基于BP神经网络算法的竖直和倾斜圆管临界热流密度预测

BP Neural Network-based Prediction of Critical Heat Flux in Vertical and Inclined Tubes

  • 摘要: 为探究海洋条件下的临界热流密度(CHF)的变化规律,并揭示海洋条件下CHF变化机理,本文开展轴向均匀加热的竖直及倾斜圆管内CHF实验研究。采用BP神经网络(BPNN)算法构建CHF预测模型,并基于预测结果定义强化因子CHFRI(倾斜CHF/竖直CHF)以量化倾斜对CHF的影响程度。基于出口条件获取了实验数据点共498个,随机划分75%数据集放入神经网络进行训练。研究结果表明,BPNN模型在实验参数范围内对CHF预测有着较高的精度,均方差为3.06%,最大相对误差为10.93%。通过分析训练后的BPNN模型输出的CHF参数变化趋势与先验理论一致:倾斜会对CHF产生恶化影响(CHFRI≤1),且恶化程度受系统压力、质量流速、含汽率等热工参数的耦合作用影响。

     

    Abstract: Offshore nuclear power plants (NPPs) hold significant promise for broad application. However, owing to the influence of marine conditions, the thermo-hydraulic parameters in the reactor exhibit substantial differences from those of land-based NPPs. To investigate the variation characteristics of critical heat flux (CHF) under marine conditions and to elucidate its underlying mechanisms, experimental studies were conducted on CHF in stationary, uniformly heated vertical and inclined tubes. The test section consisted of an Inconel 690 tube with an inner diameter of 8 mm and an outer diameter of 10 mm. The heating length could be flexibly adjusted via two copper electrodes (upper and lower), while the entire test section was mounted on a motion platform to simulate the inclination and rolling motions encountered in marine environments. To characterize the CHF characteristics representative of typical pressurized water reactor (PWR), the experimental matrix was configured following the operational parameters of small modular reactors and Katto’s fluid scaling principles. The experiments covered pressures ranging from 1.6 to 2.7 MPa, mass fluxes of 1 000 to 3 000 kg·m−2·s−1, and inlet temperatures of 20 to 60 °C. The CHF onset was identified as the point where the slope of the temperature-heat flux curve changes significantly. The experimental results indicate that the CHF decreases under inclined conditions compared to vertical, suggesting that inclination induces a deterioration effect. This adverse impact intensifies with increasing inclination angles. At low thermal equilibrium quality, the influence of inclination becomes more pronounced as pressure and mass flux decrease. In contrast, at thermal equilibrium quality, CHF remains nearly unaffected by the inclination angle. A back propagation neural network (BPNN) was employed to predict the CHF. Based on the prediction results, a CHF enhancement factor, CHFRI (defined as the ratio of inclined CHF to vertical CHF), was introduced to quantitatively evaluate the influence of inclination on CHF performance. A total of 498 experimental data obtained under outlet conditions were used to train the BPNN. The prediction results show that the BPNN achieved high accuracy within the experimental range, with a mean square deviation of 3.06% and a maximum relative error of 10.93%. The parameter trends extracted from the trained BPNN model indicate that inclination has a deteriorating effect on CHF, and the extent of deterioration is governed by the coupled interaction of relevant thermal-hydraulic parameters such as system pressure, mass flux, and quality. The above research provides important data support and theoretical basis for the engineering design involving two-phase flow thermal safety under marine conditions.

     

/

返回文章
返回