Scanning Electron Microscopes(SEMs)are widely used in experimental science laboratories,often requiring cumbersome and repetitive user analysis.Automating SEM image analysis processes is highly desirable to address th...Scanning Electron Microscopes(SEMs)are widely used in experimental science laboratories,often requiring cumbersome and repetitive user analysis.Automating SEM image analysis processes is highly desirable to address this challenge.In particle sample analysis,Machine Learning(ML)has emerged as the most effective approach for particle segmentation.However,the time-intensive process of manually annotating thousands of SEM images limits the applicability of supervised learning approaches.Self-Supervised Learning(SSL)offers a promising alternative by enabling knowledge extraction from raw,unlabeled data.This study presents a framework for evaluating SSL techniques in SEM image analysis,focusing on novel methods leveraging the ConvNeXtV2 architecture for particle detection.A dataset comprising 25,000 SEM images is curated to benchmark these proposed SSL methods.The results demonstrate that ConvNeXtV2 models,with varying parameter counts,consistently outperform other techniques in particle detection across different length scales,achieving up to a34%reduction in relative error compared to established SSL methods.Furthermore,an ablation study explores the relationship between dataset size and SSL performance,providing actionable insights for practitioners regarding model selection and resource efficiency.This research advances the integration of SSL into autonomous analysis pipelines and supports its application in accelerating materials science discovery.展开更多
Ferrograph-based wear debris analysis(WDA)provides significant information for wear fault analysis of mechanical equipment.After decades of offline application,this conventional technology is being driven by the onlin...Ferrograph-based wear debris analysis(WDA)provides significant information for wear fault analysis of mechanical equipment.After decades of offline application,this conventional technology is being driven by the online ferrograph sensor for real-time wear state monitoring.However,online ferrography has been greatly limited by the low imaging quality and segmentation accuracy of particle chains when analyzing degraded lubricant oils in practical applications.To address this issue,an integrated optimization method is developed that focuses on two aspects:the structural re-design of the online ferrograph sensor and the intelligent segmentation of particle chains.For enhancing the imaging quality of wear particles,the magnetic pole of the online ferrograph sensor is optimized to enable the imaging system directly observe wear particles without penetrating oils.Furthermore,a light source simulation model is established based on the light intensity distribution theory,and the LED installation parameters are determined for particle illumination uniformity in the online ferrograph sensor.On this basis,a Mask-RCNN-based segmentation model of particle chains is constructed by specifically establishing the region of interest(ROI)generation layer and the ROI align layer for the irregular particle morphology.With these measures,a new online ferrograph sensor is designed to enhance the image acquisition and information extraction of wear particles.For verification,the developed sensor is tested to collect particle images from different degraded oils,and the images are further handled with the Mask-RCNN-based model for particle feature extraction.Experimental results reveal that the optimized online ferrography can capture clear particle images even in highly-degraded lubricant oils,and the illumination uniformity reaches 90%in its imaging field.Most importantly,the statistical accuracy of wear particles has been improved from 67.2%to 94.1%.展开更多
基金funded by the German Research Foundation(DFG)under Project ID 390874152(POLiS Cluster of Excellence)N.J.S.,A.G.,G.C.,O.D.,A.J.were supported by the D2S2 program within the U.S.Department of Energy,Office of Basic Energy Sciences,Materials Sciences and Engineering Division under Contract No.DE-AC02-05-CH11231(D2S2 program,KCD2S2)A.G.acknowledges support from the Swiss National Science Foundation(SNSF,project#P500PN_222166).
文摘Scanning Electron Microscopes(SEMs)are widely used in experimental science laboratories,often requiring cumbersome and repetitive user analysis.Automating SEM image analysis processes is highly desirable to address this challenge.In particle sample analysis,Machine Learning(ML)has emerged as the most effective approach for particle segmentation.However,the time-intensive process of manually annotating thousands of SEM images limits the applicability of supervised learning approaches.Self-Supervised Learning(SSL)offers a promising alternative by enabling knowledge extraction from raw,unlabeled data.This study presents a framework for evaluating SSL techniques in SEM image analysis,focusing on novel methods leveraging the ConvNeXtV2 architecture for particle detection.A dataset comprising 25,000 SEM images is curated to benchmark these proposed SSL methods.The results demonstrate that ConvNeXtV2 models,with varying parameter counts,consistently outperform other techniques in particle detection across different length scales,achieving up to a34%reduction in relative error compared to established SSL methods.Furthermore,an ablation study explores the relationship between dataset size and SSL performance,providing actionable insights for practitioners regarding model selection and resource efficiency.This research advances the integration of SSL into autonomous analysis pipelines and supports its application in accelerating materials science discovery.
基金the National Natural Science Foundation of China(Nos.51975455,52105159 and 52275126)the China Postdoctoral Science Foundation(No.2021M702594)the Open Foundation of State Key Laboratory of Compressor Technology(Compressor Technology Laboratory of Anhui Province),No.SKL-YSJ202102.
文摘Ferrograph-based wear debris analysis(WDA)provides significant information for wear fault analysis of mechanical equipment.After decades of offline application,this conventional technology is being driven by the online ferrograph sensor for real-time wear state monitoring.However,online ferrography has been greatly limited by the low imaging quality and segmentation accuracy of particle chains when analyzing degraded lubricant oils in practical applications.To address this issue,an integrated optimization method is developed that focuses on two aspects:the structural re-design of the online ferrograph sensor and the intelligent segmentation of particle chains.For enhancing the imaging quality of wear particles,the magnetic pole of the online ferrograph sensor is optimized to enable the imaging system directly observe wear particles without penetrating oils.Furthermore,a light source simulation model is established based on the light intensity distribution theory,and the LED installation parameters are determined for particle illumination uniformity in the online ferrograph sensor.On this basis,a Mask-RCNN-based segmentation model of particle chains is constructed by specifically establishing the region of interest(ROI)generation layer and the ROI align layer for the irregular particle morphology.With these measures,a new online ferrograph sensor is designed to enhance the image acquisition and information extraction of wear particles.For verification,the developed sensor is tested to collect particle images from different degraded oils,and the images are further handled with the Mask-RCNN-based model for particle feature extraction.Experimental results reveal that the optimized online ferrography can capture clear particle images even in highly-degraded lubricant oils,and the illumination uniformity reaches 90%in its imaging field.Most importantly,the statistical accuracy of wear particles has been improved from 67.2%to 94.1%.
