碳纤维质量控制AI模型研究

打开文本图片集

Research on carbon fiber quality control AI model

LIU Yiyin 1* ，ZHANG Juan1，LIU Yu² （1.AVIC Composite Co.，Ltd，Beijing 101300；2.Beihang University，Beijing 100080）

Abstract：Inthisstudy，acarbonfiberSEMimageclasification model was establishedbasedontheResNet-5O，whichautomaticallydistinguisheshigh-qualityand inferiorcarbonfibersamplesand providesasistanceforcarbonfiberqualitycontrol.The performanceof the model was improved byaplying specific image preprocesing techniques （forinstance： contrast enhancement and data enhancement）.The classification accuracy of the modelin thevalidation setreaches 98.8% ，indicatingthatthemodelcancapturethestructural characteristicsofcarbon fiberSEMimages.Inaddition，basedonthe transfer learningof the SEM images generated underdiferentoperating conditions，the clasificationaccuracyofthenew model in thevalidation set reaches 74.11% ，which proves that the model also has a certain robustness and generalization ability for the new dataset under different imaging conditions.

Keywords：carbonfiber;SEMimage;AI model

1引言

近年来，碳纤维预浸料及其复合材料因高强度、耐腐蚀、质量轻以及可设计性强等优点，在航空航天、轨道交通等领域广泛应用[1-3]，碳纤维原材料质量控制的重要性也大幅凸显。（剩余5292字）