Heat transfer characteristics of supercritical cryogenic fluids play important role in thermal equipment for liquid air energy storage, liquid hydrogen, etc. In this study, a low-temperature and high-pressure experimental device was established to investigate the heat transfer behavior of supercritical nitrogen. The combined effects of forced convection and drastic changes in physical properties should be considered to investigate the heat transfer behavior, especially the abnormal heat transfer phenomenon in the medium / range where the heat transfer coefficient (HTC) rises with /. A new HTC correlation considering pseudo-boiling number was proposed, and a mean absolute relative deviation (MARD) of 12.7% was achieved. The results showed that the bulk fluid at the inlet had a high heat absorption capacity because it was close to , which inhibited the onset of heat transfer deterioration (HTD) together with the inlet effect. Machine learning has high accuracy in the HTD prediction and avoids the prediction of over-HTD in existing models. A support vector machine model considering the inlet state and heating conditions was trained and tested in this work. Several parameters were used to identify HTD cases from multiple latitude features. The machine learning model has a high prediction accuracy of 96%.

中文翻译：

---

超临界氮气立管传热行为及传热恶化预测实验研究

超临界低温流体的传热特性在液态空气储能、液氢等热工设备中发挥着重要作用。本研究建立了低温高压实验装置来研究超临界氮气的传热行为。研究传热行为时应考虑强制对流和物理性质剧烈变化的综合影响，特别是传热系数（HTC）随/升高的介质/范围内的异常传热现象。提出了考虑伪沸腾数的新 HTC 相关性，平均绝对相对偏差 (MARD) 为 12.7%。结果表明，入口处的散装流体具有较高的吸热能力，因为它接近 ，这与入口效应一起抑制了传热恶化（HTD）的发生。机器学习在HTD预测方面具有较高的准确性，并且避免了现有模型中过度HTD的预测。在这项工作中，对考虑入口状态和加热条件的支持向量机模型进行了训练和测试。使用多个参数从多个纬度特征中识别 HTD 病例。机器学习模型的预测准确率高达96%。