Sekundarni jezik: |
Angleški jezik |
Sekundarni naslov: |
ǂThe ǂprediction of landslide movements with artificial neural networks |
Sekundarni povzetek: |
The thesis deals with the problem of the prediction of landslide movements with artificial neural networks (ANN). At the beginning of the thesis the landslides are introduced in general and especially the Macesnik landslide. Later on geodetic methods for observing movements of landslides are described from referential geodetic methods to geodetic methods for mass collection. The list of content of the project is proposed, which should be used for geodetic observations of the Macesnik slide. The description of all past geodetic observations of the Macesnik slide is given and analysis of the effect of rainfall on the landslide movements is presented. In the fifth part artificial neural networks are presented, the training principles of artificial neural networks and a detailed explanation of artificial neural network with error back propagation algorithm. In the experimental part we have presented the use of artificial neural networks for the prediction of landslide movements. It is shown that the use of ANN can be a successful alternative to other methods, which require thorough geological, hydrological and geomechanical evaluation. It is usual for landslides that none of influential factors alone can reliably explain the sliding mechanism of the slope. Also, with statistical tools it is difficult to correctly determine dependency between individual influential factors. To use artificial neural networks we do not need to know all influential factors and we also do |
Sekundarne ključne besede: |
geodesy;graduation thesis;Macesnik's landslide;artificial neural network;geodetic measurements;movements;rainfall;analysis; |
Vrsta datoteke: |
application/pdf |
Vrsta dela (COBISS): |
Diplomsko delo |
Komentar na gradivo: |
Univ. v Ljubljani, Fak. za gradbeništvo in geodezijo |
Strani: |
XI, 89 str. |
Vrsta dela (ePrints): |
thesis |
Naslov (ePrints): |
The prediction of landslide movements with artificial neural networks |
Ključne besede (ePrints): |
Macesnikov plaz;umetna nevronska mreža;geodetske meritve;premiki;padavine;analiza |
Ključne besede (ePrints, sekundarni jezik): |
Macesnik’s landslide;artificial neural network;geodetic measurements;movements;rainfall;analysis |
Povzetek (ePrints): |
Diplomsko delo obravnava problem napovedovanja plazenja zemljin z umetnimi nevronskimi mrežami. V njem najprej predstavimo plazove na splošno, ter posebej Macesnikov plaz. V nadaljevanju opišemo geodetske metode za spremljanje premikov zemeljskih plazov od referenčnih geodetskih metod do geodetskih metod za masovni zajem. Naštejemo vsebine projekta, ki naj bi ga uporabljali izvajalci geodetskih opazovanj na Macesnikovem plazu. Sledi opis vseh dosedanjih opazovanj na Macesnikovem plazu in analiza vpliva padavin na plazenje. V petem delu predstavimo umetne nevronske mreže, pravila učenja umetnih nevronskih mrež ter podrobno razložimo umetno nevronsko mrežo z vzvratnim razširjanjem napake. V eksperimentalnem delu je prikazana uporaba umetnih nevronskih mrež pri napovedovanju plazenja zemljin. Predvidevamo, da njihova uporaba predstavlja alternativo klasičnim prognoznim metodam, ki jih pridobimo z geološkimi, hidrološkimi in geomehanskimi meritvami. Za plazove je značilno, da noben vplivni dejavnik ne more zanesljivo razložiti obnašanja plazenja zemljine, prav tako je s statističnimi orodji težko pravilno ugotoviti odvisnosti med posameznimi vplivnimi dejavniki. Za uporabo umetnih nevronskih mrež ne potrebujemo poznavanja vseh vplivnih dejavnikov, prav tako ne potrebujemo poznavanja odvisnosti med njimi. Umetna nevronska mreža se nauči teh relacij iz dovolj velikega števila vhodno – izhodnih parov. Za vhodne podatke smo uporabili meritve padavin, za izhodne pa premike plazenja zemljine. V sedmem delu podamo zaključek, ki vsebuje kratek povzetek vseh glavnih ugotovitev. |
Povzetek (ePrints, sekundarni jezik): |
The thesis deals with the problem of the prediction of landslide movements with artificial neural networks (ANN). At the beginning of the thesis the landslides are introduced in general and especially the Macesnik landslide. Later on geodetic methods for observing movements of landslides are described from referential geodetic methods to geodetic methods for mass collection. The list of content of the project is proposed, which should be used for geodetic observations of the Macesnik slide. The description of all past geodetic observations of the Macesnik slide is given and analysis of the effect of rainfall on the landslide movements is presented. In the fifth part artificial neural networks are presented, the training principles of artificial neural networks and a detailed explanation of artificial neural network with error back propagation algorithm. In the experimental part we have presented the use of artificial neural networks for the prediction of landslide movements. It is shown that the use of ANN can be a successful alternative to other methods, which require thorough geological, hydrological and geomechanical evaluation. It is usual for landslides that none of influential factors alone can reliably explain the sliding mechanism of the slope. Also, with statistical tools it is difficult to correctly determine dependency between individual influential factors. To use artificial neural networks we do not need to know all influential factors and we also do |
Ključne besede (ePrints, sekundarni jezik): |
Macesnik’s landslide;artificial neural network;geodetic measurements;movements;rainfall;analysis |
ID: |
8307942 |