中文摘要：

为了使用蓝宝石晶体高温传感器对工业烟气温度进行长期在线监测，设计了蓝宝石晶体高温传感器的解调系统，并对其进行了标定。对蓝宝石晶体高温传感器的白光偏振干涉测温原理进行了理论分析，并利用离散腔长变换（DGT）解调算法对光程差信息进行解调。在此基础上建立了一套以热电偶为参照的标定系统，使用S型高温热电偶采集温度数据，得到了光程差一温度样本。分别利用二次多项式拟合法与BP神经网络法对传感器的输出曲线进行了拟合与泛化，并进行了对比。实验结果表明：在800～1300℃温度范围内，与二次多项式拟合方法相比，BP神经网络的拟合精度较高，拟合残差均值达到0．33℃；泛化能力强，多次逆化结果误差均值为0．56℃，均方误差为0．55oC。最终使用BP神经网络方法对传感器进行标定，使得传感解调系统满足了工业测高温的精度要求。

英文摘要：

In order to measure the high-temperature long-line in the industrial boiler flue gas environment, a high-temperature sensor based on single crystal sapphire was presented. In this paper, the temperature measurement principle based on the white light polarization interferometer theory was analyzed and Discrete Gap Transform demodulation algorithm was used to get the Optical Path Difference （OPD） information. On this basis, a calibration system with S thermelectric couple as reference was established to get the OPD-temperature samples. Then, the quadratic polynomial fitting method and the Back-Propagation （BP） neural network were adopted to fit curves and generalize separately. Experimental results indicate that the BP method is preciser than the cubic polynomial fitting method, for its residual error mean is 0.33℃. Moreover, the BP method has strong generalization capability, for its generalization error mean is 0.56 ℃ and the MSE is 0.55℃. So the BP method was used for the sensor calibration. The results show that the sensor system can satisfy the precision requirements of industrial high-temperature measurement.