Abstract: Aiming at the black-box operation problem in ferrosilicon smelting, as well as the difficulty in improving the qualification rate of high-grade products caused by manual experience reliance and insufficient quantitative analysis in traditional quality control, this paper proposes a full smelting process data governance and precise quality composition traceability solution. Firstly, a full smelting process data acquisition system based on end-edge-cloud collaborative architecture is constructed, which realizes the spatio-temporal alignment and fusion of multi-source heterogeneous data including high-frequency electrical parameters and low-frequency assay data under high-temperature and strongly coupled conditions. Secondly, mathematical traceability models covering raw materials, electrical parameters and quality indicators are established based on material conservation and statistical process control principles, converting empirical smelting into accurate data mapping. Finally, an intelligent system consisting of four core modules, namely data management, quality analysis, traceability query and system management, is designed and developed. Engineering applications verify that the system supports transparent data storage and visual monitoring throughout the full smelting process. The traceability model shortens the quality anomaly tracing cycle from several days to real time, and improves product quality stability and batch consistency. This study provides a practical technical reference for the digital transformation of the silicon-based alloy smelting.