Neural network-based estimation of time-varying city bus mass
| Autor: | Miković, Matija |
|---|---|
| Přispěvatelé: | Deur, Joško |
| Jazyk: | chorvatština |
| Rok vydání: | 2021 |
| Předmět: | |
| Popis: | U ovom radu predstavljena je neuronska mreža kao moguća metoda procjene mase vozila. Objašnjeni su osnovni pojmovi neuronskih mreža. Opisan je postupak prikupljanja i obrade velike količine podataka potrebnih za neuronske mreže. Proučena je već postojeća metoda procjene mase korištenjem fizikalnog modela uzdužne dinamike vozila. Provedeno je učenje neuronskih mreža uz prethodno odabranu strukturu i definirane parametre učenja. Predložene su četiri metode procjene mase korištenjem neuronskih mreža. Analizom rezultata usporedile su se različite neuronske mreže i predložene metode procjene mase. Također su se za usporedbu prikazali i rezultati fizikalnog modela. Naposljetku su dane smjernice za moguća poboljšanja. In this thesis, the neural network is presented as a possible method of vehicle mass estimation. Basic concepts of neural networks are explained. The process of collecting and processing a large amount of data required for neural networks is described. The already existing method of mass estimation using a longitudinal vehicle dynamics model was studied. Neural network was trained with a previously selected structure and defined training parameters. Four methods of neural network-based estimation of mass have been proposed. Different neural networks and proposed mass estimation methods were analysed and compared. The results of vehicle dynamics models were also presented for comparison. Finally, guidelines for potentional improvements are given. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|