基于云计算的框架结构参数并行辨识算法

大型复杂结构健康监测（Structural Health Monitoring，SHM）系统的安装产生了海量监测数据，传统结构分析与数据处理技术使得监测数据得不到实时分析处理，导致不能及时评估结构工作状态并进行危险预警。为了解决这一问题，该文对传统多粒子群协同优化（Multi-Particle Swarm Coevolution Optimization，MPSCO）算法进行分布式并行化改进，开发了基于云计算的PMPSCO算法。在此基础上，提出了基于PMPSCO算法的框架结构物理参数辨识方法，并在MATLAB分布式云计算平台上对一个15层框架数值试验和一个7层钢框架实验室试验进行结构物理参数辨识，探讨了接入不同分布式并行节点数时该算法的加速关系。辨识结果表明:PMPSCO算法具有良好的精度、稳定性和拓展性，可通过增加接入的分布式并行节点数灵活提高算法运算速度，以满足结构监测数据实时处理的要求。

A PARALLEL IDENTIFICATION ALGORITHM ON PHYSICAL PARAMETERS OF FRAME STRUCTURES BASED ON CLOUD COMPUTING

The installation of large-scale complex Structural Health Monitoring (SHM) systems causes huge amount of data. Traditional structural analysis and data processing techniques cannot real-time process the data from SHM systems. Timely status assessment and risk warning cannot be available to these large-scale complex structures. To solve this problem, parallel improvement is conducted in the traditional multi-particle swarm coevolution optimization (MPSCO) algorithm, thus a new PMPSCO algorithm based on cloud computing is developed. On the basis of the PMPSCO algorithm, a parallel physical parameter identification method is proposed for frame structures, and a numerical experiment of a 15-story frame and a laboratory test of a 7-story steel frame are analyzed to validate the proposed method. Furthermore, the relationship between speedup ratio and number of parallel nodes are explored on MATLAB distributed cloud computing platform. The identification results indicate that the proposed PMPSCO algorithm has high precision, good stability and expandability. In addition, the speed of algorithm can be increased greatly by adding more parallel nodes to meet the needs of real-time processing monitoring data.