Research On Risk Assessment Method Of Human-Machine Interaction In Civil Aircraft Cockpit During Landing Phase

Landing is the high stage of civil aircraft flight accidents.With the development of automation technologies,most of the flight accidents are caused by pilot error during human-machine interaction in the cockpit.The research on risk assessment method of human-machine interaction in the cockpit during landing phase is of great significance to reduce flight accidents.By analyzing the cognitive characteristics of pilot in the cockpit human-machine interaction process,combined with SRK,COCOM and ACT-R cognitive control principles,a multi dimensionl cognitive model of pilot was established,and the cognitive mechanism of pilot from three levels: surface,modularization and physiology.The pilot’s human error pattern recognition method was studied.Through the improvement of the HET method and combing the standard operating procedures,the HTA analysis of the multi-scene tasks in the landing phase of the civil aircraft was carried out,and the comprehensiveness of the HET method analysis was strengthened.Based on the SC method,a complexity measurement method for the cockpit human-machine interaction tasks was proposed,which improves the accuracy and objectivity of the HET method.Combined with the example of landing phase,the feasibility of modified HET method was demonstrated.Then,a pilot operation reliability inference model was established.Based on the HTA analysis of the multi-scenario task during the landing of the civil aircraft,the UML model of the pilot workflow sequence diagram was established,and the human-machine interaction process between the crew and the cockpit was visually and clearly described on the time axis.Combining the Hidden Markov,the UML-HMM-based pilot operation reliability inference model was established by using the forwardand-backward algorithm.The validity of the model was demonstrated by the example.Finally,based on fuzzy inference and BP neural network,the human-machine interaction risk prediction model of civil aircraft landing phase was constructed.Three evaluation indicators of cockpit human-machine interaction risk were given.The subjective evaluations were converted into quantitative datas through fuzzy inference,and the quantified datas were trained through neural network,which improved the sensitivity of model and the accuracy of the datas.The calculation method of relative risk coefficient was proposed as well.Combined with the operation example of the pilot in the landing phase,the feasibility and applicability of the model were demonstrated.