Analysis of Operating Status of the Operator in the MCR of NPPs Using Facial Feature Analysis Techniques

The daily tasks of the main control room (MCR) operator are to monitor the entire state of nuclear power plants (NPPs) by instruments and control system and complete the relevant operations such as startup, power operation, testing, shutdown, etc. The main task of the operator is to monitor the parameters of the nuclear power plant, which results in the operator's monotonous, highly repetitive, and prone to fatigue. This paper uses emotional recognition technology, including facial expression analysis, fatigue analysis, and attention analysis. CNN-LSTM (Convolutional Neural Networks, Long Short-Term Memory) method is used to recognize the facial expression that includes six basic emotions. The SVM method is used for the data fusion of the operator’s expressions in unit time. The operator’s fatigue state is analyzed by the eyelid state and mouth movement. The concentration of the operator’s attention on the screen at any moment is analyzed by analyzing the head rotation. Moreover, an evaluation model of fatigue is established for the early warning of the operation state.