334. Derivative-Free Constrained Optimization in Hydraulics: Augmented Lagrangian Method with BOBYQA
Invited abstract in session TB-1: Zeroth-Order Optimization Methods for Stochastic and Noisy Problems, stream Zeroth and first-order optimization methods.
Tuesday, 10:30-12:30Room: B100/1001
Authors (first author is the speaker)
| 1. | Fabio Fortunato Filho
|
| Department of Applied MathematicsIMECC, The State University of Campinas (UNICAMP) | |
| 2. | José Mario MartÃnez
|
| Dept. Applied Mathematics, University of Campinas |
Abstract
This work proposes a Derivative-Free Safeguarded Augmented Lagrangian method for solving constrained optimization problems. The method was developed with a solid theoretical foundation and applied to the estimation of the hydraulic coefficient in river flow modeling, a problem widely studied in the literature. Additionally, an analysis of the neighborhood of each point generated in the minimization of subproblems was conducted to evaluate the behavior of each solver.
The channel modeling was performed using the Saint-Venant equations, applied to the East Fork River, with data collected over 31 days. These equations were solved using the finite difference method with artificial diffusion, chosen for its computational efficiency given the need to solve the equations repeatedly during the optimization process.
To minimize the error between observed and simulated data, derivative-free methods, such as PRIMA and BOBYQA, were tested in a box-constrained optimization problem.
The results demonstrate that the proposed approach is efficient and viable for all tested methods, reducing computational time and providing accurate estimates of the hydraulic coefficient. The study also highlights the advantages of optimization over traditional methods, pointing to potential future improvements and the integration of the method into a project compatible with HEC-RAS software, which is widely used for this type of problem.
Keywords
- Derivative-free optimization
- Black-box optimization
- Applications of continuous optimization
Status: accepted
Back to the list of papers