A novel method for predicting delamination of carbon fiber reinforced plastic (CFRP) based on multi-sensor data. (August 2021)

• Record Type: Journal Article
• Authors: Cui, Jiacheng; Liu, Wei; Zhang, Yang; Gao, Changyong; Lu, Zhe; Li, Ming; Wang, Fuji
• Is Part Of: Mechanical systems and signal processing. Volume 157(2021)
• Publication Date: 2021-08
• Subjects: Multisensor measurement -- Delamination evaluation -- Machine learning -- In situ prediction -- Drilling of CFRP; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2021.107708
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A method for predicting delamination of carbon fiber reinforced plastic (CFRP). The force, torque, temperature, vibration during drilling and hole exit images were employed. Statistical delamination factor Fs for quantification of delamination damage. Singular spectrum analysis (SSA) to eliminate randomness. Experimental verification of the proposed delamination factor Fs and the prediction model. Abstract: Carbon fiber reinforced plastic (CFRP) has been widely used in many fields such as in the aerospace and automotive industries. Drilling of CFRP is a key process in the manufacture of CFRP components. The existing quality control and tool change decision methods are mainly based on delamination damage. However, estimating delamination damage in situ is still a challenge in the process of continuous drilling. To solve this problem, a comprehensive delamination prediction method based on multi-sensor data is proposed in this paper. In process of the drilling, the force, torque, temperature, vibration and hole exit images were collected, and the delamination was quantified by a proposed statistical delamination factor F s . Singular spectrum analysis (SSA) is used to smooth the F s sequence to reduce randomness. Then, a XGBoost-ARIMA model is constructed for rolling prediction of F s . Finally, drilling experiments were carried out to verify the effectiveness of the proposed method. The experimental results showed that compared with traditional delamination evaluation factors, F s reduced the mean square error (MSE) of prediction by more than 50%. Compared with that of traditional machine learning models such as an SVM and ANN, the MSE of the model's regression part is decreased by more than 39%. The proposed method can provide a solution for real-time and in situ prediction of delamination damage in the continuous drilling process of CFRP components.