The Research Of Flame Detecting And Combustion Diagnosing Based On Digital Image Processing For Coal-fired Utility Boilers

Flame detection plays an important role in Burner Management System for avoiding boilers exploding. In order to ensure the safety of boilers, flame signals provided by flame detection system must be reliable and precise.The main purpose of this research is to study the method of flame detecting and combustion diagnosing for 300MW coal-fired utility boilers. The research is based on digital image processing according to coal-fired utility boilers characteristics and flame image feature characteristics. The system can detect the flame more reliably, and monitor the course of burning at the real time.The research is carried on the basis of image processing and the pattern recognition using BP neural-network. This system helps dynamic checking flame combustion status.The main work of the research is as follows:1. After consulting the flame detecting technology of overseas, considering 300MW coal-fired utility boilers characteristics and the field condition, the function of the system is analyzed. Aiming at the disadvantages of former flame detector, such as the pollutant on thesurface of flame sensor, aging caused by high temperature, and little visual angle, image transmission optical fiber and CCD with blowing and cooling system are used as flame sensor in this system. This kind of sensor has larger visual angle to overcome flame drifting, and the structure of the hardware system for flame image collecting is designed for 300MW coal-fired utility boilers.2.To facilitate that computer recognizes the state of boilers, flame images captured by camera are processed on the basis of characteristics of flame images. A processing method which is optimal to extract characteristics of flame images is gotten by experiments.3.On the basis of flame characteristics, aiming at mutually jamming between adjoining burners (named as "peeking") causing wrong alarm by former flame detectors, flame feature characteristics of the native burner as the inputs of BP to enhances anti-jamming ability effectively in this system. And the pattern recognition using BP neural-network helps dynamic checking flame combustion status, ensures the system more adaptability. The flame diagnosis algorithm, which is self-learning and self-adapting, is put forward according to the characteristic of flame image, to avoid failure diagnosis of former flame detecting. The reliability of the system is higher.4. The structure of the software system for flame detecting isdesigned for 300MW coal-fired utility boilers.