Prédiction de distorsions induites lors du procédé de trempe de brames d’acier de fortes dimensions : modélisation et validation expérimentale

Abstract

The forging industry utilizes quench as the ultimate heat treatment process for steel hardening through phase transformation. This process allows companies to provide the best mechanical properties to their products, such as a well-defined hardness profile across the thickness as well as the microstructure distribution. The huge amount of variability as well as the involved parameters in quenching parts at industrial scale (up to 40 inches in thickness) requires using predictive tools that help control the process fluctuations. The quench modelling of a large forging was carried out in this work using a finite element method, in order to predict separately the evolution of all physical domains involved in calculation during this process (thermal, microstructural and mechanical) up to the end of quenching. The first part used an optimization method that combines a symmetric three-dimensional finite element model with a progressive neural network to predict the heat transfer coefficient. The used FEM model was considering for analysis a thermal model coupled with a metallurgical model that considers the phase transformation to improve the temperature predictions with the consideration of latent heat release. The FEM model was initially integrating different heat transfer coefficients (HTC) from the literature to simulate the effect of HTC variability on the thermal behavior of the forging steel block. The developed artificial neural network model (ANN) uses initially the temperatures calculated by the FE model as Inputs as well as their corresponding HTC (used as boundary conditions) as outputs for the training. The steel block instrumentation was carried out to compare the validity of the simulated thermal response of the block, through the calculated temperature at the locations of the thermocouples follows to the different HTCs relative to the measured temperature. The simulation of the trained ANN was performed by using the measured temperature in order to predict a new HTC, that will be used again as a boundary condition for modeling the evolution of temperature inside the steel block. This progressive method seems improve the temperature predictions by supplying the ANN model by Inputs and Outputs with higher quality after each iteration (simulation of the ANN). The results show that the progressive HTC predictions by ANN could be improved after each iteration and establish a recurrent profile. Finally, the HTC was considered optimized if it indirectly minimizes the percentage of average error between the FEM calculated temperature and the measured one up to 1.5%. In the second part, the HTC developed in the first part was introduced in the FE model, in order to study the phase transformation evolution for a reduced geometry of an industrial block. The block was instrumented by thermocouples to monitor the temperature evolution during quenching. The Time Temperature Transformation diagram (TTT) was first optimized so that its corresponding Continuous Cooling Transformation (CCT) diagram (computed) could represent the real CCT of the investigated steel, by using the critical cooling rate as an optimization constraint. The FE results showed a good agreement on temperature evolution with the measured ones. Extracted samples from the instrumented block were used as a reference. However, dilatometry tests were additionally carried out to physically simulate the experimental real thermal cycle experienced by the instrumented block in order to reproduce samples with similar microstructure. The predictions of phase volume fractions distribution using the FE model showed a good agreement with the experimental ones, calculated from the dilatometric curves Likewise, metallographic analysis using scanning electron microscopy confirmed the overabundance of bainite in the quarter thickness of the block in comparison to the center. In the third part, the main mechanical properties needed for FEM simulations were experimentally determined for each phase and at different temperatures. An experimental methodology was developed to measure the block distortions after heat treatment. The methodology is based on the experimental measurement of the external geometry of the steel block before and after heat treatment using a high-resolution portable 3D scanner. The scanned shapes were analyzed by CATIA V5 software to produce a continuous 3D model. The measurement objective was to assess in 3D, the measured thickness reduction for the block. The FEM results predicted similar distortion trend to the one observed in the experimental measurement. In addition, the maximum thickness reduction magnitude was predicted with acceptable accuracy, which justify using the FE model to monitor the distortion evolution over time and thus estimate the level of residual stress generated after quench. Finally, the developed FE model was able to monitor phase transformation and temperature evolution through the whole block and then predict a microstructure mapping after quench. This was also confirmed by the observed agreements between the FE predicted distortion and experimentally measured as well as hardness values.

Date	7 Jul 2020
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Mohammad Jahazi (Supervisor) & Henri Champliaud (Co-supervisor)