A machine learning-based process operability framework using Gaussian processes

The objective in this work is to develop a machine learning-based framework for process operability using surrogate responses based on Kriging (also known as Gaussian Process Regression). Currently, the available approaches nonlinear systems are limited by problem dimensionality that they can address, not being computationally tractable high-dimensional systems. proposed approach will use Kriging-based models substitute developed first-principles or simulation-based models. built generate comparable terms of accuracy, while reducing computational effort. To achieve goal, systematic analysis highly nonlinear, large-dimensional at steady state developed. benchmarked against current methods and provides new direction field employing Two case studies associated with natural/shale gas conversion addressed illustrate effectiveness methods, namely membrane reactor direct methane fuels chemicals natural combined cycle power plant. It shown time calculations significantly decreased when approach, reductions up four orders magnitude, relative errors respect output below 0.3% worst-case scenario considering all cases. This thus contributes learning formulations algorithms enable improved design, operations manufacturing chemical energy