Research On Visual Inspection Method And Optimized Control Technology For Rotary Kiln

An industrial rotary kiln is a large-scale thermal equipment for processing bulk or slurry materials. It is widely employed in many fields such as metallurgical and cements industry and plays an important role in the national economy. Due to the lack of viable online contact sensors under high combustion temperature, contactless image-based visual inspection technology has become an effective method for estimating some key quality parameters of the working condition inside the kiln. Since existing techniques mainly focus on extracting the visual features from the flame and utilizing these features for demarcating the kiln temperature and identifying the working condition, it still needs further investigation on the technology of visual inspection for clinker and the effective method to deal with the complex and dynamic calcinations process. This thesis is supported by National Natural Science Foundation of China, and the clinker image of industrial rotary kiln is choosen as study object. The main purpose of this research is to ensure the normal sinter for clinker and stable control for rotary kiln. Thus, the technology of visual inspection and the optimal control for rotary kiln are deeply studied and the major achievments of this thesis are described as follows:Firstly, a texture analysis of the clinker image based on GLCM is proposed for predicting the sinter mode of clinker. According to the variety of displacement operator and description features of GLCM, the Fisher coefficient is used to extract the best displacement operator of the GLCM and to reduce the texture measurement sets. Then C4.5decision tree is applied to classify the sampled clinker images into three categories:over-sintering, under-sintering and normal-sintering, whose detailed classification rule was listed.Secondly, the visual recognition of upper repose angle of clinker is developed to quantitatively analyze the sinter degree of clinker. The repose angle of single image sample is calculated by edge detection and linear least square regression. Corresponding to the real-time video captured under industrial rotary kiln, the sequence analysis is needed for the curve of repose angle in order to extract the upper repose angle of each motion period. Noises are filtered and smoothed from the original curve, and then the abnormal and missed data can be removed. Finally, all positive peaks of curve are detected by the method of Guassion-based muti-peak curve fitting. As a result, the upper repose angles are achievmented with these peaks.Thirdly, the visual-based measurement of filling percentage of clinker is produced. The value of filling percentage of clinker can be used to detect the feeding rate of slurry and the mutation of sinter zone temperature. An innovative FCM method is developed for segmenting the clinker from the whole image accurately and quickly. This segmentation mothod applies local spatial neighborhood information with an adaptive window to generate the linear-weighted sum image, in which the noise is eliminated while the edge is preserved. It is worth noting that the clustering segmentation is based on the number of gray-level rather than pixels of the whole image to reduce the computation complexity. The total clinker region is then recognized by region growing technology based on gray-level connectivity. Consequently, the filling percentage of clinker is calculated by the rate of the pixels number between the clinker region and the total image. The relationship between the curve of filling percentage and the feeding rate of slurry and zone temperature mutation is analyzed by case study and simulation.Fourthly, the cascade control strategy is proposed to deal with the mismatch change delay between the temperature of the rotary kiln and the sintering status of the clinker. The cascade control consists of two controllers, which are the HISC for the clinker sinter and the DRSPC for the temperature control of rotary kiln. In the automatic control of clinker sinter, the sinter degree of clinker is determined by the fuzzy fusion result of the texture feature and the upper repose angle. HISC switches to different control modes following the variation of the sinter degree error, such as the amplitude, the direction and the trend of the variation. Then, the temperature of the rotary kiln, which is the output response of the HISC, is transmitted to DRSPC as input. The adjustments of the rotation speed of coal burner pipe are used to respond to temperature changes in the kiln. The DRSPC can rapidly provide corresponding optimal and robust outputs to serve the rotation speed of coal burner pipe. In the DRSPC, the measurement errors of the kiln temperature can be statistical eliminated. Two response surface models, one for the mean value and the other for the standard deviation of the kiln temperature, can achieve optimal and robust outputs because of the appropriate constraint given by each other. The objective function of controller is designed for minimizing the variation of the rotation speed of the coal burner pipe, which also considers the requirement of accurate and stable control for the kiln temperature. The weight coefficients are used in the objective function to keep the trade-off between system cost and efficiency. Case study and experiment results validate the efficiency of the controller, which is designed for dynamic and complex induatrial work conditions of roatry kiln.This thesis is based on the basic theory of image processing, and focuses on the key technology of visual inspection for rotary kiln, provides vision-based optimal control methods, demonstrates important scientific significance and broad application prospect of improving automatic control technology of rotary kiln and expanding the vision application area.