变系数瞬态热传导问题边界元格式的特征正交分解降阶方法?
- 大连理工大学工程力学系,大连,116024
- 大连理工大学航空航天学院,工业装备结构分析国家重点实验室,大连 116024
摘要: 提出了一种基于边界元法求解变系数瞬态热传导问题的特征正交分解(POD)降阶方法,重组并推导出变系数瞬态热传导问题适合降阶的边界元离散积分方程,建立了变系数瞬态热传导问题边界元格式的POD降阶模型,并用常数边界条件下建立的瞬态热传导问题的POD降阶模态,对光滑时变边界条件瞬态热传导问题进行降阶分析。首先,对一个变系数瞬态热传导问题,建立其边界域积分方程,并将域积分转换成边界积分;其次,离散并重组积分方程,获得可用于降阶分析的矩阵形式的时间微分方程组;最后,用POD模态矩阵对该时间微分方程组进行降阶处理,建立降阶模型并对其求解。数值算例验证了本文方法的正确性和有效性。研究表明:1)常数边界条件下建立的低阶POD模态矩阵,能够用来准确预测复杂光滑时变边界条件下的温度场结果;2)低阶模型的建立,解决了边界元法中采用时间差分推进技术求解大型时间微分方程组时求解速度慢、算法稳定性差的问题。
Reduced order mo del analysis metho d via prop er orthogonal decomp osition for variable co e?cient of transient heat conduction based on b oundary element metho d
- 大连理工大学工程力学系,大连,116024
- 大连理工大学航空航天学院,工业装备结构分析国家重点实验室,大连 116024
Abstract: Boundary element method (BEM) is widely used in engineering analysis, especially in solving the transient heat conduction problem because of the advantage that only boundary of the problem needs to be discretized into elements. The general procedure of solving the variable-coe?cient transient heat conduction problem by using the BEM is as follows. First, the governing differential equations are transformed into the boundary-domain integral equations by adopting the basic solution of the linear and homogeneous heat conduction problem—Green function. Second, domain integrals in the integral equation are converted into boundary integrals by the radial integral method or the dual reciprocity method. Finally, the time difference propulsion technology is used to solve the discrete time differential equations. A large number of practical examples verify the correctness and validity of the BEM in solving the variable coe?cient of transient heat conduction problem. However, two deficiencies are encountered when the system of time differential equations is solved with the time difference method, i.e., one is the stability of the algorithm, which is closely related to the time step size, and the other is time-consuming when the freedom degree of the problem is large and all specified time steps are considered, because a system of linear equations needs to be solved in each time step. Therefore, in this paper we presenta reduced order model analysis method of solving the variable-coe?cient transient heat conduction problem based on BEM by using the model reduction method of proper orthogonal decomposition (POD). For variable-coe?cient transient heat conduction problems, the discrete integral equations which are suitable for order reduction operation are deduced by using the BEM, the reduced order model is established by using the model reduction method of POD, and a low-dimensional approximate description of the transient heat conduction problem under time-varying boundary condition is obtained by projection of the initial discrete integral equations on some few dominant POD modes obtained from the problem under constant boundary conditions. First, for a variable coe?cient transient heat conduction problem, boundary-domain integral equations are established and the domain integrals are transformed into boundary integrals by using the radial integration method. Second, the time differential equations with discrete format which is suitable for order reduction operation are obtained by reorganizing the integral equations. Third, the POD modes are developed by calculating the eigenvectors of an autocorrelation matrix composed of snapshots which are clustered by the given results obtained from experiments, BEM or other numerical methods for transient heat transfer problem with constant boundary conditions. Finally, the reduced order model is established and solved by projecting the time differential equations on reduced POD modes. Examples show that the method developed in this paper is correct and effective. It is shown that 1) the low order POD modes determined under constant boundary conditions can be used to accurately analyze the temperature field of transient heat conduction problems with the same geometric domain but a variety of smooth and time-varying boundary conditions; 2) the establishment of low order model solves the problem of heavy workload encountered in BEM where a set of large linear equations will be formed and solved in each time step when using the time difference method to solve the large time differential equations.