Yi, Chen-PeiChen-PeiYiTsai, I-HaurI-HaurTsaiLin, Yi-JenYi-JenLinYeh, Hsin-TienHsin-TienYehSHIH-CHIN YANG2026-04-142026-04-142026https://www.scopus.com/record/display.uri?eid=2-s2.0-105030673243&origin=resultslisthttps://scholars.lib.ntu.edu.tw/handle/123456789/737172This paper presents a deployable motor health monitoring framework for inverter-fed industrial motors, integrating online acquisition of voltage and current signals. The proposed approach exploits the physical correspondence between voltage excitation and current response and adopts a one-dimensional convolutional neural network–based Encoded Voltage–Current Conversion (EVCC) model to learn normal electromagnetic behavior. By reconstructing current responses from voltage inputs through a voltage-to-current cross-domain model, the framework overcomes the limited interpretability and weak sensitivity of conventional current-only autoencoder (AE)–based anomaly detection methods. To enable practical deployment, an attachable and non-intrusive external voltage sensor (E-VS) is introduced for online PWM voltage measurement in industrial environments. An observer-based operating-condition labeling scheme and a noise-suppression strategy are further incorporated to ensure robust anomaly detection under varying speeds, loads, and measurement noise. The proposed sensing and diagnostic pipeline are validated through laboratory experiments on six bearings under nine operating conditions, demonstrating improved diagnostic accuracy compared with conventional AE approaches. In addition, long-term field deployment is conducted on two 15-kW cooling pump motors in a real industrial facility. The proposed EVCC framework successfully detects an abnormal operating event which cannot be identified by current-only AE method. These results demonstrate that voltage-guided current reconstruction provides enhanced visibility of electrical anomalies and offers a practical and effective solution for industrial motor condition monitoring.truejournal article10.1109/ACCESS.2026.36655602-s2.0-105030673243