Artificial intelligence and digital twin enabled intelligent grinding: Progress and opportunities

Traditional grinding is affected by random abrasive cutting, time-varying contact, abrasive tool wear, and force-thermal coupling, resulting in bottlenecks in process-state perception, abnormality and risk identification, quality prediction, and closed-loop parameter regulation. Consequently, it is difficult to meet the requirements of high-quality and high-efficiency intelligent manufacturing characterized by process self-generation, state self-perception, parameter self-decision-making, and process self-adaptation. First, focusing on four typical task-dominant scenarios, namely difficult-to-machine materials, complex surfaces, low-rigidity structures, and high-consistency manufacturing, this paper systematically reviews the research progress of intelligent grinding enabled by Artificial Intelligence (AI) methods and Digital Twin (DT) technology across the three-stage full task chain of pre-grinding planning, in-process execution, and post-grinding feedback. Second, five common supporting capabilities for intelligent grinding are discussed, including multi-source data infrastructure and unified state representation, mechanism-data fusion modeling, virtual-physical synchronization and digital thread, industrial deployment and trustworthy closed-loop operation, and system integration architecture. Furthermore, future opportunities for intelligent grinding are discussed from four aspects: open data and unified evaluation, small-sample learning and cross-condition generalization, safety and trustworthiness with human-in-the-loop collaboration, and platform-based operation and maintenance and capability development. Finally, the research trend of AI- and DT-enabled intelligent grinding is summarized as an evolution from single-source perception, local modeling, and offline optimization toward systematic closed-loop intelligence covering four typical scenarios, spanning three stages, and supported by five common capabilities. This review provides a reference for intelligent perception, stability prediction, closed-loop regulation, and continuous evolution of grinding processes under complex working conditions.