摘要
Process monitoring plays a vital role in ensuring safe, efficient, and sustainable industrial operations. While prior surveys have advanced the field, most are fragmented, focusing narrowly on either specific algorithms or application domains. This review provides a comprehensive and integrative perspective by classifying monitoring methods into three broad categories: data-driven, model-driven, and fusion-driven approaches. The distinctive contribution of this work lies in moving beyond conventional comparisons. First, we present a balanced taxonomy that clarifies the scope and overlaps of different approaches. Second, we highlight emerging directions underrepresented in earlier surveys, including reinforcement learning, graph-based methods, topological data analysis, Bayesian optimization, hybrid deep architectures, and advanced state observers. Third, we emphasize practical implementation aspects, covering uncertainty quantification strategies and the deployment of both commercial and open-source monitoring software platforms. Finally, we connect process monitoring to the future of autonomous operations within Industry 4.0 ecosystems, outlining challenges related to integration, scalability, and interpretability. By consolidating methodological advances, industrial platforms, and forward-looking challenges, this review provides researchers and practitioners with a clear roadmap for designing monitoring systems that are accurate, interpretable, and resilient, while identifying open questions that will guide the next generation of autonomous process industries.