Eventos Anais de eventos
DINAME2019
DINAME2019
Fuzzy inference of oil furnace combustion state through computer vision information
Submission Author:
Gustavo Cunha da Silva Neto , AM
Co-Authors:
Danilo de Santana Chui, Flávio Trigo, Flavius Martins, Agenor Fleury
Presenter: Danilo de Santana Chui
doi://10.26678/ABCM.DINAME2019.DIN2019-0198
Abstract
Regular furnace operation systems require continuous monitoring of air/fuel ratio, oil and water temperatures, combustion byproducts emissions, etc. Experts analyze these data to detect anomalies and act to prevent the system to reach critical or undesired conditions. PID controllers may control parameter reference levels, however human decision is still crucial to the control process. A first step on human decision is to recognize flame patterns that constitute anomalous behavior and then, through experience, change parameters to stabilize the combustion process. This research focus on this first step and proposes a method that infers different anomalous states from images captured by a digital camera from an experimental oil furnace. Different image processing algorithms extract information through features vectors that are analyzed by a previously trained “artificial expert”. State of the combustion processes are then obtained through fuzzy inference together with estimated input values. Results show that the proposed “artificial expert” is able identify most different anomalous states as desired.
Keywords
combustion diagnostics, Digital Image Processing, fuzzy inference
