关键词: 引水隧洞施工, 施工仿真参数, 差分进化支持向量机, 自适应缩放因子, 混沌理论

Abstract: Update of construction simulation parameters has a great impact on the accuracy of simulation results. Most previous studies use Bayesian theory to update these parameters but have to introduce an assumption of their probability distributions. Such methods cannot generate the series of predicted parameters necessary for describing their dynamic variations. To avoid these shortages, this paper presents a schedule simulation parameter update method for diversion tunnel construction, based on an adaptive chaos differential evolution-support vector machine (ACDE-SVM). First, we develop an ACDE algorithm by adopting an adaptive scaling factor and the chaos theory to improve the differential evolution, so that it can not only reduce the searching time but overcome the shortcoming of the differential evolution that is easy to become premature. Then, we optimize the SVM parameters using ACDE and the time series of onsite parameters, and construct an ACDE-SVM algorithm for predicting the schedule simulation parameters that can overcome the inefficiency of parameter selection and the weakness in generalization capability. Last, we adopt certain error indexes to evaluate this parameter update method and compare it with the conventional simulation and Bayesian method to demonstrate its consistency and superiority. A case study shows that the method can fit well the time trends of parameters and improve simulation accuracy.

Key words: diversion tunnel construction, construction simulation parameters, differential evolution-support vector machine, adaptive scaling factor, chaos theory