Computerized Numerical Control (CNC) machines are the building block of industrial manufacturing, while CNC machines (CNCM) are designed to work relentlessly, breakdown may occur due to ageing, wrong operation, malfunctioning of sensors/actuators etc. Breakdown of CNCM hampers the production, supply chain and occurs financial loses. To improve the availability of the CNCM, predictive maintenance is considered a gold standard which requires continuous monitoring of the state of CNCM. The state of the CNC machine is temporal and requires a continuous data stream (CDT) to represent the correct present state. To achieve CDT, we propose and evaluate the Self-Adaptive Digital Twin (SADT) model of the CNC Machine (CNCM) using open-source industry-standard OPCUA protocol in compliance with RAMI4.0 architecture. We tested our approach on actual CNC machines in an industrial setup in production environment. We found that the proposed SADT with data analytics can identify potential issues in the CNCM, various bottlenecks in production and provides real-time notifications for fault prediction, which leads to the optimization of spares and improved predictive maintenance. We presented a few applications of the proposed SADT and found that it helps in optimising the axis closed loop control, thereby reducing the vibrations and the following error by five times. It also predicts the runtime of the part program accurately in real-time by analysing the current feed rate. We observed a 15% increase in Mean Time Between Failure (MTBF), 11% decrease in Mean Time to Repair (MTTR), 21% reduction in the percentage of breakdowns and 14.2% reduction in the power consumption of the CNCM after the implementation of the proposed system.