This paper introduces a non-invasive device designed for real-time monitoring of electric motors’ working conditions using IoT technology and deep learning algorithms (DLA). The device gathers vibration data from an electric motor and detects anomalies through DLA. Additionally, a dedicated application was developed to monitor the real-time operational status of electric motors. The anomaly prediction process utilizes two types of vibration data: one in the time domain and the other in the frequency domain. Various feature extraction models in DLA were employed to assess their accuracy in anomaly prediction. To evaluate the device’s performance, experiments were conducted on a grinding machine operating under different grinding conditions. The results indicate that using time-domain vibration data results in more accurate predictions of an electric motor’s condition compared to frequency-domain data. Among the tested models, the Serenest26d_32x4d and ResNet34 feature extraction models demonstrated superior training performance with time-domain vibration data. The ResNet34 model achieved the highest accuracy, attaining an F1-score of 1 in predicting the grinding machine’s working condition. Furthermore, the computation time for all prediction models was under 0.02 s, confirming the device’s capability for real-time applications.