工程模拟自学法在基坑开挖中的应用

在基坑开挖工程中，常常可以利用数值模拟方法计算基坑场地的变形。然而，常规的数值模拟一般没有充分结合经验分析和场地观测数据来计算出较为精确的开挖变形。工程模拟自学法是一种基于神经网络的数值模拟方法，它属于一种可以综合有限元与人工智能的反分析技术。工程模拟自学法能充分发挥神经网络的自适应性、自组织性及学习、联想、容错及抗干扰能力，揭示出历史资料分析和场地测量数据中所蕴含的非线性关系,提炼出更接近土体实际状态的本构模型进行数值分析，改善变形计算精度。工程模拟自学法的具体应用可通过基坑开挖例子说明。

The Approach of Self Learning in Engineering Simulations for Excavations

In excavation engineering designers often estimate potential deformations by numerical modeling. However the conventional numerical modeling can’t fully integrate case histories and observation of performance to get more factual and accurate deformation of excavation. A numerical simulation introduced in the paper is called self learning in engineering simulations. It is an inverse analysis technique that incorporates finite analysis with artificial intelligence. It can fully exert its abilities of self-adaptability, self-organization, powerful learning, association, fault tolerant and anti-interference and improve the calculated deformation with the extract soil constitutive models based on the data training of case histories or case observation. The application of the approach of self learning in engineering simulations is demonstrated using a simulated excavation.