nlwxpearo
Joined: 13 Dec 2010
Posts: 520
Read: 0 topics
Warns: 0/5
Location: England
|
 Posted: Thu 23:16, 16 Dec 2010 Post subject: Ductility of steel joints in the artificial intell |
|
|
Ductility of steel joints in the artificial intelligence prediction
Identify a suitable neural network model for structural engineering is the key to solving the problem. Because it affects not only the network structure, learning and computing speed, but also affect the forecast is accurate. After analysis, were taken to determine the input layer 6, the ductility of the main factors to the node are: concrete strength X. (MPa),[link widoczny dla zalogowanych], the node stirrup X2 (%), reinforced with steel ratio X3 (%), hoop strength X4 (MPa), axial compression ratio x5, steel strength X6 (MPa). (1) to determine the ductility of the output unit with a displacement ductility ratio is generally expressed,[link widoczny dla zalogowanych], and sometimes also used to express the ultimate displacement angle, so, the paper output unit is divided into two: the limit displacement ratio yl, ultimate displacement angle than y2,[link widoczny dla zalogowanych], ( 2) The number of nodes in the hidden layer according to the spreadsheet, find the first hidden layer take six nodes, the second hidden layer nodes take four small systematic errors, therefore, the network structure is 6-4-2. 3 DUCTILITY forecast due to the current strength of steel reinforced concrete frame of less relevant information less,[link widoczny dla zalogowanych], therefore we selected eight high-strength steel reinforced concrete specimen samples for learning,[link widoczny dla zalogowanych], training [5-9】, with the tests to predict the 5 samples, the results shown in Table 1. Table 1 Comparison of predicted and measured results can be seen from Table 1, the application I built ductility steel high strength concrete frame prediction neural network model to predict DUCTILITY, error of less than 1O% in If the samples further increased, possible to predict higher accuracy to meet the engineering requirements, but in the forecast, we must first obtain a certain number of samples to train the network, be predicted. Meanwhile, according to the analysis, we can see on the test specimen in this forecast is basically accurate.
The post has been approved 0 times
|
|
