Flame Detection Technology Based On Image Processing

As the modem thermal energy generating set develops toward to large capacity and high parameter, the equipment is getting more and more complex, and it requires higher controlling quality of production process. The detection technique based on image processing for furnace flame is very meaningful to monitor and diagnosis the combustion status of power plant pulverized-coal boiler for security, economic and environment protection. Digital images are the most direct reflection of the boiler furnace's burning states and playing a more and more important role in combustion monitoring. Using the CCD camera and image processing technology, the visual combustion monitoring is studied to obtain more direct and more adequate information of flame detection.The main purpose of this paper is based on digital image processing and combustion flame detection method of diagnosis, to achieve the flame visualization and intelligent diagnosis of combustion state, the main results and conclusions of this paper are listed as follows: (1) Flame image pretreatment and its features extraction are achieved using image processing algorithm;(2) Analyzing the boiler combustion state with the characteristic parameters of flame image;(3) Recognition model of boiler combustion state based on the flame image is build by using intelligence theory.Firstly, this paper introduces composition of combustion state monitoring system of pulverized-coal boiler and the principle of image acquisition, researches image processing technology for image noise removal and local enhancement and image segmentation, and compares the results of image processing methods by experiment test. Because of the importance of accurate edge detection for flame image, a segmentation method based on C-V active contour model is proposed, and it is verified with simulation. The results show that C-V model initialization is simple, the requirement of initial curve position and shape is less, C-V model is robust to noise, and finally a continuous edge of the goal can be obtained. Compared with the classical edge detection operator, C-V model is more suitable for power plant boiler furnace flame edge extraction, and obtain the satisfactory image segmentation, which will lay a good foundation for subsequent feature extraction of flame image.Secondly, based on analyzing pulverized-coal boiler combustion process, the boiler's combustion state and usual characteristic parameters reflected the combustion state and calculation methods are defined. Extracted characteristic parameters of flame image from the multiple flame images, the simulation results show that the relationship between the flame image characteristic parameters and combustion state. In addition, these parameters reflect the state of flame images, on the one hand, they can be used to judge the overall combustion trend, and generate the intelligent diagnosis model of combustion state; On the other hand, it also can be used to judge the combustion states in real time by diagnostic model combined with operating parameters, so it has practicality value in project.Finally, in-depth analysis and study is performed for support vector machines theory which is used for furnace combustion state diagnosis, combined with four data sets, the importance of support vector machine parameters selection is verified. To solve these problems that support vector machine parameters is difficult to determine and the use of cross-validation, this paper proposed genetic algorithm to optimize parameters of support vector machines to get the best support vector machines model. Compared with LIBSVM toolbox, optimized support vector machine is used for combustion state diagnosis, the experimental results show the feasibility and superiority of this method.