摘要
曲线拟合是传感器非线性校正的重要方法,但热偶规真空传感器输出特性具有明显的分段性且各段特性差异大,采用单一模型难以取得满意的拟合结果.本文提出基于多支持向量回归机(MSVR)的曲线拟合方法,将输入样本空间分割成多个子空间,为每一子空间建立一个支持向量回归机(SVR)来映射该局部空间的非线性关系,合成各SVR的输出即为传感器全量程输出特性.用所提出的方法拟合热偶规真空传感器输出特性,仿真结果表明了该方法的有效性.
Curves fitting is an important approach of nonlinear correction for sensor complex output characteristics, it is difficult for a thermocouple vacuum sensor to obtain s. But due to high-precision fitting with a single model. Here we propose a curve fitting method based on multiple support vector regression （MSVR）. The input samples are divided into some sub-spaces, and mapping relation is set up for each sub-space by using a support vector regression （SVR） which has parameters different from others. Each SVR approximates the sensor characteristic of the corresponding area. The sum of all SVR output is the full-scale characteristic output of the sensor. The method was applied to fit the output characteristic curve for the thermocouple vacuum sensor. The simulation results show that this approach is valid.
出处
《测试技术学报》
2010年第1期29-33,共5页
Journal of Test and Measurement Technology
关键词
多支持向量回归机
样本空间分割
拟合
非线性校正
multiple support vector regression （MSVR）
samples space division
fitting
nonlinear correction
