Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
INVERSE MODEL BASED ON NEURAL NETWORKS TRAINED WITH EXTREME LEARNING MACHINE AND LEVENBERG-MARQUARDT ALGORITHM
Submission Author:
Ádamo Oliveira , PR , Brazil
Co-Authors:
Ádamo Oliveira, Gideon Leandro
Presenter: Ádamo Oliveira
doi://10.26678/ABCM.COBEM2019.COB2019-0349
Abstract
In this paper, inverse models are obtained using single-hidden layer feedforward neural networks (SLFN) trained with Extreme Learning Machine (ELM) and Levenberg-Marquardt Algorithm (LMA). Three case studies are presented: buck converter, electric heater and cascaded tank system. The algorithms performances are compared through the SLFN training time and the coefficient of determination for the inverse models. Since these algorithms use random number generation steps, the Kolmogorov-Smirnov test and Wilcoxon rank sum test were performed to analyze the data from simulation experiments in order to draw more reliable conclusions. The results of the tests indicate that ELM tends to achieve better performance to obtain inverse models for the studied systems.
Keywords
inverse models, Extreme learning machine, Levenberg-Marquardt Algorithm, hypothesis test
