Find in Library

Several potential network structures are chosen to do a large number of experimental analysis, historical data is divided into training sample and testing sample, and the corresponding neural network model is established with BP learning algorithm. After checking the testing sample, a superior network integration model which can be applied for hydraulic metal structure health grade diagnosing is determined. By plenty of experimental tests and verification analysis, it is concluded that the two-hidden-layer neural network model suits hydraulic metal structure health diagnosing better. As for the gate health diagnosing, based on Bagging technology, the BP neural network integration model for hydraulic metal structure health diagnosing is researched and constructed. The analysis of the sample showed that its accuracy rate (78%) is obviously better than the single neural network model(67%). The BP neural network integration model will work together with the FAHP model the author studied, that can make the diagnosis results more reasonable and reliable.