Options
結構非線性反應之系統識別
Date Issued
2004
Date
2004
Author(s)
鄭弘偉
DOI
zh-TW
Abstract
Earthquakes make a lot of damages in Taiwan. It is really important to make sure that how the structures behave during earthquakes. Recently the neural network is a powerful tool to identify the nonlinear properties. So I use the neural networks to identify a 3-story RCS structures property. The 3-story RCS structures prototype structure is designed for a highly seismic location either in Taiwan or California. Pseudo dynamic test was conducted to this frame structure. The total test consists of four excitations in the following order: 1) 50% chance of exceeding in 50 years TCU082, 2) 10% chance of exceeding in 50 years LP89g04, 3) 2% chance of exceeding in 50 years TCU082, and 4) 10% chance of exceeding in 50 years LP89g04. These ground motions were followed by a monotonic push out to 8% roof drift. And then I identify the frequency response function of the structure.
And then I use the 2-input and 1-output neural network to identify the Bitan bridge’s dynamic property such as frequency response function and multi-input single output Volterra kernels.
Finally, I use the Kalman filter and recursive least square method to make the procedures of finding the weighting of the neural network easily. It shows that even if we simplify the procedures ,it still has a great good performance to identify the 3-story RCS structure’s hysteretic force.
And then I use the 2-input and 1-output neural network to identify the Bitan bridge’s dynamic property such as frequency response function and multi-input single output Volterra kernels.
Finally, I use the Kalman filter and recursive least square method to make the procedures of finding the weighting of the neural network easily. It shows that even if we simplify the procedures ,it still has a great good performance to identify the 3-story RCS structure’s hysteretic force.
Subjects
類神經網路
Neural Network
Type
thesis
File(s)
No Thumbnail Available
Name
ntu-93-R91521237-1.pdf
Size
23.31 KB
Format
Adobe PDF
Checksum
(MD5):9662e60f67934707dbc8ed6ef7cf624e