The emergence of highly productive Cyber-Physical Manufacturing systems has motivated researchers to focus on minimizing communication overheads and saving computing resources. Traditional building virtual machine methods using the web-service and sensor techniques require a lot of system resources and incur high overheads, which make these techniques unsuitable for Cyber-Physical Cloud Manufacturing (CPCM) systems. We propose a novel model which builds digital twins (virtual machines) so that they can be adapted to the CPCM platform. To achieve this goal, we design a framework that can incorporate with MTconnect (a manufacturing protocol to retrieve process information from machine tools) to extract data from physical machines. In addition, we also design information models to represent the structure and reflect the real-time state of physical machines.