半圓頂型屋蓋結構風壓頻譜之類神經網路模擬

ANN MODELING OF WIND PRESSURE SPECTRA ON HEMISPHERICAL DOME ROOFS

王人牧(Jenmu Wang) ； 劉博溢(Bo-Yi Liu) ； 羅元隆(Yuan-Lung Lo)

《中國土木水利工程學刊》 31卷8期 (2019/12) Pp. 739-746

https://doi.org/10.6652/JoCICHE.201912_31(8).0007

摘要

結構物的耐風設計通常需要經由風洞實驗，取得風壓頻譜的實驗數據，其過程相當耗時且費用昂貴。使用回歸公式來整理分析實驗數據，常無法得到準確的風壓頻譜值，因此，如何更有效的利用風洞實驗氣動力資料庫是一個重要的課題。本研究利用半圓頂型結構風壓資料庫，著重於半圓頂模型其曲率與結構高度的變化，對於子午線上風壓頻譜的影響，利用隨機選取法撰寫幅狀基底函數類神經網路（RBFNN）程式，在訓練、驗證與測試網路的過程中，尋找符合理論且準確的估算模型，最後得到的ANN預測模型與前人之回歸公式做進一步的比較探討，更突顯了其準確性與適用性的優勢。

關鍵字

輻狀基底函數類神經網路 ； 結構風工程 ； 風壓頻譜 ； 大跨徑結構

並列摘要

Wind resistant design of buildings often needs to acquire wind spectra from wind tunnel tests. Using regression formulas to process and analyze experimental data of wind spectra usually is not very accurate. Therefore, one of the most important issue is how to use experimental wind load aerodynamic database more effectively. A wind pressure database for hemispherical dome roofs was collected. The emphases of the research were on the study of wind pressure spectra on the meridian with the change of curvature and height as well as the establishment of an Artificial Neural Network (ANN) prediction model. Random center selection method was used to write Radial Basis Function Neural Network (RBFNN) programs to train, validate and test the ANNs. The estimation models found not only accurate but also theoretically consistent. ANN Models were also compared with previous regression formula showing better accuracy and applicability.

並列關鍵字

RBFNN ； structural wind engineering ； wind pressure spectrum ； large span structure