Co-evolution Research And Design In Foundation Pit Based On Genetic Algorithm

Deep foundation pit' s optimization design concerns with many factors and all these factors have complicated relationship. Its scheme and detail design have hierarchic sequence with discrete design data. For these reasons traditional optimization is not adaptive to foundation pit optimization.Genetic Algorithm is a global algorithm which uses natural law "survival of the fittest, extinction of the unfitness" and natural selection mechanism to guide and determinate search direction. Comparing with other optimization methods, it needs few mathematical requirements, so it appropriates to deep foundation pit optimization.Parallel implementation of Genetic Algorithm provides a feasible method to speed up convergence rate of Genetic Algorithm. After contrasting with common parallel genetic methods, this paper suggests a distributed parallel method: Distributed Co-evolution Algorithm. This algorithm bases on the parallelism of population fitness computation and the parallelism of population evolution, and can adapt to complicated engineering application. Its foundation absorbs the cooperatively evolution idea of Hybs&Gero's disjunction of problem space and solution space, and the ideology of collaboration in parallel engineering. It divides engineering optimization task to scheme design and detail design. After decomposition of fitness function and adjustment of constraint condition, scheme and detail evolutes respectively and hierarchically in various processors in local network. Here scheme design evolutes in system optimizer; detail designconsists of various parts, and each part evolutes in different sub-optimizer. Distributed co-evolution introduces "fittest survival" law of genetic algorithm, and adds hierarchical co-evolution of problem space and solution space. It provides variant scheme with variant religion of detail, and can adaptive to discrete and hierarchical complicated engineering problems such as deep foundation pit optimization problem.After analyzing factors of deep foundation, the relationship between these factors, and their contribution or effect to deep foundation, a scheme-detail two-level co-evolution model and algorithm is proposed.But experiment shows that the two-level model takes too much time on detail design, so basing on two-level co-evolution model, we add one level of detail parts to two-level co-evolution model and form three-level co-evolution model. And according to Distributed Co-evolution, a distributed three-level co-evolution implement method is built.Finally, cheap personal computers are used to construct distributed environment and design a three-level co-evolution model system. The system includes main server program algorithm, sub-server algorithm and client algorithm. It solves communication problem between all the processors and evolutes scheme design and detail design hierarchically and synergically. Experiment shows that three-level co-evolution is feasible and useful.