A parallelized multi-degrees-of-freedom cell mapping method

作者：Eason R P; Dick A J^*

来源：Nonlinear Dynamics, 2014, 77(3): 467-479.

DOI：10.1007/s11071-014-1310-8

• Multi-DOF cell mapping (MDCM)
• Global dynamic analysis
• Basin of attraction
• Bi-stability

Cell mapping methods, in general, provide a computationally efficient way to analyze the long-term global dynamics of lower-dimensional systems. The multi-degrees-of-freedom cell mapping (MDCM) method, in particular, overcomes the scaling limitations of other cell mapping methods, allowing efficiency benefits to be realized for higher-dimensional systems. Unfortunately, the sequential structure of the MDCM algorithm limits the ability to utilize the parallel processing capabilities of modern computers. In this paper, the parallelized multi-degrees-of-freedom cell mapping (PMDCM) method is introduced. The PMDCM method features a restructured algorithm that employs parallel computation to streamline one of the most time-consuming elements: numerical integration. The PMDCM algorithm is described in detail and is demonstrated by comparing results produced by the PMDCM method to those produced by MDCM and the grid-of-starts. By using the PMDCM method on a quad-core processor with 100 simultaneous integrations per mapping step, the total computation time is reduced by 93 %, as compared with the MDCM method. With PMDCM, the global integrity measure also agrees more closely with the computationally intensive grid-of-starts method when compared with the MDCM method.

• 出版日期2014-8