论文部分内容阅读
由于工程结构的复杂性和引起结构损伤原因的不确定性,结构早期微弱和潜在的损伤难以识别和预测。为此提出了基于聚类经验模式分解(EEMD)和支持向量机回归(SVR)的结构健康状态趋势预测方法。首先对多自由度结构渐进损伤的加速度振动信号进行聚类经验模式分解(EEMD);再进行希尔伯特变换(HT)计算瞬时频率;然后用回归支持向量机对反映结构健康状态的瞬时频率进行趋势预测。详细分析了各种参数对回归和预测精度的影响,提出了这些参数的选用方法和一般原则。研究表明:该方法具有训练样本少的特点;在采用二阶多项式核函数、回归步长m=3~5、误差惩罚因子C=100、敏感因子ε=0.01时,可以准确地和高精度地预测结构状态趋势,预测精度达到0.24781%。
Due to the complexity of the engineering structure and the uncertainty of the cause of structural damage, the early weakness and potential damage of the structure are difficult to identify and predict. For this reason, a structural health trend prediction method based on clustering empirical mode decomposition (EEMD) and support vector machine regression (SVR) is proposed. Firstly, EEMD is performed on the acceleration vibration signal of gradual damage in multi-degree-of-freedom structure. Then the instantaneous frequency is calculated by Hilbert transform (HT). Then the instantaneous frequency Make trend forecast. The effects of various parameters on regression and prediction accuracy are analyzed in detail, and the selection methods and general principles of these parameters are proposed. The results show that this method has the characteristics of fewer training samples. When using the second-order polynomial kernel function, the regression step m = 3 ~ 5, the error penalty factor C = 100 and the sensitivity factor ε = 0.01, the method can accurately and accurately Prediction structure trend, forecast accuracy reached 0.24781%.