<p>In this study a method for minimizing the energy consumption in the CNC machining process by combining the energy phase diagrams, anomaly detection, and energy loss due to anomaly is developed. Anomaly based energy waste estimation is the process of detecting and quantifying the unexpected or abnormal energy consumption in a system. In the proposed approach, we begin by constructing energy phase diagrams to visualize and identify the energy-efficient and high-consumption operational zones in different machining phases. Autoencoders, which are a type of unsupervised deep learning model, are then used to detect the anomalies in real-time energy consumption patterns. This helps the system to flag the deviations from the expected behavior, which indicate inefficiencies or faults. The moving average method was added to the autoencoders to find the total energy waste due to the anomalies. This approach enables adaptive, data-driven, and interpretable energy optimization, resulting in a reduction in energy waste and an improvement in operational efficiency.</p>