Intelligent Recognition of Artificial Joint Wear Debris by Radar Graph Fractal Feature Prior Convolutional Network

doi:10.3901/JME.260296

Abstract

Abstract: Artificial joint wear debris generate from the frictional motion of joint pairs, carrying mount of micro scale wear information. However, most of the artificial joint wear debris research is mainly by manual recognition, which lacks objective evaluation criteria. Classification and recognition of wear debris belongs to the image recognition field, intelligent classification such as convolutional neural network may solve the drawbacks of manual analysis. Therefore, radar graph fractal feature prior convolutional network is constructed to realize the intelligent recognition of artificial joint wear debris. Firstly, fractal dimension values of wear debris are calculated by the improved radar method, and the fractal dimension is used as the fixed characteristic of the wear debris to perform prior judgment. The judgement results are imported into the convolutional neural network models AlexNet and deep residual network (ResNet) to recognize the wear debris class. The models are trained in the wear debris picture dataset collected from literatures, and the generalization ability of the models is verified in the experimental dataset. The data augmentation method is also employed to expand the dataset and balance the uneven distribution among debris classes. Moreover, to compare the performance difference between the convolutional neural networks and ordinary machine learning models, a support vector machine (SVM) model is also used to classify the debris classes for the training data. The results showed that the radar graph fractal dimension had a good correlation with the shape distribution of the wear debris. The recognition accuracy and precision of convolutional neural network models are significantly higher than that of SVM. The fractal dimension prior judgment and data augmentation can obviously improve the recognition accuracy of the original model. The recognition accuracy of models based on AlexNet and ResNet networks reach 83.70% and 89.26%, respectively, which is higher than 80.74% and 82.59% of the AlexNet and ResNet original models of networks. The recognition accuracy models based on AlexNet and ResNet networks on the validation data are 86.60% and 88.66% respectively, which are close to the models on the training data, showing good generalization ability.

Key words: artificial joint wear debris, radar graph fracture dimension, convolutional neural network, intelligent identification

CLC Number:

TH117
R684

ZHANG Weipeng, GU Yongqing, LUO Yong, QU Luping. Intelligent Recognition of Artificial Joint Wear Debris by Radar Graph Fractal Feature Prior Convolutional Network[J]. Journal of Mechanical Engineering, 2026, 62(9): 254-267.

