An inference model for combustion diagnostics in an experimental oil furnace

Abstract

AbstractThe continuous monitoring of the air/fuel ratio, oil/water/air temperatures, and gas/particulate emissions of combustion processes in oil‐based furnaces allows experts to detect anomalies and act to prevent faults and critical conditions. These important but tedious tasks can be performed by an expert system designed to mimic the human abilities of recognizing relevant patterns and finding their most likely causes. In this article, we present the architecture of an expert system that uses flame images grabbed during the combustion process in an experimental oil furnace as input parameters. Computational processing of those images provides feature vectors for analysis by “artificial experts” that correlate changes in flame appearance with typical combustion states. The Dempster–Shafer method is used to build the knowledge base and the inference engine. The results of tests in which flame conditions are suddenly modified by altering the physical variables of the combustion process revealed that the method can properly combine measures from various flame image characteristics to issue diagnostics. Such diagnostics are similar to those given by human experts. This suggests that the proposed approach may fill the gap between models based on features extracted from images and real‐world operating conditions. This is the intended contribution of this work.