Research Of The Application Of Data-mining In Engineering

Nowadays there are amounts of data in manufacturing with the rapid growing of database technique, which troubles the enterprises a lot, but there're useful information in them. While the last few years knowledge discovery tools have been used in such new fields as manufacturing, fault diagnosis, and so on. In this paper, methods and development have been provided to analyze some production problems including fault diagnosis, quality prediction, and technical parameter deployment.The classification algorithm applyed in decision-tree module is C4.5, which can deal with massive mounts of numerical value. It has been applied successfully in car CD fault diagnosis, that is, a decision tree has been got from the history data and useful rules have been extracted from tit. According to the result, estimation become more easy and more convenient than before, and the manufacturer may give a plan of after service ahead of schedule. In this paper, we also introduce other algorithms in classification and make a comparison between them, so that the reasonable choice taken from so many algorithms can be given specifically.The association algorithm used in the module is Apriori, in which hashtree is traditional used and the operation rate is always unsatisfying. While in the paper the operation rate is greatly improved via another method which is a method of comparison, in stead of hashtree. This paper also proposes a method to dispose ANN (Artificial Neural Network) training parameters on the basis of association algorithm which belongs to data-mining. It improves the precision of prediction to the yarn breaking rate in ANN model effectively. Through carrying on Associationrules drawing to the record that tests the parameter in production, this method can rapidly get rid of the training parameters exerting a negative influence, and choose those improving prediction. Thus the goal of improving the ANN performance can be achieved.ANN (Artificial Neutral Network) is usually used in quality prediction especially in manufacturing. In the system mentioned in the paper, the quality of yarn products is predicted, and on the support of Association module, satisfying parameter deployment can be accomplished. Recent forecasting work indicates that the parameter deployment helps to obtain an optimal result.The data-mining tool developed with VS.NET includes three parts: decision-tree module, association module and ANN module, and they all are put into use and provide satisfied results.