions) in a self-portrait by Anthony van Dyck. To our knowledge, this represents the first application of a deep knowledge representation model to the layered narrative of pentimenti, moving beyond visual documentation to create a truly computable semantic twin.