改进遗传神经网络在传感器温度补偿中的应用  

Application of Improved Genetic Neural Network in Sensor Temperature Compensation

作　　者：唐炜[1] 徐晓苏[1] 

机构地区：[1]东南大学仪器科学与工程学院

出　　处：《电子测量与仪器学报》2008年第1期62-66,共5页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金资助项目(编号:60374046)

摘　　要：精确辨识传感器温度补偿模型对于提高系统测试精度具有重要的意义。神经网络具有良好的自学习、自适应和非线性映射能力,但往往训练速度慢、易陷入局部极小值,而遗传算法具有很强的全局寻优能力,但其局部搜索能力却不足。本文探讨了利用改进遗传算法优化函数链神经网络,以获得全局最优解的方法,并根据多温度条件下的实测数据,对电涡流传感器温度补偿模型进行了有效辨识。结果表明,该方法运算快速、精度高、通用性强,在智能传感器建模与补偿等领域具有良好的应用前景。Precise identification of sensor temperature compensation model is significant to improve its measurement accuracy. Neural network has the abilities of self-learning, self-adaptation and non-linear mapping, but its convergence speed is slow and it is liable to fall into some local minimum. Genetic algorithm has high global optimization ability, but its local search ability is weak. In this paper, improved genetic algorithm （IGA） is adopted to optimize the functional link neural network （FLNN） in order to find global optimization by using their respective merits, then the temperature compensation model of eddy current sensor was identified effectively based on the practical test data under multi-temperature conditions. Test result shows that the proposed method has the advantages of fast calculation, high accuracy and high universality, which has extensive application prospect in modeling and compensation fields of smart sensors.

关 键 词：传感器 温度补偿 改进遗传算法 函数链神经网络 

分 类 号：TP212.6[自动化与计算机技术—检测技术与自动化装置] TH701[自动化与计算机技术—控制科学与工程]