面向工业场景的碳纤维质量评估模型研究

打开文本图片集

Research on carbon fiber quality evaluation model for industrial applications

WANG Ziqi 1 ，LIUYiyin²，ZHAOHongyuan'，LIU Yu*1 ，KANG Peng'，ZHANG Juan *2 （1.Beihang University，Beijing 100080；2.AVIC Composite Co.，Ltd.，Beijing 101300）

Abstract：This studyproposesacarbonfiberqualityevaluation methodbasedonnanoindentationtestdataandanintegrated machinelearning model.Through nanoindentation experiments，indentation morphologycharacteristics（includingcarbon fiberdiameterand indentation pointpositions）and load-displacementcurve mechanical parameters （covering indicators such as elasticdeformationworkandhardness）ofcarbon fiberfilaments wereobtained.Adatasetcomprising12multidimensional physicalfeatures wasconstructed，withqualitativecompliance labeledaspositive/negativeclasses toestablishamapping relationship betweencarbon fiberfeatures andquality.This feature sampling method，incorporatingsample preparationinformation，issuitableforengneeringapplications.Duringmodeldevelopment，anutomatedmachinelearning framework （Auto Gluon）wasemployed toconductautomatedalgorithmsearchingand hyperparameteroptimization，ultimatelyyieldingareliablecarbon fiberqualityassessmentmodel.Theresultsdemonstrate thattheintegrated modelachievesaclasificationaccuracyofO.74onthetestset，validating itsefectivediscriminationcapabilityforcarbonfiberquality.Through feature mportanceanalysis，thecarbonfiberdiameter，hardnes，andspacinguniformitywereidentifiedasthecorequalityindicatorsinfluencingmodeldecisions.Thresholdsensitivityanalysisenableddynamicadjustmentofthemodel'sdecisionboundaries，allowing the asssment framework toadapt todiffrentiated scenariossuch asaerospace and civilian manufacturing.This approach provides a reliable basis for quality control and process optimization of carbon fiber products

Keywords： carbon fiber; nanoindentation testing;machine learning; integrated mode

1引言

碳纤维增强复合材料凭借其优异的比强度、比模量及耐腐蚀性能，已成为航空航天、高端装备制造等战略领域不可或缺的关键材料[1-3]。（剩余10442字）