Research On Furnace Flame Detection Based On Msif

For large-scale power plant coal-fired boilers, the process of pulverized coal combustion is complex suspension combustion, and its condition is very unstable. So monitoring the state of suspension combustion is very important to boiler's reliability, security and economy. Firstly, CCD as well as relative image acquisition and processing equipment are adopted as measurement tool to acquire the image of burner flame, based on analysis of the traditional methods and characteristics of flame detector. Secondly, the characteristic of BP neural network is detailed expatiated. Based on BP neural network, the judgement method of flame burning state is put forward. Result shows that using neural network information fusion is an effective way. Last, a two-step fusion framework based on combination of BP neural network and the D-S evidence theory is put forward, which aims at the uncertainty in flame detection. Result shows that, this method is verified feasibility and effective, and can effectively improve the flame detection accuracy and boiler combustion safety.