Abstract: The new industrial revolution is driving the manufacturing industry towards increasingly digital and intelligent evolution, with the key feature being the integration of the digital and physical worlds in Cyber Physical System (CPS). However, due to constraints in multiple areas of development, it is not possible to achieve fully autonomous CPS in the short term, and human remain an important part of the system. The co-existence and symbiosis between human and CPS has become an urgent problem to solve. Workload is an important variable that describes the overall performance of the system and the human-computer collaboration relationship, and can be identified and predicted through observation of the operator's mental workload. This paper explores the problem of automatically identifying the mental workload of operators in industrial control environments based on smart manufacturing domain operator interactions with industrial software and production systems, using electroencephalography (EEG) collection technology. Since mental workload indicators and model performance are affected by factors such as task type, difficulty, standards, and time requirements, this paper established an experimental paradigm simulating industrial control tasks, extracted key EEG sensitivity parameters in task scenarios, and achieved a mental workload model based on single EEG modality from multiple task perspectives. The model's accuracy in identifying operator mental workload can reach 100%. The average recognition accuracy of the model in the monitoring and verification, control operation, and communication scenarios is 97.85%, 96.95%, and 89.88%, respectively. To further understand the relationship between operator physiology and system workload, promote bidirectional adaptation support between industrial complex systems and operators, and optimize the design of a passive Brain-Computer Interface (pBCI) based human-in-the-loop adaptive system, this paper provides a new perspective and theoretical basis.