A Boiler Flame Detection Based On Bayesian Criteria

In power industry of our country, coal-fired thermal power generating units occupy the leading position of power supply. Relating to the safety of property, the safe operation of boiler equipment is all concerned by all large-scale enterprises. This is because the boilers are huge in size and during the process of coal burning in its internal combustion, improper operation leads to instability of coal burner flame. If such a situation is not detected and the burner powder is continuously fed. the furnace will be accumulate large amounts of untreated combustion of fuel which mixed with air. Then any slight of fire will light the mixture and add the pressure in furnace, which lead explosion or deflagration. This will be a serious threat to the security of furnaces and equipment. Therefore, in order to ensure safety in production, it is necessary to monitor not only the flame of inside the boiler, but also the flame state of single burner.The boiler flame detection project based on DSP system is designed to detect the state of the burner. This dissertation introduces firstly the hardware of the system of boiler flame monitoring, analyses the existing boiler flame detection algorithm, and takes a test over the data of burned pulverulence flame in coal boiler which are provided by Nanchang Longyuan State Power Control Technology Co.Ltd. Then, we put forward the cumulative frame difference algorithm, and have an experiment both in coal boiler flame video and oil boiler flame video. The results show that the algorithm is simpler and more efficient. In order to improve the algorithm, the dissertation focuses on the characteristics and distribution of burning flame. Combing with Bayesian theory, we put forward an automatic thresholding algorithm to obtain the best results which solves the problem; the threshold value of the experience is not accurate. We take experiments and select a threshold effectively. Practice has proved that this algorithm greatly reduces the probability of false alarm. We use Visual C++ 6.0 based on Microsoft Windows XP compiling software, and use Code Composer Studio(6000)to compile DSP software.