基于EMD与IHHO-LSTM的神经网络时延预测方法  

Neural Network Delay Prediction Method Based on EMD and IHHO-LSTM

作　　者：游达章[1,2] 袁野 张业鹏 YOU Dazhang;YUAN Ye;ZHANG Yepeng(College of Mechanical Engineering,Hubei University of Technology;Hubei Province Key Laboratory of Modern Manufacturing Quality Engineering)

机构地区：[1]湖北工业大学机械工程学院 [2]湖北省现代制造质量工程重点实验室

出　　处：《仪表技术与传感器》2025年第3期108-115,共8页Instrument Technique and Sensor

摘　　要：基于Ether CAT控制多轴系统存在的同步误差问题,时延预测对于提高Ether CAT网络中主从同步精度十分重要。Ether CAT总线传输时延具有不确定性和非线性等特性,针对该问题,提出了基于经验模态分解(EMD)与多策略哈里斯鹰(IHHO)的长短期记忆神经网络(LSTM)时延预测方法。首先,将时延序列利用EMD算法处理,以降低时延序列的非平稳性;其次,针对哈里斯鹰种群初始分布不均匀的问题,增加改进的ICMIC混沌映射来提高种群分布多样性,并引入反S型能量控制因子,平衡算法的全局搜索和局部开发能力,修正寻优模型;由于随机向量的不确定性,采用衰减机制更新随机向量,得到规则的向量变化;最后,利用IHHO-LSTM模型对分解的固有模态函数建模,并将各分量的预测值叠加得到最终的预测值。实验结果表明:文中提出的EMD-IHHO-LSTM时延预测模型相较于传统HHO-LSTM模型在MAE,RMSE,MAPE分别提高了58.39%,69.85%,69.66%,可有效预测网络时延,验证了所提模型的有效性和优越性。Time delay prediction is crucial for improving the master-slave synchronization accuracy in Ether CAT networks due to the synchronization error problem in multi axis systems controlled by Ether CAT.The transmission delay of Ether CAT bus has characteristics of uncertainty and nonlinearity.To address this issue,a long short term memory(LSTM)delay prediction method based on empirical mode decomposition(EMD)and multi strategy Harris Hawk(IHHO)was proposed.Firstly,the delay sequence was processed using the EMD algorithm to reduce the non stationary of delay sequence.Secondly,to address the issue of uneven initial distribution in the Harris Eagle population,an improved ICMIC chaotic mapping was added to enhance population diversity,and an anti S-shaped energy control factor was introduced to balance the algorithm's global search and local development capabilities,and to modify the optimization model.Due to the uncertainty of random vectors,a decay mechanism was used to update the random vectors and regular vector changes were obtained.Finally,the IHHO-LSTM model was used to model the decomposed intrinsic mode function,and the predicted values of each component were superimposed to obtain the final predicted value.The experimental results show that the EMD-IHHO-LSTM delay prediction model proposed in this paper improves MAE,RMSE,and MAPE by 58.39%,69.85%,and 69.66%respectively compared to the traditional HHO-LSTM model,and can effectively predict network delay,verifying the effectiveness and superiority of the proposed model.

关 键 词：Ether CAT控制系统 时延预测 改进ICMIC混沌映射 能量控制因子 衰减机制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]