桥梁预警系统的数据预处理

针对影响桥梁预警系统有效工作的数据获取问题，提出了融合神经规则、数据分类和数据评估的预警系统数据预处理方法.通过训练集将人工完成数据处理过程中所用到的知识融入到神经规则中，并应用神经规则剔除噪声数据；根据拟定的相似性指标对神经规则输出的数据进行分类，合并相似度较大的信息，以大幅度降低数据量；应用曲率模态对各测点保留的信息进行剖分，并与桥梁结构各测点的标准曲率模态比较，提取与初始数据信息不一致的采样数据，为预警系统的损伤识别提供依据.模拟分析表明，该方法能够在大幅度降低预警系统数据量的基础上保留结构状态发生变化的关键信息，具有一定的应用价值.

Data Preprocess for Bridge Damage Alarming System

Bridge alarming system data sets are often interspersed with noise. If it does not consider noise, the data process algorithm in alarming system may not provide accurate results. A new hybrid approach comprising of neural rules, data classification and data evaluation was proposed to carry out data selection. Specifically, the neural rules remove noise data by applying the principles of an intelligent reasoning process. Data classification classifies the patterns in the reduced datasets based on the similarity index. The data evaluation applies curvature model to extract kernel information of bridge structure. The simulation results demonstrate that the proposed approach produces good classification accuracy and a higher level of consistency to bridge damage condition.