Cyber-Physical Systems (CPS) are essential to sectors including manufacturing, energy, and transportation because they combine physical processes with computing algorithms. Nevertheless, because of their complexity, CPS are vulnerable to a variety of abnormalities that may result in system malfunctions, performance deterioration, or even security breaches. We give a thorough introduction to CPS in this book chapter and walk through a variety of abnormalities, emphasizing their importance for system safety and reliability. We concentrate on how machine learning may be used to find anomalies in CPS, highlighting how it can spot minute variations that conventional techniques would miss. To illustrate real-world uses we gathered sensor data from an industrial flexible manufacturing system based on pneumatics. To guarantee consistency, the data was pre-processed using scaling and standardization methods. Class imbalance, a frequent problem in anomaly detection data, due to which model often favor the majority class; to solve this problem we employed three different data imbalance technique and compared their performance. Finally, we employed Deep Neural Network for detecting anomalies and to identify any threat in the system. This chapter offers a useful manual for scholars and practitioners in the field by shedding light on the procedures and methods used when implementing machine learning for anomaly detection in CPS.