Application And Research On Data Mining On Human Error

With the upgrading of Aviation manufacturing industrial technology, the fault record of aviation machinery and equipment has been reduced to very low percentage. However, in the worldwide with the rapid increase in air traffic accident rate for air also reached the highest point in recent decades. After ten years of research, ICAO and national aviation researchers point out that the frequent accidents caused by flying is the main reason for air traffic control personnel, maintenance personnel and maintenance staff in routine maintenance, flight management of human error all of which known as human error in aviation.In this paper, human error as the research object, combined with the existing airlines definition and classification errors, build the aviation sector in the classification model of human errors. Then defined error model to the end of the basic error of type event, the accident occurred as the top event, the middle class the other error events, the application of logic gates to connect the various errors into the fault tree form, which is converted to a fault tree model. Then we apply the FAT on human error for the qualitative and quantitative analysis, the purpose of qualitative analysis is to identify failure of the model, to test the correctness of the results of association rule mining provides a standard, the purpose of quantitative analysis is to calculate the error and the importance two indicators of the probability of error. Finally, select MLAR-FP algorithm to analysis error, while in the candidate selection process will be set quantitative analysis of the fault tree are an important degree of error and error probability into one of two targets. The purpose is to reduce the number of candidate itemsets and to achieve different levels of error correlation compare the size, in order to find the internal correlation between the errors. So as to supply the internal mechanism of human error for Aviation managers and provide theoretical guidelines for the better customized routine maintenance and flight standards.The results achieved are:First, constructe human error model in the field of aviation maintenance; through the human error model could be divided into several broad categories; meanwhile, it could also describe the basic errors in each category.Second, introduce important degree of error into the analysis of multi-level association rules; the advantage is that in the process of candidate set pruning we can cut a greater level of non-frequent item sets at the same time to analyze the credibility of the rules.The innovations are further analyzes using the fault tree analysis's result to the human error. The advantage is not only found a significant error factor error factor can also find hidden factor, while application of multilevel association rules technology combined with the hierarchical model of the fault tree, digging out the different levels of internal rrors.