A Hybrid Method of Theoretical Model and Machine Learning for Predicting Friction Force Between Metal Rod and Carbon Fiber for Manufacturing Three‐Dimensional ( 3D ) Composite Preforms

ABSTRACT In the process of replacing the guide with yarn in flexible‐guided 3D weaving, the frictional force of guide extraction exhibits unpredictable magnitude. This hinders the design of the automated device, reduces the progress efficiency, and restricts the universality of product application. This study developed a hybrid theory‐GBR (gradient boosting regressor) model for friction force prediction during the process of replacing the guide with yarn. The theoretical model accurately explained the mechanism between the friction force and the structure parameters (the number of fabric layers and fabric layer density), and the theoretical model was validated through experiments. GBR model was trained to capture the nonlinear residual between the theoretical friction force and the experimental friction force. Training data were obtained from experiments. The training method was leave‐one‐out cross‐validation. The prediction performance of the hybrid theory‐GBR model was compared with that of machine learning models, GBR and support vector regression, and a theoretical model. The results showed that the experimental results were consistent with the theoretical model that friction force increased linearly with the number of layers and nonlinearly with fiber volume fraction. The hybrid theory‐GBR model achieved better performance in predicting the pull‐out friction force. It combines the advantages of both theoretical and data‐driven approaches. This research provides a new method for predicting the friction force of guide rod removal, and it also provides data support for flexible‐guided 3D weaving.