摘要
比较遗传算法与神经网络的特点 ,并对将遗传算法用于函数连接型神经网络 (FLNN)的优点进行了研究 .对遗传算法的编码方法、交换和变异操作做了改进 ,提出了一种融合改进遗传算法的FLNN用于微硅加速度传感器动态性能补偿的新方法 .该方法不依赖于传感器的动态模型 ,可根据传感器的动态响应数据 ,建立补偿模型 ,采用改进遗传神经网络搜索和优化补偿模型参数 ,既保留了遗传算法的全局搜索能力 ,又具有神经网络的鲁棒性和自学习能力 .介绍补偿原理及算法 ,给出动态补偿网络的数学模型 .结果表明 ,该补偿方法能克服FLNN收敛速度慢、容易陷入局部极小的缺陷 ,具有网络训练速度快、实时性好、良好的全局搜索能力、精度高。
The characteristics of neural network (NN) and genetic algorithm (GA) are described. The advantages of the application of genetic algorithm to the functional link neural networks (FLNN) are discussed. The coding method and the operator of crossover and mutation for GA are improved. A kind of neural networks compensation method is proposed in which genetic algorithm and neural network are mixed for micro-silicon accelerometer. In this method, a dynamic compensation model can be set up according to measurement data of dynamic response of micro-silicon accelerometer without knowing its dynamic model. The dynamic compensation model parameters are trained by improved genetic neural network. So the method remains both the global searching ability of GA and the robustness and self-learning ability of NN. The compensation principle and algorithm are introduced and the mathematical model is founded. The results show that the proposed method can overcome the shortcoming of FLNN, such as slow speed in training and easy convergence to the local minimum points, and has the advantages of fast training process, good real time, good global searching ability, high precision, strong robustness and easy realization of dynamic compensation device.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2004年第4期455-458,共4页
Journal of Southeast University：Natural Science Edition
