標題: 類神經網路在結構系統識別上之應用
The application of Artificial Neural Network in system identification of structure
作者: 駱國陽
Kuo-Yang Luo
鄭復平
Fu-Ping Cheng
土木工程學系
關鍵字: 模態指數、模態曲率指數、模態振頻指數;MODAL INDEX, MODAL CURVATURE INDEX, MODAL FREQUENCY INDEX
公開日期: 1994
摘要: 結構物施工完成後,常因許多原因而導致結構物發生變異或是結構物的質 量,勁度偏離當初設計的安全範圍,在這種情形下,如果遇到地震或其他外 力因素,就可能會對結構物造成危險的威脅。如果有適當的結構系統識別 的方法,能夠識別出結構物是否已經產生勁度、質量上的變異,變異程度 多少,我們就可以據此來修正原先設計時的結構物數學模型,重新評估結 構物的安全性;必要時加以適當的補強與修護,以減少結構之損害,達到 預防勝於治療的目的。本文是以面積慣性矩(AREA MOMENT OF INERTIAL) 及密度之變異模擬結構桿件勁度及質量之變異,並以PC版ANSYS5.0A套裝 軟體,建構出金屬桿件有限元素法模型,求得該結構模型之動態特性(自 然頻率、振態),加上訓練倒傳遞類神經網路,建立金屬桿件元素變異型 式(位置、面積慣性矩、密度)與動態特性之關連性,爾後一旦獲取未知 變異金屬桿件之動態特性時,即可藉類神經網路做桿件之元素變異位置、 勁度、密度之判別。本文研究發現,以有限元素法配合類神經網路做為結 構系統識別之工具,將是一個可行的方法。 The mass and stiffness of the structures might be varied after its completion due to various reason. These variations might be dangerous when earthquake occur. The system identification method is needed to identify these variations for the reference of the repairing and strengthening work. The building structure was simulated by the cantilever beam. ANSYS FEM package PC version was used to analysis the dynamic characteristics of the beam. After the relationship between the variations of the structure and its dynamic characteristics was established, the artificial neural network was used to be an identification system. This approach was shown to be appropriate in this research.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT830015019
http://hdl.handle.net/11536/58709
Appears in Collections:Thesis