The purpose of this research programme is to develop quantitative models for the prediction of mechanical properties (fracture toughness) using experimental data collected from the literature, together with a powerful computational technique known as neural network. Creating a truly general model requires a combination of available data and metallurgical knowledge. This model is proposed for martensitic and ordinary bainitic steels in addition to the more recent class of non-structural super-bainitic steels. Super-bainitic steels are
attractive for many applications such as armour. The model of fracture toughness, based on chemical composition, heat treatment an
d mechanical properties is proposed. The predictions of fracture toughness are generally acceptable but the uncertainties
are high and more input data need to be collected for super-bainitic steels when available in the future to improve the predictions of this model.