一种加速的参数化模型降阶方法

参数化模型降阶(PMOR)方法离线阶段训练数据的过程比较耗时,因此提出了一种加速方法。在基于矩阵插值的PMOR方法的基础上,采用组合近似的重分析技术对基于振型向量的模型降阶(MOR)过程进行加速,利用初始刚度分解矩阵生成迭代计算基向量以对系统矩阵降阶,通过降阶矩阵生成参数化模型的振型向量,对参数空间上的采样点重复整个加速计算过程生成离线数据库。并以电磁振动台动圈为例,采用普通方法和加速方法对均布采样样本点展开仿真研究,结果表明,在保证所构建的离线模型数据库准确度的前提下,此方法能减少80%以上的MOR计算时间,且随着采样点的增多,增速越明显,可以大幅度提高参数化降阶模型离线训练效率。 

Accelerated parametric model order reduction method

Given the relatively time-consuming offline data trainingphase of the parametric model order reduction (PMOR) method, an accelerated method was proposed. Based on the PMOR method by matrix interpolation, the combined approximation reanalysis technology was applied to accelerate the repeated model order reduction (MOR) procedure. The factorization matrices of the initial stiffness matrix were utilized to generate the basis vectors to reduce the system matrices, calculated the final modal shape vectors via reduced matrices and repeated the calculation procedure for the sampling points in parameter space to build the offline database. Then the moving coil of the electrical-dynamic shaker was used as an example. The regular and improved methods were also used to perform the simulation for the uniformly distributed sample points. The results show that the proposed method can significantly reduce the MOR times more than 80% and improve the efficiency of building offline parametric database with guaranty of the reduced models’ accuracy.