Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
DESIGN OF EXPERIMENTS COMBINED WITH ARTIFICIAL NEURAL NETWORKS APPLIED ON THE CONTROL PARAMETERS OF A REAL STEAM GENERATOR
Submission Author:
Lara Werncke Vieira , RS
Co-Authors:
Lara Werncke Vieira, Barbara Pacheco da Rocha, Paulo Smith Schneider
Presenter: Lara Werncke Vieira
doi://10.26678/ABCM.COBEM2019.COB2019-0216
Abstract
The present paper aims to use statistical analysis through DoE methodology to select and rank the most influential parameters of a steam generator modeled by an ANN. The goal is to focus the efforts only on the significant operational input parameters according to their influence on the three responses: steam generator efficiency, electric power generation and flue gas outlet temperature. The main controllable parameters are: primary air flow, pulverized coal outlet temperature, velocity of the dynamic classifier, excess O2, primary air’s crossover duct pressure, secondary air’s crossover duct pressure and coal flow. The ANN is trained with operational data provided by the PECEM power plant (2x360 MW), from Fortaleza, Brazil. The metrics that evaluate the ANNs performance are the mean absolute error, the mean square error and the mean percent error. According to the DoE results the primary air flow and the velocity of the dynamic classifier are not important to the flue gas outlet temperature output. A simplified ANN is developed with the selected parameters. For the two other outputs, all input parameters remained significant.
Keywords
coal-fired power plant, design of experiments, Artificial neural networks, Response Surface Methodology, Steam Generator
