TY - JOUR
T1 - Evaluation of hole quality in drilling CF/BMI composite via machine learning
T2 - Multi-defects analysis and fatigue life prediction
AU - Zhang, Shengguo
AU - Wang, Wenhu
AU - Zhang, Tianren
AU - Xiong, Yifeng
AU - Huang, Bo
AU - Jiang, Ruisong
N1 - Publisher Copyright:
© 2025 Elsevier Ltd
PY - 2025/7
Y1 - 2025/7
N2 - CF/BMI (carbon fiber/bismaleimide) composite has emerged as one of the most promising of the high-performance carbon fiber-reinforced polymer (CFRP) composites owing to its excellent mechanical strength and high temperature resistance. Although drilling in CF/BMI composite is a common machining procedure, the BMI resin tends to become brittle after curing, which means that drilling holes can lead to serious defects such as delamination, tearing, and scratches on the hole-wall. In this paper, to predict fatigue life under multi-defects and calculate the sensitivity of defects to fatigue life, a machine learning model for comprehensively evaluating the quality of drilled holes was proposed. Firstly, the quantitative characterization methods of tearing, burr, and delamination in uniform dimensions were presented, and the hole-wall surfaces were sampled and characterized with three-dimensional surface roughness. The comparative effect of tool type, ultrasonic-vibration assisted drilling (UVAD), and drilling parameters on defects was analyzed. Through quasi-static tensile and fatigue tests, the effects of multi-defects on the mechanical properties of open-hole laminates were investigated. An ANN model was developed to predict the correlation between drilling-induced defects and fatigue life. The model was optimized by tuning parameters and hyperparameters, the accuracy error of the model was a MAPE value of 1.263 % and an R2 value of 0.913.
AB - CF/BMI (carbon fiber/bismaleimide) composite has emerged as one of the most promising of the high-performance carbon fiber-reinforced polymer (CFRP) composites owing to its excellent mechanical strength and high temperature resistance. Although drilling in CF/BMI composite is a common machining procedure, the BMI resin tends to become brittle after curing, which means that drilling holes can lead to serious defects such as delamination, tearing, and scratches on the hole-wall. In this paper, to predict fatigue life under multi-defects and calculate the sensitivity of defects to fatigue life, a machine learning model for comprehensively evaluating the quality of drilled holes was proposed. Firstly, the quantitative characterization methods of tearing, burr, and delamination in uniform dimensions were presented, and the hole-wall surfaces were sampled and characterized with three-dimensional surface roughness. The comparative effect of tool type, ultrasonic-vibration assisted drilling (UVAD), and drilling parameters on defects was analyzed. Through quasi-static tensile and fatigue tests, the effects of multi-defects on the mechanical properties of open-hole laminates were investigated. An ANN model was developed to predict the correlation between drilling-induced defects and fatigue life. The model was optimized by tuning parameters and hyperparameters, the accuracy error of the model was a MAPE value of 1.263 % and an R2 value of 0.913.
KW - Carbon fiber/bismaleimide (CF/BMI) composites
KW - Drilling-induced defects
KW - Fatigue life
KW - Machine learning
KW - UVAD
UR - http://www.scopus.com/inward/record.url?scp=86000726966&partnerID=8YFLogxK
U2 - 10.1016/j.tws.2025.113189
DO - 10.1016/j.tws.2025.113189
M3 - 文章
AN - SCOPUS:86000726966
SN - 0263-8231
VL - 212
JO - Thin-Walled Structures
JF - Thin-Walled Structures
M1 - 113189
ER -