摘要
使用演化算法求解MEMS继电器参数优化主要瓶颈在于算法运行时间过长,而算法运行时间过长主要由于电磁仿真软件进行建模和分析需要耗费大量的计算时间。针对该问题,采用主从并行模式,对演化算法个体适应值计算阶段并行化处理。在充分考虑计算机资源的使用效率与负载均衡等因素下,使服务器尽量少地参与任务计算及减少与客户机的通信以增强并行模式的分布能力,并且增加了客户端掉线处理,任务重分配等操作以增强并行模式的容错能力。经过测试,该并行演化算法在MEMS微波继电器参数优化上加速比接近线速,具有良好的并行效率且容错性较高。
Time-consuming by using evolutionary algorithm to optimize MEMS microwave relay parameters is the manly problem. To solve this problem, this paper presents a master-slave parallel model to parallel the fitness calculation during the evolutionary algorithm. Considering the efficiency of the use of computer resources and load balancing, and other factors, the parallel model allows the server as little as possible to participate in mission computing and to reduce the communication between clients to enhance the distribution capacity. Besides, the model can complete the client exception handling, task redistribution to enhance the model’s fault-tolerant capability. By comparing the testing result, it proves the parallel evolu-tionary algorithm for MEMS microwave relay parameters optimization has a good performance.
出处
《计算机工程与应用》
CSCD
2014年第6期200-204,共5页
Computer Engineering and Applications
基金
国家自然科学基金(No.61103145)
中央高校基本科研业务费专项资金资助项目(No.CUG100314,No.CUG120409)
关键词
并行计算
演化算法
微机电系统(MEMS)微波继电器
parallel computing
evolutionary algorithm
Micro-Electro-Mechanical System（MEMS） microwave relay
