[论文解读] Multifidelity Surrogate Modeling of Depressurized Loss of Forced Cooling in High-temperature Gas Reactors
该论文评估多保真机器学习代理，用于预测 HTGR DLOFC 過载失去强制冷却的时间到发生以及 ONC 后温度，在输入与保真度上比较两保真与三保真策略。
High-fidelity computational fluid dynamics (CFD) simulations are widely used to analyze nuclear reactor transients, but are computationally expensive when exploring large parameter spaces. Multifidelity surrogate models offer an approach to reduce cost by combining information from simulations of varying resolution. In this work, several multifidelity machine learning methods were evaluated for predicting the time to onset of natural circulation (ONC) and the temperature after ONC for a high-temperature gas reactor (HTGR) depressurized loss of forced cooling transient. A CFD model was developed in Ansys Fluent to generate 1000 simulation samples at each fidelity level, with low and medium-fidelity datasets produced by systematically coarsening the high-fidelity mesh. Multiple surrogate approaches were investigated, including multifidelity Gaussian processes and several neural network architectures, and validated on analytical benchmark functions before application to the ONC dataset. The results show that performance depends strongly on the informativeness of the input variables and the relationship between fidelity levels. Models trained using dominant inputs identified through prior sensitivity analysis consistently outperformed models trained on the full input set. The low- and high-fidelity pairing produced stronger performance than configurations involving medium-fidelity data, and two-fidelity configurations generally matched or exceeded three-fidelity counterparts at equivalent computational cost. Among the methods evaluated, multifidelity GP provided the most robust performance across input configurations, achieving excellent metrics for both time to ONC and temperature after ONC, while neural network approaches achieved comparable accuracy with substantially lower training times.
研究动机与目标
- Motivate rapid surrogate modeling to reduce CFD cost in HTGR DLOFC safety analysis.
- Assess how input sensitivity and fidelity pairing affect multifidelity surrogate performance.
- Compare two-fidelity and three-fidelity strategies under equivalent computational budgets.
- Investigate input selection (dominant vs non-dominant) on surrogate accuracy.
- Provide guidance on method selection for HTGR transient prediction.
提出的方法
- Develop HF, MF, and LF CFD datasets for an HTGR DLOFC scenario using Ansys Fluent across three mesh-based fidelity levels (HF ~70k elements, MF ~35k, LF ~17.5k).
- Train and compare multiple multifidelity surrogate architectures: delta NN, flag NN, intermediate NN, GPmimic, two-step NN, three-step NN, and MF-GP, in bi- and tri-fidelity configurations.
- Validate methods on MF2 analytical benchmarks (Forrester, Booth, Park91A, Hartmann, Borehole) before applying to ONC data.
- Perform hyperparameter tuning via grid search on benchmarks and apply to ONC data, with loss weighting and regularization; evaluate on time to ONC and temperature after ONC.
- Partition inputs into All, Dominant, and Non-Dominant sets based on prior sensitivity analysis to study performance under dominant vs non-dominant inputs.]
- research_questions: ["How do two-fidelity versus three-fidelity configurations perform for predicting time to ONC and temperature after ONC under the HTGR DLOFC scenario?","Which fidelity pairings maximize predictive accuracy relative to computational cost for this problem?","Does input parameter relevance (dominant vs non-dominant) influence multifidelity surrogate performance?","Which multifidelity architectures provide robust performance across input configurations in this application?","Can benchmark validation on analytical functions predictably translate to the ONC CFD dataset?"]
- key_findings: ["Multifidelity GP generally offers robust performance across input configurations for both time to ONC and post-ONC temperature.","Neural network surrogates achieve comparable accuracy with substantially lower training times.","Two-fidelity configurations often match or exceed three-fidelity counterparts at equivalent computational cost.","Dominant inputs identified by prior sensitivity analysis consistently improve surrogate performance when used for training.","Low- and high-fidelity pairings yield stronger performance than configurations including only a medium fidelity level.","Models trained with dominant inputs perform better than models trained on the full input set under comparable budgets.]
- table_headers: []
- table_rows: []
更好的研究，从现在开始
从论文设计到论文写作，大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成，并经人工编辑审核。