2017 Hui Xin Tian A New AdaBoost.IR Soft Sensor Method for Robust Operation Optimization of Ladle Furnace Refining isijinternational.isijint 2016 371

LF (Ladle Furnace) refining plays a critical role in the secondary metallurgic process, where traditionally, operations depend heavily on worker experience, affecting stability, product quality, and energy efficiency. This study proposes a novel robust operation optimization method for molten steel temperature utilizing the AdaBoost.IR soft sensor in the LF refining process. The methodology begins with establishing an intelligent model based on a BP (Back Propagation) neural network, which serves as a sub-model to analyze energy changes throughout the refining process. Subsequently, an AdaBoost.IR model is crafted to accommodate the characteristics of industrial data, facilitating industrial soft sensor modeling. This ensemble soft sensor model allows for real-time, online measurement of molten steel temperature with enhanced accuracy. The robust optimization model of operation is analyzed using the AdaBoost.IR soft sensor, with the HPOS-GA algorithm employed to derive the optimal operation solution. Implementation of this robust optimization for temperature in a 300 t LF at Baosteel Company demonstrates the soft sensor’s superior predictive accuracy for temperature, leading to greater stability in the end temperature following robust optimization.