In the smelting process of fused magnesium furnace, semi-molten is the abnormal working condition that burns the furnace wall to red because of the uneven impurities in material. If it is not detected and dealt with timely,the furnace can be burnt through. At present, the detection of semi-molten mainly relies on experienced operators by"observing fire"at the scene of the fused magnesium production. The environment of scene is hostile and the working intensity is high. The human observation may cause safety issues and can lead to overlook and mistakes. This work introduces a detection technology for the semi-molten working conditions of fused magnesium furnace based on the deep convolutional neural network(CNN) model trained using historical images of visible and infrared thermal sensors.A prototype system is developed based on this technology. An industrial camera and an infrared thermal imager are wed to acquire images of the fused magnesium productive process, and the deep learning technology is combined with the working condition of workers’ experience to build the detection and recognition model. With the system, on-line identification of semi-molten condition through real-time image analysis is achieved. The proposed technology is tested in a factory of electric-fused magnesium, which can demonstrate its effectueners.
Control and Decision
fused magnesium furnace
working condition detection
generative adversarial networks