基于PCA和GA融合算法的湿度传感器校准实验研究  

Experimental study of humidity sensor calibrating based on fusion algorithm of PCA and GA

作　　者：陶佰睿[1,2,3] 李春辉[1] 苗凤娟[1] 张景林[3] 刘艳菊[3] 王欢[1] 郭琴[1] 

机构地区：[1]齐齐哈尔大学通信与电子工程学院,黑龙江齐齐哈尔161006 [2]中国科学院上海技术物理研究所红外物理国家重点实验室,上海200083 [3]齐齐哈尔大学计算中心,黑龙江齐齐哈尔161006

出　　处：《实验技术与管理》2015年第2期42-46,81,共6页Experimental Technology and Management

基　　金：黑龙江省高等教育教学改革项目(JG2012010671);黑龙江省高等教育教学改革项目(JG2013010025);齐齐哈尔大学教育教学研究项目(2014082)

摘　　要：针对湿度传感器因湿滞或温度漂移等因素引起的非线性问题,提出一种基于主成分分析（PCA）和遗传算法（GA）融合优化神经网络（BP）传感器校准改进措施。以HR202电阻型湿度传感器为例,分别在5个温度和10种饱和盐溶液标准湿度环境下进行测试,对采集的数据首先利用PCA算法＂降维＂,通过提取温度和湿滞补偿的主要信息,在原信息损失较小的前提下,将BP传感器校准多维问题简化,然后采用GA-BP对样本进行训练。经实验测试和Matlab平台仿真研究结果表明,基于PCA和GA融合算法处理的神经网络对湿度传感器定标校准数据相比该产品示值校准精度从±5%RH提高到±3%RH,训练速度相比传统BP方法提高3~5倍。该方法在湿度传感器校准测量精度提高方面有参考价值。To solve the nonlinear problems caused by the factors of humidity sensor hysteresis or temperature drift,a method of sensor calibration and improvement measures is proposed,using fusion algorithm of neural network（BP）optimized by principal component analysis（PCA）and genetic algorithm（GA）.Taking the HR202 humidity sensor as an example,the tests were made respectively in the 5temperature levels and 10 kinds saturated salt solution humidity environment.First,the measured data samples are reduced by the PCA algorithm.Then,the BP multidimensional sensor calibration is simplified with extracting the main information of temperature and humidity hysteresis compensation under the premise of original information loss smaller.Then the samples are trained by GA-BP.After experimental testing and simulating with MATLAB platform,the experimental results show that the neural network based on PCA and GA fusion algorithm for humidity sensor calibration data compared to product value,calibration precision is improved from ± 5%RH to ± 3%RH,the training speed compared with the traditional BP method is improved by 3-5 times.The method has an important reference value in improving the measurement accuracy in the humidity sensor calibration.

关 键 词：湿度传感器 主成分分析 遗传算法 神经网络 非线性校准 

分 类 号：TP212.1[自动化与计算机技术—检测技术与自动化装置]