The Application Of Data Mining In The Monitoring System Of One Certain Test Bed

Data mining can extract useful information from a large number of data, it’s a new research direction in the research field of database system and can provide a strong basis for decision making. This paper takes a type of test bed system as the foundation, makes a tentative study on application of data mining technology in the test database of aero engine components. This paper introduced the basic concepts of data mining technology and general knowledge, the related algorithm in data mining algorithms are improved and been applied to the test bed monitor system. The main research contents of this paper are as follows:In order to find the relationship between speed, amplitude, temperature and other parameters of a test bed system, the association rules have been introduced to mine the rules of 28 parameters in monitoring system database. The high confidence and support of the association show the mined rules are effective. To improve the efficiency of the classical Apriori algorithm in the mining of monitoring system, an improved Apriori algorithm which can effectively reduce the candidate has been put forward. The results show that, compared with the classical Apriori algorithm, the candidate set size is effectively reduced, operation time significantly reduced under the same degree of support and confidence.To solve the problem that linear mode can’t predict the relationship between amplitude,which is the main monitoring parameter, flow, pressure and other parameters of Aero engine accessory test bed, a 7-8-1 Genetic Back Propagation(BP) neural network structure is established, whose threshold and weight optimized by genetic algorithm(GA). A prediction model is built by using the output speed, booster pump entrance flow, booster pump entrance temperature, booster pump entrance pressure, booster stage outlet flow, booster stage outlet temperature and booster stage outlet pressure as the network input and radial amplitude of aircraft accessory as the network output. The trained network is used to predict the vibration trend of a certain type of aircraft accessory and the error between the acquired prediction value and the actual value is under the requirement. Compared with the traditional linear regression model, the genetic BP neural network model has a higher accuracy, which is suitable and feasible for the prediction of vibration trend.