BOF steelmaking endpoint real-time recognition based on flame multi-scale color difference histogram features weighted fusion method

Abstract

BOF (Basic Oxygen Furnace, BOF) steelmaking endpoint prediction is one of the most important steps in the blowing process. The flame recognition has proven as a useful method for endpoint prediction. But the previous methods are not suitable to extract the flame characteristic because of flame randomness and multi-scales problems. A multi-scale color difference histogram features weighted fusion method is proposed to describe the flame changes during the blowing process. The segmented images are converted into the L*a*b* space; the color difference histogram is built based on the defined calculation model and the features are calculated to describe the histogram; in order to fuse multi scales characteristics as a whole feature vector and contains each scale features in a reasonable level, a multi-scale features weighted function is defined; finally, the GRNN (General Regression Neural Network, GRNN) recognition model is built to realize the blowing stage prediction according to the flame features. The experimental and comparisons results show that the proposed method has a better recognition rate and high calculation speed, and have a bright practical value in the BOF endpoint control.