Explainable Modeling in Digital Twin

Abstract

Stakeholders' participation in the modeling process is important to successful Digital Twin (DT) implementation. The key question in the modeling process is to decide which options to include. Explaining the key question clearly ensures the organizations and end-users know what the digital models in DT are capable of. To support successful DT implementation, we propose a framework of explainable modeling to enable the collaboration and interaction between modelers and stakeholders. We formulate the modeling process mathematically and develop three types of automatically generated explanations to support understanding and build trust. We introduce three explainability scores to measure the value of explainable modeling. We illustrate how the proposed explainable modeling works by a case study on developing and implementing a DT factory. The explainable modeling increases communication efficiency and builds trust by clearly expressing the model competencies, answering key questions in modeling automatically, and enabling consistent understanding of the model.