基于RSM和SVR-IDBO的柔性压力传感单元灵敏度性能预测与优化  

Sensitivity performance prediction and optimisation of flexible pressure sensing unit based on RSM and SVR-IDBO

作　　者：唐颖 陈垲 何宋西莹 严少秋 雷建勇 林远长 何国田 TANG Ying;CHEN Kai;HE Songxiying;YAN Shaoqiu;LEI Jianyong;LIN Yuanchang;HE Guotian(School of Artificial Intelligence,Chongqing Institute of Green Intelligent Technology,Chinese Academy of Sciences,Chongqing 500109,China;School of Mechanical Engineering,Chongqing University of Technology,Chongqing 500113,China;The University of Melbourne,Victoria 3010,Australia;Chongqing Luban Institute of Robot Technology,Chongqing 500109,China;School of Mechanical Engineering,Chongqing University,Chongqing 500106,China;School of Mechanical Engineering,Chongqing Jiaotong University,Chongqing 500108,China)

机构地区：[1]中科院重庆绿色智能技术研究院人工智能学院,重庆500109 [2]重庆理工大学机械工程学院,重庆500113 [3]The University of Melbourne,Victoria 3010 [4]重庆鲁班机器人技术研究院,重庆500109 [5]重庆大学机械工程学院,重庆500106 [6]重庆交通大学机械工程学院,重庆500108

出　　处：《复合材料学报》2025年第2期885-899,共15页Acta Materiae Compositae Sinica

基　　金：国家重点研发计划(2018YFB1306601);重庆市在渝高校与中科院所属院所合作项目(HZ2021011);重庆市技术创新与应用发展专项(cstc2021jscx-cylhX0009)。

摘　　要：为优化柔性压力传感单元制备工艺,提升传感器灵敏度特性。本文通过建立灵敏度解析模型,确定了影响其性能表现的主要因素,采用机械共混的方式,通过调节碳纳米管(CNT)、多层石墨烯(MLG)、搅拌时间、成型温度等参数优化了柔性传感单元的灵敏度性能。首先在单因素分析的基础上,应用实验设计(DOE)中的中心复合实验方法(CCD)进行多因素实验设计,通过响应面法(RSM)和支持向量机(SVR)对多因素的交互影响进行了分析,并分别建立了灵敏度预测模型。其次根据决定系数(R^(2))、均方根误差(R_(mse))和平均误差率(M_(ae))对两种模型进行评估与定型,模型性能对比结果表明,通过超参优化后的SVR模型表现出更高水平的准确性和可预测性。然后基于改进的蜣螂优化算法(IDBO)对模型进行迭代优化,得到了比早期实验更好的灵敏度性能。仿真结果显示,在0~30 kPa的单轴压力下,当CNT含量为2.3wt%、MLG含量为1.9wt%、混合时间15 min、成型温度78℃时,灵敏度达到0.5512 kPa^(-1),经过实验验证,与实际灵敏度(0.5371 kPa^(-1))的相对误差为2.625%,且与同类型研究相比较,本文的传感单元灵敏度性能也处较高水平。证明该方法有助于寻找最佳的传感器含量配比与制备工艺,提升实验效率,节约实验成本,为快速制备高性能电容式柔性压力传感单元提供了新思路。To enhance the preparation process of flexible pressure sensors and improve their sensitivity characteristics,objective measures must be taken.This study identified the main factors affecting performance through a sensitivity analysis model.The sensitivity performance of the flexible sensing unit was optimized by adjusting the parameters of carbon nanotubes(CNTs),multilayered graphene(MLG),mixing time,and molding temperature using mechanical co-mingling.The study utilized the central combinatorial method(CCD)in design of experiments(DOE)for multi-factor experimental design.The interaction effects of the multi-factors were analyzed using response surface methodology(RSM)and support vector machine(SVR),based on single-factor analysis.Sensitivity prediction models were established accordingly.The two models were evaluated and finalized based on the coefficient of determination(R^(2)),root mean square error(R_(mse)),and mean error rate(M_(ae)).The results indicate that the SVR model,optimized by hyperparameterization,exhibits a higher level of accuracy and predictability.The model was then iteratively optimized using the improved dung beetle optimization(IDBO)algorithm,which yielded better sensitivity performance than earlier experiments.The simulation results show a sensitivity of 0.5512 kPa^(-1) at a uniaxial pressure of 0-30 kPa when the CNTs content is 2.3wt%,the MLG content is 1.9wt%,the mixing time is 15 min,and the molding temperature is 78℃.The experimentally verified relative error to the actual sensitivity(0.5371 kPa^(-1))is 2.625% and the sensitivity performance of the sensing unit in this study is also at a high level when compared with similar studies.It is proved that this method helps to find the optimal sensor content ratio and preparation process,improve the experimental efficiency,save the experimental cost,and provide a new idea for the rapid preparation of high-performance capacitive flexible pressure sensing units.

关 键 词：柔性压力传感器 灵敏度 机械共混 响应面法 支持向量机 蜣螂算法 

分 类 号：TP3-05[自动化与计算机技术—计算机科学与技术] TB332[一般工业技术—材料科学与工程]