Investigation of process influences on the amount of single-fiber siliconization in C/C–SiC samples by machine-learning methods

Abstract

Ceramic Matrix Composites are an interesting option for high-temperature combustive environments as often encountered in aerospace applications. In the past a lot of research was conducted in order to find the right process parameters for optimal performance of these materials. The mechanical properties of CMCs are vastly dependent on their microstructure. Therefore, a lot of past research focused on finding correlations between process parameters and microstructure of CMCs, most of which was based on empirical trial and error methods.In this paper we use several data-driven, probabilistic machine-learning models to quantify the microstructural composition of C/C–SiC based on the process parameters and choice of raw materials. As a ground truth 123 samples of C/C–SiC with varying process parameters and microstructures were used. The predictive capabilities of the models were demonstrated by the use of the R² metric. By this analysis density in siliconized state as well as open porosity and mass change during siliconization proved to be the parameters with the highest impact on microstructural formation. If siliconization was taken out of the equation the porosity in CFRP state and fiber type were found to be the most influential factors.

Keywords

• Ceramic matrix composites
• Machine-learning
• Artificial intelligence
• Process optimization