Numerical simulations are frequently utilized to investigate and optimize the complex and hardly in situ examinable Physical Vapor Transport (PVT) method for SiC single crystal growth. Since various process and quality-related aspects, including growth rate and defect formation, are strongly influenced by the thermal field, accurately incorporating temperature-influencing factors is essential for developing a reliable simulation model. Particularly, the physical material properties of the furnace components are critical, yet they are often poorly characterized or even unknown. Furthermore, these properties can be different for each furnace run due to production-related variations, degradation at high process temperatures and exposure to SiC gas species. To address this issue, the present study introduces a framework for efficient investigation and calibration of the material properties of the PVT simulation by leveraging machine learning algorithms to create a surrogate model, able to substitute the computationally expensive simulation. The applied framework includes active learning, sensitivity analysis, material parameter calibration, and uncertainty analysis.