Research On Data Mining-Based Prognostic Models For Aircraft Fault

With the rapid development of aviation technology, the flight equipment is becoming to be more and more complex, so aircraft fault diagnosis and prediction have to suffer new challenges. In recent years, although the stability of the aircraft component has been improved gradually, aircraft fault prediction remains a tough mission. Any failure or mistake in flight would cause terrible losses e. Therefore, it has significant social and economic value to predict airplane fault in a more accurate way.Data mining is an emerging technology with broad application prospect, which has been developed in the recent few years, and can be applied in every field. Therefore, the combination between aircraft fault prediction and data mining will has great significance. This paper focuses on two problems, one is how to predict the attrition rate of aircraft bearing by using gray prediction model and support vector machine prediction model, the other is how to predict the failure rate of main fuel control system of aircraft by using gray prediction model and neural network model.1. Prediction of the attrition rate of aircraft bearing. Firstly, the data of the attrition rate of aircraft bearing were collected and the data feature was analyzed. By comparison of several analysis models, we selected the gray model and support vector machine model to measure the attrition rate of aircraft bearing. To obtain the data, the gray model based on different modeling length and different support vector machine parameters were used. Then the advantages and disadvantages of these two models were valued and the possibility of combination of them was discussed. Next, gray support vector machine fusion model was used for the prediction. Finally, the prediction results of three models were analyzed and compared, the results demonstrated that gray support vector machine model needed less data, and showed greater ability to support vector machine nonlinear mapping and to predict “poor” information. Comparing to the two kinds of single models, it performs better in prediction.2. Prediction of aircraft main fuel control system failure rate. Firstly, the data of main fuel control system failure rate was analyzed, the results showed that the gray model and neural network model were more appropriate to predict the failure rate, and the best parameters and a suitable network structure were selected. Then the possibility of merging the models together was analyzed, which were linear weighting model and direct integration model to predict the failure rate. Following the analysis and comparison of the four models, gray neural network model achieved the best prediction ratio, and showed the characteristics of needing less sample data, greater uncertain information prediction ability, neural network with nonlinear mapping and self-learning, self-adaptive and so on. Finally, the structure of the aircraft maintenance management system is showed, and some pages are described and demonstrated.