Abstract The conventional particle swarm optimization is improved by composing a nonlinear inertia weight function and importing in an acceleration factor,which can enhance the convergence and efficiency of computation. At the same time,the improved particle swarm optimization is improved again by message passing interface(MPI)-based master-slave parallel framework. The back analysis process of large-scale underground engineering which based on the ordinary computer fleet system distributed-storage parallel mode is compiled with Fortran language. According to distributed-memory parallel mode,the parallel computation can be conducted and completed using computer cluster networks;thus considerably reduce the cost and enhance the efficiency of computation. The results indicate that the improved particle swarm optimization is efficient. Moreover,the influences of excavation damage of surrounding rock mass,reliability of measured data,parallel granularity and load balance on computational efficiency and accuracy of back analysis in underground engineering are briefly discussed. The proposed improved method and the rational recommendations provide the back analysis of parameters and dynamic optimal design of underground engineering with a new idea.
|
Received: 07 September 2012
|
|
|
|