DIGITAL TWIN AGGREGATES AND BIG DATA ANALYTICS FOR PREDICTIVE RISK MANAGEMENT IN CONTINUOUS MANUFACTURING: A SYSTEMATIC LITERATURE REVIEW

Digital Twin (DT) technology is a virtual representation of a physical asset, process or system that is continuously updated from operational data and is a key enabling technology for Industry 4.0 manufacturing research (Grieves andVickers, 2017; Kritzinger et al., 2018). Most of the current studies are related to the single asset, such as condition monitoring and fault diagnosis/predictive maintenance, using DTs (Tao et al., 2019; Tekinerdogan et al., 2021; Catal et al., 2022; Chen et al., 2023). The asset-level approach is not entirely appropriate for continuous manufacturing where machines and processes are highly inter-dependent and a disturbance at one step can make it to the flow, pressure, product quality and downstream processes. A way forward from asset level to system level risk prediction is through Digital Twin Aggregates (DTAs) that integrate information from several DTs into a system level view, along with big data analytics (Lu et al., 2020). While the implementation does have challenges of data integration, computational requirements, interoperability and standardisation (Shao and Helu, 2020; Perno, Hvam and Haug, 2022), these are heightened in resource-constrained, developing industrial economies (Muhla et al., 2025a; Mashava et al., 2026). This paper is a systematic literature review (SLR), performed following the PRISMA 2020 statement (Page et al., 2021) and the guidelines provided by Kitchenham and Charters (2007). Records published between January 2016 and September 2026 were obtained from Scopus, Web of Science, IEEE Xplore and ScienceDirect and a total of 88 records were screened to identify 50 studies with an average quality score of 4.12 out of 5. All studies included were classified based on a multidimensional taxonomy of DT architecture, manufacturing level, data sources, analysis techniques, risk-management function, and validation environment. Among the 50 studies, there are 22 that are reviews or surveys, and only 19 that include empirical studies, 11 of which involve continuous or process manufacturing, and none of which have been validated with industrial data in a continuous process, in the case of multi-twin or aggregated architectures. The review summarizes existing applications, emerging technologies, and industrial implications, and identifies a research roadmap for practical, interoperable and system-level predictive risk management for continuous manufacturing.