Injection molded rubber parts are used in a broad range of consumer devices and technical equipment as functional elements such as seals, damping elements or flexible mounts. Premature failure of these elements can lead to breakdown of the entire system. However, it is not feasible for suppliers to establish 100 percent control routines for their parts. Fisher Discriminant Analysis can be used to detect process abnormalities and thus reduce scrap and poor quality parts. In this work, parts were manufactured from an industrial NBR rubber compound, which was stored at different environmental conditions prior to processing. During the test runs, we switched between these two portions and also altered the dosing settings to introduce process fluctuations. By using multivariate statistical process monitoring, especially FDA, on process signals readily available from modern injection molding equipment, we could successfully and immediately identify these fluctuations. When training the FDA on historical data, we could also automatically classify the disturbances into types. By performing multivariate process monitoring, part suppliers could detect unwanted changes in the process automatically, while fault classification can be used to efficiently counteract these disturbances.