压力传感器数据融合算法研究

Algorithmic Study on Data Fusion of Pruessure Sensor

传感器存在的交叉灵敏度,运用BP网络对数据进行融合,可以有效降低非目标参量对输出特性的影响.MATLAB中提供了多种针对BP网络的训练方法,运用其中的变梯度法、拟牛顿法和LM法分别对压力传感器进行数据融合,比较其结果可以得出:LM算法可以有效地克服局部收敛,更好地提高压力传感器的稳定性和可靠性.

Sensor is the core parts in apparatus and measure system, is the most tache in process control. Transducer＇s output, which is affected by surroundings,directly effect the whole capability. It is significant to improve the sensor＇s precision. BP neural network is usually used to reduce or eliminate the errors caused by temperature cross sensitivity of pressure sensor. Many algorithms are offered by MATLAB to train BP. Conjugate gradient backpropagation （CGBP）, Quasi-Newton algorithms and levenberg-marquardt algorithm （LM algorithm） are common to train BP. They can handle data perfectly. Levenberg-marquardt algorithms is the best algorithm by comparing the three algorithms.