Study On Flame Combustion State Detection Method For The Boiler

Industrial boiler is a kind of common device in industrial production.The combustion state of flame is the key factor that affects the start and stop of the boiler system and the timing of fuel input.Therefore,flame combustion state detection is of great significance to the safety,efficiency,energy saving,environmental protection and combustion efficiency of the production process.Because of the state of diverseness,complicating and noising in the Industrial boiler,it is difficult to detect the flame combustion state.In this paper,the flame combustion state detection method is researched and the flame combustion state detection system is developed based on the characteristics of the flame combustion state.Firstly,the flame characteristics are analyzed by using the video data of boiler combustion in the industrial field.The flame characteristics are extracted by using the image processing method.The flame combustion experiment platform is found in the laboratory,which was used for flame image acquisition and preprocessing.Processes are to extract the characteristics of flame intensity,area,rate of change and flicker frequency.The structure of the BP neural network is the three layer network model for identifying flame status.The method of the flame image collection is the real-time collection and sliding updates for a continuous period of flame.The state of flame in this part of time was judged through the collection of the continuous flame image.The judgment accuracy is 95%by using the BP neural network.Aiming at the shortage of the BP neural network which is failure in the situation of the fault state,the fuzzy recognition method was considered for judging the combustion state of the flame.In this method,the fuzzy membership functions and fuzzy rules were defined through combining with a variety of flame characteristics.Comparing with the method of BP neural network,the situation of the fault state can be judged in this method.And the ANPIS,combining BP neural network and fuzzy recognition,is used to judge the combustion state of the flame.The accuracy is 99.82%by ANPIS,which is better than the state detection method of BP neural network.Aiming at the fact that it is difficult to consider all flame features fully,the flame combustion image is used directly to build the convolutional neural network model to judge the flame combustion state.The recognition rate reaches 100%by this method.And there is no need to pre-extract the characteristics of the flame combustion state.The flame combustion status detection system including the upper computer monitoring interface was developed based on the recognition method of convolutional neural network.The detection system was embedded into the raspberry PI platform,in which the flame combustion process was monitored,the flame intensity,area and flicker frequency characteristics were detected,and the historical data were saved.At the same time,the analog output and digital output of flame brightness characteristics are realized.And the communication protocol is designed to realize the transmission of all flame characteristics detected by serial port.Finally,an intelligent instrument for detecting flame combustion status is formed.