TR2021-128

Simulation Failure-Robust Estimation of Black-Box Model Parameters via Bayesian Optimization


    •  Chakrabarty, A., Bortoff, S.A., Laughman, C.R., "Simulation Failure-Robust Estimation of Black-Box Model Parameters via Bayesian Optimization", IEEE International Conference on Systems, Man, and Cybernetics, October 2021.
      BibTeX TR2021-128 PDF
      • @inproceedings{Chakrabarty2021oct2,
      • author = {Chakrabarty, Ankush and Bortoff, Scott A. and Laughman, Christopher R.},
      • title = {Simulation Failure-Robust Estimation of Black-Box Model Parameters via Bayesian Optimization},
      • booktitle = {IEEE International Conference on Systems, Man, and Cybernetics},
      • year = 2021,
      • month = oct,
      • url = {https://www.merl.com/publications/TR2021-128}
      • }
  • MERL Contacts:
  • Research Areas:

    Machine Learning, Multi-Physical Modeling, Optimization

Abstract:

Advances in modeling and computation have resulted in high-fidelity digital models capable of simulating the dynamics of a wide range of industrial systems. These models often require calibration, or the estimation of an optimal set of parameters in some goodness-of-fit sense, to reflect a system's observed behavior. These models are often not designed for the efficient solution of the calibration problem, however, and the application of existing methods often results in repeated model evaluations over parameters that can cause the simulations to be unreasonably slow or fail altogether. In general, the shape of these parameter regions that result in simulation failure is unknown. In this paper, we propose a novel failure-robust Bayesian optimization (FR-BO) algorithm that learns these failure regions from online simulations and informs a Bayesian optimization algorithm to avoid failure regions while searching for optimal parameters. This results in acceleration of the optimizer's convergence and prevents wastage of time trying to simulate parameters with high failure probabilities. The effectiveness of the proposed failure-robust Bayesian optimization algorithm is demonstrated via a well-known system that exhibits numerical stiffness and a second, practical example related to parameter estimation for buildings with integrated HVAC.