Résumé: Abstract Surrogate models, both polynomial and ANN-based (artificial neural networks), are developed to predict the rolling load in cold rolling of flat metals. An accurate but fast model was developed to serve as high-fidelity model for the training of the machine learning algorithms, allowing for large sampling sizes (up to 1000 samples) with different sampling methods, a number of eight input parameters, and various configurations of surrogate models. The ANN-based models have shown excellent predictive abilities provided that the training sampling is sufficiently large (more than 500 elements). In contrast, polynomial models converge much rapidly to their optimal accuracy (samplings of tens of elements) but their predictive ability is more limited, unless the order of the polynomials is increased. The latin hypercube sampling was more efficient than the random sampling in all cases.

Résumé: During the metal forming process, the avoidance of ductile fracture has been of great interest to the scientific and engineering communities over the past decades. Hence, ductile damage prediction remains a key issue for achieving defect-free products. In this paper, the elastoplastic damage behaviour of DC04 steel has been studied and simulated to predict the fracture during the deep drawing process and reduce the industrial trial cost. In this context, a fully coupled elastoplastic damage model has been developed and implemented in the Abaqus explicit code using the VUMAT subroutine, knowing that the used elastoplastic and the damage parameters were identified by experimental tests. Numerical simulations have been performed to validate this model, followed by comparisons with the experimental results. These comparisons show a good correlation between the experimental and simulation results and good agreement with the empirical observations. Thus, the initiation of damage and its evolution leading to ductile fracture can be predicted using this model.

Résumé: Abstract The present work is a contribution in improving the hydrodynamic method proposed by Li et al. (Steel Res Int 87(9999) 88:2053–2059, 2016; Int J Adv Manuf Technol. Springer-Verlag, London, 2016) used to predict the pressures and the rolling speeds during hot rolling of aluminum strips. The hydrodynamic model gives good prediction. However, it is based on empirical coefficients which must be identified for each rolling case. Therefore, a critical analysis of the Si Li’s method has been first made and then a methodology for improving it has been presented. The improvement consists in coming out of the empiric coefficients and considering the variation of viscosity as a function of pressures. Finite element simulations have been conducted to validate the improved method. Much reliable results have been obtained that are in good agreement with the experimental data. The proposed approach is rapid and much easier to use within the industrial application.

Résumé: In this paper, a calculation technique for solving the problem of regulating inter-stand tension in a tandem cold rolling is proposed. Based on the slices method, the proposed technique develops a computational model for a single stand, and then generalizes it for five stands. The effectiveness of this technique is evaluated using experimental data acquired from tandem rolling mill of IMittal steel complex of El-Hadjar-Algeria. By taking into account the elasticity of the rolls and using Newton’s method; the developed model can be used to calculate, successfully, the tensions correction of the five stands. Compared with the LAM3 software, the obtained results indicated that the proposed technique is effective and can be used to produce better performance of tandem cold rolling.

Résumé: Abstract The present paper introduces a simplified two-dimensional model for the prediction of pressures and velocities in hot strip rolling process through the combination of squeezing and shearing models. First, the squeezing model based on Karman equation is used to obtain a predictive model of stresses and velocities on the roll-strip contact along the longitudinal direction referred as ROL1D. Secondly, the shearing model based on thin film assumption is developed to predict pressures in the longitudinal direction while shearing stresses and velocities along normal and longitudinal directions referred as ROLXZ. Then, the two former models have been combined in order to obtain a squeeze-shear model referred as ROL2D. The latter model permits to assess the rolling process with respect to squeezing and shearing. Results show that where squeezing is most dominant, the ROL2D model is conforming the reality of the rolling process and in good agreement with experimental and literature data.

Résumé: In metal forming, the most widely used method today is that of finite elements. However, it requires a slow runtime. Other techniques are currently used as the application of genetic algorithms by examples [3]. The approach focused in this work is the development of simple models in rolling. These models must be fast running and able to integrate recent advances in metal forming.