非线性摩擦失谐叶片排序并行退火算法

基于叶片模态实验获得的叶片失谐参数,基于非线性摩擦阻尼分析提出一种失谐叶盘结构的集中参数模型.综合模拟退火算法的局部搜索能力和遗传算法的全局搜索能力,并引入禁忌表作为搜索记忆表,提出了一种应用模拟退火进化算法附加禁忌表的叶片排序优化方法.同时结合图形处理器运算设计了基于compute unified device architecture(CUDA)的并行算法并且分析其算法性能.研究表明:基于并行框架寻优可以获得4.5倍的加速,同时按照优化后方案安装,可以明显降低叶盘系统振动响应幅值与方差. 

Mistuned blade arrangement with non-linear friction based on parallel annealing algorithm

Based on mistuned blades parameters obtained by modal experiment, a lumped parameter model of mistuned bladed disk system was established considering the non-linear friction damping. Blade arrangement optimization method using simulated annealing evolutionary algorithm additional tabu list was proposed by combining the local search ability of simulated annealing algorithm and global searching ability of genetic algorithm. At the same time, a parallel algorithm was designed based on the computation unified device architecture (CUDA)in combination with graphic processing unit. The results show that based on parallel framework optimization, the acceleration of 4.5 times can be obtained. Meanwhile bladed disk system vibration response of the largest amplitude and variance can be significantly reduced when using optimized arrangement.