Research On The Human Error Recognition Of Oil And Gas Production Process Operators Based On Eye Tracking

During Oil and gas production process,most of its raw materials are flammable,explosive,toxic and corrosive.It will produce unpredictable losses and irreversible influence once the accident occurs based on the continuity of production process,technically complexity and the characteristics of equipment.Accidents will lead to personal injury and death,property damage,endanger public safety,and produce unimaginable economic losses and irreparable consequences.In this paper,a process simulation operation platform is established for possible operational behavior errors in the operation of oil and gas production processes.In this paper,the eye tracking technology is used to monitor the operator's abnormal cognitive behavior,and features such as eye movement data and eye movement heat maps are introduced into the operator's human error pattern recognition,which can effectively avoid oil and gas production accidents caused by process operation errors.(1)In order to solve the problem of real-time perception of cognitive behavior in the production process,which is based on the situation that too much reliance on historical data or subjective judgment,the method of real-time collection of eye movement data of oil and gas production process operators based on eye tracking technology is proposed.A simulation platform of oil and gas production process operation experiments is established by dynamic simulation of industrial ethanol production processes.In the simulation experiment,the eye tracker collected the eye movement data of the operator in the interference suppression task in real time.The result shows,it can realize the cognition of the operator by comparing the process parameter changes during the experiment with the operators' area of interest at the same time.This result provides a data basis for the following research on the human error recognition.(2)In order to solve the problem that the traditional errors are characterized by the cognitive psychological characteristics and the behavioral operation characteristics as the index,fault mode cannot be judged objectively.This paper proposes a method for human error pattern recognition based on learning vector quantization neural network and using the characteristics of operator's eye movement data as indicators.The operator's error patterns are summarized as unfamiliarity,distraction and intense stress level three categories by statistically summarizing eye movement experimental data,combining cognitive state and safe behavioral scientific principles.By dividing the experimental operation interface into areas of interest,a learning vector quantification model was established to identify the error modes of operators,which using 13 statistical data obtained from the operator's eye movement experiments as the identification indicators.The result shows that the accuracy of the methods described in this chapter for identifying process operator human errors has reached 86.36% and 92%,respectively.Compared with BP neural network,this method has improved by 22.73%and 8%.It reflects the stability and accuracy of LVQ neural network for operator's human error pattern recognition.(3)In order to solve the problem of qualitative judgment of human error feature,a method for evaluating operators' human error based on image characteristics of eye movement heat map is put forward.This method extracts the GLCM features and HOG features of eye movement heat maps and fuses them to construct the HOG?GLCM feature.A recognition model of operator human errors based on support vector machines is established.The results show that the recognition accuracy of the HOG?GLCM fusion feature described in this chapter has reached 90%,which is 15%higher than the single HOG feature and 35% higher than the single GLCM feature.This method can effectively identify the operator's human error and is helpful for the on-site supervision of the operator.