| 摘要: |
| 【目的】对墙土系统动测模型进行研究，同时提出了基于模态参数的土体附加参数识别方法，为墙土系统的损伤识别和健康诊断提供一种简单、适用的模型。【方法】建立了墙土系统动测模型，假定附加刚度和附加质量成三角形分布，采用附加刚度和附加质量模拟土体对墙土系统模态特性的影响。基于模态频率构造目标函数，建立模态频率和土体附加参数之间的BP神经网络结构，识别得到不同土体弹性模量对应的土体附加参数。将三维有限元整体模型分析得到的模态频率作为真实值，计算真实值和识别值之间的相对误差，判定识别方法的可行性。【结果】基于模态频率和BP神经网络的土体附加参数识别方法能够满足工程精度要求，模态频率相对误差均在2%以内；随着土体弹性模量的提高，土体的附加刚度呈双指数函数增大，附加质量呈双指数函数减小。【结论】提出的动测模型及基于模态参数和BP神经网络的附加参数识别方法是可行的，为墙土系统的健康诊断提供了一种行之有效的方法。 |
| 关键词: 墙土系统 动测模型 附加参数 模态频率 神经网络 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目（51027004，50878218);长江学者和创新团队发展计划项目(IRT1045) |
|
| Dynamic-detection model of soil-wall system and the identification method of virtual parameter |
|
|
| Abstract: |
| 【Objective】To provide a simple and applicable model for the health diagnosis and damage identification of soil-wall system,the dynamic-detection model of soil-wall system is studied,and the identification method of virtual parameter based on modal parameters is offered.【Method】The dynamic-detection model of soil-wall system is established,the influence of soil on modal parameters of soil-wall system is simulated by a set of added spring and added mass,the triangle distribution of added spring and added mass is assumed.The objective function based on modal frequency is constructed,the BP neural network structure between the virtual parameter and modal frequency is established to identify the virtual parameter.Modal frequency is determined based on the three dimensional finite element model as actual value.Meanwhile,in order to determine the feasibility of the identification method,relative error between actual value and identification value is calculated.【Result】The identification method of virtual parameter based on modal frequency and BP neural network could satisfy the engineering precision,the relative error of modal frequency is less than 2%.with the increases of the elastic modulus of soil,the added stiffness is double exponential function increases,and added mass is double-exponential function decreases.【Conclusion】The proposed dynamic-detection model and the identification method of virtual parameter are feasible,thus a simple and effective model is proposed for the health diagnosis of soil-wall system. |
| Key words: soil-wall system dynamic-detection model added parameter modal frequency neural network |