Add to citation manager EndNote|Reference Manager|ProCite|BibTeX|RefWorks

URL: http://www.cjmenet.com.cn/EN/10.3901/JME.260296

http://www.cjmenet.com.cn/EN/Y2026/V62/I9/254

References

[1] 刘玉梅. 载阿仑膦酸钠超高分子量聚乙烯磨屑的生物学反应研究[D]. 成都：西南交通大学，2011. LIU Yumei. Study on biological response to wear particles of ultra high molecular weight polyethylene loaded with alendronate sodium[D]. Chengdu： Southwest Jiaotong University，2011.
[2] KHAN M，KUIPER J，RICHARDSON J. Can cobalt levels estimate in-vivo wear of metal-on-metal bearings used in hip arthroplasty?[J]. Proceedings of the Institution of Mechanical Engineers Part H-Journal of Engineering In Medicine，2007，221(H8)：929-942.
[3] YAN X，XU X，SHENG C，et al. Intelligent wear mode identification system for marine diesel engines based on multi-level belief rule base methodology[J]. Measurement Science and Technology，2018，29(1)：015110.
[4] SONG J，CHEN F，LIU Y，et al. Insight into the wear particles of PEEK and CFRPEEK against UHMWPE for artificial cervical disc application：Morphology and immunoreaction[J]. Tribology International，2020，144：106093.
[5] 朱玉苹. 不同人工关节材料配副的滑动摩擦腐蚀及磨损颗粒研究[D]. 徐州：中国矿业大学，2015. ZHU Yuping. Study on sliding friction and corrosion behaviors and wear particles of different joint materials combinations[D]. Xuzhou：China University of Mining and Technology，2015.
[6] 崔文. CoCrMo/UHMWPE磨损颗粒的分形表征及形态分类研究[D]. 徐州：中国矿业大学，2017. CUI Wen. Study on fractal characterization and morphological classification of CoCrMo/UHM-WPE wear debris[D]. Xuzhou：China University of Mining and Technology，2017.
[7] ZHANG D，LIU H，WANG J，et al. Wear mechanism of artificial joint failure using wear debris analysis[J]. Journal of Nanoscience and Nanotechnology，2018，18(10)：6805-6814.
[8] 付松卿，王松，陈凯，等. 人工膝关节不同体外磨损周期的UHMWPE磨屑特征及细胞毒性研究[J]. 摩擦学学报，2023，43(9)：986-994. FU Songqing，WANG Song，CHEN kai，et al. Characterization and cytotoxicity of UHMWPE Wear debris from different wear simulating cycles of the artificial knee joint in vitro[J]. Tribology，2023，43(9)：986-994.
[9] 刘洪涛. 人工关节磨屑表征及磨屑生物反应研究[D]. 徐州：中国矿业大学，2008. LIU Hongtao. Description to wear debris of artificial joint and its bioreaction[D]. Xuzhou：China University of Mining and Technology，2008.
[10] 王松，付松卿，田徐腾越，等. 人工关节CoCrMo-UHMWPE配副在不同接触应力下的摩擦、磨损和磨屑特征[J]. 材料保护，2022，55(12)：33-39，69. WANG Song，FU Songqing，TIAN Xutengyue，et al. Friction，wear and debris characteristics of artificial joint CoCrMo-UHMWPE pair under different contact stresses[J]. Materials Protection，2022，55(12)：33-39，69.
[11] MYSHKIN N，KOVALEV A. Adhesion and surface forces in polymer tribology-A review[J]. Friction，2018，6(2)：143-155.
[12] 张峻铭，杨伟东，李岩. 人工智能在复合材料研究中的应用[J]. 力学进展，2021，51(4)：865-900. ZHANG Junming，YANG Weidong，LI Yan. Application of artificial intelligence in composite materials[J]. Advances in Mechanics，2021，2021，51(4)：865-900.
[13] 王伟华，殷勇辉，王成焘. 基于磨粒分析的设备状态监测技术[J]. 润滑与密封，2003(4)：22-24，26. WANG Weihua，YIN Yonghui，WANG Chengtao. The condition monitoring technology based on wear debris[J]. Lubrication Engineering，2003(04)：22-24，26.
[14] PENG Y，CAI J，WU T，et al. WP-DRnet：A novel wear particle detection and recognition network for automatic ferrograph image analysis[J]. Tribology International，2020，151：106379.
[15] 陈海燕，王成国. 分形理论及其在摩擦学研究中的应用[J]. 材料导报，2002，16(12)：6-8. CHEN Haiyan，WANG Chengguo. Fractal theory and its application in tribological research[J]. Materials Reports，2002，16(12)：6-8.
[16] GE S，CHEN G，ZHANG X. Fractal characterization of wear particle accumulation in the wear process[J]. Wear，2001，251：1227-1233.
[17] LIU H，GE S. Description to wear debris boundaries by radar graph fractal method[J]. Chinese Science Bulletin，2007(16)：2285-2290.
[18] LI J，WANG S，FU S，et al. Study on wear debris characteristics of different polyethylene under reciprocating sliding and compound motion modes[J]. Wear，2023，530：205062.
[19] WANG Q，LIU J，GE S. Study on biotribological behavior of the combined joint of CoCrMo and UHMWPE/BHA composite in a hip joint simulator[J]. Journal of Bionic Engineering，2009，6(4)：378-386.
