In order to identify the mental load of operators under repetitive high-precision tasks effectively, 36 subjects were recruited in this study. The fine tasks on the electronic assembly line were simulated in the laboratory, and operators' behavioural performance (completion time, error rate) and the changes of Oxyhaemoglobin (O2Hb) in prefrontal channels 2, 8, 12, 17, 20 and 21 of the brain were characteristic factors for mental load recognition. A model of recognition of mental load state of operators based on BP neural network was constructed and the fatigue state of operators was divided into four levels. Finally, the recognition rate of the mental load state of operators was 86.81% by combining the experimental data. The combination of behavioural performance indicators and physiological measurement indicators can effectively identify mental work state of operators, which provides a new idea for classification and recognition of mental load state of operators under repetitive high-precision operation tasks, and provides a reference for the establishment of effective labour organization.