激光技术, 2010, 34 (2): 173, 网络出版: 2010-05-06
基于人工神经网络脉冲激光强化镀层形貌预测
Prediction of the pattern of electroless deposit after pulse laser heating via artificial neural network
激光技术 人工神经网络模型 脉冲激光强化 化学复合镀 laser technique artificial neural network model pulse laser hardening electroless deposit
摘要
为了探索脉冲激光强化镀层的规律,采用误差反向传播神经网络对脉冲激光参量与镀层形貌(强化层深度、宽度及熔化状态)之间的关系进行建模,并选取带动量的自适应学习率算法对网络进行改进,以增加网络稳定性,提高训练速度与精度。结果表明,该网络模型对激光处理后镀层形貌的预测值与实际值接近,其相对误差在±8.33%以内,可以有效地对激光强化镀层形貌进行预测。该方法为探索脉冲激光强化镀层的规律提供了一条新的途径。
Abstract
A model of the relationship between pulse-laser parameters and the pattern of electroless deposit composite coatings (taking into account hardened width,depth and melting state) with a back propagation neural network was constructed in order to explore the theoretical principles underlying pulse-laser reinforcement of plating coatings.The momentum-adaptive learning rate algorithm was selected to increase network stability,training speed and accuracy.The appearance of composite coating was effectively predicted with ±8.33% relative error.This method is a new way of exploring the theoretical principles of pulse-laser coating-reinforcement.
张文博, 张群莉, 姚建华. 基于人工神经网络脉冲激光强化镀层形貌预测[J]. 激光技术, 2010, 34(2): 173. ZHANG Wen-bo, ZHANG Qun-li, YAO Jian-hua. Prediction of the pattern of electroless deposit after pulse laser heating via artificial neural network[J]. Laser Technology, 2010, 34(2): 173.