Injection molding processes are influenced by internal and external disturbances such as machine wear or batch fluctuations. The goal is to detect these fluctuations and, in the best case, to correct them through quality prediction and process control using machine learning algorithms. However, machine learning models are only capable of doing this if the effects are included in the training data. If the process changes, the models lose prediction performance until they are no longer suitable for the new situation. This is called concept drift. In injection molding, sooner or later a process change occurs, so reliable detection is required. The objective of this paper is to evaluate different concept drift detection methods for injection molding applications. Nine different algorithms for concept drift detection were evaluated. For this purpose, disturbances were intentionally added to a stable process, which had to be detected by the algorithms. For example, the viscosity in the material was affected by adding material with higher residual moisture to represent batch variations. A total of 1600 injection molding cycles were carried out in three different experiments.