Diagnostics and Control of Natural Gas-Fired furnaces via Flame Image Analysis using Machine Vision & Artificial Intelligence Techniques

Abstract

OBJECTIVESA new approach for detection of real-time properties of flames was used in this project to develop sensors to improve diagnostics and controls for natural gas fired furnaces. Camera images along with advanced image analysis and artificial intelligence techniques were used to provide high speed virtual sensors suitable for integration with the plant control system. The output of these sensors provides guidance for balancing air/fuel ratios. Identifying and correcting fuel rich burners would result in improved fuel efficiency. It is anticipated that this on-line diagnostic and control system will offer great potential for improving furnace thermal efficiency, lowering NO x and carbon monoxide emissions, and improving product quality.This project addresses the need for improved diagnostics and burner-balancing control of natural gas fired glass furnaces in order to improve fuel efficiency and reduce emissions. The project involved two Phases. Phase I was a feasibility study with one year duration and was completed in This report starts with a summary of accomplishments during phase I (February 2000(February -2001 followed by a summary of accomplishments during each year of phase II (February 2001-December 2004. 3 SUMMARY OF ACCOMPLISHMENTS DURING PHASE IPhase I activities included: 1) Camera and spectrometer equipment selection and acquisition, 2) Data acquisition at a pilot-scale glass furnace in Rolla, Missouri, and at a research combustion facility in Pennsylvania, 3) spectroscopic data analysis and temperature calculation based on black body radiation model, and 4) Flame image analysis including image processing and feature extraction and classification by two ART neural network algorithms. At the end of Phase I a flame classification simulation game was developed as a marketing tool. Highlight of Phase I activities:• Equipment Acquisition:• Periscope, camera, Spectrometer, computer, and laptopSeveral experiments were performed for collecting data using both spectrometer and camera at the following two laboratory scale furnaces. A simulation game module was also developed for demonstration of the concept. Figure 1 shows the cover page of this module which can be played on the web using JAVA Media Frame Player and is currently located at: http://www.missouri.edu/~keyvans/public_html/demo/cover.html Pilot Scale Glass Pilot Scale Glass FurnaceThis pilot scale furnace was a 0.16 Million Btu/hr glass furnace designed to carry out simulation and parametric studies of industrial glass tanks. The burner of this furnace was of the diffusion flame type, typically used in glass furnace applications. The furnace was controlled using a Labview hardware and software control system. Figure 2 shows the equipment set-up at this furnace. Figure 3 shows the computer screen of the furnace Labview software. Combustion FacilityCombustion experimentations were conducted using a 2 Million Btu/hr combustion facility (see Figure 4) at the Pennsylvania State University in September and October 2000. The objective was to per...

Multivariate Image Analysis of Flames for Product Quality and Combustion Control in Rotary Kilns

“…Flame imaging has already been investigated in the past, but most of these contributions were performed on laboratory-scale combustion systems and premixed flames, and their approach was to compute geometrical and luminous properties of the flame extracted from gray-scale images and to use them to either classify the flame into arbitrarily defined states , or to predict various quantities such as flicker rate, unburnt carbon, CO 2 and NO x emissions, − or fuel and air flow rates . Only a few past investigations were extracting the flame features from RGB color images , and were taking advantage of the three wavelengths to estimate the flame temperature distribution using a bicolor method. Finally, a few research works analyzed the flames using spectrometers, − from which it is possible to extract more precise chemical information about the radicals present in the flame.…”

“…Only a few past investigations were extracting the flame features from RGB color images , and were taking advantage of the three wavelengths to estimate the flame temperature distribution using a bicolor method. Finally, a few research works analyzed the flames using spectrometers, − from which it is possible to extract more precise chemical information about the radicals present in the flame. A limitation of these approaches is related to extracting flame visual features directly from the image space.…”

Rotary Kiln Product Quality Forecasting Based on Flame Imaging

“…Flame imaging has already been investigated in the past, but most of these contributions were performed on laboratory scale combustion systems and premixed flames, and their approach was to compute geometrical and luminous properties of the flame extracted from gray scale images and use them to either classify the flame into arbitrarily defined states (Bertucco et al, 2000;Victor et al, 1991) or to predict various quantities such as flicker rate (Huang et al, 1999), unburnt carbon, CO 2 and NO x emissions (Shimoda et al, 1990;Lu et al, 1999;Yan et al, 2002) or fuel and air flow rates (Tao and Burkhardt, 1995). Only a few past investigations were extracting the flame features from RGB color images (Wang et al, 2002;Keyvan, 2003) and were taking advantage of the three wavelengths to estimate the flame temperature distribution using the bicolor method. Finally, a few research works analyzed the flames using spectrometers (Keyvan, 2003) from which it is possible to extract more precise chemical information about the radicals present in the flame.…”

“…Only a few past investigations were extracting the flame features from RGB color images (Wang et al, 2002;Keyvan, 2003) and were taking advantage of the three wavelengths to estimate the flame temperature distribution using the bicolor method. Finally, a few research works analyzed the flames using spectrometers (Keyvan, 2003) from which it is possible to extract more precise chemical information about the radicals present in the flame. An additional limitation of these approaches consists of extracting the flame visual features directly from the image space.…”