基于双权值神经网络的焊缝跟踪自适应控制设计  

Design of Welding Tracking Adaptive Controllor Based on Double-Weights Neural Network

作　　者：李华[1] 田国华[1] 

机构地区：[1]浙江广播电视大学萧山学院,浙江杭州311200

出　　处：《测控技术》2015年第6期74-76,共3页Measurement & Control Technology

摘　　要：基于双权值神经网络设计了一个模型参考自适应控制系统,以实现焊缝跟踪的自适应控制。通过分析双权值神经网络控制器在最陡下降法训练下的收敛和稳定条件,保证系统达到稳定运行的状态。最后仿真实验证明,由于存在方向权值和核心权值双权值的调整,基于双权值神经网络的焊缝跟踪控制系统的性能远优于BP网络控制系统的性能,可以更好地实现焊缝的实时跟踪控制。A model reference adaptive control system based on double-weights neural network is designed to re- alize welding tracking adaptive control. Analyzing the convergence and stability conditions of double-weights neural network controller, trained by the method of steepest descent, can ensure the system to achieve stable operation state. Finally, the simulation experiment shows that due to the presence of directional weight and core weight, performance of double-weights neural network welding tracking control system is much better than BP network control system, real time welding tracking can be realized better.

关 键 词：双权值神经网络 焊缝跟踪 模型参考自适应控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]