類神經網路模式用於推估邊坡破壞面積之研究

Estimation of Slope Failure Area byUsing Neural Networks Model

曾智鴻(TSENZ-CHIH-HUNG)

指導教授 ： 柯亭帆

國立屏東科技大學/工學院/土木工程系所/碩士(2014年)

https://doi.org/10.6346/NPUST.2014.00032

摘要

台灣地區受到颱風襲擊，或是遇到強降雨，常常導致山區邊坡崩塌形成土石流，河川溪水氾濫並且堤防潰堤及平地淹水，莫拉克颱風後之影響更鉅。本研究之目的在於了解屏31縣道往德文村及大社村沿線，及台24線26.8K處的三德檢查哨到谷川大橋兩處邊坡因遭災害而破壞之情形調查，並利用類神經網路推估邊坡破壞面積。本研究以人員現地調查方式進行邊坡現況調查，發現有49處有邊坡破壞之現象，而從調查中發現有四項主要造成破壞之自然因子，亦即(1)坡度、(2)坡高、(3)坡長、及(4)雨量。本研究將調查搜集到之資料，利用類神經網路建立上述破壞因子與邊坡破壞範圍面積之相關模式。研究結果顯示在類神經網路模式中輸入層參數為雨量、坡長、與坡高三項，隱藏層3個神經元，及輸出層為邊坡破壞面積，可以得到相對最佳的模式。推估結果與實際測量值之相關係數值平方達0.802，由此可知本研究所建立之推估模式具有不錯之預測能力，而亦可應用到其他地方，因此可提供相關工程領域一些有用之參考資訊。

關鍵字

山區道路 ； 邊坡破壞 ； 破壞因子 ； 破壞面積 ； 類神經網路模式

並列摘要

Taiwan is the place which easily suffers the devastation of typhoon or strong downpour, by such frequently happened natural disaster, it would lead the slop of mountain to collapse and turnout to be mudslide , deluge or serious cataclysm.After typhoon morocco , the situation becomes even worse, so the purpose of this study is to realize the damage caused by catastrophe at the location of This study takes road line Ping31 from Dewentodashe Along the line and road line Tai24 from sande Checkpoint to gu chuan River Bridge Along the line as the objects.And uses the method of on-site investigation to examine current slope situation in these road sections. it is observed that there are 49 slopes have a phenomenon of slope failure, The investigation results are also shown that there exist four main natural reasons to cause the damages, which are: (1) slope, (2) slope height, (3)slope length, (4) rainfall. we are going to take this measurement to accomplish a mode which is related between neural network approach and slope-wrecking area by our collecting-data form previous investigation.
The results show that the parameters in the input layer neural network model for rainfall, slope length, and slope height of three .Hidden layer 3 neurons, and output layer slope failure area, you can get a relatively optimal model. Squared correlation coefficient estimate with the actual measured values of 0.802,The research reveals the perfect match :If we input the 4 major factors mentioned above, we could get an accurate data of wrecking-zone as our result thereby, the prediction-mode not only provides us a good prefiguration of area-dilapidation, but also we might be able to offer some useful references to certain related-construction.

並列關鍵字

mountain roads ； slope failure ； damage factor ； damage area ； Neural network model

參考文獻

15. Choobbasti, A.J., Farrokhzad. F. ,Barari, A. 2009,‟Prediction of slope stability using artificial neural network(case study: Noabad, Mazandaran, Iran) ”, Arabian Journal of Geosciences, Volume 2, Issue 4, pp 311-319.

17. Biswajeet Pradhana, Saro Lee,2010, ”Landslide susceptibility assessment and factor effect analysis: backpropagationartificial neural networks and their comparison with frequency ratio andbivariate logistic regression modelling”, Environmental Modelling & Software 25, pp. 747–759.

19. Dawson, C.W. & Wilby, R.L. , 2001 , Hydr- ological modelling using artificial neural networks. Prog. pp., 80-108.

21. F. Farrokhzad, A. Bararib,A.J.Choobbastia, .B.Ibsenb,2011, ”Neural network-based model for landslide susceptibility and soillongitudinal profileanalyses:Twocasestudies”,JournalofAfricanEarthSciences,Volume 61, Issue 5, Pages 349–357.

22. Fookes, P.G., Sweeney, M., Manby, C.N.D., and Martin, R.P. ,198“Geological and geotechnical engineering aspects of low-cost roads inmountainous terrain”, Engineering. pp. 1-152.