Infrared Thermographic Fault Detection Using Machine Vision With Convolutional Neural Network for Blast Furnace Chute
Xiaoman Cheng, Shusen Cheng
Abstract
The metallurgical industry is the manufacturing industry with the highest carbon emissions in China. Carbon emission in blast furnace accounts for 73.6% of the total long process flow. The working state of chute directly affects the longevity, safety, efficiency and low-carbon of blast furnace. However, at present, there is no monitoring method for chute damage. In this paper, an intelligent chute fault detection method based on numerical simulation, machine vision and deep learning is proposed. Specifically, discrete element method (DEM), gray level co-occurrence matrix (GLCM) and kernel correlation filter (KCF) modules are introduced to screen the best detection frames. Comprehensive experiments show that the proposed algorithm is superior to other traditional classification methods and other deep learning classification network (SVM and LetNet-5), with an accuracy of 96.1%. This method not only contributes to understand the wear mechanism of rotating chute, but also provides a reference for automatic wear monitoring of chute.