The increasing demands for safety, comfort, and reliability of automotive transmission systems have highlighted the need for effective in-service health management of multi-disc clutches. Under severe thermo-mechanical conditions, friction components experience progressive degradation and irreversible damage, including wear, thermal deterioration, structural instability, and fatigue cracking, which may ultimately lead to failure. This paper systematically reviews recent advances in friction-wear behavior, thermal load, dynamic response, and damage-related failure mechanisms from a damage-oriented perspective, with particular attention to damage classification, evolution, and quantitative indicators. The development and representative applications of digital twin (DT) technology are subsequently reviewed. On this basis, a conceptual self-evolving digital twin (SEDT) framework comprising a physical entity, virtual model, data and software system, and knowledge base is synthesized for multi-disc clutches. The framework integrates damage-aware sensing and modeling with multi-source data fusion, diagnosis, prognosis, and maintenance decision making, and its prospective application to wear prediction under varying loads is illustrated. Finally, current limitations, implementation challenges, and future research directions for SEDT-based clutch health management are discussed.