[20] STRATTON-POWELL A，WILLIAMS S，TIPPER J，et al. Mixed material wear particle isolation from periprosthetic tissue surrounding total joint replacements[J]. Journal of Biomedical Materials Research Part B-Applied Biomaterials，2022，110(10)：2276-2289.
[21] STRATTON-POWELL A，WILLIAMS S，TIPPER J L，et al. Isolation and characterisation of wear debris surrounding failed total ankle replacements[J]. Acta Biomaterialia，2023，159：410-422.
[22] MATSUMAE G，KIDA H，TAKAHASHI D，et al. Determination of optimal concentration of vitamin E in polyethylene liners for producing minimal biological response to prosthetic wear debris[J]. Journal of Biomedical Materials Research Part B-Applied Biomaterials，2022，110(7)：1587-1593.
[23] 张涛. 人工关节材料的混合磨屑分离提取方法及磨损机理研究[D]. 徐州：中国矿业大学，2021. ZHANG Tao. Research on isolation method of mixed debris and wear mechanism of artificial joint materials[D]. Xuzhou：China University of Mining and Technology，2021.
[24] VISENTIN M，STEA S，SQUARZONI S，et al. A new method for isolation of polyethylene wear debris from tissue and synovial fluid[J]. Biomaterials，2004，25(24)：5531-5537.
[25] 贾玉梅. 超声场作用下人工关节磨屑提取及分离研究[D]. 徐州：中国矿业大学，2015. JIA Yumei. Study on the extraction and isolation of artificial joint wear debris under ultrasonic irradiation[D]. Xuzhou：China University of Mining and Technology，2015.
[26] NINE M，CHOUDHURY D，HEE A，et al. Wear debris characterization and corresponding biological response：Artificial hip and knee joints[J]. Materials，2014，7(2)：980-1016.
[27] 黄传辉. 人工髋关节的磨损行为及磨粒形态研究[D]. 北京：中国矿业大学，2004. HUANG Chuanhui. Study about wear behavior of prostheses and characterization of wear particles[D]. Beijing：China University of Mining & Technology，Beijing，2004.
[28] 魏慧慧，刘洪涛，曹希传，等. 人工关节磨屑的提取及表征[J]. 中国组织工程研究，2012，16(26)：4776-4780. WEI Huihui，LIU Hongtao，CAO Xichuan，et al. Isolation and description of the artificial joint wear debris[J]. Chinese Journal of Tissue Engineering Research，2012，16(26)：4776-4780.
[29] 周磊. 人工关节用超低磨损聚乙烯生物相容性、摩擦学性能与磨屑表征研究[D]. 北京：北京协和医学院，2014. ZHOU Lei. Study on biocompatibility，tribological property and wear debris characterization of ultra-low-wear polyethelene as artificial joint materials[D]. Beijing：Peking Union Medical College，2014.
[30] LIU H，GE S，CAO S，et al. Comparison of wear debris generated from ultra high molecular weight polyethylene in vivo and in artificial joint simulator[J]. Wear，2011，271(5-6)：647-652.
[31] YAMAMOTO K，CLARKE I C，MASAOKA T，et al. Microwear phenomena of ultrahigh molecular weight polyethylene cups and debris morphology related to γ radiation dose in simulator study[J]. Journal of Biomedical Materials Research，2001，56(1)：65-73.
[32] AFFATATO S，FERNANDES B，TUCCI A，et al. Isolation and morphological characterization of UHMWPE wear debris generated in vitro[J]. Biomaterials，2001，22(17)：2325-2331.
[33] MABREY J，AFSAR-KESHMIRI A，ENGH G，et al. Standardized analysis of UHMWPE wear particles from failed total joint arthroplasties[J]. Journal of Biomedical Materials Research，2002，63(5)：475-483.
[34] HU X，SONG J，LIAO Z，et al. Morphological residual convolutional neural network (M-RCNN) for intelligent recognition of wear particles from artificial joints[J]. Friction，2022，10(4)：560-572.
[35] 张长桃. 织构化金属—高分子材料摩擦副磨损机理的研究[D]. 南京：南京航空航天大学，2017. ZHANG Changtao. Study on the wear mechanism of friction pair made of textured metal and polymer[D]. Nanjing：Nanjing University of Aeronautics and Astronautics，2017.
[36] WU J，YAN X，JIN Z，et al. Application of principal component analysis and a fuzzy C-means clustering algorithm to wear debris morphology classification[J]. Proceedings of the Institution of Mechanical Engineers Part J-Journal of Engineering Tribology，2009，223(J7)：1059-1066.
[37] KANDEL S，SU S，HALL R，et al. An automated system for polymer wear debris analysis in total disc arthroplasty using convolution neural network[J]. Frontiers in Bioengineering and Biotechnology，2023，11：1108021.
