摘要
针对传统纱线质量的正演、反演模型中存在收敛速度慢、精度低等问题,以及标准粒子群算法存在陷入局部极值的缺陷,提出一种粒子群遗传混合算法。使用该算法优化BP神经网络的权值和阈值并建立纱线条干正演模型。在此基础上,以纱线条干CV值为对象构建了粒子群遗传算法反演模型;使用历史生产数据对生产过程参数进行反演。结果表明:各生产过程参数反演结果的平均相对误差均低于4%。认为:该反演方法具有较高的可行性与准确性。
In order to solve the problems of slow convergence and low precision in the traditional forward and inversion models of yarn quality,and the defect of local extremum in standard particle swarm optimization algorithm,a particle swarm genetic hybrid algorithm was proposed.The algorithm was used to optimize the weights and thresholds of BP neural network and establish the forward yarn evenness model.On this basis,the inverse model of particle swarm genetic algorithm was constructed based on the CV value of yarn evenness.Historical production data was used to inverse production parameters.The results showed that the average relative error of the inversion results of each production parameter was kept below 4%.It is considered that the inversion method is feasible and accurate.
作者
梁棋
张立杰
LIANG Qi;ZHANG Lijie(Xinjiang University,Urumqi,830017,China)
出处
《棉纺织技术》
CAS
2024年第6期1-7,共7页
Cotton Textile Technology
基金
新疆维吾尔自治区科技重大专项(2022A01008-1)。
关键词
粒子群算法
遗传算法
生产过程参数反演
纱线条干
BP神经网络
particle swarm optimization
genetic algorithm
production parameter inversion
yarn evenness
BP neural network
