基于贝叶斯的CRTS I 型板式无砟轨道CA充填层的损伤识别

China Railway Track System (CRTS) I slab track is one of the most popular ballastless tracks in the world. Due to the complexity of geological and climatic condition along Chinese high-speed railway, the infrastructure deterioration of the high speed railway can be observed and the damage of the cement asphalt (CA) mortar layer is the most obvious. Due to the complex property of the CA motor and the uncertainties from the measurement and modelling error, there are a lot of difficulties in detecting the damage of the CA motor layer. Firstly, the proposed project will utilize the Bayesian model class selection method to select the most probable viscoelastic calculation model class of the slab track based on the measured vibration data. Secondly, the project will develop an advanced Markov chain Monte Carlo-based Bayesian model updating method to fast detect the damage of the CA mortar layer and quantify the corresponding uncertainties. Moreover, the applicability and feasibility of the proposed new detection methodology of the CA motor layer of the slab track based on the vibration data and Bayesian approach are fully validated by conducting theoretical analysis, numerical simulations as well as laboratory experiments and field tests. The research in this project provides new ideas for the research areas of the safe cooperation of the high-speed railway and has profound significance to the development of the discipline itself or the engineering application.

中国铁路轨道系统I型板式无砟轨道是目前世界上最主要的无砟轨道类型之一。由于我国高速铁路沿线地质和气候条件复杂，运营线上基础结构的劣化已进入显现期，其中沥青砂浆（CA）充填层的损伤最明显。由于CA充填层自身的复杂特性以及测量和模型误差导致CA充填层损伤识别的研究存在诸多困难。本项目首先通过贝叶斯模型选择的方法从拟建的粘弹性数学模型群中选择合适的基于振动数据的板式无砟轨道粘弹性计算模型；然后，开发出一种基于改进的马尔科夫链蒙特卡洛贝叶斯的模型修正方法快速高效的进行砂浆充填层的损伤识别以及不确定性分析；通过理论分析、数值模拟、实验室模型及现场试验研究，探讨基于振动数据和贝叶斯的无砟轨道CA砂浆充填层的损伤识别新方法的可行性。本项目的研究成果为实现高速铁路的安全运营提供全新的研究思路, 无论对学科自身的发展，还是对工程应用都具有深远的意义。

我国高速铁路沿线地质和气候条件复杂，运营线上轨道结构的劣化已进入显现期，其中沥青砂浆（CA）充填层的损伤最明显。此外，线路里程长、轨道结构承受高强高频荷载，且CA砂浆和弹性扣件的刚度存在较大的不确定性，导致高速铁路运营维护出现较大困难。本项目首先在贝叶斯框架下，采用贝叶斯模型修正和模型选择方法，对CA砂浆层的贯通和非贯通空洞损伤进行损伤识别。通过理论分析、数值模拟以及实验验证，探究了基于时域贝叶斯的无砟轨道CA砂浆充填层的损伤识别方法的可行性。研究结果表明基于时域贝叶斯方法不仅能够识别无砟轨道CA砂浆层的空洞损伤，还可提供损伤程度的概率分布信息，表达识别结果的不确定性。此外，基于不同传感器数量对模型参数的后验不确定性研究结果表明，即使采用一个传感器，参数的后验不确定性也在可接受的范围内。本项目的研究成果为不确定性的损伤识别方法应用于高速铁路安全运营领域奠定了基础。