The objective of the thesis is to implement a neural network capable of estimating, in a sufficiently accurate way, the magnetic model of a synchronous electric motor through the use of an industrial system, a microcontroller board, characterized by a cost of some orders of magnitude smaller than a professional computer. The first part of the paper will describe the main types of neural networks and training algorithms, to better understand the nature of the choices made in the various stages of the research. In the second part, the theoretical notions will then be applied to the case in question, reporting the structure of the tests performed and the data obtained, as well as the main problems encountered. Finally, we will proceed with a post-processing of the results in order to refine what has been obtained for a more precise motor control.
L’obiettivo della tesi è quello di implementare una rete neurale in grado di stimare, in modo sufficientemente accurato, il modello magnetico di un motore sincrono mediante l’utilizzo di un sistema di tipo industriale, una scheda a microcontrollore, caratterizzato da un costo di alcuni ordini di grandezza inferiore rispetto ad un elaboratore professionale. Nella prima parte dell’elaborato si andranno a descrivere le principali tipologie di reti neurali e di algoritmi di addestramento, per meglio comprendere la natura delle scelte effettuate nelle varie fasi della ricerca. Nella seconda parte si andranno poi ad applicare le nozioni teoriche al caso in esame, riportando la struttura dei test eseguiti e i dati ottenuti, oltre alle principali criticità riscontrate. Infine, si procederà con una post-elaborazione dei risultati al fine di affinare quanto ottenuto per un controllo del motore più accurato.
Implementazione di una Rete Neurale su un Microcontrollore Industriale per la Stima del Modello Magnetico di Motori Sincroni
PARISE, LORENZO ALESSANDRO
2021/2022
Abstract
The objective of the thesis is to implement a neural network capable of estimating, in a sufficiently accurate way, the magnetic model of a synchronous electric motor through the use of an industrial system, a microcontroller board, characterized by a cost of some orders of magnitude smaller than a professional computer. The first part of the paper will describe the main types of neural networks and training algorithms, to better understand the nature of the choices made in the various stages of the research. In the second part, the theoretical notions will then be applied to the case in question, reporting the structure of the tests performed and the data obtained, as well as the main problems encountered. Finally, we will proceed with a post-processing of the results in order to refine what has been obtained for a more precise motor control.File | Dimensione | Formato | |
---|---|---|---|
Parise_LorenzoAlessandro.pdf
accesso aperto
Dimensione
23.8 MB
Formato
Adobe PDF
|
23.8 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/10203