An interval uncertainty optimization algorithm based on radial basis function network differentiation

A new interval uncertainty optimization algorithm is proposed to replace two-layer nested optimization, owing to the low efficiency of the latter. The radial basis function network is established to obtain the first-order differential, which is difficult to achieve in practical engineering problems. The results obtained by this network differential method are verified by a mathematical example. The network differential method is combined with the interval perturbation method to compute the bounds of uncertain objective functions and constraints, and the subinterval method is introduced to address the large level of uncertainty. The example of a compression spring shows the feasibility of this interval analysis method. The interval uncertain optimization problem is transformed into a deterministic one through the interval order relationship and probability model, and solved using the genetic algorithm or non-dominated sorting genetic algorithm-II. A numerical example and electromagnetic buffer model demonstrate the accuracy, efficiency and practicability of this new method.