Improving Blast Furnace Raceway Blockage Detection. Part 3: Visual Detection Based on Tuyere Camera Images

Abstract

The present paper is part 3 of a paper series discussing raceway blockages and various approaches for automated detection based on blast furnace (BF) plant data. While part 1¹⁾ gave an overview on the different appearances of raceway blockages and part 2²⁾ focussed on signal processing of hot blast data, this third part discusses various approaches for image processing of tuyere camera data. The visual impression of raceway blockages strongly varies between different events. This makes automated detection based on digital image processing of tuyere cameras a difficult task. On one hand the image processing algorithm should be robust and easy to tune for different tuyeres or different blast furnaces, on the other hand it should be fast enough, so that all tuyeres of a blast furnace can be processed on-line with a sufficiently high image frame rate. While algorithms optimized for motion detection fail due to the lack of a homogeneous background, adaptive thresholding of the grey-level histograms delivers useful results. Due to the nature of chaotic motion of coke particles inside the raceway also line based processing methods can extract the information from tuyere images in a sufficient manner and are very fast with regards to online implementation in a process control system. However, image processing of tuyere camera data has some disadvantages compared to the signal processing of hot blast data as discussed in part 2 of this paper.