自适应与附加动量BP神经网络的ECT流型辨识  

Identification of Flow Regimes Based on Adaptive Learning and Additional Momentum BP Neural Network for Electrical Capacitance Tomography

作　　者：王莉莉[1] 刘洪波[1] 陈德运[1] 冯其帅 

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2018年第1期105-110,共6页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(60572153;60972127);黑龙江省博士后资助项目(LBH-Z11109);黑龙江省青年科学基金(QC2012C059);黑龙江省教育厅科学技术研究项目(11541040;12511097)

摘　　要：传统BP神经网络是解决电容层析成像系统流型辨识经典的算法,虽然在一些简单问题上达到了工业实际应用的要求,但如果解决复杂工业问题时就会暴露出很多缺陷。针对传统BP神经网络算法的不足,为降低误差震荡现象,引入了自适应调节学习速率和附加动量因子。通过输入电容值进行训练,得到适合流型识别神经网络。仿真实验结果表明,该算法不仅继承传统BP神经网络的优点,而且还提高了ECT系统流型辨识中的收敛速度慢,解决了容易陷入局部极小值的问题。Traditional BP neural network is a typical mehtod to solve ECT system of flow pattern identification.It is applied to the simple problems in industrial applications,but there are many defects in solving complex industrial problems.In this paper based on the analysis of deficiency of BP neural network,for reducing the error oscillation,the adaptive learning rate adjustment factor and the additional momentum is introduced.In this method,the electrical capacitance values are input to train a network to identify the flow patterns.The simulation results show the algorithm not only inherits the advantages of traditional BP neural network,but also improve slow convergence and solve being prone to fall into local minimum problems in flow pattern identification of ECT system.

关 键 词：电容层析成像 流型辨识 BP神经网络 局部极小值 收敛速度 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]