多因素下混凝土冻融劣化的粒子滤波寿命预测评估及优化

Affected by many factors such as material properties and chloride ion erosion, the freeze-thaw degradation process is of high uncertainties, which means great difficulty in modeling. As a result, service life of concrete structures under freeze-thaw cycles is often shorter than design life. Therefore, the demand for eliminating the uncertainty in the process of degradation and realizing the life prediction of freeze-thaw is increasing day by day..This project is based on the background of freeze-thaw failure of concrete pavement such as expressway and airport pavement in rainy and snowy days. With the realization of standard particle filter-based life prediction of freeze-thaw under the influence of single factor, this project intends to carry out freeze-thaw test under multiple factors and to analyze the freeze-thaw failure mechanism by microscopic technology such as SEM and X-ray. Meanwhile, relevant theory and life prediction model are used to build the evolution equation of multi-factor freeze-thaw degradation process. Comparative study of mass loss rate, relative dynamic elastic modulus as well as other assessment methods are conducted to build a suitable observation equation. Employing the evolution and observation equation, a state space model describing the multi-factor degradation process of concrete freeze-thaw is set up to widen the applying area, and to predict the remaining useful life of the concrete based on standard particle filter. To fight against sample impoverishment problem of the standard particle filter, many improved particle filter methods are to be used to improve the life prediction accuracy..The results can provide data support for the construction of expressways, airport pavements, offshore and coastal projects, as well as other construction projects, which have important application value and economic benefits for improving the design, extending the service life and enhancing the maintenance and protection of concrete structures.

受材料属性、氯离子侵蚀等多因素影响，冻融劣化过程具备高度不确定性，建模难度大，冻融循环下的服役寿命常低于设计寿命。故消除劣化过程中的不确定性，实现冻融损伤寿命预测愈发重要。.本项目以机场、高速公路等混凝土路面在雨雪天气下的冻融破坏为研究背景，在单因素冻融的标准粒子滤波寿命预测的基础上，开展多因素冻融实验，结合SEM、X射线等，微观分析冻融破坏机理；应用相关理论与寿命预测模型，构建多因素冻融劣化过程的状态方程；对比质量损失率、相对动弹性模量等监测方法，构建适用于量化冻融损伤的观测方程；基于状态与观测方程，建立多因素冻融劣化过程的状态空间模型，拓展其适用范围，并实现多因素冻融的标准粒子滤波寿命预测；针对粒子多样性匮乏问题，应用改进粒子滤波算法，提升单/多因素下冻融寿命预测精度。.研究可为机场、公路等建设提供数据支撑，对改进混凝土结构设计、延长使用寿命和增强维护保障具有重要的应用价值与经济效益。

在多种内外部因素的影响下，混凝土冻融损伤的劣化过程具有高度不确定性，因此其精确建模较为困难。如考虑众多不确定性因素，开展可靠性实验，预测其寿命，将耗费大量的实验成本。粒子滤波算法可通过大量的“粒子”来表征各种不确定因素下冻融损伤的劣化过程，对模型复杂度要求较低，不需要大量的训练数据，且与合适的观测手段相结合，可及时对预测结果进行修正。本项目结合粒子滤波及其进改进算法，实现了多因素下混凝土冻融的粒子滤波寿命预测，对指导混凝土结构设计，优化维护策略具有重要意义，项目主要开展了以下研究：.一、以相对动弹性模量随冻融循环的单段损失模式衰减规律构建状态方程，以超声波声时与相对动弹性模量的关系构建观测方程，最终建立描述冻融劣化过程的状态空间模型，该模型适用于多种不确定因素的影响，具备通用性；.二、以状态空间模型为基础，提出了一种基于标准粒子滤波的多因素混凝土寿命预测方法，克服了材料组成、氯盐浓度和测试误差等多种不确定性因素的影响，提升了冻融寿命预测精度，并利用C30和C50混凝土于3%、5%和20%氯盐溶液的冻融破坏实验，验证了算法有效性；.三、从重要性函数选择优化的角度出发，提出了基于辅助粒子滤波的多因素混凝土寿命预测方法，缓解了标准粒子滤波方法中存在的粒子多样性匮乏现象，并在冻融破坏实验中进行了验证，其预测精度得到了提高；.四、从重采样技术的角度出发，提出了基于确定性重采样粒子滤波的多因素混凝土寿命预测方法，克服了粒子多样性匮乏现象，并通过冻融破坏实验进行了验证，对比标准粒子滤波与辅助粒子滤波算法，具备更高的预测精度；.五、从重采样技术的角度出发，针对标准粒子滤波算法存在的离散性误差，引入采样方差作为代价函数，提出了基于最小采样方差粒子滤波的多因素混凝土寿命预测方法，最大程度地降低了重采样过程中的信息损失，有效地缓解了粒子多样性匮乏现象，并通过冻融破坏实验进行了验证。对比标准粒子滤波与辅助粒子滤波算法，具备更高的预测精度，同时该算法保留了维度自由的特性，是对确定性重采样粒子滤波无法应用于复杂系统的缺点的有效补充。