| Popis: |
In this paper, the cavitation process on a grain boundary in polycrystalline Alloy 247 is simulated and modelled. The cavitation process includes mechanisms such as nucleation, growth and coalescence of creep pores and was used for characterization of polycrystalline Alloy 247 in terms of creep damage and especially creep pore sizes. Two main parameters were defined that are either responsible for pore nucleation or pore growth and that are calculated from simulation variables and material coefficients. Therefore, a Design of Experiment (DOE) was carried out to identify the effect of each material coefficient and simulation variable on the cavitation process as well as the ranges of the main parameters for pore nucleation and pore growth. As a result of the investigation, a series of Monte Carlo simulations was run with designated combinations of main parameters within the identified range. The simulation results were used as training and test data to create a model by machine learning methods. Different machine learning methods such as neural network, random forest tree and k-nearest neighbor were applied and compared to determine the best fitted model. Based on metallographic images, a first calibration of the model’s parameters has also been carried out. The resulting machine learning model allows the prediction of creep pore sizes in the grain boundary of Alloy 247 for any given temperature and stress. |
