非线性系统
理论(学习稳定性)
计算机科学
不稳定性
应用数学
动力学(音乐)
数学
数学优化
物理
机械
量子力学
机器学习
声学
出处
期刊:Journal of Applied Mechanics
[ASME International]
日期:1975-06-01
卷期号:42 (2): 464-470
被引量:232
摘要
The behavior of linear multistep methods has been evaluated for application to structural dynamics problems. By examining the local stability of the currently popular methods as applied to nonlinear problems, it is shown that the presence of historical derivatives can cause numerical instability in the nonlinear dynamics even for methods that are unconditionally stable for linear problems. Through an understanding of the stability characteristics of Gear’s two-step and three-step methods, a new method requiring no historical derivative information has been developed. Superiority of the new method for nonlinear problems is indicated by means of comparisons with currently popular methods.
科研通智能强力驱动
Strongly Powered by AbleSci AI