Polymer O-rings are an essential part of many designs, including mission and safety critical systems. Currently, there are no accurate destructive tests for measuring the polymer properties of O-rings (e.g. durometer), let alone nondestructive methods. As such, it is difficult to identify substandard, nonconforming, improperly processed or counterfeit O-rings. This work combines resonant ultrasound spectroscopy (RUS) with machine learning and predictive analytics to sort O-rings based on material and durometer (multinomial classification) and to accurately estimate the mass and durometer with an ultrasonic examination that takes less than 10 seconds. Results from a population including eight materials and six durometers are presented and discussed.