Efficient cellular automata model for prediction of damage of hot forging tools due to thermal fatigue

Paweł Nowak, Łukasz Rauch

AGH – University of Science and Technology, Mickiewicza 30, 30-059 Kraków, Poland.

DOI:

https://doi.org/10.7494/cmms.2014.4.0491

Abstract:

The paper presents design and implementation of the cellular automata model, which predicts damage of forging tools due to fatigue. The transition rules for the model were developed on the basis of known information regarding crack initiation and propagation. The coefficients in the model were determined by the inverse analysis of the thermal fatigue tests performed on the Gleeble 3800 simulator. The model was connected with the finite element (FE) program, which simulates stresses in tools during forging. This multiscale approach is extremely demanding as far as computing times are considered, therefore, an efficient implementation of this model was the main objective of the work. Cellular Automata (CA) model of fatigue fracture was realized for heterogeneous hardware architectures. The source codes were implemented in form of the CA framework and algorithm on the basis of API (Application Programming Interface), delivered by the OpenCL library. The framework which supports CA computations was divided into library and application parts. The former supplies API to handle kernels of the program in the OpenCL technology dedicated to both CPU and GPU devices. Application part uses API offered by the library to supply and control input parameters for calculations. This part creates also CA space, introduces parameters of cells and sends them to calculations on accelerators. Implementation of the model was described in the paper and an example of calculation for the hot forging die was presented. High speed-up was obtained due to application of four GPU cards, while a good scalability was maintained.

Cite as:

Nowak, P., & Rauch, Ł. (2014). Efficient cellular automata model for prediction of damage of hot forging tools due to thermal fatigue. Computer Methods in Materials Science, 14(4), 197 – 205. https://doi.org/10.7494/cmms.2014.4.0491

Article (PDF):