基于主成分残差的结构损伤识别方法

为消除环境温度对结构损伤识别的影响,提出了一种基于主成分分析与小波包系数节点能量谱的损伤识别方法,该方法把环境温度当成影响结构动态响应的潜在变量,通过计算结构动态响应的主成分残差来消除环境温度对动态响应的干扰,并通过此残差小波包系数节点能量谱计算结构损伤敏感特征,通过对比结构未知状态和参考状态的损伤敏感特征判别结构是否...     

独立成分分析技术综述

独立成分分析是一种将源信号从混合信号分离出来的信号处理技术。本文首先概述了独立成分分析与盲信号分离,主成分分析的关系,然后系统地介绍了该技术的基本概念和数据模型,对现今比较流行的各种独立成分分析算法进行了剖析和总结。最后对独立成分分析的发展及应用前景作了展望。     

基于主成分分析和学习向量量化神经网络的制动工况路面识别与验证

开展车辆制动时路面类型识别的研究，提出一种基于主成分分析-学习向量量化神经网络 （Principal Component Analysis - Learning Vector Quantization，PCA-LVQ） 的制动工况路面识别方法。利用主成分分析对多维度驾驶数据降维处理，提取能表征路面特征的主要成分，采用学习向量量化神经网络对降维处理后的驾驶数据进行训练，并用于路面特征分类，使用制动工况下实车试验数据和硬件在环仿真数据进行验证。结果表明，所提出的 PCA-LVQ算法能准确识别路面类型特征，路面识别的精度达到 97%，与传统 BP神经网络的路面类型特征识别精度提升 7%；同时，在不同车速下，基于PCA-LVQ算法也能较准确地识别路面类型特征。     

Fatigue assessment of ship structures based on equivalent wave probability (EWP) concept (1st report): Proposal of EWP concept and its verification by 8600TEU container ship's onboard hull monitoring

A long-term fatigue assessment method based on EWP concept is proposed. ‘Equivalent wave probability (EWP)’ is the fictitious (H_S, T_m)'s joint probability distribution function (JPDF), for which the frequency distribution of the stress variance R², f(R²), calculated by spectral fatigue assessment agrees with the observed one. By choosing probability function p(R²) to fit f(R²), the R²'s statistical model (R2SM) which represents the relation between the EWP parameters and R²'s population parameters is developed, and the Bayesian inference, which can estimate the EWP parameters from the measured R² data is developed. The EWP at the reference position (RP) can be determined by Bayesian inference from the measured R² through the R2SM at RP. To accurately estimate the measured f(R²) at the target position (TP) from the EWP at RP, an R2SM correction factor at TP, denoted by α_TP, is introduced in the process of assimilating R2SM. The resulting R2SM, which has been assimilated by Bayesian inference using measured data, is referred to as data-assimilated R2SM (DAR2SM). The fatigue assessment using EWP at RP as the input of DAR2SM at TP is called Bayes-EWP-DAR2SM analysis. The validity of Bayes-EWP-DAR2SM analysis is verified by using the long-term (about four years) multi(12)-position hull monitoring (HM) data of an 8,600TEU container ship. The fatigue damages estimated by Bayes-EWP-DAR2SM based solely on the stress history of a single sensor are in agreement with measurements with sufficient accuracy, independent of the chosen data assimilation period. This demonstrates that the multi-position fatigue assessment solely through HM at one RP based on EWP concept is realized.