Prediction and validation of through coil final mechanical properties of high strength hot rolled coil using artificial neural networks

Abstract

In the manufacture of rolled steel from a hot strip mill, the final mechanical properties, such as yield strength, ultimate tensile strength and elongation to fracture, are important requirements specified by the customer. With the advent of ever more powerful computational hardware and software, non-linear modelling techniques such as artificial neural networks can result in the creation of acceptably accurate predictive models. In the present paper, the application of known metallurgical knowledge on the selection of model input data, along with the excellent non-linear mapping capabilities of neural networks, combines to create a solution for the prediction of final mechanical properties throughout the length of rolled coils from the Corus Port Talbot hot strip mill. Predictive models are trained on two data sets (one from 2003 and the other from 2005) of a structural steel grade S275 from the high strength steel family. Actual through coil tensile test results from a validation coil (rolled in 2005) were used to certify the models. The model results clearly show that when an optimised neural network architecture is trained on coil samples taken from the specific grade of steel and within a reasonable timeframe relative to the unseen validation data, an acceptably accurate model is created for the prediction of final mechanical properties.