为提高热负荷预测的精度,提升基于负荷预测的供热系统调控效果,提出一种基于特征融合的供热系统预测调控方法。首先,采用偏自相关函数(partial autocorrelation function,PACF)、Pearson相关系数和最大信息系数(maximum information coe...为提高热负荷预测的精度,提升基于负荷预测的供热系统调控效果,提出一种基于特征融合的供热系统预测调控方法。首先,采用偏自相关函数(partial autocorrelation function,PACF)、Pearson相关系数和最大信息系数(maximum information coefficient,MIC)相结合的特征选择方法来确定预测模型的基本特征;然后,使用线性回归融合、指数融合和主成分分析融合对基本特征进行融合,应用递归MLR预测确定最佳融合方法,进一步对比在最佳融合策略下递归MLR、PSO-SVR、CNN和XGBoost中效果最优的预测方法;最后,将辨识出的融合方法和预测模型方法用于实际热力站调控。结果显示,基于线性回归融合的XGboost预测方法效果最好,可以提升训练精度并减少计算时间,同时可以有效指导调控,节热率达到4%以上。展开更多
目的针对飞机货舱配载方案评估中多属性决策的复杂性,以及现有评估方法在权重确定上过度依赖专家经验导致的主观偏差,或单纯依赖客观数据忽略决策者偏好的局限性,提出一种融合主观先验与客观数据驱动的混合赋权评估模型,以提供更为合理...目的针对飞机货舱配载方案评估中多属性决策的复杂性,以及现有评估方法在权重确定上过度依赖专家经验导致的主观偏差,或单纯依赖客观数据忽略决策者偏好的局限性,提出一种融合主观先验与客观数据驱动的混合赋权评估模型,以提供更为合理、可靠的配载方案择优决策支持。方法首先,引入大型语言模型(Large language model,LLM),构建“虚拟专家委员会”,通过精心设计的提示词工程,获取多维度、多情境下的主观权重。其次,针对传统熵权法对数据分布敏感、难以有效区分指标优劣等问题,提出一种改进的数据预处理熵权法(Improved data preprocessing entropy weighting method,IDPEW),该方法结合指标值的辨识度和信息熵的均衡性来确定客观权重。最后,将LLM生成的主观权重与IDPEW计算的客观权重进行加权组合,构建综合评价函数,对飞机货舱配载方案进行全面评估和排序。结果实验结果表明,LLM模拟专家意见时最关注“装载率”(主观权重0.2250),而IDPEW方法从数据中识别出“横向不平衡度”最具区分力(客观权重0.2481)。混合赋权模型(α=0.5)有效平衡了主客观偏好,在24个方案中精准识别出综合性能最优的方案,验证了模型在复杂情境下的稳定性。结论创新性地利用LLM低成本构建“虚拟专家”获取先验知识,并通过耦合指标辨识度与均衡性的IDPEW方法,提升了客观赋权精度。该模型克服了单一赋权的局限,为飞机货舱配载方案的科学评估提供了一种兼具可解释性和实用性的新范式。展开更多
Studies of wave-current interactions are vital for the safe design of structures.Regular waves in the presence of uniform,linear shear,and quadratic shear currents are explored by the High-Level Green-Naghdi model in ...Studies of wave-current interactions are vital for the safe design of structures.Regular waves in the presence of uniform,linear shear,and quadratic shear currents are explored by the High-Level Green-Naghdi model in this paper.The five-point central difference method is used for spatial discretization,and the fourth-order Adams predictor-corrector scheme is employed for marching in time.The domain-decomposition method is applied for the wave-current generation and absorption.The effects of currents on the wave profile and velocity field are examined under two conditions:the same velocity of currents at the still-water level and the constant flow volume of currents.Wave profiles and velocity fields demonstrate substantial differences in three types of currents owing to the diverse vertical distribution of current velocity and vorticity.Then,loads on small-scale vertical cylinders subjected to regular waves and three types of background currents with the same flow volume are investigated.The maximum load intensity and load fluctuation amplitude in uniform,linear shear,and quadratic shear currents increase sequentially.The stretched superposition method overestimates the maximum load intensity and load fluctuation amplitude in opposing currents and underestimates these values in following currents.The stretched superposition method obtains a poor approximation for strong nonlinear waves,particularly in the case of the opposing quadratic shear current.展开更多
Based on the split hopkinson pressure bar(SHPB)tests results,the cubic specimens have been numerically modeled in this paper to investigate the impact of key factors,such as the rise time,duration,and incident pulse s...Based on the split hopkinson pressure bar(SHPB)tests results,the cubic specimens have been numerically modeled in this paper to investigate the impact of key factors,such as the rise time,duration,and incident pulse shape,on achieving stress uniformity.After analysis,the paper provides actionable methods aimed at optimizing the conditions for stress uniformity within the cubic specimen.Finally,the lateral inertia effect of cubic specimen has been scrutinized to address the existing gap in this academic area.展开更多
