摘要
针对压铸机压射时压射速度与插装阀阀芯位移量关系理论模型误差大的问题,提出了一种基于时间序列数据驱动的压铸机压射速度系统的机理模型,利用长短期记忆(Long short-term memory, LSTM)神经网络处理时间序列的优势,建立了压射速度与阀芯位移量的内在联系,通过小波阈值去噪的方法降低了噪声对模型精度的影响。所建立模型的均方根误差为0.336%,相关系数R^(2)为0.998,可较准确地预测阀芯位移量。通过与理论公式模型的计算结果进行对比,该模型预测结果的平均误差为0.11%,优于理论公式模型。
Aiming at the problem of great error in the theoretical model of the relationship between injection speed and cartridge valve spool displacement during injection of die casting machine,a mechanism model of injection speed sys⁃tem of die casting machine based on time series data-driven was proposed.Taking advantage of the long short-term memory(LSTM)neural network to process the time series,the intrinsic relationship between the injection speed and the displacement of the valve spool was established,and influence of noise on the model accuracy was reduced by the wavelet threshold denoising method.The root mean square error(RMSE)of the established model is 0.336%and the correlation coefficient(R^(2))is 0.998,which can accurately predict the displacement of valve spool.By comparing with the calculation results of theoretical formula model,the average error of the prediction results of the model is 0.11%,which is superior to the theoretical formula model.
作者
钟建辉
娄军强
冯光明
胡奖品
刘亚刚
彭文飞
余军合
ZHONG Jianhui;LOU Junqiang;FENG Guangming;HU Jiangpin;LIU Yagang;PENG Wenfei;YU Junhe(College of Mechanical Engineering and Mechanics,Ningbo University,Ningbo 315211;Ningbo L.K.Tech-nology Co.,Ltd.,Ningbo,315806)
出处
《特种铸造及有色合金》
CAS
北大核心
2024年第7期995-1000,共6页
Special Casting & Nonferrous Alloys
基金
国家重点研发计划资助项目(2022YFB3706800)
宁波市重点研发计划暨“揭榜挂帅”项目(2023Z037)。
关键词
压铸机
压射系统
LSTM
数据驱动
小波分解
Die Casting Machine
Injection System
LSTM
Data-driven
Wavelet Decomposition