Use of computational vision to estimate the weight of steel slab places on the rolling mill reheating table

Resumo

The fuel consumption of the slab rolling reheating furnace is one of the largest expenses of the hot
rolling mill production process. Decreasing the time between slab leakage and creation is the scenario to be
achieved. It is then desired to ensure conditions for the slabs to be hung increasingly hot by eliminating a “toll” in
this process, which is the weighing of the slab at the oven inlet. In a continuous hot hanging process, weighing
may cause a loss of 150 ° in the temperature of the slabs. The purpose of this paper is to present a new proposal
for the detection of steel slab dimensions in the hanging table using computational vision and convolutional neural
networks. To achieve this goal, the Convolutional Neural Networks regression technique will be used. The
regressive technique is able to identify a numerical value from a set of information, in the case of images. Networks
recognized in this application will be used, such as: VGG16, DenseNet, InceptionV3, etc., in addition to networks
built and trained for this process, with existing datasets and generated in this production process. At the end of the
work, we want to have an intelligent system capable of detecting the width, length and thickness of the slabs in
order to obtain the weight of the material, eliminating the need for other forms of dimensional measurements.