Journal of Materials Processing Technology 177 (2006) 315–318 On the use of artificial intelligence tools for fracture forecast in cold forming operations Rosa Di Lorenzo ∗ , Giuseppe Ingarao, Fabrizio Micari Dipartimento di Tecnologia Meccanica, Produzione e Ingegneria Gestionale, Universit` a di Palermo, Viale delle Scienze, 90128 Palermo, Italy Abstract The design of cold forming processes requires the availability of a procedure able to deal with the prevention of ductile fracture. In fact, the ability to predict fracture represents a powerful tool to improve the production quality in mechanical industry. In this paper, artificial intelligence (AI) techniques are applied to ductile fracture prediction in cold forming operations. The main advantage of the application of AI tools and in particular, of artificial neural networks (ANN), is the possibility to obtain a predictive tool with a wide applicability. The prediction results obtained in this paper fully demonstrate the usefulness of the proposed approach. © 2006 Elsevier B.V. All rights reserved. Keywords: Ductile fracture; Artificial neural networks; Bulk forming; Artificial intelligence; Forming process design 1. Introduction The design of cold forming processes requires the availabil- ity of a procedure able to deal with ductile fracture prevention. In fact, the ability to predict fracture represents a powerful tool to improve the production quality in mechanical industry since it allows to properly design forming processes in order to avoid the manufacture of unsafe components. The technical literature presents many different approaches for ductile frac- ture prediction [1]. The most important approaches are focused on the assessment of fracture criteria, based on the considera- tion that fracture occurrence is dependent on stress and strain conditions during a given process. Generally, such criteria are expressed by determining a critical value of a damage func- tion, which depends on stress and strain paths: ductile fracture is assumed to occur when such critical value is reached dur- ing the analyzed process. Some criteria are based on critical values of tensile strain energy per unit of volume or on the hydrostatic stress [2], while others take into account a void growth model. The latter criteria assume that cracks occur after void nucleation, growth and coalescence and generally, they are based on macro variables related to the fracture mechanism [3]. A different vision due to the authors, is a proposed dam- age mechanics formulation for fracture investigation; such ∗ Corresponding author. E-mail address: rosanna@dtpm.unipa.it (R. Di Lorenzo). approaches take into account damage insurgence and evolu- tion by including micro structural variables into the constitutive equations of the materials [4] and they are very effective from a theoretical point of view. Nevertheless, such models are very complex and their proper calibration is quite difficult. In recent years some alternative techniques were proposed to design forming processes with the aim to prevent fracture; such techniques are based on artificial intelligence tools (namely artificial neural networks) aimed to obtain damage probability in forming operations [5,6]. This approach seems very promis- ing since it overcomes the main drawbacks of ductile frac- ture criteria which are often reliable only for certain process conditions. An artificial neural network was built which can predict the occurrence of fracture, receiving as input data on stress and strain histories. The result of this approach is the availability of a pre- diction technique which is “process kind insensitive”, i.e. it can be utilized to predict fracture for a wide range of bulk forming operations. 2. Problem formulation Neural networks proved their ability to deal with experimen- tal or numerical data and to provide predictive capability, in particular in classification problems. The developed neural net- work was a feedforward backpropagation network, which was trained by a set of stress and strain data and provided as an output the prediction of fracture occurrence. 0924-0136/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2006.04.032