Investigation of Multistage Oxidation Behavior of Al4SiC4 Powders with Aid of Back Propagation Artificial Neural Network

In order to investigate the multistage oxidation behavior of aluminum silicon carbide (Al4SiC4), back propagation artificial neural network (BP-ANN) has been trained and employed considering the oxidation temperature, time, and aspect ratio. The results denote that the BP-ANN model can accurately and efficiently simulate the oxidation behavior of Al4SiC4 powders with different reaction laws. In addition, the extrapolation ability suggests that the BP-ANN model can maintain a high accuracy with the coefficient of determination >= 0.801 to expand the experimental data to 1.2 times the original range. By incorporating a real physical picture model developed by the research group, the experimental data can be further expanded to 2.2 times. This study can provide a new path to recognize the oxidation behavior of materials with multistage oxidation.

Automated summary

By training and employing BP-ANN, the multistage oxidation behavior of aluminum silicon carbide (Al4SiC4) was investigated, considering the oxidation temperature, time, and aspect ratio. The results showed that the BP-ANN model can accurately and efficiently simulate the oxidation behavior of Al4SiC4 powders with different reaction laws. In addition, the extrapolation ability of the BP-ANN model allows for maintaining high accuracy and expanding the experimental data to 1.2 times the original range with a coefficient of determination >= 0.801. By incorporating a real physical picture model developed by the research group, the experimental data can be further expanded to 2.2 times. This study provides a new path for understanding the oxidation behavior of materials with multistage oxidation.