Reliability-based design of tapered composite tubes under torsional loadings: An ANN-assisted approach

The performance of composite load-bearing structures is highly sensitive to uncertainties in material properties, emphasizing the need for design-for-reliability approaches. Traditional structural reliability methods are computationally intensive. To address this, a novel methodology combining Artificial Neural Networks, Monte Carlo simulation, and the Finite Element Method is proposed to evaluate the reliability of tapered composite tubes under torsional loading. The finite element method is used to determine the torsional response of the tapered composite tube, while an approximate analytical model based on Donnell-Mushtari-Vlasov shell theory calculates the angle of twist, validating the finite element results. The propagation of material property uncertainty across micro-, meso-, and macro-scales is quantified. The influence of laminate stacking sequence and taper angle on uncertainty propagation and mechanical response is analyzed. Reliability-based angle of twist values are determined using the proposed approach, and its accuracy is benchmarked against direct Monte Carlo simulation. Probability density functions are used to quantify structural reliability. Results indicate that stacking sequence significantly affects reliability, while taper angle has a relatively smaller impact. Designs based on mean material properties offer only 50% reliability, equivalent to a 50% probability of failure, highlighting the necessity of incorporating material property variability into composite tube design.