基于分段融合的压阻式传感器温度补偿方法

Temperature Compensation Method of Piezoresistive Sensor Based on Piecewise Fusion

针对硅压阻式压力传感器在工程应用中受环境温度和压力的影响产生漂移,影响测量精度等问题。提出一种基于粒子群优化RBF神经网络与最小二乘法融合的温度补偿模型。使用粒子群算法对常规RBF神经网络的权值和阙值进行优化,提高神经网络的泛化性能和训练效率,增强传感器非线性段温度补偿的效果;使用最小二乘法对线性段进行温度补偿,提高整体模型的补偿效率。以飞思卡尔24 PC型压力传感器进行补偿实验,结果表明:对比优化前的神经网络和最小二乘方法,利用本文方法进行温度补偿,耗时短,总体误差低于其他两种方法。传感器在整个温度区间和压力测试点下的输出基本不受影响,补偿效果明显,数据精度符合课题实验的要求。

For the silicon piezoresistive pressure sensor in engineering applications by the ambient temperature and pressure to produce drift,affecting the measurement accuracy and other issues.A temperature compensation model based on improved particle swarm optimization RBF neural network and least squares method is proposed to realize compensation of the pressure sensor.The weights and thresholds of conventional RBF neural networks are optimized by particle swarm optimization algorithm.Improve the generalization performance and training efficiency of neural network,and enhance the effect of nonlinear temperature compensation of sensor.Application of Freescale 24PC pressure sensor for temperature compensation studied.The results showed that comparison of not improved neural network and least square method,using this method for temperature compensation overall time-consuming and the average error is lower than the other two methods.The output of the sensor over the entire temperature range and the pressure test point is substantially unaffected,compensation effect is obvious,data precision meets the requirements of the experiment.