Cyber-manufacturing system (CMS) is a vision of smart factories where manufacturing processes are fully integrated with computational components. In CMS, an effective intrusion detection system (IDS) is essential in protecting manufacturing operations from cyber-physical attacks. Current IDS analyses data from cyber and physical domains but produces reports separately for cyber domain and physical domain. To utilize connections between cyber and physical alerts, this paper presents a cyber-physical alert correlation method. To evaluate the method, four case studies have been developed and carried out on a CMS testbed. The experimental results demonstrate that the method can effectively reduce the number of false alerts, improve the detection accuracy, and identify root causes.