| The air preheater is one of the main components of a power station boiler.The main function is to use the waste flue gas emitted by the boiler to preheat the air that is about to enter the boiler.Because there are many gaps in the heat storage element of the air preheater,when it is operated under conditions of low load or insufficient fuel combustion,"secondary combustion"is prone to occur,and if it cannot be processed in time,huge losses will occur.The main content of this topic is to design a set of high-performance hotspot detection device for hot spot detection problems in air preheaters.Based on the structure of the top-view scanning system,the selection of the hardware platform,the preparation of the image stitching algorithm,and the decision-making of the SVM are completed.First of all,the use of the top-view scanning system enables temperature detection and image acquisition in the upper high-temperature zone of the air preheater.This study provides a new idea for hot spot detection using image features.In order to meet the real-time requirements of the preheater hot spot detection,considering the image processing speed of the selected hardware platform and the convenience of program debugging,the DragonBoard 410c was selected as the development platform of the subject.Secondly,the hotspot scanning system is composed of a top view scanning device,a camera,and a Dragboard 410c.When a hot spot image is captured in the upper high temperature region of the air preheater,the cross-sectional area of the preheater is large and the camera captures an independent image,which needs to be determined.The shape and size of the hotspot area,this subject adopts SIFT algorithm and SURF algorithm to splicing the preheater image collected,and compare their running duration,and finally found that the SURF algorithm can meet the requirements of the subject.Finally,due to the low probability of fire hot spots in the air preheater at the industrial site,a simulation experiment environment was set up in the laboratory.The electric heating furnace was used to heat the steel wire mesh,and the number of layers covered by the steel wire mesh simulated the depth of the fire.A thin steel plate was placed on the electric heat rate to block the occurrence of fire hot spots by heating the electric heating wire.The infrared temperature meter and the visible light camera were respectively used for image acquisition and temperature acquisition,and the bright spot distribution area was extracted from the collected images.The luminance average value is extracted,and the temperature value measured by the infrared thermometer is used as the input of the SVM,The model is trained to classify the presence or absence of hotspots.The obtained model has a classification accuracy of 1.0.In the case of hot spots,the steel wire mesh is used.The number of layers then determines the depth of the fire. |
PDF Full Text Request