Cyber-physical systems (CPSs), a new computing system to control industrial infrastructures, are widely used in many key areas such as manufacturing, energy, and safety management systems. The rapid involvement of CPS in industry has considerably expanded the range of cyber threats. Many machine-learning methods have been employed for the design of effective anomaly detectors in CPSs. Currently, federated learning methods have been applied to distributed machine learning. The distributed nature of some CPS systems creates a potential to use federated learning in this ecosystem. In this paper, we have applied three federated learning methods, in various scenarios, over three datasets obtained from SWaT (Secure Water Treatment) data. This dataset is obtained from an operational water treatment testbed. We have done a sort of preprocessing and feature selection on SWaT, to get a better and clean dataset. Then federated and non-federated (central based) methods are applied in various scenarios. The results of using federatedbased methods are very promising and in many cases even better than central-based methods.