遗传神经网络在GMI传感器设计中的应用

 巨磁阻抗（GMI）微磁传感器具有灵敏度高、响应速度快等突出优点，但其输出信号呈高度非线性特性。利用交流偏置方法产生非对称巨磁阻抗效应（AGMI），对磁场传感器的线性度有一定改善，但仍存在线性范围小、线性误差较大的缺点。BP神经网络具有良好的自学习、自适应和非线性映射能力，但通常训练速度较慢、易陷入局部极小值；遗传算法有很强的全局寻优能力，但其局部搜索能力不足。为充分发挥二者优点，本研究提出一种基于遗传神经网络的传感器非线性误差校正方法，并针对所设计的GMI传感器，设计了适合本系统的遗传神经网络，可通过Matlab软件实现。结果表明，经过训练的网络输出结果有序，网络的非线性映射性能良好，能精确反映该传感器系统的函数关系。该方法运算快速、精度高，对智能GMI传感器的设计具有一定工程应用价值。

Application of Genetic Neural Network in GMI Sensor Design

The GMI sensor enjoys many advantages, such as high sensitivity, fast response, but its response characteristics are highly nonlinear. Although by introducing ac bias, with the AGMI effect, the degree of sensor's linearity can be improved to some extent, the linear range and error are still not satisfactory. The BP neural network has the abilities of self-learning, self-adaptation and non-linear mapping, but its convergence is slow and it is easy to fall into a local minimum. Genetic algorithm has a high global optimization ability, but its local search ability is weak. To give full play to the advantages of the two methods, a genetic neural network is proposed to solve the problem of non-linear correction in sensor systems, and according to the designed GMI sensor, using the software of Matlab, we have implemented the designed genetic neural network. Test result shows that the trained network has an ordered data structure and good nonlinear mapping properties, which can accurately reflect the function relation of the sensor system. The proposed method has the advantages of fast calculation and high precision, which may find important applications in designing smart GMI sensors.