Research On The Detection Method Of Rotary Kiln Working Condition Based On Kiln Image

In the production process of rotary kiln,temperature is one of the key factors affecting product quality.Too high or too low temperature not only causes waste of materials,but also affects the service life of rotary kiln.Therefore,temperature monitoring plays a very important role in the production process of rotary kiln.Temperature change is closely related with the combustion state of flame in the furnaces,through the observation of flame state can determine the current kiln combustion state,currently in the process of rotary kiln production,however,most of them adopt artificial ways observation to watch the fire flame,due to operating personnel technical level is uneven,rotary kiln production process caused by technical and economic index volatile,unstable running condition,energy consumption per unit product,raw material consumption and harmful gas emissions is higher,seriously restricts the improvement of enterprise energy conservation and emissions reduction.For this reason,the paper points out the problems of manual fire detection,and proposes a method to detect the working condition of rotary kiln based on the image in kiln.First for rotary kiln flame image from the scene of the rotary kiln industry,industrial site conditions,but the rotary kiln dust particles into the stove to make innovation shoot video suffered serious pollution,at the same time,the flame area change of volatile kiln is larger,the number of these cases,camera image resolution is poor,flame,video image fuzzy,so the paper first step to the original flame image preprocessing,pretreatment including the original image noise reduction and segmentation,noise reduction method selection: use a variety of noise reduction method in the process of removing images to obtain freedom generated by the noise,and analyze the results,the method to choose: Different image segmentation algorithms such as fuzzy c-means clustering method and KMEANS method were used to compare the flame image segmentation.Through the comparison of experimental results,the pretreatment methods selected are:mask denoising method and KMEANS segmentation methodThe second step of flame image feature extraction and classification recognition: the paper according to the state of the flame will be divided into "owe to burn",is "burning","burnt" three kinds of states,In this paper,thefeature extraction method of these three states is directional gradient histogram,which is used to extract the feature matrix and compare with the feature extraction data of rotation invariant local binary method,after complete the rotary kiln flame image feature extraction,extracted by need to let the computer to learn the characteristics of the data,and classifying the data,the method is: The extreme learning machine algorithm and support vector machine algorithm were used to study the acquired flame image feature data,and another group of unclassified images were used to test the classification results of the two algorithms.Finally,the classified results were fused with the field industrial data,and the final fusion data were used to detect the working conditions of the rotary kiln.The experiment shows that the method has a good performance in the testing of working conditions.