Application of Artificial Intelligence for Theoretical Prediction of Mechanical Properties in Filament-Wound Composite Tubes

This paper presents a theoretical and experimental framework for predicting the tensile strength of composite tubes manufactured using filament winding technology, by applying Design of Experiments (DoE) and simulated analysis based on artificial intelligence (AI) principles. In the experimental phase, eight composite tube specimens were produced and tested, following a full factorial DoE matrix with three technological parameters: winding speed, fiber tension, and winding angle. Tensile strength was measured for each specimen using standardized mechanical testing procedures. Based on the obtained results, a simulated AI prediction was conducted, proposing theoretical tensile strength values within ±5% variation, as a conceptual representation of a possible machine learning model (e.g., Random Forest or Artificial Neural Network). The purpose of this simulation is to demonstrate the potential of AI-based models to predict mechanical properties for new combinations of input parameters—without the need for additional physical testing. The findings suggest that, with a valid DoE-based dataset, machine learning could significantly reduce the number of experiments, time, and resources required for the development and optimization of composite structures.

Keywords: Composite tubes, Filament winding, Artificial intelligence, Design of experiments, Tensile strength