A Nonlinear Interval Programming Model based on Fuzzy Neural Networks

Compared to traditional deterministic problem research methods, interval analysis methods have a larger computational load and longer computational time due to the presence of uncertainty. Therefore, for models with complex structures or large scales, there is a lot of uncertain information, which requires a significant amount of computation and reduces the efficiency of the algorithm. The current optimization algorithms also have such shortcomings. In order to improve algorithm efficiency, this article focuses on researching parallel computing methods for interval optimization, designing efficient parallel algorithms, decomposing optimization problems into multiple subtasks that can be parallelized, and analyzing the feasibility and effectiveness of these algorithms in detail. The results indicate that multiple artificial neural network models are used to establish mapping relationships between design variables and uncertain variables, corresponding objective function values, and constraint function values. The objective function and each constraint function are trained with one network each, and a heuristic algorithm is selected as the inner and outer layer optimization solver.