Abstract
Probabilistic design methods may be used to account for inherent variability in the manufacturing quality of aerospace structures. Two different probabilistic approaches are compared alongside a deterministic method for the aeroelastic design of composite plate wings with uncertain ply orientations. The instability speed is found to be a discontinuous function of the lamination parameters, which are themselves functions of the ply orientations. A surrogate modeling approach is presented in which Gaussian processes are combined with support vector machines to emulate the discontinuous instability speed to efficiently propagate uncertainty through the model. The surrogate model is used to calculate the objectives of two optimization strategies: a reliability-based design, in which the probability of aeroelastic instability occurring within the design envelope is minimized, and a robust design, in which the mean and standard deviation of the instability speed are traded off. A genetic algorithm is used to optimize the layup, and reductions in the probability of failure of 94% are achieved in the reliability-based design using a 145 m/s design speed, relative to benchmark deterministic designs. A 74% reduction in the standard deviation is achieved in the robust design, at the expense of a 1%reduction in the mean.
Original language | English |
---|---|
Pages (from-to) | 3539-3552 |
Number of pages | 14 |
Journal | AIAA Journal |
Volume | 55 |
Issue number | 10 |
DOIs | |
Publication status | Published - 2017 |
Externally published | Yes |