基于神经网络的MEMS红外热电堆自校准响应分析

Self-calibration response analysis of MEMS infrared thermopile based on neural network

为克服MEMS红外热电堆原有自校准响应分析方法存在的考虑参数单一、故障覆盖率低、校准精度低等缺点,全面考虑多个相关参数影响,采用RHPNN神经网络提高其故障覆盖率,再用小波神经网络提高校准精度,最后用FPGA实现了该算法。与传感器结合实验结果表明,本方案故障覆盖率达到92%,自校准后的测温绝对误差降到0.03 K。

In order to overcome the shortcomings of the original self-calibration response analysis method of MEMS infrared thermopile,such as considering the single parameter,low fault coverage,and low calibration precision,in this paper,the RHPNN neural network is used to improve the fault coverage,and then the wavelet neural network is used to improve the calibration precision. Finally,the algorithm is implemented by FPGA. The experimental results show that the fault coverage of this scheme is 92%,and the temperature measurement error of the sensor is reduced to 0. 03 K after self-calibration.