利用响应面法优化人工神经网络的座椅频响函数预测模型与分析

打开文本图片集

中图分类号：TP393 文献标志码：A DOI：10.16385/j.cnki.issn.1004-4523.202311064

Abstract：Artificialeuralnetworkmodelinghasbeenprelimnarilyemployedtoinvestigateeffectsonthebiodynamicresponses. Inordertoevluatethevibrationtransmissoncharacteristicsoftheseatoupantsystemfurtherquantitativeresearchisded. Drawing fromalowfrequencyexperimentalinvestigationintowholebodyvibration，thisstudyisaimed todevelopanANNmodel withtheresponsesurface methodoptimization.Theage，stature，siting height，kneeheight，butock-to-knee，weight，gender， BMI，cushionthicknessandfrequencyareusedas network input toexplore thatthesehowtopredict transmissbilityfromtheseat baseto the seat pan.Based on the interaction between hyperparameters，the mapping relationshipbetween model hyperparameters and predictionperformance indexes wasestablished，andtheoptimalcombinationof hyperparameters wasoptimizedandobtained. Theresultssowthatthersonanefrequenciesitheverticalinlineandtheforeandaftcrosaxis ransmiibiliesfromsatbase toseatpandecreasedwithincreasingthicknessoffoamattheseatpan.BP-ANNmodelhas goodperformanceinestablishingthenonlinearrelationshipbetween theanthropometric，seat structurecharacteriticsandvibration transmissioncharacteristicsof seat-occupant system.Compared with BP-ANN model， the error of RSM-BP-ANN model is reduced by 25% and 18% respectively inpredictingvertialin-letransmsibilityandforanaftrosaxistrasmisslityfromseatbasetoeatpan.Andtslsopo videsanidea foradjustingtheparametersofneuralnetwork models toimprovethe predictionaccuracyofseat transmssibility.

Keywords：seat-occupant system;seat transmissibilities;aritificial neural network;response surface method交通运输环境中的各种振动会影响乘坐人员的 驾乘舒适性，高强度、长时间的振动暴露甚至会引岁健康问题[1]。（剩余14366字）