Abstract: To eliminate the influence of ambient temperature on thermal sensor in gas detection, the authors put forward a new calibration method for sensor temperature characteristic based on data fusion and Canonical Particle Swarm Optimization-Support Vector Machine (CPSO-SVM). The method adopted SVM to fuse the data of sensor pair composed of a thermal sensor and a temperature sensor, and applied CPSO and the principle of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) minimization of test samples set to tune the parameter vector of SVM. The experimental results of H2 detection show that the proposed method can effectively improve the temperature quality of thermal sensor, and realizes accurate detection of gas concentration.

Key words: thermal sensor, temperature characteristic calibration, Support Vector Machine (SVM), data fusion, Canonical Particle Swarm Optimization (CPSO)

摘要： 为了消除环境温度对热导气体传感器的影响,提出了一种热导传感器温度特性的经典粒子群优化--支持向量机(CPSO-SVM)数据融合校正方法。该方法将热导传感器和温度传感器构成传感器组,利用支持向量机对传感器组的输出信号进行数据融合,采用经典粒子群优化算法和测试样本集均方根误差与平均绝对百分比误差同时最小原则选择和优化支持向量机的参数向量。对氢气浓度的检测实验表明,该方法能有效地改善传感器的温度特性,实现了气体浓度的精确检测。

关键词: 热导传感器, 温度特性校正, 支持向量机, 数据融合, 经典粒子群优化