Building Trust in Digital Twins Through Verification and Validation

Abstract

Digital Twin (DT) is a concept of growing interest, driven by the technological advancements related to Industry 4.0. A DT integrates innovative technologies to create a virtual model that replicates a physical system and enables bi-directional data flow, facilitating real-time monitoring, analysis, and control. These capabilities make DTs increasingly attractive for supporting decision-making in complex environments, especially in manufacturing, where cost and safety are critical. However, the effective adoption of DTs depends on the ability to trust the model and all its components throughout development and deployment. Verification and Validation (V&V) have traditionally been employed to assess the credibility of models, providing a venue for the development of trust. Therefore, V&V are essential foundations for developing DTs that can support decisions in real-world environments. Through a systematic literature review, this research investigated whether and how V&V practices were employed in DTs developed for manufacturing applications. The findings revealed that relatively few studies reported performing both verification and validation activities, indicating a significant gap in DT credibility methods. The review also examined commonly used V&V techniques, their alignment with DT capability levels, and application domains. Results indicated inconsistencies in V&V terminology, execution, and objectives, as well as the absence of a standard framework to guide V&V in DT development. To address these challenges, this research introduces the Digital Twin V (DTV) framework. The DTV integrates V&V activities across the DT development lifecycle, emphasizing iterative and recursive processes tailored to the unique characteristics of DTs. A case study applying the DTV framework to a fused deposition modeling additive machine illustrates its applicability. The framework’s scalability and potential for adaptation across various DT maturity levels are demonstrated, offering actionable guidelines for practitioners and researchers to enhance the trust and adoption of DTs. Finally, this research also explores the integration of uncertainty quantification techniques into the DTV framework to enhance the understanding of the relationship between virtual and physical systems, as well as to provide a tool to measure the uncertainties in the DT output and how these uncertainties compromise model performance. Together, this work bridges the gap between theory and practice by uncovering the gaps in current V&V use within DT research, offering a practical methodology to guide future development of trusted DTs with uncertainty quantification considerations.