This PhD project was proposed and supported during a Digital Twins seminar arranged by SUBPRO in November 2020. Digital twins are believed to play an important role in maintenance optimization and maintenance engineering, since they could continuously estimate the state and trend of process systems and help the optimization of maintenance plans. There are still some challenges to be met until digital twins may be implemented in the industry, such as ensuring the quality of models and assuring that the output from the models is continually trustable. DNV has published a Recommended Practice document “DNV-RP-A204 Qualification and assurance of digital twins” in 2020. The document gives a guideline to the designers and operators regarding the method and structure of qualification and assurance for digital twins. The document is a general instruction for a broad range of digital twins’ applications, while digital twins for maintenance have their own specific properties different from other applications. Hence there is a need for a specialized method for qualification and assurance of such systems. Figure 1 shows the capability level for Digital Twin for maintenance [1]. 
There are three types/areas of maintenance models: 
•    Fault diagnosis
•    Failure prognosis 
•    Maintenance planning/optimization
All these models are designed based on probability theory, and the underlying algorithms for diagnosis and failure prediction are not visible to the user. Therefore, trustworthiness of the models is difficult to assess. The same goes for digital twins for maintenance which are based on such maintenance models. The purpose of the PhD project is to develop methods for qualification and assurance of digital twins for maintenance.