The overarching goal of this project is to synthesize a digital twin testbed that can provide credible data for data analytics investigation. The testbed will then be subject to multi-scale modeling and analysis, and machine learning for fault diagnosis. Digital Twin is a virtual image of an asset, maintained throughout the lifecycle and should be easily accessible at any time. It is an essential part of our digital asset ecosystem, and will enable a new generation of advanced analytics and understanding of the product. Despite rapid progresses in data analytics for system diagnosis and prognosis, the research community as a whole lacks methodology validation using a controllable testbed that can exhibit various failure modes under different operational conditions. The scarcity of data and the general lack of systematic methodology of experimentation hinders the further advancement of digital twinning as the results obtained may not be easily extrapolated. In this project, leveraging an on-going naval research, we plan to synthesize an experimental testbed with representative machinery components and data acquisition system. Various failure modes will be injected into the testbed system. Preliminary data analysis will be conducted for modeling validation and fault analysis that leverage state-of-the-art machine learning techniques.