Font Size: a A A

Prediction Methods Based On Neural Network Device Status And On The Blower

Posted on:2005-01-15Degree:MasterType:Thesis
Country:ChinaCandidate:H F QinFull Text:PDF
GTID:2192360125955423Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
D350 Rapid-Speed Fan is the key equipment in Pingguo Aluminium Co.Ltd. It is very important for equipment's reliability and security by developing fault diagnose and state prediction. As an important part of fault diagnose, Prediction Technology (PT) can be used to predict working state and developing trend. PT can be regarded as a higher level fault diagnose technology.In this paper, author summarized completely fault diagnose field -development and trend of PT. Prediction method based on optimized neural Network is put forward. Whether the connotative node number of neural Network is reasonable or not make a great influence on neural Network's performance. Based on reduction ability of Rough Sets, a new method is proposed, that taking advantage of Rough Sets decision and reduction, connotative layer structure of Neural Network is optimized. And so, a more reasonable network structure is made as possible. What's more, prediction result is much preciser.Combined prediction based on Neural Network is also being put forward. On the foundation of multinomial regression, GM(1,1) prediction and feed-forward Neural Network prediction, all of which have their own characters, combined prediction based on Neural Network is assembled. Taking advantage of great nonlinear and terribly strong self-learning ability of Neural Network, the difficulty of confirming weight coefficient of each prediction method in combined prediction is solved; Data is pretreated by means of principal component analysis technology. Precision of combined prediction is further improved. At the same time, based on optimized feed-forward Neural Network, combined prediction is carried out successfully in alarm time prediction of D350 Rapid-Speed Fan.Relating practice engineering project, applying the above research achievements, our project team developed on-line prediction and fault diagnosing system of D350 Rapid-Speed Fan. At present, this system has been checked, accepted and has brought good economical and social profits for the corporation.
Keywords/Search Tags:Neural Network, iterative arithmetic, Rough Sets, GM(1, 1), Combined Prediction, D350 Fan
PDF Full Text Request
Related items