【摘要】 针对铠装热电偶传感器进行精度补偿中对辨识建模精度要求高的问题,提出运用支持向量机(SVM)辨识建模,并结合量子遗传算法(QGA)全局寻优能力强的特点对SVM核函数参数进行优化的方法,以减小建模误差。采用各种检验信号对SVM模型的输出和校验误差进行了推广能力测试,并与递推最小二乘估计法辨识建模进行建模误差比较,结果表明SVM辨识建模与QGA结合的方法在对铠装热电偶传感器辨识建模方面具有良好的推广能力和较高的建模精度。通过精度补偿实验,进一步验证了该方法能使模型精度达到铠装热电偶传感器对精度补偿的要求。更多还原

【Abstract】 Aiming at the high accuracy requirement of identification modeling in sheathed thermocouple sensor accuracy compensation,an approach for the parameter optimization of support vector machine( SVM) kernel function is proposed,which adopts SVM identification modeling combined with quantum genetic algorithm( QGA) that has powerful global searching ability to minimize modeling error. The generalization performance of SVM identification model output response and checking error was tested,and the SVM identification modeling error was compared with that of recursive least square estimation method. The results show that the SVM identification modeling method based on QGA has better generalization performance and modeling precision for sheathed thermocouple sensor identification modeling. The related error compensation experiment proves that the proposed method can meet the requirement of sheathed thermocouple sensor error compensation in modeling precision.更多还原