橋梁在跨距長的情況下容易遭受風力之影響,因此在設計橋梁時,如何決定斷面寬深比是需要考量的因素,為了探求斷面寬深比對於氣動力之效應而進行風洞實驗(Wind Tunnel Test)。顫振導數(Flutter Derivatives)是影響氣動力穩定性最重要之參數,而本文採用類神經網路識別法(IDNN),是在類神經網路訓練後而獲得權值,權值即代表結構反應特性,最後建立出識別模式而求得氣動力顫振導數值。在類神經網路識別時,考慮各個平板斷面模型,並且分別在平滑流場(Smooth Flow)與紊流場(Turbulence Flow)兩種流況下,經由風洞實驗在各風速時獲得垂直向和扭轉向之位移歷時反應,即為識別所採用之資料。 由於要經由風洞實驗以獲得輸入資料後才可使用識別法,因此本文為了不藉由實驗過程而要獲得某一平板斷面模型之顫振導數值,所以應用類神經網路(Artificial Neural Network),在各種平板斷面模型之無因次風速作為網路訓練資料,訓練完成時,此時網路已具有結構之特性,最後進行預測時輸入未知平板斷面模型資料,而得到顫振導數值。本文結果發現,利用類神經網路識別法以及類神經網路預測求得顫振導數值是為一種可行性高之方法。
This investigation develops an artificial neural network (ANN) algorithm to identify aeroelastic parameters of cable-supported bridge section models in smooth flow and turbulent flow in a wind tunnel test. The ANN approach method uses observed dynamic responses to train a back-propagation (BP) neural network frame. The characteristic parameters of the section model for various wind velocities are estimated using weight matrices in the neural network. The eight flutter derivatives can then be determined precisely. The procedure can be applied to process experimental data obtained from wind tunnel tests involving flat plate section models given various width/depth (B/D) ratios. Finally, the flutter characteristics of various bluff bodies are examined, as they are very sensitive to geometry and structural dynamics.