摘要
针对压力传感器对温度存在交叉灵敏度这一具体问题,常采用BP神经网络对其进行数据融合。但BP神经网络方法训练收敛速度慢,易陷入局部最优。采用PSO全局优化算法训练多层前向神经网络权值,使网络训练误差比BP方法降低了两个数量级,并且收敛速度明显加快。融合结果表明基于PSO神经网络方法更有效地消除了温度对压力传感器的影响,显著提高了传感器的稳定性和准确度。
BP neural network is usually adopted to solve the problem of intercross sensitivity of pressure sensor to temperature, but its training speed is too slow and its results are local optimized results. The global optimization called PSO algorithm is applied to train the weights of neural multi-layer forward neural network, which makes the network training error two orders lower than the method of BP in quantitatively. And the results of the data fusion show that the method of neural network based on PSO algorithm could remove the influence of temperature on pressure sensor effectively. In addition, the stability and accuracy of the sensor are improved greatly.
出处
《传感技术学报》
CAS
CSCD
北大核心
2006年第4期1284-1286,1289,共4页
Chinese Journal of Sensors and Actuators
关键词
神经网络
PSO算法
传感器
数据融合
neural network
PSO（particle swarm optimizer）
sensor
data fusion
