Desarrollo de un modelo de regresión con redes neuronales artificiales para estimar la resistencia rotórica de un motor de inducción.

Published in: Engineering for a Smarter Planet: Innovation, ITC, and Computational Tools for Sustainable Development: Proceedings of the 9th Latin American and Caribbean Conference for Engineering and Technology
Date of Conference: August 3-5, 2011
Location of Conference: Medellin, Colombia
Authors: Juan Carlos Jaimes Jauregui
Oscar Eduardo Gualdrón Guerrero
Jorge Luis Diaz Rodriguez
Refereed Paper: #208

Abstract

This paper presents the implementation of Neural Networks in a programmable logic device such as the FPGA, the aim is to develop a model in Sysgen able to estimate the rotor resistance in induction motors. To create a neuron in Sysgen is from the standard model of an artificial neuron, which is the sum of weights minus the threshold, all multiplied by an activation function, in this case the activation function was used for Tansig Purelin hidden layer to output layer. The neural network uses decimal data input/output, and the FPGA is digital, therefore Sysgen was designed in a fitting capable of handling binary data on the card, for this we used a Digital- Analog Converter, and was created as suppress the number of pins used in the FPGA, designing a stage of registers used to store the desired data and thus send it to one of the inputs of the neural network. The data frame becomes progressively each record is used to store the decimal value present at the output of Digital-Analog converter and each time you change the data logging is activated for the input of the neural network.