随机共振与神经网络结合的针刺材料强力预测  

Needle Punching Nonwoven Strength Prediction on Combination of SR and Neural Network

作　　者：邓辉[1,2] 周浩[1] 杨森[1] 朴锦兰 钱晓明[1,2] 

机构地区：[1]天津工业大学纺织学院,天津300387 [2]天津工业大学先进纺织复合材料教育部重点实验室,天津300387

出　　处：《测试技术学报》2016年第4期277-283,共7页Journal of Test and Measurement Technology

基　　金：国家自然科学基金资助项目(51005168);天津市应用基础与前沿技术研究计划重点资助项目(15JCZDJC38500);纺织工程国家重点学科优秀青年教师计划资助项目(13-06-01)

摘　　要：采用最优多级随机共振方法对针刺非织造材料生产过程中采集到的原始数据进行信息预处理,以RBF神经网络为训练方法,通过分析对针刺非织造材料生产有影响的各种因素,选择合适的输入层参数,完成对针刺非织造材料强力的预测.通过涤纶和锦纶两种纤维实验证明：经过最优多级随机共振预处理的RBF网络预测模型明显优于没有最优多级随机共振预处理的预测模型,相对误差缩小了1.7~9倍,大大提高了预测准确度.The optimal multistage stochastic resonance is used to preprocess the raw data gathered in needle punching nonwoven production process,RBF neural network is selected as the training method and at the same time the appropriate input layer parameters is selected too,through analyzing a variety of the influential factors of the needle punching nonwovenproduction,strength performance of needle punching nonwoven is predicted.Experiments which the polyester and nylon fibers are usedshow that the RBF network prediction model through optimal multistage stochastic resonance pretreatment is significantly better than the prediction model without optimal multistage stochastic resonance pretreatment,and the relative error is reduced by 1.7to 9times.This method can greatly improve the prediction accuracy.

关 键 词：针刺非织造材料 强力性能 RBF神经网络 最优多级随机共振 

分 类 号：TS171.9[轻工技术与工程—纺织材料与纺织品设计]