文摘为提高热负荷预测的精度,提升基于负荷预测的供热系统调控效果,提出一种基于特征融合的供热系统预测调控方法。首先,采用偏自相关函数(partial autocorrelation function,PACF)、Pearson相关系数和最大信息系数(maximum information coefficient,MIC)相结合的特征选择方法来确定预测模型的基本特征;然后,使用线性回归融合、指数融合和主成分分析融合对基本特征进行融合,应用递归MLR预测确定最佳融合方法,进一步对比在最佳融合策略下递归MLR、PSO-SVR、CNN和XGBoost中效果最优的预测方法;最后,将辨识出的融合方法和预测模型方法用于实际热力站调控。结果显示,基于线性回归融合的XGboost预测方法效果最好,可以提升训练精度并减少计算时间,同时可以有效指导调控,节热率达到4%以上。
文摘目的针对飞机货舱配载方案评估中多属性决策的复杂性,以及现有评估方法在权重确定上过度依赖专家经验导致的主观偏差,或单纯依赖客观数据忽略决策者偏好的局限性,提出一种融合主观先验与客观数据驱动的混合赋权评估模型,以提供更为合理、可靠的配载方案择优决策支持。方法首先,引入大型语言模型(Large language model,LLM),构建“虚拟专家委员会”,通过精心设计的提示词工程,获取多维度、多情境下的主观权重。其次,针对传统熵权法对数据分布敏感、难以有效区分指标优劣等问题,提出一种改进的数据预处理熵权法(Improved data preprocessing entropy weighting method,IDPEW),该方法结合指标值的辨识度和信息熵的均衡性来确定客观权重。最后,将LLM生成的主观权重与IDPEW计算的客观权重进行加权组合,构建综合评价函数,对飞机货舱配载方案进行全面评估和排序。结果实验结果表明,LLM模拟专家意见时最关注“装载率”(主观权重0.2250),而IDPEW方法从数据中识别出“横向不平衡度”最具区分力(客观权重0.2481)。混合赋权模型(α=0.5)有效平衡了主客观偏好,在24个方案中精准识别出综合性能最优的方案,验证了模型在复杂情境下的稳定性。结论创新性地利用LLM低成本构建“虚拟专家”获取先验知识,并通过耦合指标辨识度与均衡性的IDPEW方法,提升了客观赋权精度。该模型克服了单一赋权的局限,为飞机货舱配载方案的科学评估提供了一种兼具可解释性和实用性的新范式。
基金Supported by the Development and Application Project of Ship CAE Software.
文摘Studies of wave-current interactions are vital for the safe design of structures.Regular waves in the presence of uniform,linear shear,and quadratic shear currents are explored by the High-Level Green-Naghdi model in this paper.The five-point central difference method is used for spatial discretization,and the fourth-order Adams predictor-corrector scheme is employed for marching in time.The domain-decomposition method is applied for the wave-current generation and absorption.The effects of currents on the wave profile and velocity field are examined under two conditions:the same velocity of currents at the still-water level and the constant flow volume of currents.Wave profiles and velocity fields demonstrate substantial differences in three types of currents owing to the diverse vertical distribution of current velocity and vorticity.Then,loads on small-scale vertical cylinders subjected to regular waves and three types of background currents with the same flow volume are investigated.The maximum load intensity and load fluctuation amplitude in uniform,linear shear,and quadratic shear currents increase sequentially.The stretched superposition method overestimates the maximum load intensity and load fluctuation amplitude in opposing currents and underestimates these values in following currents.The stretched superposition method obtains a poor approximation for strong nonlinear waves,particularly in the case of the opposing quadratic shear current.
基金Funded by the National Natural Science Foundation of China(Nos.52278518 and 51938011)the Natural Science Foundation of the Jiangsu Higher Education Institutions of China(No.24KJB560021)。
文摘Based on the split hopkinson pressure bar(SHPB)tests results,the cubic specimens have been numerically modeled in this paper to investigate the impact of key factors,such as the rise time,duration,and incident pulse shape,on achieving stress uniformity.After analysis,the paper provides actionable methods aimed at optimizing the conditions for stress uniformity within the cubic specimen.Finally,the lateral inertia effect of cubic specimen has been scrutinized to address the existing gap in this academic area.
