Crashworthiness Prediction of Perforated Foam-Filled CFRP Rectangular Tubes Crash Box Using Machine Learning

Area of Science:

• Materials Science and Engineering
• Mechanical Engineering
• Computational Mechanics

Background:

• Carbon fiber-reinforced polymer (CFRP) tubes offer high specific strength and energy absorption, making them suitable for automotive crash boxes.
• Optimizing crashworthiness requires understanding the impact of design parameters like perforations and internal filling.
• Traditional experimental testing for crashworthiness is time-consuming and costly.

Purpose of the Study:

• To investigate the axial crashworthiness of rectangular CFRP tubes with varying hole configurations and polyurethane foam (PUF) filling.
• To evaluate the influence of hole diameter, number, and placement, as well as PUF filling, on crash performance.
• To assess the feasibility of using machine learning (ML) for predicting CFRP crash box performance to reduce experimental efforts.

Main Methods:

• Quasi-static axial compression tests were performed on designed CFRP tubes.
• Crashworthiness indicators, including initial peak force (P_ip), mean crushing force (P_m), and energy absorption (EA), were recorded.
• Multiple ML algorithms (DTR, LR, RR, LAR, ENs, MLP) were employed to predict crashworthiness indicators using experimental data.

Main Results:

• PUF-filled tubes exhibited significantly enhanced crashworthiness, with P_m and EA increasing nearly threefold compared to unfilled tubes.
• In unfilled tubes, holes had variable effects based on diameter and placement; in PUF-filled tubes, holes reduced performance.
• The Decision Tree Regressor (DTR) model demonstrated the highest prediction accuracy, with RMSE of 1251 and MAPE of 11.37%.

Conclusions:

• Polyurethane foam filling is crucial for enhancing the crashworthiness of CFRP tubes.
• Perforation design significantly impacts the crash performance of both filled and unfilled CFRP tubes.
• Machine learning models, particularly DTR, offer a viable and efficient approach for optimizing CFRP crash box designs.