Three-phase induction motors (TPIMs) are a common asset that can be found in a variety of industries, as they are economical, rugged and reliable. TPIMs suffer faults that are caused by normal operation, among which stator winding damage accounts for the second most important defect. There is an increasing demand for early fault identification on TPIMs without the need for interrupting operation, so Preventive Maintenance can be performed. There are multiple techniques that demonstrated effectiveness for early stator winding fault identification. These techniques utilize sensors of various types to measure features on TPIMs, combined with Machine Learning (ML) algorithms for fault identification. Measured features include vibration, current, temperature, flux, voltage, power and acoustic emission. ML techniques include Support Vector Machine (SVM), Random Forest (RF), Multi-Layer Perceptron (MLP) and Deep Learning (DL). An experiment was set up where various sensors gathered data from a TPIM. This data was gathered for both Healthy and Faulty states of the TPIM. The sensors used were current transducers, piezoelectric film, vibration sensor (internal inductance coil) and temperature sensor. The features measured were current, vibrations and temperature. Then, the gathered data was subject to a MLP classifier for stator winding fault identification. Finally, sensor cost-effectiveness was evaluated according to the classifier score, practicality assessment and cost. The results showed that, among the sensors compared, current sensors proved to be the most cost-effective sensor for stator winding damage fault identification in a TPIM.