期刊文献+
共找到6篇文章
< 1 >
每页显示 20 50 100
守恒型差分格式及保真模型的构造问题研究
1
作者 李磊 叶英 《海洋预报》 北大核心 2003年第3期68-75,共8页
综合评述了守恒型差分格式及保真模型的研究进展,评价了守恒型差分格式及保真模型在地球流体动力学数值模拟中的作用。
关键词 守恒型差分格式 保真模型 地球流体动力学 数值模拟
下载PDF
自适应多保真数据融合的神经网络模型
2
作者 陈柏宁 谢芳芳 孟旭辉 《气体物理》 2024年第4期1-8,共8页
数据驱动的深度学习建模在力学、材料等不同学科中得到了较多应用。深度学习建模的精度依赖大量高保真数据。在实际应用中,高保真数据往往是少量且昂贵的,而低保真数据却是成本低廉且数量较多的。当高保真数据量过少时,深度学习建模精... 数据驱动的深度学习建模在力学、材料等不同学科中得到了较多应用。深度学习建模的精度依赖大量高保真数据。在实际应用中,高保真数据往往是少量且昂贵的,而低保真数据却是成本低廉且数量较多的。当高保真数据量过少时,深度学习建模精度较低。近期发展的多保真深度神经网络,通过融合不同保真度的数据,在高保真数据较少时,依然保持了较高的建模精度。然而,已有的多保真深度神经网络模型的精度较为依赖针对模型参数的正则化调节。当添加的正则化过强时,网络对非线性关联式的拟合能力不足;当添加的正则化强度不够时,在学习多保真数据间的线性关联关系时又会出现过拟合现象。两者都会严重影响模型的预测精度。在缺乏高保真验证数据集时,较难得到最优的正则化系数。为此,通过改进已有多保真网络模型的损失函数,引入一个与线性关联式相关的参数,提出了自适应多保真数据融合的神经网络模型。该模型能根据给定数据自适应地拟合不同保真度数据间的线性或非线性关系,对正则化依赖较小,从而提高了建模的鲁棒性。在多个标准测试案例及实际应用的翼型气动参数的预测中,该模型均能表现出较高的精度和稳定性。 展开更多
关键词 深度学习 保真深度神经网络 保真建模 自适应多保真数据融合的神经网络模型 气动分布
下载PDF
基于加权最小二乘法的变保真代理模型 被引量:2
3
作者 韩畅阳 王硕 +1 位作者 刘富文 宋学官 《组合机床与自动化加工技术》 北大核心 2023年第12期21-24,29,共5页
构建代理模型所用样本点可以分为高保真(high-fidelity,HF)和低保真(low-fidelity,LF)两类,前者准确率高,后者计算成本低,为了同时发挥两者的优点,提出了一种基于加权最小二乘法的变保真代理模型(MFS-WLS),该模型利用少量HF样本和大量L... 构建代理模型所用样本点可以分为高保真(high-fidelity,HF)和低保真(low-fidelity,LF)两类,前者准确率高,后者计算成本低,为了同时发挥两者的优点,提出了一种基于加权最小二乘法的变保真代理模型(MFS-WLS),该模型利用少量HF样本和大量LF样本,计算LF修正因子和差异函数的未知系数以表示高保真响应。将MFS-WLS与工程优化领域常用的两种变保真和一种单保真代理模型进行比较,结果表明MFS-WLS具有较高的预测精度和鲁棒性,最后通过压力机上梁轻量化设计案例进一步验证了模型的实用性。 展开更多
关键词 保真代理模型 加权最小二乘法 优化设计 轻量化
下载PDF
空间映射与响应面法相结合的结构优化 被引量:3
4
作者 张立新 高学仕 +1 位作者 付志远 郝小龙 《石油大学学报(自然科学版)》 EI CSCD 北大核心 2005年第3期92-95,105,共5页
应用空间映射与响应面方法实现了由低保真模型到高保真模型的响应面映射,在原有空间映射方法的基础上将一类结构优化问题构造为标准二次规划模型,有利于高保真模型的优化求解,且解决了空间映射方法中低保真模型设计变量更新困难的问题... 应用空间映射与响应面方法实现了由低保真模型到高保真模型的响应面映射,在原有空间映射方法的基础上将一类结构优化问题构造为标准二次规划模型,有利于高保真模型的优化求解,且解决了空间映射方法中低保真模型设计变量更新困难的问题。此方法可大大减少应用序列RSM方法进行结构多学科优化时的重分析计算量。数值算例表明此方法可获得较好的求解效率和精度。 展开更多
关键词 空间映射 响应面法 替代模型 结构优化 响应面映射 序列RSM方法 重分析近似法 保真模型 二次规划
下载PDF
Simulation of Biogas Production from Solid Organic Wastes
5
作者 Ignacio Contreras-Andrade Jonathan Parra-Santiago Carlos Alberto Guerrero-Fajardo 《Journal of Chemistry and Chemical Engineering》 2015年第2期107-112,共6页
The design and building of new alternative fuel plants is an increasing necessity to replace old technology and non-renewable fossil fuels. To optimize the performance of these plants and to obtain an economically fea... The design and building of new alternative fuel plants is an increasing necessity to replace old technology and non-renewable fossil fuels. To optimize the performance of these plants and to obtain an economically feasible production of these types of fuels, it is necessary to have a total control of each variable involved in the process of production and how these factors affect the yield of fuel production. In this paper it is proposed a model of a digester to generate gas using a Vensim software designed to generate simulations in dynamic state. This simulation was developed using differential equations to model the behavior at each stage of the process and auxiliary conditions to complement the mathematical description of the model. The main factors in the biogas production are the retention time and the methanogen mortality ratio. For retention time lower than 10 h the process loses effectiveness due to bacterial growth is not completed efficiently, but a high retention time involves a bigger reactor and the yield of production decreases considerably for retention time higher than 40 h. The best yields were obtained for a mortality ratio in methanogen and acidogenic bacteria lower than 0.2 and a retention time of 30 h with a final production of 3.33 L by each kilogram of biomass. 展开更多
关键词 Bio-digester bio-digestion gas synthesis renewable energy.
下载PDF
Simulation Models for Evaluation and Selection of Depreciation Policy for Agricultural Companies in the Republic of Bulgaria
6
作者 Ivanka Borisova Dimitrova Yordanka Velcheva 《Journal of Modern Accounting and Auditing》 2013年第10期1322-1333,共12页
In the present paper, the results from research on the influence of criteria (factors) such as depreciation method, materiality threshold, and inflation rate on the selection of depreciation policy are shown. The re... In the present paper, the results from research on the influence of criteria (factors) such as depreciation method, materiality threshold, and inflation rate on the selection of depreciation policy are shown. The research has been carried out with the help of specifically devised simulation models. The simulation is realized in accounting for the unevenness of money in time and the profitability of the company, the tax rates for corporate tax, and the inflation rate. The models have been devised after analyzing methodology issues, related to depreciation policy and investigating current practices in its implementation in specific agricultural companies in present-day Bulgaria. It has been determined that there is a clear-cut relation between the profitability of the companies and the materiality threshold and depreciation method chosen. The main conclusion is that when the companies develop their depreciation policy, at ceteris paribus they should select higher materiality thresholds and diminishing balance depreciation methods due to the inequality of money in time, the acceleration of obsolescence, and the inflation processes. 展开更多
关键词 depreciation policy materiality threshold methods of depreciation future value inflation rate
下载PDF
上一页 1 下一页 到第
使用帮助 返回顶部