Industrial process plants have a lifecycle of several decades, and only some of the most modern plants have digital, machine-readable design information available. For all other plants, the information is often available in PDF and other human-readable formats. Based on this information, a digital twin could be constructed only with considerable human effort. There is a need for a methodology for the semi-automatic generation of digital twins for brownfields with such source information. The objective of this paper is to propose a roadmap towards a methodology for the semi-automatic generation of digital twins for brownfields with such source information as can be expected to be available for brownfields. The purpose of the roadmap is to: conceptualize the methodology, position relevant previous work along this methodology and identify further research challenges to develop the industrial applicability of the methodology. It was discovered that numerous relevant works exist, some of which do not specifically address brownfields. However, there is a lack of research to integrate such research to a methodology for the generation of digital twins.