项目名称: 奇异摄动数值模拟的多谱多尺度有限元特征值分解的研究
项目编号: No.11301462
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 江山
作者单位: 扬州大学
项目金额: 22万元
中文摘要: 针对奇异摄动问题研究多谱多尺度有限元法处理特征值子问题,利用有效的正交分解方案,得到高效数值模拟结果。通过多尺度基函数构造多尺度泛函空间,利用多尺度基捕捉边界层的局部性态,探究一致收敛的、稳定的自适应多尺度模拟,以期获得高精度、优效率、超收敛的数值结果。 多谱多尺度有限元法通过求解局部微分方程得到多尺度基函数,通过有效的正交分解处理复杂多尺度的特征值逼近,能够获取解的微观特征值信息,从而节约计算资源在宏观尺度得到高效的多尺度数值解。针对奇异摄动边界层现象,通过改进误差和残量校正确定多尺度自适应方案,建立新型的自适应误差指示子,获得与摄动参数无关的一致收敛方法,将方法推广到Petrov-Galerkin类型以期获得最优泛函空间。拓展相关理论研究和数值实践,实现并行化的多谱多尺度有限元法应用于多尺度问题的理论进展和计算优势,对于奇异摄动与特征值问题的研究具有重要理论意义和实际应用价值。
中文关键词: 多谱多尺度有限元法;奇异摄动;特征值分解;边界层;数值模拟
英文摘要: This project studies the properly orthogonal decomposition for the spectral multiscale finite element eigenvalue problem in the singularly perturbed simulations. For effectively approximating the numerical solutions of singular perturbation, we propose the novel spectral multiscale finite element method for the eigenvalue sub-problem, and consider the properly orthogonal decomposition for it. We construct the multiscale function space spanned by the multiscale basis functions, and make use of these multiscale bases to capture the local information of the boundary layers, and we investigate the consistently convergent and stable adaptive multiscale simulation for the singular perturbation. And there are many advantages of the proposed spectral multiscale approximations such as good accuracy, high computation efficiency and super-convergence, and so on. The spectral multiscale finite element method constructs the multiscale basis functions in the localized homogenous or non-homogenous problem. With the properly orthogonal decomposition to approximate the complicated multiscale eigenvalue, the multiscele bases have abilities to capture the properties of the microcosmic model, and thus the novel method can naturally reduce the computation costs to obtain the accurate and efficient solutions of the macroscopical prob
英文关键词: spectral multiscale finite element method;singular perturbation;eigenvalue decomposition;boundary layers;numerical simulation