Overall temperature levels eg., from EAF tapping to LF exit can be reduced by up to 10K. The whole temperature guidance is not only optimized to save energy and CO2 emissions but to supply the liquid in the best suitable condition for prime-quality solidification to the casting machine. Therefore, the SST Temperature AI optimizes either possible route in even complicated secondary metallurgy processes, i.e. considers treatment and purging stands, ladle furnaces, vacuum degassers.
Lower energy consumption, enhanced quality, and therefore increased yield, directly translate into the reduction of carbon dioxide emissions. The result is permanent energy and CO2 savings.
Our models compute optimal values for continuous and discrete casting as well as melt shop parameters that lead to the best quality, e.g., minimization of slivers, while considering highly complex constraints originating from both business requirements and physical limits of the casting equipment.The parameters are automatically fed back into the casting planning system. This results in an optimized sequence that minimizes quality deviations, and downgrading and reallocation. The yield increase translates directly into reduced Greenhouse gas emissions.
Applying the SST Casting AI achieves a permanent reduction of the rate of sliver defects for automotive exposed grades by up to 50% and more, as well as consequently less downgrading. The models compute optimal values for continuous and discrete casting as well as melt shop parameters that lead to the best quality, e.g., minimization of slivers, while considering highly complex constraints originating from both business requirements and physical limits of the casting equipment.
Correctly classified surface defects from the automated surface inspection system are the mandatory basis for optimization. We have achieved highest classification accuracy using our deep convolutional neural network (CNN) classifiers specifically designed for steel surface images taken at individual steel processing steps. The network topologies of our Surface AI are fine-tuned with plant-specific training and test data and beat any other method in terms of labeling accuracy. Integrating the full material genealogy enables potential defects to be cross-referenced to preceding manufacturing steps to achieve even higher labeling accuracy. Reliable surface defect classification and mapping enables producers to implement data driven casting optimization with the SST Casting AI, leading to less scrap, rejects and reallocation.