Abstract: The truck mounted concrete pump boom health monitoring scheme whose fatigue damage is calculated by using strain signal is established. But the service life of the strain gauge is short and not reliable, so it cannot meet the need of truck mounted concrete pump long-term health monitoring through direct measurement of strain signal by strain gauge. So the soft sensor modeling method for strain signal based on LS-SVM is given. After analyzing strain features of truck mounted concrete pump boom, the secondary variables are selected. To improve the precision of the model, decoupling the strain signal, the soft measurement model of static and dynamic strain whose parameters are optimized by genetic algorithms are respectively established, and then the total strain is acquired. Separate modeling effect is compared with overall modeling. The emulation result indicates that soft-sensing technique provides a feasible method for the realization of truck mounted concrete pump health monitoring in engineering, and respectively establishing the soft sensor model of the static and dynamic strain is more accurate than the overall modeling.

Key words: truck mounted concrete pump boom, structure health monitoring, soft sensor, Least Square Support Vector Machines（LS-SVM）, genetic algorithm

摘要： 利用混凝土泵车臂架应变信号计算其疲劳累积损伤的健康监测方案，由于应变片的使用寿命较短且不可靠，通过应变片直接测量泵车臂架应变信号不能适用于泵车臂架结构长期健康监测。采用软测量技术，基于最小二乘支持向量机（LS-SVM）建立软测量模型来间接获得泵车臂架应变信号。分析泵车臂架应变信号的特点，进而选择辅助变量。为了提高模型精度，对应变信号进行解耦，分别建立静态应变和动态应变的软测量模型进而得出总应变，利用遗传算法对模型参数进行了优化，与总体建模结果进行了比较。仿真分析结果表明，软测量技术为泵车臂架结构健康监测的工程实现提供了一种可行的方法，并且分别建立静态应变和动态应变的软测量模型比总体建模精度更高。

关键词: 泵车臂架, 健康监测, 软测量, 最小二乘支持向量机, 遗传算法