Predicting the Hardness of Pipe Steels Using Machine Learning Methods

Abstract

One of the main stages in the pipes production from low-carbon steel grades is hardening. In the course of quenching with standard modes, the metal structure changes and, as a consequence, the mechanical properties change. Comparing the indicators values of these properties, such as hardness, strength, ductility, etc., makes it possible to judge the choice correctness of hardening modes. Therefore, it is important to select in advance the optimal heat treatment modes in order to obtain metal with desired mechanical properties. There are many theories and approximations that allow predicting the values of mechanical properties under given initial conditions, such as chemical composition, grain size, heating temperature, cooling rate, etc., however, in most cases they are either not accurate or tied to a specific production unit and, as a result, are not suitable for use in other (different) conditions. The purpose of this work is to build a model for predicting the hardness of pipe steels after quenching using modern machine learning methods. As an initial sample, an aggregated array of experimental data is used, collected mainly from supercooled austenite decomposition diagrams, as well as tables and other data obtained from various sources. This paper describes in detail the stage of data preprocessing, model construction and validation, as well as the comparison of the simulation result with the results of similar work.