Reliable detection of mechanical wear is essential for maintaining operational stability and reducing unplanned downtime in industrial extrusion systems. This study investigates non-invasive detection of screw wear using operational electrical measurements acquired from a single-screw industrial extruder. Electrical parameters were recorded under steady-state processing conditions for healthy and worn screw configurations to determine whether measurable differences in electromechanical behavior could support condition assessment. The collected signals were segmented into 1429 labeled samples and evaluated using statistical and time–frequency analyses. Mean electrical parameters were compared between technical states, and independent samples Welch t-tests confirmed statistically significant differences in phase voltage for all monitored phases (p < 0.001). Continuous wavelet transform was applied to capture non-stationary signal characteristics, enabling extraction of energy- and entropy-based descriptors associated with variations in mechanical load. The derived features were subsequently used for automated classification of machine condition. The results revealed consistent increases in phase voltage for the worn screw ranging from 0.50% to 0.61%, indicating a stable shift in the electrical operating characteristics of the drive system. Supervised classification achieved an accuracy of 96.2% (289 of 300 samples correctly classified in the testing subset), demonstrating reliable separability between technical states without the need for additional vibration instrumentation. These findings confirm that operational electrical signals provide diagnostically relevant information for screw wear detection and support scalable implementation of electrical condition monitoring in industrial extrusion systems.