Research On Safety Recommendations In Aviation Accident Investigation From The Perspective Of Big Data

After an aviation accident/incident happens,the relevant agencies will propose corresponding safety recommendation for the hidden dangers in the aviation system found during investigation.Safety recommendation is an important part of accident investigation,is the core of accident investigation,can play an important role in preventing the recurrence of similar accidents.Safety recommendation play an extremely important role in safety management,providing safety managers with management focus and direction,so it is increasingly important to drill down on safety recommendation.At the same time,safety recommendation is characterized by shortness,large quantity,and complex content.In order to dig out more information in the safety recommendation,this study is based on the existing safety recommendation data,using word frequency analysis method,cluster analysis method and mind map to analyze and study the safety data.The results are to provide aviation related personnel for the characteristics of content,and ultimately improve China Airlines' safety level.The main work of this paper is as follows:Firstly,keywords in Chinese and American safety recommendation are studied by word frequency analysis Results show that the four aspects of safety recommendation mostly involved are ‘pilot/unit',‘procedure/regulation',‘machine/equipment',‘practice/training',which are the key orientations of aviation safety management.Most safety recommendation given by CAAC aims at airlines,while the US aims more at FAA;in terms of procedure and regulation,CAAC emphasizes mostly the compliance of regulation,while the US emphasizes the revision of regulation.Secondly,the cluster analysis method is used to study the safety recommendation of China and the United States,including the process of safety recommendation word segmentation,stop word filtering,and the establishment of vector space model.When selecting clustering algorithm,k-means are used widely.The entire process of clustering is done by writing Python code.Then,based on the results of the clustering,the classification criteria for safety recommendation is established,the safety recommendation is classified into six categories of ‘flight operation',‘cabin air service',‘machine quality assurance',‘air traffic operation',‘ground transportation operation';‘design and manufacture';twelve sub-items of ‘procedures and regulations',‘training practice',‘regulatory review',‘technical development',‘airborne equipment',‘cultural style',‘medical security',‘emergency rescue',‘human resources,CRM,cooperation and exchange,and other projects.According to this standard,the safety recommendation of this paper is divided and quantified,and analyzed from the horizontal and vertical aspects respectively,revealing the safety situation contained in the safety recommendation.Finally,the mind map method is used to visualize the safety recommendation.