The paper proposes methodological approaches to development of a human-machine interface for the specialized software supporting operator activities of industrial facilities in emergencies. As an example, development of a system is considered to support activities of the oil refinery operators in eliminating various failures and their consequences in accordance with the adopted regulations of the accident elimination plan. Features of the operator actions in the emergency were analyzed, and requirements for the operator support program interface were formulated. The proposed approach is based on introduction of design patterns for human-machine interfaces in the safety-critical systems. Main patterns were analyzed, and recommendations were given on the use of specific patterns in creating a software interface to support the operator actions. Results of the experimental study of possibilities of the proposed software practical application are presented showing significant reduction in the time spent by the operator on actions to execute the emergency response plan requirements and decrease the number of errors. This confirms effectiveness of the developed methodology in practice. As the area for further improvement of the industrial facilities operator actions, it is advisable to consider methods for estimating the operator state according to the data obtained from heterogeneous information channels including speech, analysis of the number of blinks, evaluation of emotions, analysis of the head tilt, direction of gaze and others using the convolutional neural networks of deep learning
Read full abstract