Surrogate-based prediction of pressure limit violations in drum boilers: A failure-prevention perspective

• Surrogate predicts pressure and drum-level with <1% steady-state error. • Transient validation: max drum-level deviation ±0.27 m with Ts ≈10–12 h. • Real-time (<5 ms) safety-map generation for operational decision support. • Doubling drum volume enlarges safe region by ∼65% in design space. • Framework enables risk-informed design and predictive reliability assessment. Drum-boiler pressure regulation plays a critical role in ensuring operational safety and reliability in industrial energy systems. This study develops a lightweight surrogate framework trained using first-principles data from an enhanced four-state Åström–Bell drum-boiler model. The surrogate achieves <1% steady-state prediction error on pressure and drum level and <3% error in the energy-storage constant. Transient validation further shows a maximum drum-level deviation of only ±0.27 m under startup ramp conditions with stable recovery. The model executes within <5 ms, enabling real-time computation of pressure contours and safety envelopes for decision support. Parametric analysis reveals that doubling drum volume from V to 2V (V = 67 m³) expands the safe region by ∼65% in the heat–steam–water design space. These results demonstrate that the proposed framework effectively bridges high-fidelity thermodynamic modeling with rapid surrogate prediction, supporting risk-informed design, operational guidance, and predictive reliability assessment for pressure-boiler systems.