摘要
针对经编机产能预测的高度非线性和不确定性,采用BP(back propagation)神经网络和RBF(radial basis function)神经网络对其进行求解。同时,针对BP神经网络的权值和阈值以及RBF网络隐节点中心向量值、基宽向量和连接权值难确定的问题,采用PSO(particle swarm optimization)算法和GA(genetic algorithm)对其优化求解,设计PSO-BP、GA-RBF和PSO-RBF等3种预测算法实现经编机产能的预测,并采用RMSE(root mean square error)值、MAPE(mean absolute percentage error)值和预测精度作为网络性能评价指标。试验结果显示,相比较PSO-BP和GA-RBF算法,PSO-RBF算法的各项网络性能指标均为最佳,其中预测精度达到了95.71%。因此,PSO-RBF算法在解决经编机产能预测问题上具有可行性、可靠性和优势性。
For the high nonlinearity and uncertainty of warp knitting machine productivity prediction,the BP(back propagation)neural network and RBF(radial basis function)neural network were used to solve the problem.At the same time,in view of the difficulty of determining the weights and threshold of BP neural network as well as the center vector,base width vector and connection weights of hidden nodes of RBF network,PSO(particle swarm optimization)and GA(genetic algorithm)were used to optimize the solution.Three prediction algorithms,namely PSO-BP,GA-RBF and PSO-RBF were designed to realize the prediction of warp knitting machine productivity.RMSE(root mean square error)value,MAPE(mean absolute percentage error)value and prediction accuracy were used as the network performance evaluation indicators.The test results show that compared with PSO-BP and GA-RBF algorithms,PSO-RBF algorithm has the best network performance indicators,among which the prediction accuracy reaches 95.71%.Therefore,the PSO-RBF algorithm is feasible,reliable and advantageous in solving the problem of warp knitting machine productivity prediction.
作者
刘诗敬
郗欣甫
侯曦
孙以泽
LIU Shijing;CHI Xinfu;HOU Xi;SUN Yize(College of Mechanical Engineering,Donghua University,Shanghai 201620,China;China Textile Machinery Association,Beijing 100028,China)
出处
《东华大学学报（自然科学版）》
CAS
北大核心
2021年第4期35-42,共8页
Journal of Donghua University(Natural Science)
基金
国家重点研发计划资助项目(2017YFB0309800)
东华大学青年教师科研启动基金资助项目(103-07-0053079)。
关键词
经编机
产能预测
神经网络
网络优化
warp knitting machine
productivity prediction
neural network
network optimization