[38] BIAN Y，ZHOU L，ZHOU G，et al. Study on biocompatibility，tribological property and wear debris characterization of ultra-low-wear polyethylene as artificial joint materials[J]. Journal of the Mechanical Behavior of Biomedical Materials，2018，82：87-94.
[39] MORO T，TAKATORI Y，KYOMOTO M，et al. Wear resistance of the biocompatible phospholipid polymer-grafted highly cross-linked polyethylene liner against larger femoral head[J]. Journal of Orthopaedic Research，2015，33(7)：1103-1010.
[40] SUNER S，GOWLAND N，CRAVEN R，et al. Ultrahigh molecular weight polyethylene/graphene oxide nanocomposites：Wear characterization and biological response to wear particles[J]. Journal of Biomedical Materials Research Part B-Applied Biomaterials，2018，106(1)：183-190.
[41] NAKANISHI Y，NAKASHIMA Y，FUJIWARA Y，et al. Influence of surface profile of Co-28Cr-6Mo alloy on wear behaviour of ultra-high molecular weight polyethylene used in artificial joint[J]. Tribology International，2018，118：538-546.
[42] MARKHOFF J，ZIETZ C，FABRY C，et al. Influence of analyzed lubricant volumes on the amount and characteristics of generated wear particles from three different types of polyethylene liner materials[J]. Journal Of Biomedical Materials Research Part B-Applied Biomaterials，2018，106(3)：1299-1306.
[43] RICHARDS L，BROWN C，STONE M，et al. Identification of nanometre-sized ultra-high molecular weight polyethylene wear particles in samples retrieved in vivo[J]. Journal of Bone and Joint Surgery-British Volume，2008，90B(8)：1106-1113.
[44] PLUMLEE K，SCHWARTZ C. Investigating UHMWPE wear mechanisms by decomposing wear debris distributions[J]. Wear，2011，271(9-10)：2208-2212.
[45] SIEVING A，WU B，MAYTON L，et al. Morphological characteristics of total joint arthroplasty-derived ultra-high molecular weight polyethylene (UHMWPE) wear debris that provoke inflammation in a murine model of inflammation[J]. Journal of Biomedical Materials Research Part A，2003，64A(3)：457-464.
[46] LIU T，WOOD W，LI B，et al. Effect of reinforcement on wear debris of carbon nanofiber/high density polyethylene composites：Morphological study and quantitative analysis[J]. Wear，2012，294：326-335.
[47] LIU T. Graphitic polymer nanocomposites：Wear performance and wear debris analysis[D]. Washington State：Washington State University School，2014.
[48] SHANBHAG A，JACOBS J，GLANT T，et al. Composition and morphology of wear debris in failed uncemented total hip replacement[J]. The Journal of Bone and Joint Surgery British Volume，1994，76(1)：60-67.
[49] VISENTIN M，STEA S，SQUARZONI S，et al. Isolation and characterization of wear debris generated in patients wearing polyethylene Hylamer inserts，gamma irradiated in air[J]. Journal of Biomaterials Applications，2005，20(2)：103-121.
[50] HYDE P J，TIPPER J，FISHER J，et al. Wear and biological effects of a semi-constrained total disc replacement subject to modified ISO standard test conditions[J]. Journal of the Mechanical Behavior of Biomedical Materials，2015，44：43-52.
[51] LIU H，GE S. Description to wear debris boundaries by radar graph fractal method[J]. Chinese Science Bulletin，2007，52(16)：2285-2290.
[52] ZHANG X，ZHANG T，CHEN K，et al. Wear mechanism and debris analysis of PEEK as an alternative to CoCrMo in the femoral component of total knee replacement[J]. Friction，2023，11(10)：1845-1861.
[53] IOFFE S，SZEGEDY C. Batch normalization：Accelerating deep network training by reducing internal covariate shift[C]//Proceedings of the 32nd International Conference on Machine Learning，Lille，FRANCE，2015.
[54] ZOU K，O'MALLEY A，MAURI L. Receiver-operating characteristic analysis for evaluating diagnostic tests and predictive models[J]. Circulation，2007，115(5)：654-657.
[55] SUN K，LIU X，CHEN G，et al. Wer debris recognition and quantification in ferrography images by instance segmentation[J]. Tribology Transactions，2022，65(3)：508-518.
[56] PENG Y，YUE H，Wang S，et al. Probability-weighted ensemble support vector machine for intelligent recognition of moving wear debris from joint implant[J]. Tribology International，2023，186：108583.