Parametric sensitivity through optimization under uncertainty approach

by

Parametric sensitivity through optimization under uncertainty approach

Kishalay Mitra

Department of Chemical Engineering, Indian Institute of Technology Hyderabad, Yeddumailaram, Hyderabad 502205, India.

DOI:

https://doi.org/10.7494/cmms.2013.1.0418

Abstract:

While mimicking a physical phenomenon in a computational framework, there are tuning parameters quite often present in a computational model. These parameters are generally tuned with the experimental data to capture the process behavior as close as possible. Any optimization study based on this model assumes the values of these tuning parameters as constant. However, it is known that these parameters are subjected to inherent source of uncertainties such as errors in measurement or model tuning etc. for which they are not tuned for. Assuming these parameters constant for rest of the optimization is, therefore, not realistic and one should ideally check the sensitivity of these parameters on the final results. In this study, we are going to use approach based on the paradigm of optimization under uncertainty that allows a decision maker to carry out such an analysis. Additionally, this study captures the tradeoff between solution quality and solution reliability that is captured here using non-dominated genetic algorithm II. The generic concept has been applied on a grinding process model and can be extended to any other process model.

Cite as:

Mitra, K. (2013). Parametric sensitivity through optimization under uncertainty approach. Computer Methods in Materials Science, 13(1), 107 – 112. https://doi.org/10.7494/cmms.2013.1.0418

Article (PDF):

Keywords:

Optimization, Uncertainty, Parameter sensitivity, Grinding, Genetic algorithm, Multi-objective optimization, Pareto

References: