A State Monitoring Method for Czochralski Silicon Single Crystal Growth Process Based on HC-NPE-CVA

To address the critical challenge of real-time anomaly detection in Czochralski silicon single crystal (CZ-SSC) growth processes, this study proposes a novel process monitoring framework integrating Hybrid Cosine-Neighborhood Preserving Embedding with Canonical Variate Analysis (HC-NPE-CVA). The HC-NPE algorithm is first employed to extract essential features from high-dimensional process data through nonlinear dimensionality reduction while preserving local geometric structures. Subsequently, the CVA method is utilized to establish a state-space monitoring model that extracts canonical variables representing dominant process dynamics, with residual components capturing abnormal process variations. The developed model enables effective real-time monitoring of the CZ-SSC growth process. The experimental results demonstrate that the proposed method exhibits high monitoring accuracy and effectively identifies abnormal data during the silicon single crystal growth process.