Parameter Identification Of Parallel Computing. Deep Foundation Excavation

The high-speed development of Chinese economy promotes the constructing of high-rise building and the utilization of city underground space in large quantity, which leads to the excavation of the large-scale deep foundation, and the deep foundation excavated is larger and deeper than before. Consequently, it is important to compute soil physical property parameters reasonably. Then, we can use it to direct the excavation engineering, and reduce the influence during the deep foundation excavation process on the surrounding environment.Theoretical analysis method of various projects must be based on reasonable soil physical property parameters. Determination of the parameters in geological engineering has been traditionally carried out on the basis of the results obtained from the laboratory or hi situs tests. However, because of the disturbance of the sample of soil hi the course of excavation and transmission, the difference of boundary condition between the locale and the laboratory, the precision of the instrument etc., parameters determined this way often show great difference from the actual values, which leads to the unreliability of the analysis results.A novel method based on artificial neural network BP algorithm to perform the parametric identification in deep foundation excavation is proposed hi the paper. Taking in situs measurements as network input and parameters to be identified as network output, the network is trained with the samples obtained from FEM computation. Using the first order and the second order derivative information of the network error function to the learning rate factor// and the momentum factor ?, dynamic optimization of the learning rate factor// and the momentum factor ? is obtained during the network training process, which efficiently speeds up the network learning rate. Finally, a parallel program is coded and compiled on the Dawning Supercomputer. The examples the paper presents show that the difference between the identification value and the theoretical value is minute, and the mapping relationship between the soil displacement and soil physical property parameters founded by three-layer artificial neural network model is suitable. The parallel program also shows certain accelerator.