This paper reports on results starting with laboratory research, developing a prototype and transferring the capability to a railroad service center. Inspection of locomotive wheels is often done by magnetic particle/ dye penetrant technique, which requires careful cleaning of the surfaces and can be time consuming. Using a non-contact sensor such as the electromagnetic acoustic transducer EMAT, to scan the wheels in an automated manner offers greater inspection speed at lower manpower. This paper reviews the basic concept of EMATs, introduces a recently developed technique for simulating EMAT performance by Finite Element calculation and features bench top results of waveform acquisition and signal-to-noise ratio dependence on lift-off. Next presented are calibration results for spark-eroded flaws in wheel sections for a variety of locations and sizes. Finally data are on flaw detection in a railroad service facility on several locomotives with wheels spinning at speeds up to 40 meters/minute. Results for both artificial and actual flaws are shown. In conclusion, the EMAT non-contact sensor provides improved inspection for locomotive wheels without their removal.