Predictive Analytics Approach for Steel Billets Quality Control System
The paper deals with the problem of improving the quality of metal products. Nowadays destructive methods of quality control of the steel billets prevail at metallurgical enterprises. This approach to assessing the quality of the steel billets is wasteful, which increases its cost. One of the ways to reduce the cost of production of metal products is to decrease the use of the destructive control methods through automatic certification of metals. The paper proposes to use an algorithm for predicting the mechanical properties of the final product based on the analysis of data obtained during the production of the steel products.
The prediction algorithm is chosen based on the classical deep machine learning models. The target model is the one that shows the highest accuracy. This paper presents the results of applying modern machine learning algorithms for predicting the mechanical properties of steel billets and automatic certification of metal according to predicted values. The results of the study are planned to be implemented at the Metallurgical Complex Mill-5000 of OJSC Vyksa Metallurgical Plant.